FOREWORD – 3

Foreword
The gap between rich and poor keeps widening. In the decades before the
Great Recession, economic growth benefited disproportionally higher income groups
while lower-income households were left behind. Since the crisis, disparities widened and
in many OECD countries inequality is today at its highest since data collection started.
This long-run increase in income inequality does not only raise social and political but
also economic concerns: income inequality tends to drag down GDP growth, and it is the
rising distance of the lower 40% from the rest of society which accounts for this effect.
Debates how to best curb this trend and promote opportunities for all have moved to the
top of the policy agenda in many countries.
This report is the third OECD flagship publication on trends, causes and remedies to
growing inequalities. The 2008 report Growing Unequal? documented and analysed the
key features and patterns of trends in income inequality in OECD countries. The 2011
publication Divided We Stand: Why Inequality Keeps Rising analysed the deep-rooted
reasons for rising inequality in advanced and most emerging economies. In It Together:
Why Less Inequality Benefits All highlights the key areas where inequalities originate
and where new policy approaches are required. It takes a fresh look at the question how
trends in inequality have affected economic growth and through which channels; looks
at the consequences of the recent period of crisis and fiscal consolidation on household
incomes; analyses the impact of structural labour market changes, such as rising nonstandard work, job polarisation and profound changes in women’s employment and
earnings; documents levels of wealth concentration and indebtedness; and discusses the
role for redistribution policies in OECD and emerging economies. The report also
discusses a range of promising policy practices to tackle high inequality and promote
equality of opportunities.
This report is the outcome of a collective effort with contributions from a team of
policy analysts largely from the OECD Social Policy Division of the Directorate for
Employment, Labour and Social Affairs. The overview Chapter 1 was prepared by the
entire team. Principal authors of the other chapters were Wen-Hao Chen (Social Policy
Division at the time of writing, currently Statistics Canada): Chapters 4 and 5;
Federico Cingano (Employment Analysis Division): Chapter 2; Christine Le Thi
(Statistics Directorate): Chapter 6; Ana Llena-Nozal (Social Policy Division): Chapters 4,
5, 6 and 7 (Section 7.2); Horacio Levy (Social Policy Division): Chapters 3, 4 and 7
(Section 7.2); Nora Lustig (Tulane University): Chapter 7 (Section 7.3); Fabrice Murtin
(Statistics Directorate): Chapter 6; Céline Thévenot (Social Policy Division): Chapter 7
(Section 7.2). Guillaume Cohen, Pauline Fron and Maxime Ladaique (all Social Policy
Division) contributed to all chapters and provided research assistance.
Michael Förster (Social Policy Division) led the team and co-ordinated the project
and the publication. Monika Queisser, Head of the OECD Social Policy Division,
supervised the preparation of this report and provided useful comments on various drafts.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4 – FOREWORD
Marlène Mohier prepared the manuscript for publication and Patrick Hamm contributed
to the editing of the report.
The analyses in this report rely partly on the OECD Income Distribution Database
prepared by national experts, many of whom have also provided advice on countryspecific results. They are too numerous to mention here but details can be found on the
OECD inequality website (www.oecd.org/social/income-distribution-database.htm). The
report makes use of many other data, in particular the micro data from the Luxembourg
Income Study (LIS) (www.lisproject.org) and a new set of data on household wealth
collected by the OECD Statistics Directorate. Discussions of data methodology and other
supporting material for this report can be found on the website
www.oecd.org/social/inequality-and-poverty.htm.
We are very grateful to Stefano Scarpetta and Mark Pearson, Director and Deputy
Director of Employment, Labour and Social Affairs at the OECD for their guidance and
extensive comments on various versions of the report. The report also benefited from
comments received by colleagues in and outside the OECD. We gratefully acknowledge
the many suggestions provided by members of the Working Party on Social Policy and
the Employment, Labour and Social Affairs Committee of the OECD as well as by
experts of the European Commission. The report also benefited from a close collaboration
with the OECD Statistics Directorate in the update and development of the Income
Distribution Database. We also thankfully acknowledge the comments and suggestions
provided by colleagues from OECD Directorates: the Development Centre, the
Economics Department, the Directorate for Education and Skills, the Directorate for
Employment, Labour and Social Affairs, the Public Governance and Territorial
Development Directorate and the Statistics Directorate. Finally, we are indebted to
Professors Brian Nolan, Ive Marx and Tim Smeeding for their comments and suggestions
on the first drafts of several chapters of this report.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 TABLE OF CONTENTS – 5

Table of contents
Acronyms and abbreviations ...................................................................................................... 13
Executive summary .................................................................................................................... 17
Chapter 1. Overview of inequality trends, key findings and policy directions ................... 21
1.1. Introduction and key findings ........................................................................................ 22
1.2. Inequality increased in good times, and it continued increasing in bad times ............... 25
1.3. Higher inequality drags down economic growth ........................................................... 28
1.4. Why increased non-standard work can lead to more inequality ..................................... 30
1.5. Tackling job and wage discrimination of women will boost growth and equality ......... 34
1.6. Wealth concentration widens disparities further and limits investment opportunities
for many ....................................................................................................................... 36
1.7. Designing sustainable policy packages to promote equality of opportunities ............... 38
Notes ..................................................................................................................................... 53
References ............................................................................................................................. 54
Annex 1.A1. Key indicators of income inequality and poverty ............................................. 57
Chapter 2. The impact of income inequality on economic growth ...................................... 61
2.1. Introduction and key findings ........................................................................................ 62
2.2. How inequality may affect economic growth ................................................................ 62
2.3. The impact of inequality on growth ............................................................................... 65
2.4. Inequality, social mobility and human capital accumulation ......................................... 73
2.5. Conclusion ..................................................................................................................... 81
Notes ..................................................................................................................................... 83
References ............................................................................................................................. 87
Annex 2.A1. Estimating the inequality, social mobility and growth nexus ........................... 93
Chapter 3. Income inequalities during the crisis and fiscal consolidation ........................ 103
3.1. Introduction and key findings ...................................................................................... 104
3.2. Income inequality continued to increase through the crisis ......................................... 105
3.3. Income poverty increased, no matter how it is measured ............................................ 111
3.4. Taxes and benefits cushioned the impact of the crisis on household income,
but mostly during the first years ............................................................................... 116
3.5. Tax-benefit measures during the crisis: A summary of detailed analyses
of ten countries ......................................................................................................... 118
3.6. Conclusion ................................................................................................................... 130
Notes ................................................................................................................................... 132
References ........................................................................................................................... 133

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6 – TABLE OF CONTENTS
Chapter 4. Non-standard work, job polarisation and inequality ....................................... 137
4.1. Introduction and key findings ...................................................................................... 138
4.2. A snapshot of non-standard work ................................................................................ 140
4.3. The role of non-standard employment in overall employment growth and job
polarisation ............................................................................................................... 145
4.4. Wage gaps between standard and non-standard workers and their impact
on the distribution of earnings .................................................................................. 154
4.5. The impact of non-standard work on household income and poverty ......................... 172
4.6. The role of taxes and benefits with regard to non-standard work ................................ 181
4.7. Conclusion ................................................................................................................... 190
Notes ................................................................................................................................... 193
References ........................................................................................................................... 196
Annex 4.A1. Measuring job polarisation ............................................................................. 201
Annex 4.A2. Data sources for Sections 4.4 and 4.5 ............................................................. 203
Annex 4.A3. Additional tables and figures .......................................................................... 205
Annex 4.A4. Tax-benefit simulations .................................................................................. 209
Chapter 5. Women, work and income inequality ................................................................ 211
5.1. Introduction and key findings ...................................................................................... 212
5.2. Trends in employment and earnings gaps between men and women .......................... 213
5.3. Trends in individual earnings inequality among male and female workers ................. 216
5.4. From individual to household earnings and the income distribution ........................... 219
5.5. Conclusion ................................................................................................................... 228
Notes ................................................................................................................................... 229
References ........................................................................................................................... 230
Annex 5.A1. Additional figures ........................................................................................... 232
Annex 5.A2. Decomposition of individual earnings inequality by subgroup ...................... 236
Annex 5.A3. A counterfactual decomposition analysis of changes
in income inequality ................................................................................................. 238
Chapter 6. How does the concentration of household wealth compare across
countries? ........................................................................................................................ 241
6.1. Introduction and key findings ...................................................................................... 242
6.2. The distribution of household wealth across OECD countries .................................... 243
6.3. Demographic characteristics of households with positive wealth ............................... 258
6.4. The composition of assets ............................................................................................ 262
6.5. The incidence of debt and over-indebtedness .............................................................. 268
6.6. Trends in household wealth since the Great Recession ............................................... 273
6.7. Conclusion ................................................................................................................... 276
Notes ................................................................................................................................... 278
References ........................................................................................................................... 281
Annex 6.A1. Methodological features of household wealth data ........................................ 284
Annex 6.A2. Long-term determinants of private wealth in eight OECD countries ............. 287
Annex 6.A3. Estimations of future pension wealth ............................................................. 290

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 TABLE OF CONTENTS – 7

Chapter 7. Inequality and fiscal redistribution in emerging economies ........................... 295
7.1. Introduction and key findings ...................................................................................... 294
7.2. Inequality trends and policy responses in emerging economies (EEs) ........................ 296
7.3. Fiscal policy and income redistribution in Brazil, Chile, Colombia, Indonesia,
Mexico, Peru and South Africa ................................................................................. 309
Notes ................................................................................................................................... 323
References ........................................................................................................................... 324
Annex 7.A1. The CEQ project and underlying household survey data ............................... 329
Annex 7.A2. Lambert’s conundrum ..................................................................................... 334
Tables
Table 1.A1.1. Key indicators on the distribution of household disposable income and poverty,
2007, 2011 and 2013 or most recent year ............................................................................ 58
Table 2.1. The negative impact of inequality on growth in OECD countries ............................ 68
Table 2.2. Inequality at the bottom and at the top of the income distribution ............................ 72
Table 2.A1.1. Years of schooling, family background and inequality ....................................... 98
Table 2.A1.2. Numeracy scores, family background and inequality ....................................... 100
Table 2.A1.3. Literacy scores, family background and inequality ........................................... 101
Table 3.1. Public budget deficits during the crisis ................................................................... 118
Table 3.2. Countries by category of fiscal consolidation .......................................................... 119
Table 3.3. Fiscal consolidation instruments reported by governments .................................... 121
Table 3.4. Summary assessment of growth and equity effects of fiscal consolidation
instruments ....................................................................................................................... 122
Table 3.5. Simulated overall effect of tax-benefit measures, 2008-13, by policy type ............ 125
Table 3.6. Simulated overall effect of tax-benefit measures by year ....................................... 126
Table 3.7. Simulated overall effect of tax-benefit measures by family composition, 2008-13 ... 127
Table 4.1. Estimates of (log) hourly wage gap between standard and non-standard workers,
pooled OLS: Men ............................................................................................................. 159
Table 4.2. Estimates of (log) hourly wage gap between standard and non-standard workers,
pooled OLS: Women ........................................................................................................ 160
Table 4.3. Fixed effects estimate of wage gaps ........................................................................ 162
Table 4.4. Change of employment contract and earnings mobility, average over
panel periods ..................................................................................................................... 163
Table 4.5. Distributional position of non-standard workers in quintiles of household
equivalent income, by quintile of individual earnings, pooled 15 EU countries, 2012 .... 174
Table 4.6. Statutory benefit differences between non-standard and non-standard work,
by benefit, 2010 ................................................................................................................ 183
Table 4.A1.1. Decomposition of changes in employment shares by task and sector,
EU-23 countries average, 1995/98-2010 .......................................................................... 201
Table 4.A3.1. Household income components by household work pattern,
pooled 15 EU countries, 2010 .......................................................................................... 208
Table 4.A3.2. Impact on Gini by income source ..................................................................... 208
Table 5.A1.1. Decomposition of individual earnings inequality (Theil index)......................... 237
Table 6.1. Ratios of household asset and liabilities between micro sources
and national accounts ....................................................................................................... 247
Table 6.2. Growth rate of asset prices and savings rate, for the periods 1970-2012
and 2000-13 ...................................................................................................................... 248
Table 6.3. Selected indicators of the distribution of household net wealth .............................. 250
Table 6.4. The joint distribution of income and wealth in the United States, 2013 ................. 256
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 8 – TABLE OF CONTENTS
Table 6.5. Household portfolio composition ............................................................................ 266
Table 6.6. Changes in mean net wealth between 2006 and 2012 ............................................. 274
Table 6.A1.1. Data sources and characteristics ........................................................................ 286
Table 6.A2.1. Long-term determinants of private wealth in eight OECD countries,
1970-2010 ......................................................................................................................... 288
Table 7.A2.1. Lambert’s conundrum ....................................................................................... 322
Figures
Figure 1.1. Income inequality varies greatly across OECD countries and emerging
economies ........................................................................................................................... 22
Figure 1.2. Lower and lowest incomes were increasingly left behind ....................................... 23
Figure 1.3. Income inequality increased in most OECD countries ............................................ 26
Figure 1.4. Taxes and benefits cushioned the effect of the crisis on household incomes .......... 27
Figure 1.5. The risk of income poverty has shifted from the elderly to the young .................... 27
Figure 1.6. Inequality lowers skills of the poor .......................................................................... 30
Figure 1.7. More than half of all jobs created since 1995 were non-standard jobs .................... 31
Figure 1.8. In most countries, the majority of temporary workers stay on such jobs ................. 34
Figure 1.9. In all OECD countries, women's employment put a brake on increasing inequality.... 36
Figure 1.10. Wealth is more concentrated at the top than income .............................................. 37
Figure 1.11. Half of all households have debts and one tenth is over-indebted ......................... 38
Figure 2.1. Inequality, enrolment and mobility across OECD countries ................................... 74
Figure 2.2. Probability of educational attainment by parent educational background (PEB)
and inequality ..................................................................................................................... 77
Figure 2.3. Average numeracy score by parent educational background (PEB) and inequality ... 79
Figure 2.4. Average numeracy score conditional on education by parent educational
background (PEB) and inequality .............................................................................................. 79
Figure 2.5. Probability of not being employed over working life by parent educational
background (PEB) and inequality ....................................................................................... 80
Figure 2.A1.1. Years of schooling by parent educational background (PEB) and inequality..... 99
Figure 2.A1.2. Literacy scores by parent educational background (PEB) and inequality ....... 100
Figure 3.1 Income inequality during the crisis ......................................................................... 106
Figure 3.2. Decomposition of changes in the Gini coefficient of labour income .................... 107
Figure 3.3. Changes in household real disposable income by income group ........................... 108
Figure 3.4. Income trends at the bottom, top and median of the distribution .......................... 109
Figure 3.5. Trends in household disposable income by income group over the last 25 years ..... 110
Figure 3.6. Changes in relative and anchored poverty rates during the crisis .......................... 112
Figure 3.7. Relative poverty rates by age group in 2007 and 2011 .......................................... 113
Figure 3.8. Changes in relative poverty among the elderly during the crisis ........................... 114
Figure 3.9. Changes and relative differences in household disposable income by age group ..... 115
Figure 3.10. Trends in taxes, benefits and market income during the crisis ............................ 117
Figure 3.11. Trends in taxes, benefits and market income across countries ............................ 117
Figure 3.12. Fiscal consolidation plans, 2009-15 ..................................................................... 120
Figure 3.13. Simulated overall effect of tax-benefit measures implemented in 2008-13,
by labour market circumstance and earning level ............................................................ 128
Figure 4.1. Share of non-standard employment by type, 2013 ................................................ 142
Figure 4.2. Share of women in non-standard employment, 2013 ............................................ 143
Figure 4.3. Incidence of non-standard employment by age group, 2013 ................................. 144
Figure 4.4. Incidence of non-standard employment by educational attainment, 2013 ............. 144
Figure 4.5. Incidence of non-standard employment by firm size, 2013 ................................... 145
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 TABLE OF CONTENTS – 9

Figure 4.6. Trends in shares of non-standard employment, shares in total employment,
1985-2013 ......................................................................................................................... 146
Figure 4.7. Employment growth by type of employment ........................................................ 148
Figure 4.8. Contribution of non-standard work/standard work to changes in employment
share by task, 1995/98-2010 ............................................................................................. 150
Figure 4.9. Percentage-point change in employment share due to non-standard work/standard
work by job decile, mid-1990s to 2010 ............................................................................ 152
Figure 4.10. Earnings ratio between standard and non-standard workers (standard workers = 1),
2012 .................................................................................................................................. 155
Figure 4.11. Influence of previous labour market status on the probability of having
a standard employment ..................................................................................................... 167
Figure 4.12. Non-standard work and earnings by decile of hourly wages, OECD-14 average ... 171
Figure 4.13. Share of non-standard workers as the main/secondary income earner
in a household ................................................................................................................... 173
Figure 4.14. Distributional position of non-standard workers in household income
quintiles, by quintile of individual earnings, 2012 ........................................................... 175
Figure 4.15. Household employment patterns and household equivalised income,
2012 or most recent year ................................................................................................... 176
Figure 4.16. Gini coefficient of equivalised household earnings for households
with different employment patterns, 2012 or most recent year ........................................ 177
Figure 4.17. Breakdown of household equivalent income inequality (Gini) by income
source, 2012 ...................................................................................................................... 178
Figure 4.18. Poverty rates of households for various household employment patterns,
2012 or most recent .......................................................................................................... 180
Figure 4.19. Distribution of poor households by household employment pattern,
2012 or most recent year ................................................................................................... 180
Figure 4.20. The impact of taxes and transfers on poverty reduction ...................................... 181
Figure 4.21 Differences in tax-benefit amounts between part-time/self-employment
workers and workers in standard jobs, 2010 .................................................................... 184
Figure 4.22. Tax-benefit adequacy ........................................................................................... 186
Figure 4.23. Participation tax rates for part-time workers ....................................................... 188
Figure 4.24. Participation tax rates for self-employed workers ............................................... 188
Figure 4.25. Transition tax rates from part-time to full-time employment .............................. 189
Figure 4.26. Transition tax rates from part-time to full-time self-employment ....................... 190
Figure 4.A3.1. The marginal effect of previous labour force/contract status
on the probability of transition to unemployment ............................................................ 205
Figure 4.A3.2. Effect of non-standard work on log hourly wages, by decile .......................... 207
Figure 4.A3.3. Household work patterns by standard work/non-standardemployment,
working households, 2010 or most recent year ................................................................ 208
Figure 5.1. Employment gender gap in 1992 and 2013 ........................................................... 214
Figure 5.2. Employment/population ratios (EPR) and full-time equivalent
employment rates (FTER), 15-64 years old, 2013 ........................................................... 214
Figure 5.3. Gender pay gaps over time .................................................................................... 215
Figure 5.4. Changes in annual earnings dispersion, workers with positive earnings ............... 216
Figure 5.5. Percentage changes in real annual earnings and annual hours worked
by earnings quintile .......................................................................................................... 217
Figure 5.6. Contribution to changes in individual earnings inequality by gender ................... 219
Figure 5.7. Female earnings as a share of total disposable household income,
mid-1980s/early 1990s – mid/late 2000s .......................................................................... 220
Figure 5.8. Changes in male/female labour participation among households .......................... 221
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 10 – TABLE OF CONTENTS
Figure 5.9. Changes in male/female employment intensity among households ...................... 222
Figure 5.10. Changes in male/female job skill nature among households ............................... 223
Figure 5.11. Wives’ employment rates, by husband’s earnings decile, OECD average .......... 224
Figure 5.12. Correlation between the earnings of husbands and wives, couple households
with at least one person working ...................................................................................... 224
Figure 5.13. Contribution to percentage point changes in Gini of household disposable
income, RIF decomposition results ................................................................................... 227
Figure 5.A1.1. Wives’ employment rates, by husband’s earnings decile ................................ 232
Figure 5.A3.1. Composition effect attributed to men’s work characteristics ........................... 239
Figure 6.1. Mean and median net wealth per household in micro data ..................................... 245
Figure 6.2. Mean and median net wealth relative to household income ................................... 246
Figure 6.3. Net wealth per household in micro data and national accounts .............................. 246
Figure 6.4. Median net wealth and change in house prices ....................................................... 249
Figure 6.5. Distribution of household disposable income and net wealth across deciles ........ 249
Figure 6.6. Ratio of mean and median net wealth per household in survey data ...................... 251
Figure 6.7. Wealth shares of top percentiles of the net wealth distribution .............................. 252
Figure 6.8. Wealth share of the top 10% and annual growth rate of real stock prices .............. 252
Figure 6.9. Mean net wealth and top wealth inequality ............................................................ 255
Figure 6.10. Income distributions of the bottom and top wealth quintiles
across OECD countries...................................................................................................... 257
Figure 6.11. Rank correlation between the distribution of net wealth and disposable income ..... 258
Figure 6.12. Mean net wealth per household by age of the reference person
and by broad component ................................................................................................... 260
Figure 6.13. Mean net wealth by education, housing status, household type
and main income source .................................................................................................... 261
Figure 6.14. Wealth composition and average net wealth by quintile in OECD countries ....... 263
Figure 6.15. Wealth composition and average net wealth for the top 10%, 5% and 1%
in OECD countries ............................................................................................................ 264
Figure 6.16. Percentage of indebted and over-indebted households ......................................... 269
Figure 6.17. Age composition of indebted and over-indebted households ............................... 270
Figure 6.18. Percentage of indebted and over-indebted households, by education................... 271
Figure 6.19. Percentage of indebted and over-indebted households by income quintile
and top distribution ............................................................................................................ 272
Figure 6.20. Depth of indebtedness for indebted households ................................................... 273
Figure 6.21. Change of mean net wealth between 2006 and 2012 ............................................ 274
Figure 6.A3.1. Net pension wealth for future retirees as a multiple of individual annual
gross earnings, OECD countries........................................................................................ 290
Figure 6.A3.2. Balance between public and private provision of mandatory pensions ............ 291
Figure 7.1. Gini coefficients in selected emerging economies and OECD countries, total
population ......................................................................................................................... 297
Figure 7.2. Size and composition of government budgets ....................................................... 301
Figure 7.3. Fiscal policy components and income inequality .................................................. 310
Figure 7.4. Inequality and redistribution, 2010 ........................................................................ 312
Figure 7.5. Redistribution and social spending, 2010 .............................................................. 313
Figure 7.6. Redistributive effect in Brazil, Chile, Colombia, Indonesia, Mexico, Peru,
South Africa, the European Union and the United States ................................................ 314
Figure 7.7. Fiscal policy components and income redistribution ............................................. 315
Figure 7.8. Concentration coefficients and budget shares for education and health ................ 317
Figure 7.9. Fiscal policy and reduction of extreme poverty ..................................................... 318
Figure 7.10. Winners and losers of the fiscal system ............................................................... 319
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 ACRONYMS AND ABBREVIATIONS – 11

Acronyms and abbreviations

BHPS

British Household Panel Survey

CEQ

Commitment to Equity

DFE

Dynamic fixed-effect

ECB

European Central Bank

EE

Emerging economies

EMTR

Effective marginal tax rate

EPL

Employment protection legislation

EPR

Employment/population ratio

ESA

European System of Accounts

EU-LFS

European Union Labour Force Survey

EU-SILC

European Union Statistics on Income and Living Conditions

EUROMOD

EU tax-benefit microsimulation model

FTER

Full-time equivalent rate

FTFY

Full-time full-year

GDP

Gross domestic product

GEPF

Government Employees Pension Fund (Africa’s largest pension
fund)

GMM

Generalised Method of Moments

GSOEP

German Socio-Economic Panel

HFCS

Household Finance and Consumption Survey

HILDA

Household, Income and Labour Dynamics in Australia

ICT

Information and communication technology

IDD

OECD Income Distribution Database

ISCED

International Standard Classification of Education

ISCO

International Standard Classification of Occupations

JHPS

Japan Household Panel Survey

KLIPS

Korean Labor & Income Panel Study

LIS

Luxembourg Income Study

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 12 – ACRONYMS AND ABBREVIATIONS
NSW

Non-standard work

OLS

Ordinary least square

PEB

Parent education background

PIAAC

OECD Programme for the International Assessment of Adult
Competencies

PIM

Perpetual Inventory Method

PISA

Programme for International Student Assessment

PIT

Personal income tax

PMG

Pooled mean group

PMR

Product market regulation

PPP

Purchasing power parity

PTFY

Part-time full-year

PTPE

Part-time permanent employment

PTR

Participation tax rate

PTTE

Part-time temporary employment

RIF

Recentered Influence Function

SCF

Survey of Consumer Finances

SES

Structure of Earnings Survey

SLID

Survey of Labour and Income Dynamics

SME

Small and medium-sized enterprise

SNA

System of National Account

SW

Standard work

TBM

OECD tax-benefit model

TE

Temporary work

TTR

Transition tax rate

UQR

Unconditional quantile regression

VAT

Value-added tax

VET

Vocational education and training

YS

Years of completed schooling

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 ACRONYMS AND ABBREVIATIONS – 13

ISO country codes
Argentina
Australia
Austria
Belgium
Brazil
Canada
Chile
Colombia
China
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Iceland
India²
Indonesia
Ireland
Israel
Italy

ARG
AUS
AUT
BEL
BRA
CAN
CHL
COL
CHN
CZE
DNK
EST
FIN
FRA
DEU
GRC
HUN
ISL
IND
IDN
IRL
ISR
ITA

Japan
Korea
Latvia
Luxembourg
Mexico
Netherlands
New Zealand
Norway
Peru
Poland
Portugal
Russian Federation
Slovak Republic
Slovenia
South Africa
Spain
Sweden
Switzerland
Turkey
United Kingdom
United States

JPN
KOR
LVA
LUX
MEX
NLD
NZL
NOR
PER
POL
PRT
RUS
SVK
SVN
ZAF
ESP
SWE
CHE
TUR
GBR
USA

Conventional signs
.. Not available
In figures, OECD refers to unweighted averages of OECD countries for which data
are available.
(↘) in the legend relates to the variable for which countries are ranked from left to
right in decreasing order.
(↗) in the legend relates to the variable for which countries are ranked from left to
right in increasing order.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 EXECUTIVE SUMMARY – 15

Executive summary
Income inequality increased in good times, and in bad times
In most countries, the gap between rich and poor is at its highest level since 30 years.
Today, in OECD countries, the richest 10% of the population earn 9.6 times the income
of the poorest 10 %. In the 1980s this ratio stood at 7:1 rising to 8:1 in the 1990s and 9:1
in the 2000s. In several emerging economies, particularly in Latin America, income
inequality has narrowed, but income gaps remain generally higher than in
OECD countries. During the crisis, income inequality continued to increase, mainly due
to the fall in employment; redistribution through taxes and transfer partly offset
inequality. However, at the lower end of the income distribution, real household incomes
fell substantially in countries hit hardest by the crisis.
Much of the recent debate surrounding inequality has focused on top earners,
especially the “top 1%”. Less well understood is the relative decline of low earners and
low-income households – not just the bottom 10% but the lowest 40%. This report places
a special focus on these households, investigating some of the factors that have weakened
their economic position, and the range of policy options that can help address increasing
inequality.

Higher inequality drags down economic growth and harms opportunities
Beyond its impact on social cohesion, growing inequality is harmful for long-term
economic growth. The rise of income inequality between 1985 and 2005, for example, is
estimated to have knocked 4.7 percentage points off cumulative growth between 1990
and 2010, on average across OECD countries for which long time series are available.
The key driver is the growing gap between lower-income households – the bottom 40%
of the distribution – and the rest of the population.
A main transmission mechanism between inequality and growth is human-capital
investment. While there is always a gap in education outcomes across individuals with
different socio-economic backgrounds, the gap widens in high-inequality countries as
people in disadvantaged households struggle to access quality education. This implies
large amounts of wasted potential and lower social mobility.

Rising non-standard work can create job opportunities but also contribute to higher
inequality
Temporary and part-time work and self-employment now account for about a third of
total employment in OECD countries. Since the mid-1990s, more than half of all job
creation was in the form of non-standard work. Many non-standard workers are worse off
in many aspects of job quality, such as earnings, job security or access to training. In

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 16 – EXECUTIVE SUMMARY
particular low-skilled temporary workers face substantial wage penalties, earnings
instability and slower wage growth.
Households that are heavily dependent on earnings from non-standard work have
much higher income poverty rates (22% on average), and the increase in the number of
such households in OECD countries has contributed to higher overall inequality.
Non-standard work can be a “stepping stone” to more stable employment – but it
depends on the type of work and the characteristics of workers and labour market
institutions. In many countries, younger workers, especially those with only temporary
work contracts have a lower chance of moving on to a more stable, career job.

More women in the workforce lowers inequality
Women have made substantial progress in narrowing the participation, pay and career
gap with men and this has put a brake on rising inequality. But they are still about 16%
less likely to be in paid work and earn about 15% less than men. If the proportion of
households with working women had remained at levels of 20 to 25 years ago, income
inequality would have increased by almost 1 Gini point more on average. The impact of a
higher share of women working full-time and higher relative wages for women added
another brake of 1 point.

High wealth concentration limits investment opportunities
Wealth is much more concentrated than income: on average, the 10% of wealthiest
households hold half of total wealth, the next 50% hold almost the other half, while the
40% least wealthy own little over 3%. At the same time, high levels of indebtedness
and/or low asset holdings affect the ability of the lower middle class to undertake
investments in human capital or others. High wealth concentration can weaken potential
growth.

Designing policy packages to tackle high inequality and promote opportunities for all
Policy makers have a range of instruments and tools at hand to tackle rising inequality
and promote opportunities for all. For such policy packages to be successful, solid trust in
institutions and effective social dialogue are essential. Reducing the growing divide
between rich and poor and promoting opportunities for all requires policy packages in
four main areas:


Women’s participation in economic life: governments need to pursue policies to
eliminate the unequal treatment of men and women in the labour market and to
remove barriers to female employment and career progression. This includes
measures for increasing the earnings potential of women on low salaries and to
address the glass ceiling.



Employment promotion and good-quality jobs: policies need to emphasise access
to jobs and labour market integration. The focus must be on policies for quantity
and quality of jobs; jobs that offer career and investment possibilities; jobs that
are stepping stones rather than dead ends. Addressing labour market segmentation
is an important element of enhancing job quality and tackling inequality.



Skills and education: A focus on the early years, as well as on the needs of
families with school children, is crucial in addressing socio-economic differences
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 EXECUTIVE SUMMARY – 17

in education. More must be done to provide youth with the skills they need to get
a good start in the labour market. With a rapidly evolving economy, further
efforts, with the close involvements of business and unions, should be made in
promoting a continuous up-grading of skills during the working life.


Tax-and-transfer systems for efficient redistribution: Adequately designed
redistribution via taxes and transfers is a powerful instrument to contribute to
more equality and more growth. In recent decades, the effectiveness of
redistribution weakened in many countries due to working-age benefits not
keeping pace with real wages and taxes becoming less progressive. Policies need
to ensure that wealthier individuals but also multinational firms pay their share of
the tax burden. Large and persistent losses of low-income groups underline the
need for well-designed income-support policies and counter-cyclical social
spending.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 19

Chapter 1
Overview of inequality trends, key findings and policy directions

This chapter documents the longer-term trends as well as recent developments in income
inequality and summarises the main messages from the in-depth chapters of the report. In
particular, it highlights the channels through which inequality affects growth, the impact
of women’s employment and of developments in the type of jobs on inequality, and the
extent of wealth concentration and indebtedness. It discusses individual measures but
especially policy packages that are both equality- and growth-friendly by focusing on four
main areas: women’s participation, employment promotion and good quality jobs, skills
and education, and taxes and transfers.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of
such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements
in the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 20 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS

1.1. Introduction and key findings
Over the past three decades, income inequality has risen in most OECD countries,
reaching in some cases historical highs. Today, the Gini coefficient – a common measure
of income inequality that scores 0 when everybody has identical incomes and 1 when all
the income goes to only one person – stands at an average of 0.315 in OECD countries,
exceeding 0.4 in the United States and Turkey and approaching 0.5 in Chile and Mexico.
In the main emerging economies, income inequality is higher than in the OECD area
(Figure 1.1); in some it has increased over the past decade but there are encouraging signs
of stabilisation (e.g. China) or even declines in some of them (e.g. Brazil).
This widespread rise in income inequality has been accompanied by growing public
discussion and concern over the impact of the high and often rising gap between rich and
poor on our societies. In recent years, and especially since the onset of the economic
downturn, these debates have moved to the top of the policy agenda in many countries.
Figure 1.1. Income inequality varies greatly across OECD countries and emerging economies
Level of income inequality (Gini coefficient), 2013 or latest available year
OECD Income Distribution Database

Secondary data sources - Income based

Secondary data sources - Consumption based

0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20

Note: Data refer to 2014 for China, 2013 for Finland, Hungary, Netherlands and the United States and India, 2011 for Canada,
Chile, Israel, Turkey and Brazil, 2010 for Indonesia, 2009 for Japan, and 2012 for the other countries. See note to
Table 1.A1.1.Data from secondary data sources are not strictly comparable and should be interpreted with caution. Gini
coefficients are based on equivalised incomes for OECD countries, Colombia, Latvia and the Russian Federation; per-capita
incomes for other countries; and per-capita consumption for India and Indonesia.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm, for OECD countries,
Latvia, Russian Federation and Colombia. World Bank, Poverty and Inequality Database for India. Statistics Indonesia (Susenas) for
Indonesia. SEDLAC database for Argentina and Brazil. National Bureau of Statistics of China for China. National Income
Dynamics Survey (NIDS) from Finn, A. and M. Leibbrandt (2013), “Mobility and Inequality in the First Three Waves of NIDS”,
SALDRU Working Paper, No. 120 and NIDS Discussion Paper, No. 2013/2, SALDRU, University of Cape Town, for South Africa.
12 http://dx.doi.org/10.1787/888933207690

In the United States and other advanced economies, much of the recent debate on
inequality focuses on the richest 1% and, increasingly, the 0.1%, the groups that have
enjoyed the lion’s share of income growth in recent decades. In the United States, for
example, average pre-tax income rose by 1% a year between the mid-1970s and up to the

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 21

Great Recession. But when the growth that went to the top 1% of earners is excluded, the
annual growth rate for the remaining 99% was just 0.6%, about the same level as in
France (Förster et al., 2014). The rise of “the super-rich” has led to warnings about the
risks of rent-seeking and political and economic “capture” by the economic elite.
But while the flashy lifestyles and incomes of the top 1% are certainly eye-catching,
focusing on them exclusively risks obscuring another area of growing concern in
inequality – namely the declining situation of low-income households. This is not a small
group. In recent decades, as much as 40% of the population at the lower end of the
distribution has benefited little from economic growth in many countries. In some cases,
low earners have even seen their incomes fall in real terms (Figure 1.2). Just as with the
rise of the 1%, the decline of the 40% raises social and political questions. When such a
large group in the population gains so little from economic growth, the social fabric frays
and trust in institutions is weakened.
Figure 1.2. Lower and lowest incomes were increasingly left behind
Trends in real household incomes at the bottom, the middle and the top, OECD average, 1985 = 1
Bottom 10%

Bottom 40%

middle 50-90%

Top 10%

1.60
1.50
1.40
1.30
1.20
1.10
1.00
1985

1990

1995

2000

2005

2010

Note: Income refers to disposable household income, corrected for household size. OECD is the unweighted average of
17 countries (Canada, Germany, Denmark, Finland, France, United Kingdom, Greece, Israel, Italy, Japan, Luxembourg, Mexico,
Netherlands, Norway, New Zealand, Sweden and United States). See notes to Figure 3.5.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207705

Income inequality rose not only in bad economic times but also in good economic
times (see Chapter 3). There have been periods when some countries managed to contain
the growth in income inequality, but these have rarely lasted for long (OECD, 2014a). In
2008, the OECD rang the alarm bells about the pervasive, decades-long rise in income
inequality in a report entitled Growing Unequal? (OECD, 2008). Three years later,
OECD analysis in Divided we Stand (2011) documented the deep-rooted reasons for why
inequality was rising so much, not only in advanced but also in most emerging countries.
Because the rise in inequality is so deeply embedded in our economic structures, it
will be hard to reverse it. Changing institutions, policies, and relationships between

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 22 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
economic actors that have been with us for so long will be far from easy. And, forces of
technological change and globalisation are not going away. So we need compelling
evidence to show us the way to change so that all citizens can have better lives.
The purpose of this publication is to provide the evidence for why such a shift is
urgently needed and how to implement it. Some may consider that the social and political
costs of high and rising inequality are in and of themselves sufficient to justify action.
The central argument of this publication is different. It is that, beyond its serious impact
on social cohesion, high and often growing inequality raises major economic concerns,
not just for the low earners themselves but for the wider health and sustainability of our
economies. Put simply: rising inequality is bad for long-term growth. Chapter 2 of this
report discusses these issues in detail.
The 2011 OECD report, Divided We Stand, documented that, beyond skill-biased
technological changes, some structural reforms, while raising the economic potential, and
creating new jobs, have also been associated with a widening of wage disparities.
Exacerbating this trend has been a decline, in many countries, in the effectiveness of tax
and benefit systems to redistribute market income. Such trade-offs of pro-growth policy
reforms with both equality and stability of incomes have been described recently in
OECD (2015a) and Cournède et al. (2015).
Making the rich richer, while incomes of the bottom 40% remain flat, could be seen
as sensible from an economic perspective – after all, some are better off, and none are
worse off. However, policies which lead to this outcome may not be even economically
sensible if wider inequality reduces the capacity of the bottom 40% to improve their
position and that of their children in the future. But just because inequality is bad for
growth does not mean that all policies that reduce inequality are good for growth. OECD
work in the frame of Going for Growth (OECD, 2013a, 2014b, 2015b) points to
pro-growth policy reforms that could yield a double dividend in terms of boosting GDP
per capita and reducing income inequality.
One of the messages of this report is that structural policies are needed now more than
ever to put our economies back on a path of strong and sustainable growth, but have to be
carefully designed and complemented by measures that promote a better distribution of
the growth dividends. The challenge, therefore, is to find appropriate policy packages that
are both growth-friendly and that reduce inequality. The bulk of this report is about
looking at the main policy areas that are at the intersection of growth and a better
distribution of its outcomes. It suggests that in order to reduce the growing divide
between rich and poor and restore opportunities for all requires focusing policy attention
on four main areas:


women’s participation in economic life



employment promotion and good-quality jobs



skills and education



tax-and-transfer systems for efficient redistribution.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 23

1.2. Inequality increased in good times, and it continued increasing in bad times
The long-term trend
Over the past three decades, labour markets have been profoundly transformed by the
interplay of globalisation, technological change and regulatory reforms. These changes
have had a major impact on earnings and incomes. People with skills in high demand
sectors like IT or finance have seen their earnings rise significantly, especially at the very
top end of the scale, where performance-based pay and bonuses have become widespread.
Meanwhile, at the other end of the scale, wages of workers with low skills have not kept
up. The period has also seen reforms of tax systems that have reduced marginal tax rates
for high earners. In addition, taxes and benefits have tended to redistribute less in the
period from the mid-1990s up to the crisis; this has been felt most keenly by low-income
working-age households.
These factors, along with a number of demographic and social trends, are key to
understanding the long-term rise in income inequality in OECD countries. Cyclical
factors play a role as well, as steep drops in income during downturns lead to scarring and
longer-term disadvantage when social policies are insufficiently counter-cyclical (OECD,
2014a). In the 1980s, the richest 10% of the population earned seven times more than the
poorest 10%; today they earn almost ten times more. In broad terms, this long-term trend
has been driven by two main movements: at the top end, and especially among the top
1%, a surge in incomes; at the bottom end, much slower income growth during good
times and often a fall in incomes in bad times, especially during and after the Great
Recession.
In the two decades prior to the global economic crisis, average real disposable
household income rose by an average of 1.6% across OECD countries. With the
exception of Japan, real incomes rose at both the top and bottom of the earnings ladder in
every country (OECD, 2011). But in three-quarters of countries, household incomes of
the top 10% grew faster than those of the poorest 10%, which led to widening income
inequality. Other measures of inequality also support this general picture: the Gini
coefficient of income inequality stood at 0.29, on average, across OECD countries in the
mid-1980s. But by 2013, it had increased by about ten percent or 3 points to 0.32, rising
in 17 of the 22 OECD countries for which long-time series are available1 (Figure 1.3).
Trends differed by age groups and the working-age population often bore the brunt of
inequality increases.
In emerging economies, as documented above and in Chapter 7, levels of inequality
are generally higher than in OECD countries.2 Gini coefficients often exceed levels
of 0.5, as in Brazil and many other Latin American countries, and approach 0.7 in South
Africa. Over the long-term, inequality has been on the rise in most emerging economies,
too, although data and indicators are less comparable than those from OECD countries.
That said, all available evidence shows narrowing income gaps in most Latin American
countries since the late 1990s, most notably in Brazil, and signs of a halt in the rise in
some other countries, including China and Russia, since the mid-2000s.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 24 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Figure 1.3. Income inequality increased in most OECD countries
Gini coefficients of income inequality, mid-1980s and 2013, or latest available year
1985

2013 or latest
Increase

Little change

Decrease

0.50
0.45
0.40
0.35
0.30
0.25

0.20
0.15

Note: “Little change” in inequality refers to changes of less than 1.5 percentage points. Data year for 2013 (or latest year): see
Figure 1.1. These values differ slightly from those in Figure 1.1 for some countries as they have been adjusted to be comparable
with 1985 values.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207711

Was the crisis a game changer?
Real average disposable household income stagnated or fell in most OECD countries in
the years from 2007 to 2011. The declines were particularly striking in the countries hit most
severely by the crisis. In Greece, the average household lost over 8% of its real net income
annually, and in Spain, Ireland and Iceland, average annual losses exceeded 3.5%. The
dramatic collapse was due in large part to the loss in employment, rather than falling wages.
But these averages only tell part of the story. The groups at the bottom of the income
scale lost even more in some countries. In Spain, for instance, incomes of the poorest
10% dropped by almost 13% per year, compared to only 1.5% for the richest 10%. In
about half of those countries where incomes continued to grow, the top 10% did better
than the bottom 10%. In Austria, Denmark, France and the United States, incomes at the
top increased in real terms while they fell at the bottom.
In all OECD countries, income inequality is greatly reduced through redistribution –
typically, taxes and transfers such as unemployment and other benefits. This is why “net” or
“disposable” income inequality is much lower than “market” income inequality. But the
impact of such redistribution has changed. In OECD countries, in the decade prior to the
crisis, inequality before taxes and benefits often stabilised. But income inequality was driven
upwards by weakening redistribution. In the initial years of the crisis, income inequality
before taxes and benefits increased strongly but out-of-work benefits and other redistribution
measures managed to cushion at least partially the rise (Figure 1.4). In the most recent years
of weak economic recovery, unemployment persisted and yet governments chose to shift
focus to fiscal consolidation, including curtailing unemployment benefits, education and
investment. While income inequality before taxes and benefits continued to rise, the
cushioning effect of taxes and benefits has become weaker, accelerating the overall upwards
trend in disposable income inequality.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 25

Figure 1.4. Taxes and benefits cushioned the effect of the crisis on household incomes
Percentage changes of household market income, benefits and taxes, 2007=100%, OECD, total population

x 0.01

Household market income

Cash social benefits

Personal income taxes

112
110
108
106

104
102
100
98
96

94
2007

2008

2009

2010

2011

Note: Household incomes are equivalised by household size and adjusted for inflation using changes in consumer price indices and
differences in purchasing power using 2010 PPP indices for private consumption. OECD average refers to the unweighted average
of 26 countries.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207729

The crisis period also saw a marked rise in income poverty in OECD countries,
especially when measured in terms of “anchored” poverty, i.e. when fixing the real lowincome benchmark to pre-crisis level. This gives a stronger sense of absolute changes in
the living standards of the poor than the more commonly used ‘relative’ measure of
poverty, where the benchmark also evolves. Between 2007 and 2011, the OECD
anchored poverty rate rose by just over one percentage point to 9.4%. In Greece,
anchored poverty more than doubled to 27% and in Spain it almost doubled to 18%.
There was also a shift in the age profile of poverty, with young people replacing the
elderly as the group most at risk of poverty, continuing a trend lasting for the past
30 years (Figure 1.5). Poverty rates rose among children and especially among youth
while they fell among the elderly.
Figure 1.5. The risk of income poverty has shifted from the elderly to the young
Relative poverty rate of the entire population in each year = 100, mid-1980s to 2013, or latest available year

Note: OECD unweighted average for 18 OECD countries for which data are available from the mid-1980s: Canada, Denmark,
Finland, France, Germany, Greece, Israel, Italy, Japan, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Sweden,
Turkey, the United Kingdom, the United States.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207732

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 26 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS

1.3. Higher inequality drags down economic growth
High and often growing income inequality in so many countries has renewed interest
on its possible economic effects, over and above its impact of social outcomes. This is by
no means a new debate. Economic theory has long argued that the relationship between
inequality and growth can go either way. A gap between rich and poor means that people
have strong incentives to do what they can to be rich – including working harder,
studying longer, and taking greater risks, all of which can lead to more economic activity,
efficiency and growth (the “incentives” hypothesis). On the other hand, more inequality
means that some people – the rich – are better able to take advantage of economic
opportunities than the poor. Poor families may be unable to keep their children in
education for as long as is optimal, or to afford high-quality education, thereby harming
their future earnings. And they may find it difficult to borrow to invest in new
opportunities (the “opportunities” hypothesis). As a result, economic growth is slower
than it would otherwise be and disproportionately benefits the rich.
The empirical evidence has until recently been mixed as to which of the opposite
forces dominates and in which country. But new research at the OECD, presented in
Chapter 2, finds consistent evidence that the long-term rise in inequality of disposable
incomes observed in most OECD countries has indeed put a significant brake on longterm growth. Further, it shows that efforts to reduce inequality through redistribution –
typically, certain forms of taxes and benefits – do not lead to slower growth (confirming
similar results in Ostry et al., 2014). This suggests that redistribution can be part of the
solution, but requires a serious discussion on how to promote effective and well-targeted
measures that promote a better sharing of the growth outcomes not only for social but
also for economic considerations.
The analysis in Chapter 2 which draws on data for 31 OECD countries covering the
period 1970-2010, finds that income inequality has a sizeable impact on growth. Between
1985 and 2005, for example, inequality rose by more than 2 Gini points on average across
19 OECD countries, an increase estimated to have knocked 4.7 percentage points off
cumulative growth between 1990 and 2010.
The inequality brake on growth is not restricted to developed countries (see Ostry et
al., 2014); it affects even more emerging economies. Despite often impressive growth
rates and reductions in absolute poverty, high and sometimes very high levels of income
inequalities in emerging economies are found to undermine their long-term growth
potential. Some of these economies have recently provided important examples and
insights into how income inequality can be addressed via well-targeted social and
employment programmes (e.g. Jefes y Jefas de Hogar Desocupados in Argentina; the
Bolsa Familia programme in Brazil; Prospera in Mexico).

If the bottom loses ground, everyone is losing ground
The OECD analysis indicates that the biggest factor for the impact of inequality on
growth is the growing gap between lower income households and the rest of the
population. This is true not just for the very lowest earners – the bottom 10% – but for a
much broader swathe of low earners – the bottom 40%. Countering the negative effect of
inequality on growth is thus not just about tackling poverty but about addressing low
incomes more broadly.
Since the dominant mechanism through which inequality seems to affect growth is by
curbing opportunities for the poor and lower middle classes, Chapter 2 looks at their
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 27

investment in education and skills compared with other social groups. In particular, the
analysis compares the education performance at different levels of inequality of three
social groups – people whose parents come from high, medium and low educational
backgrounds (PEB) – across three areas, namely education attainment, skills and
employment. It is no surprise to find that people from low socio-economic groups do less
well in all three of these dimensions than people from higher socio-economic groups.
However, the analysis shows that as inequality rises the outcomes of people from lower
groups decline still further.
On average, around 40% of people from a high PEB and around 30% from a medium
PEB graduate from university. These figures remain broadly consistent regardless of the
level of inequality. However, this is not the case for children from poorer educated
families. An increase in inequality of around 6 Gini points lowers the probability of
poorer people graduating from university by around four points. A similar effect was
found when it comes to the amount of time students spend in education. Increasing
inequality by around 6 Gini points cuts the length of time children from poorer families
spend in education by about half a year.
The impact of higher income inequality on children from poorly educated families
can also be seen in quality of education, by measurements of skills (e.g. indexes of
proficiency in numeracy and literacy) drawing from the OECD Skills Survey (OECD,
2013e). Once again, the numbers remain fairly constant for people from high and medium
PEBs, regardless of the level of inequality in the country, but decline markedly for
children from poorer families (Figure 1.6). This might seem like an obvious follow-on
from the previous finding – people from poorer families spend less time in education and
therefore develop weaker numeracy and literacy skills. However, the data show that even
when poorer people spend the same amount of time in education as their better-off peers
they do worse. This suggests that a large part of their lower proficiency is not because
they have less education, but rather that they get less out of their time in education. This
is probably due to education being of poorer quality, but there may be other social factors
as well, such as people from poorer backgrounds being unable to spend the same amount
of time studying as their richer counterparts.
Finally, the impact of higher levels of income inequality is also clearly visible in
employment patterns. As inequality rises, people from poorer families face much weaker
job prospects while there is little change for those from better-off families. An increase in
inequality of around 6 Gini points rises the fraction of time spent out of employment by
someone from a low PEB by around 3 percentage points.
Rising income inequality thus has a significant impact on economic growth, in large
part because it reduces the capacity of the poorer segments – the poorest 40% of the
population, to be precise – to invest in their skills and education. It has long been popular
to say that while there is no social consensus around the desirability of tackling inequality
of outcomes, for example by redistributing wealth, surely we can agree that it is necessary
to ensure that we have equality of opportunities – i.e., that all should have the same life
chances, regardless of their initial conditions. In reality, few societies come close to
ensuring such equality of opportunities. However, the importance of research showing the
links between family status and education shows that the distinction between
opportunities and outcomes is not straightforward. Higher inequality of incomes of
parents tends to imply higher inequality of life chances of their children. To achieve
greater equality of opportunities without tackling increasing inequality in outcomes will
be very difficult.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 28 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Figure 1.6. Inequality lowers skills of the poor
Average numeracy score by parent educational background and inequality
Low PEB

Medium PEB

High PEB

300

Numeracy score

290
280

270
260
250
240

0.18

0.23

0.28

0.33

0.38

Inequality (Gini coefficient)

Note: The graph plots the average predicted numeracy score for individuals from low, medium and high family (educational)
backgrounds, as a function of the degree of inequality (Gini points) in the country at the time they were around 14 years old.
Low PEB: neither parent has attained upper secondary education; medium PEB: at least one parent has attained secondary and
post-secondary, non-tertiary education; high PEB: at least one parent has attained tertiary education. The bars indicate 95%
confidence intervals. The vertical dashed lines indicate the 25th, the median and the 75th percentiles of the underlying
distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Chapter 2, Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207742

1.4. Why increased non-standard work can lead to more inequality
Promoting equality of opportunities is not just about improving access to quality
education but also ensuring that the investment in human capital is rewarded through
access to productive and rewarding jobs. Before the crisis, many OECD countries were
facing a paradoxical situation: their employment rates (employment to working-age
population) were at record-high levels and yet income inequality was on the rise.
Typically, rising employment might be expected to reduce income inequality as the
number of people earning no salary or relying on unemployment benefits falls. However,
in recent decades the potential for this to happen has been undercut by the gradual decline
of the traditional, permanent, nine-to-five job in favour of non-standard work – typically
part-time and temporary work and self-employment. More (often low-skilled) people
have been given access to the labour market but at the same time this has been associated
with increased inequalities in wages and, unfortunately even in household income.
Policy needs to focus on access to jobs and labour market integration in order to
increase both equality and growth. But only focusing on the number of jobs is not
enough; Chapter 4 of this report shows that it is about policies for quantity and quality of
jobs; jobs that offer career and investment possibilities; jobs that are stepping stones
rather than dead ends.
The development of non-standard work is related to technological changes and the
associated evolution of labour demand. In most advanced economies, there has been an
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 29

increasing job polarisation – a decline in the share of workers in the middle of the
workforce, both in terms of skills and income, and increases in the proportions of workers
in high- and low-skill jobs. The share of workers with routine-task jobs, such as
accountants, fell from 53% to 41% between 1995 and 2010. At the same time, the
employment share for high-skill abstract jobs, such as designers, grew from 28% to 38%,
and relatively low-skill non-routine manual jobs, such as drivers, increased from 18% to
21%. The emergence of this U-shaped workforce is closely matched by developments in
non-standard employment. The decline in middle-skill employment went hand in hand
with a decrease of standard work contracts; and workers taking on low and high-skill jobs
were increasingly likely to be self-employed, part-timers or temporary workers.
The spread of non-standard work is most visible when comparing its share in new
jobs created before and since the onset of the crisis. Between the mid-1990s and the start
of the Great Recession, almost half of all job creation was in the form of non-standard
work; including the crisis years brings the share up to 60% (Figure 1.7).
The most recent data shows that non-standard work accounted for around a third of
total employment in OECD countries in 2013, shared roughly equally between temporary
jobs, permanent part-time jobs and self-employment. In some eastern European countries
the proportion of non-standard workers is lower than 20% but in most southern European
countries, Australia and Switzerland it exceeds 40% and in the Netherlands more than
half of all workers are in non-standard work, largely because of a high number of parttimers.
Figure 1.7. More than half of all jobs created since 1995 were non-standard jobs
Panel A. Employment growth (%) 1995-2007, by type of employment
Standard work
70
60
50
40
30
20
10

0
-10

-20

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

Non-standard workers

 30 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Figure 1.7. More than half of all jobs created since 1995 were non-standard jobs (cont.)
Panel B. Employment growth (%) 2007-2013, by type of employment
Standard work

Non-standard workers

20
%
15
10
5

0
-5
-10
-15
-20

Note: Working-age (15-64) workers, excluding employers as well as students working part-time. Non-standard workers include
workers with a temporary contract, part-timers and own-account self-employed.
Source: European Union Labour Force Survey (EU-LFS), Labour Force Survey for Canada, Household, Income and Labour
Dynamics in Australia (HILDA) for Australia.
12 http://dx.doi.org/10.1787/888933207750

Who are the non-standard workers?
Women and youth are more likely to be in non-standard work than prime-age men. A
quarter of men have non-standard jobs, but that proportion rises to 40% among women,
mostly because many women work part-time. They represent close to 70% of nonstandard workers in Luxembourg, Austria and Switzerland, and more than 60% in most
Nordic countries, the Netherlands, Germany, France, Belgium and Japan. Youth are the
age group with the highest incidence of non-standard work at 40%. This mostly involves
younger workers on temporary contracts; among temporary workers, close to half are
under the age of 30.
Non-standard workers are also more likely to have lower levels of education and to be
found in smaller rather than larger firms. In both cases, incidence is just under 50%. Half
of all non-standard workers are the main breadwinners in a household, and a large
majority of them (80% or more) live in two-or-more-person households, often with
children.

It’s about jobs – but which jobs?
Non-standard jobs are not necessarily bad jobs. Non-standard employment is used by
employers in need of a flexible workforce that can be adjusted quickly with production, to
cut costs during downturns or as a screening device for new hires. Part-time, temporary
and self-employment arrangements can be attractive to certain workers who opt for this
type of employment to achieve a better work-family life balance, higher life satisfaction
or, in the case of self-employment, a greater sense of control.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 31

But it can also be associated with precariousness and poorer labour conditions where
non-standard workers are exempted from the same levels of employment protection,
safeguards and fringe benefits enjoyed by colleagues on standard work contracts. The
analysis in Chapter 4 shows that many non-standard workers are indeed worse off on a
range of aspects of job quality. First, and foremost, a non-standard job typically pays less
than traditional permanent work. In general, earnings levels are lower in terms of annual
and hourly wages. For part-timers, however, the differences in hourly wages tend to
disappear once other demographic and job characteristics are taken into account. But
temporary workers face substantial wage penalties, earnings instability and slower wage
growth compared to permanent workers.
These earnings gaps are especially wide among low-skill, low-paid workers; nonstandard workers in the bottom 40% of earners typically suffer wage penalties of 20%.
This gap between non-standard and standard workers narrows markedly among higherincome workers and vanishes completely among top earners (i.e. the highest earnings
quintile).
Non-standard workers suffer other penalties, too. Full-time temporary workers are
20% less likely – and part-time workers 40% less likely – to receive training than
standard workers. Non-standard workers also face higher levels of insecurity in terms of
the probability of job loss and unemployment and, in the case of temporary workers,
report significantly higher job strain.
While associated with lower job quality, non-standard work can be a “stepping stone”
to more stable employment – but it depends on the type of work and the characteristics of
workers and labour market institutions. In particular, temporary contracts can increase the
chances of acquiring a standard job compared with remaining unemployed in the short
run by some 12 percentage points on average. But this is not true of part-time work or
self-employment, which do not increase the chances of a transition to a standard job. In
many countries, prime-age and older workers have a better chance of using non-standard
jobs as “stepping stones” than younger workers, especially those with only temporary
work contracts. In addition, transition rates remain low when considering a longer time
span: less than 50% of the workers that were on temporary contracts in a given year were
employed with full-time permanent contracts three years later (Figure 1.8).
In sum, people are more likely to be poor or in the struggling bottom 40% of society
if they have non-standard work, especially if they live in a household with other nonstandard or non-employed workers. This might not matter too much if such work were a
stepping stone to a better job in the future, but too often this is not the case. Excessive
wage penalties associated with temporary jobs in particular can contribute to wider
inequality. That can also contribute to lower levels of training and skill development –
precisely those activities that can lead to higher growth in the future. Ensuring that
different employment arrangements meet the interests of both employees and employers,
and do not become a caste system of ‘good’ and ‘bad’ jobs, is one way of ensuring that
equality and growth move hand-in-hand.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 32 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Figure 1.8. In most countries, the majority of temporary workers stay on such jobs
Percentage share of temporary employees in 2008 that were employed as full-time permanent employees in 2011
70
%
60
50
40

30
20
10

0

Note: 2007-10 for the Czech Republic, France, Greece, Sweden and the United Kingdom; 2006-09 for Norway and the Slovak
Republic; and 2005-08 for Ireland.
Source: OECD (2014), OECD Employment Outlook, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2014-en.
Calculations based on the European Union Statistics on Income and Living conditions (EU-SILC) 2005-11.
12 http://dx.doi.org/10.1787/888933207768

1.5. Tackling job and wage discrimination of women will boost growth and equality
Over the past 20 years, most OECD countries have seen big increases in the numbers
of women joining the labour market and major improvements in the status of women in
the workforce. These changes have brought benefits not just for women themselves and
their families, but for the economy as a whole. A narrowing in the labour force
participation gap between men and women of 50% has been estimated to raise annual
growth in GDP per capita by 0.3 percentage points on average (OECD, 2012a). And even
though there are complex interactions between female labour force participation and
inequality the overall impact has been positive: having more women in the workforce
lowers income inequalities.

Inequalities between men and women persist
Despite substantial progress, inequalities between women and men in the workforce
persist. Over the past 20 years the employment gap in OECD countries has narrowed by
an average of 7 percentage points but it still stands at 16 percentage points. When taking
into account that women are more likely to hold part-time jobs and thus work fewer hours
than men, the hours-adjusted employment gap widens to almost 24 percentage points (see
Chapter 5).
Differences in pay also persist, due in large part to the fact that women are more
likely to work in lower-paid occupations. Still, the pay gap has narrowed: among fulltime employees, women earned, on average, 15% less than men in 2013, an improvement
of four percentage points since 2000.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 33

More women in paid work means less income inequality
A priori, higher female labour force participation could lead to more or less
inequality. On the one hand, there are cases, where the presence of more women in the
workforce may actually result in higher income inequality. This may happen, for
example, when more women take up high-skill jobs that are better paid. Nowadays people
tend to choose their spouses more often from the same socioeconomic group, meaning
that doctors are marrying other doctors rather than nurses as it was the case more than
20 years ago. This behaviour, also called “assortative mating”, results in a situation where
higher incomes become more concentrated.
In addition, earnings inequality among women is higher than among men, and one
could expect this to drive up overall inequality. However, the earnings gap between high
and low educated women has been increasing at a much slower pace than the same gap
among men since the mid-1980s, on average by 9% as compared to 17% for men. There
are two main reasons why low-income women are not slipping as far behind as lowincome men. Firstly, their real earnings have increased more – or declined by less – than
men’s in most countries. Second, women in the bottom 20% of earners worked longer
hours than in the past.
On the other hand, the mere fact of more women entering the workforce outweighs
the disequalising pressures and clearly narrows income inequality overall. This is
particularly true – although not in all countries – when more women are working
“traditional hours” in full-time jobs. Improvements in the gender pay gap also contribute
to further reducing income inequality.
Overall, the evidence in Chapter 5 shows that if the proportion of households with a
working woman had remained at around the same levels as 20 to 25 years ago (52%
rather than 61%), income inequality would have increased by almost 1 Gini point more
on average than it actually did. In addition to the general impact of women joining the
labour force, the impact of a higher share of women working full-time and higher relative
wages for women added another brake of 1 point (Figure 1.9).
For these effects to continue into the future, governments need to pursue policies to
increase the earnings potential of women on low salaries and to address the glass ceiling,
which prevents women from moving up the career ladder. Governments can also act as a
role model for other employers by ensuring equality of opportunity in the public service
and promotion the representation of women in public life.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 34 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Figure 1.9. In all OECD countries, women's employment put a brake on increasing inequality
Contribution of composition and wage structure effects (women) to percentage point changes in Gini of household disposable
income, mid-1990s to 2007 or latest available pre-crisis year
Combined effect of women's employment changes

Change in Gini (↘)

0.08
0.06
0.04
0.02
0.00
- 0.02
- 0.04
- 0.06

Note: Data refer to working-age (25-64) households. Decomposition results are based on Recentered Influence Function (RIF)
regressions. Combined effect of women's employment changes include both the composition and wage structure effects, each
combining three covariates: participation, work intensity and job skill nature. Data refer to changes from the early/mid-1990s to
the latest available pre-crisis year.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933207774

1.6. Wealth concentration widens disparities further and limits investment
opportunities for many
Understanding people’s incomes, especially after taxes and benefits, gives a strong
sense of whether or not they can meet their bills and make long-term investments in
education, housing and so on. Yet there is no doubt that wealth also matters, both in
shaping people’s individual circumstances by generating capital income and as a wider
socio-economic force. Accumulated wealth can generate capital income, which, in turn,
can deepen income inequalities. In basic terms, income is the cash that people earn every
month through work, transfers or rents; wealth is the money they accumulate over time in
bank accounts and in assets such as property and stocks. Comparable data on income are
much more abundant than data on wealth, which perhaps partly explains why wealth
inequalities have been largely neglected, up until recently.

Wealth is more unequally distributed than income
Household wealth – in particular financial assets – is much more unequally distributed
than income. The bottom 40% owns only 3% of total household wealth in the 18 OECD
countries for which comparable data are available. For comparison, their share of total
household income is 20%. At the other end of the scale, the top 10% of the wealth
distribution own half of all total household wealth and the wealthiest 1% hold 18%. The
richest 10% of income earners get around a quarter of all cash income, ranging from 20% in
Norway to 28% in the United States, and close to 40% in Chile and Mexico. But countries
with lower income inequality levels are not necessarily those with low wealth
concentration, as witnessed by the examples of Austria, Germany and the Netherlands
(Figure 1.10).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 35

Figure 1.10. Wealth is more concentrated at the top than income
Share of top 10% of household disposable income and top 10% of household net wealth, 2012 or latest available year
Top 10% income share (↘)

Top wealth share 10%

%
80
70
60
50
40
30
20
10
0

Note: Income refers to disposable household income, corrected for household size. Wealth refers to net private household
wealth. Data refer to the shares of the richest 10% of income earners (bars) and of the richest 10% of wealth holders (diamonds),
respectively.
Source: OECD Wealth Distribution Database and OECD Income Distribution Database (www.oecd.org/social/incomedistribution-database.htm).
12 http://dx.doi.org/10.1787/888933207780

If it is problematic to develop a snapshot of wealth at any one time,3 it is doubly
difficult to track trends. Piketty (2014) compiled data from eight OECD countries since
the 1970s and concluded that, like income, private wealth has tended to become more
unequally distributed in recent decades, reversing a long-term decline throughout much of
the 20th century. Several factors have contributed to this rise, most notably the increase in
stock and housing prices relative to consumer prices.
Since the crisis, there are indications that the trend towards greater wealth inequality
has deepened. Comparable data from six OECD countries indicate that the crisis has
increased wealth concentration at the top in four of them, while wealth inequality at the
bottom of the distribution increased in five (Australia, Canada, Italy, the Netherlands and
the United States).

The flipside of wealth: Debt
Real assets, typically the family residence, are the main source of wealth, both for the
wealthy and for people with low levels of wealth. They account for 75% of the value of
total assets on average. It is only among those at the top of the wealth scale that financial
assets, such as stocks, form a significant source of wealth.
The flipside of wealth is debt, and, in some countries, there are reasons to be
concerned about the scale and concentration of liabilities, which are typically created by
loans on families’ main homes. As witnessed in the recent financial and economic crisis,
the concentration of debt on some households, even those with high levels of assets, may
put them at risk if there are sudden changes in asset prices. That, in turn, could affect
consumption and investment in the entire economy.
Close to 10% of households in the 18 OECD countries for which data are available are
over-indebted, i.e. they have a debt-to-asset ratio exceeding 75% or a debt-to-income ratio
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 36 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
over 3, and these ratios exceed 16% in the Netherlands, Norway and the United States
(Figure 1.11). The largest share of indebted households are found in the middle of the
income distribution, i.e. outside the bottom fifth and the top fifth. Only about a third of
households in the bottom fifth have any debts, as many poorer households have difficulties
in securing loans.
Figure 1.11. Half of all households have debts and one tenth is over-indebted
Percentage of indebted and over-indebted households
% of indebted households (↗)

% of households with
debt-to-asset ratio above 75%

% of households
with debt to income ratio above 3

%
90
80
70
60
50
40
30
20
10
0

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933207792

In summary, wealth is much more concentrated than income, and there are reasons to
believe that wealth inequalities are deepening over time. The capital income generated by
wealth concentration is likely to deepen income inequality still further, with implications
for deteriorating economic growth. At the same time, a high level of indebtedness and/or
low asset holdings further affects the ability of the lower middle class to undertake
investments in human capital or others, and reduces risk taking. This constitutes a second
pathway for how disparities in wealth holdings can weaken potential growth.

1.7. Designing sustainable policy packages to promote equality of opportunities
There is nothing inevitable about growing inequalities. Policy makers have a range of
instruments and tools at hand to tackle rising inequality and promote opportunities for all
while also promoting growth. There is, however, no single best model or policy mix to
adopt. Each country will have to design its own package, depending on the key factors at
the origins of inequality in the national context. How countries choose to address
inequality also depends on the extent to which their societies agree on the importance of
values such as solidarity, redistribution and equality and what role they assign to
government in promoting equality (see Atkinson, 2015).
This report adds a particular dimension to discussions of policy choices to tackle
inequality by demonstrating how inequality can reduce growth. The case for addressing
inequality becomes more urgent once it is realised that successfully doing so might also
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 37

increase long-term growth. Furthermore, the report suggests that it is the position of the
bottom 40% that matters in particular for economic growth.
There are two approaches to identifying the policy choices that countries might make
in order both to reduce inequalities and increase growth.


First, identify which policies will best promote growth, and then analyse their
potential distributional effects. This approach has recently been followed in
OECD work (e.g. OECD, 2008; Causa et al. 2014), identifying, within the progrowth policy toolbox, reforms that could yield a double dividend in terms of
boosting GDP per capita and reducing income inequality. It also identifies other
policy areas where reforms could entail a trade-off between both objectives, such
as tightening benefits to long-term unemployed people, or administrative
extensions of collective wage agreements.



The second approach is to identify those policy areas which will best address
inequality, and then assess their effect on growth. Such approach builds on the
extensive OECD work on the drivers and causes of inequality since Growing
Unequal? (2008) and Divided We Stand (2011), and the most promising elements
for a policy package are discussed below.

It might be expected that the two approaches are largely compatible. However, to the
extent that some countries prioritise economic growth they will prefer some policies over
those that are prioritised by those countries where the focus is on ensuring a more
equitable distribution of the growth benefits. Whatever the starting point, it should be
stressed that there are important policy complementarities that can and should be
exploited. A government may for example go for a growth-enhancing policy but
introduce at the same time complementary measures that reduce its potential negative
distributional effects. This means, for example, that it is a good idea to accompany
measures to reduce barriers to competition (e.g. relaxing product market regulations) with
adequate policies to support workers affected by the transition from less to more
productive activities.
The most efficient policy package will address inequalities at the point where they
originate rather than trying to pick them up only at a later stage. Redistribution strategies
based on transfers and taxes alone would not necessarily be effective or financially
sustainable. Policy needs to be pitched broadly, both with respect to the target groups and
to the policy tools applied – targeting poverty alone is not the solution. The analysis in
this report shows that it is not only the situation of the very poorest section of the
population that inhibits growth but that of a much broader group of working and lower
middle class people. Policy thus needs to be directed towards the bottom 40%.
As the analysis in this report shows, some – though not all – policies to reduce
income inequalities will not only increase fairness but will also sustain growth.
Identifying such policies is of particular importance within the broader new OECD
strategy of achieving Inclusive Growth (see www.oecd.org/inclusive-growth), i.e. growth
that creates opportunities for all and distributes the dividends of increased prosperity
fairly (OECD, 2014d, 2014e).
Following the approach of identifying policies which are effective in tackling
inequality, and then assessing their impact on growth, this report proposes a strategy for
efficient policy packages to reduce the growing divide between rich and poor and restore
opportunities for all, i.e. focusing policy attention on the four areas: i) women’s

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 38 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
participation in economic life; ii) employment promotion and good-quality jobs; iii) skills
and education, and iv) tax-and-transfer systems for efficient redistribution.

Women’s participation in economic life
While gender gaps in employment and earnings have declined, they remain large and
there is a need for policies to eliminate the unequal treatment of men and women in the
labour market. This agenda is relevant for OECD and emerging economies alike as
witnessed by a joint report of international organisations for the G20 (OECD-ILO-IMFWB, 2014). Establishing clear legislative frameworks to protect all workers against all
forms of direct and indirect discrimination and harassment – including discrimination of
workers based on gender, maternity, paternity and family responsibilities – can help provide
a formal structure to remove barriers to female employment and career progression (see the
OECD Gender Recommendation, for a comprehensive set of policy principles and
commitments to promote gender equality in education, employment and entrepreneurship,
OECD, 2013b).
Examples of such measures include labour legislation which guarantees the right of
women to return to work to the same or an equivalent position paid at the same rate after
maternity leave, as introduced for example, in Canada, Korea and the United States. The
principle of equal pay for equal work or for work of equal value should also be respected
in collective bargaining as well as labour laws and practices. In Indonesia, equal
employment opportunities guidelines were developed by the government in collaboration
with social partners, which explain the principle of equal pay for men and women for
work of equal value and provide guidance on how to identify discriminatory pay
practices.
Measures supporting the reconciliation of work and family life are also critical for
men and women to participate in the labour market on an equal footing (OECD, 2012a;
OECD 2014b). The difficulty of combining work and family responsibilities very often
results in women working part-time or dropping out of the labour force altogether. A
range of policies is necessary to remove this obstacle. In many countries, governments
and businesses have implemented family-friendly policies – parental leave, childcare,
out-of-school-hours care, flexible working arrangements, etc. – to help parents with
children. France and the Nordic countries, for example, provide a continuum of publicly
provided reconciliation support for parents during the early years of their child’s life, and
they have been able to combine high female employment with high fertility rates,
carrying a demographic dividend with them into the future.
One policy option towards better sharing of unpaid work is to institute leave
arrangements for the exclusive use of fathers. Many countries have paternity leave
ranging from a few days up to two weeks as for example, in France, Spain and the United
Kingdom. In a number of countries including Germany, Italy and France, policy
encourages fathers to take leave to care for young children by granting them the exclusive
right to part of the paid parental leave entitlement for two months or more on a ‘use it or
lose it’ basis, or provide “bonus” months when couples share leave entitlements while (to
a varying degree) providing income support during the leave period. Yet, this needs to be
combined with active promotion of the policy to change mentalities (e.g. Ikumen Project
in Japan) designed to reduce the stigma associated with men taking paternity leave, since
in both Korea and Japan such leave is seldom used. Another recent policy change is the
inclusion of “Dad and Partner Pay” for up to two weeks, paid at the rate of the national

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 39

minimum wage within the Australian parental leave scheme. Paternity leave was also
introduced recently in Argentina, Brazil and Hong Kong, China.
As highlighted in Chapter 4, women are often overrepresented in low-wage or
minimum wage occupations. An appropriate minimum wage can help to address
increasing concerns about inequality and rising levels of working poverty. Several
countries have also taken steps to improve the working conditions of domestic workers,
which are predominantly women, and reduce the incidence of informality among them.
France and Canada (Quebec), for example, have introduced model contracts to facilitate
the establishment and formalisation of an employment relationship. Some emerging
economies, have also introduced fiscal measures to encourage formalisation, such as the
“Simples Law” in Brazil or the "Regimen de Incorporation Fiscal" in Mexico. Other
emerging economies, such as South Africa, have taken steps to improve the working
conditions of domestic workers through the introduction of a minimum wage and laying
down hours of work, overtime, salary increases, and leave entitlement (Basic Conditions
of Employment Act, 2002).
In addition to gender gaps in employment participation, women still face a glass
ceiling in getting through to the top of their professions. To increase women’s
representation in decision-making positions, countries have introduced mandatory quotas
(e.g. France, Germany, Italy, Spain), target-setting (Denmark), disclosure initiatives
(United States) and monitoring processes.

Employment promotion and good-quality jobs
The most promising way of promoting equal opportunities and reducing high
inequality is to boost employment and access to good-quality jobs, not only for women
but more generally. As the analysis in Chapter 4 shows, before the crisis income
inequality increased despite strong growth in employment. One of the reasons is that
more than half of the jobs created since the mid-1990s were in non-standard work. While
non-standard arrangements such as part-time work or self-employment may suit many
workers’ efforts to strike the right work-life balance, they also hollow out middle-skill,
middle-income jobs and drive down earnings for low-skilled non-standard workers, thus
contributing to rising income inequality. To break this circle of increasing employment
and higher income gaps, more inclusive labour market policies are needed which focus
not only on the quantity but also on the quality of jobs.
The new OECD conceptual and operational framework to boost job quality suggests
that policies should focus on three dimensions of jobs: earnings quality; labour market
security; and the quality of the work environment (OECD, 2014c). Earnings quality
captures the extent to which the job contributes to workers’ material living standards.
Labour market security captures labour market-related aspects of economic security. For
the advanced economies, this focuses on the risk of unemployment and the income
support workers are entitled to if unemployed. For the emerging economies, hidden
unemployment and under-employment on extremely low pay need also be taken into
account. Quality of the working environment captures non-economic aspects of job
quality, e.g. the incidence of job strain that can impinge on workers’ health and wellbeing. Job strain occurs when high demands on workers are combined with low resources
available to address them.
An important finding for advanced countries is that job quantity and the different
dimensions of job quality tend to be positively related across countries, across population
groups, and across individuals over the lifetime. This implies that with the right mix of
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 40 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
policies and institutions it is possible to do well across all dimensions of labour market
performance. To reduce inequalities in both job quantity and job quality, policies should
target those workers who tend to do poorly along several dimensions, particularly youth
and the low-skilled. By contrast, in emerging economies, job quantity and job quality
often do not go hand in hand. The main issue is generally not open unemployment, but
the lack of better paid and more protected jobs. This calls for policy actions on job
quality. Indeed, there is increasing evidence that job quality and inclusiveness play a
positive role for economic growth and social stability. In all countries, however,
promoting the quality of the work environment impacts not only on the well-being of
workers, but also has direct economic implications, e.g. higher productivity and lower
public health expenditure. In this area, policies should focus on increasing the
effectiveness of occupational health systems to prevent work-related health problems, and
on promoting social dialogue and employers’ social responsibility on these issues.

Active labour market policies
The design and implementation of benefit systems, employment and training
programmes and employment service arrangements are important drivers of
unemployment and under-employment, benefit dependency rates but also the degree of
matching of workers with jobs. Recent OECD studies of activation policies (Connecting
People With Jobs) revealed substantial differences in eligibility conditions for benefits
and their generosity; in the operation of the public and private employment services; in
resources devoted to active labour market programmes; and in the targeting of these
resources to different groups.
There is no unique formula for effective activation. Active labour market measures
for unemployed workers vary among the different target groups. The long-term
unemployed require measures to help them find paid work based on skills acquisition and
work experience (e.g. Work Experience Phase in Australia), combined with intensive and
specific case management and interventions, such as the individualised action plans and
employment service centres in Japan.
For people with disabilities, policies should focus on preventing permanent
withdrawal from the labour market by enhancing their remaining capacity for work,
rather than their incapacity. In countries such as Australia, the Netherlands, Switzerland,
United Kingdom and the European Nordic countries, reforms have been introduced to
turn passive disability benefits into more active benefits with a re-employment
perspective. Countries that have pursued structural reforms of disability benefits by
introducing stronger gate-keeping mechanisms, time-limiting benefits with a or
reassessing the eligibility of existing recipients managed to avert the risk of the long-term
unemployed drifting into disability benefit schemes and curb long-term expenditure
during the global financial crisis. Such policies have been introduced in Sweden, the
Netherlands, Switzerland, and the United Kingdom – all countries that have generally
been able to put disability benefit claims on a declining trend. By contrast, other countries
that have failed to reform disability benefit now face mounting beneficiary rates – in
Estonia and the United States they have risen by over 10% (OECD, 2010).
For older workers, policies should be designed to allow for greater choice in work and
retirement decisions (including part-time employment), and to facilitate employment and
employability at an older age (e.g. the Perspective 50plus in Germany; Targeted Initiative
for Older Workers in Canada). Enhancing employability is getting all the more important
as programmes that fully exempt older unemployed workers from job-search
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 41

requirements have now generally been abolished, and partial exemptions from intensive
activation measures are also becoming more restricted.
Direct job creation schemes can serve as a useful back stop to ensure that the longterm unemployed and other disadvantaged groups maintain a contact with the labour
market (see OECD, 2015c). India has one of the largest public works programme in the
world in terms of coverage – the National Rural Employment Guarantee
Scheme (NREGA) – which plays an important role in reducing short-term poverty and
smooth employment and income throughout the year for rural labourers. The programme
however remains little used, mainly in poorer states because of lack of funding and weak
implementation capacity.
A number of countries have engaged in reforms that simplify a myriad of social
benefit programmes in order to lower administrative costs and increase take-up and
effectiveness of the support provided. The United Kingdom, for instance, is currently
implementing a reform aiming to bring together six different means-tested transfer
programmes into a single Universal Credit, though it is currently too early to tell what the
distributive outcomes of this reform will be.

Targeting young people
More must be done to provide youth with the skills and help their need to get a better
start in the labour market and progress in their career to avoid high drop-out rates and
careers in low-paid work. The OECD Action Plan for Youth recommends a set of
measures to tackle the current situation of high youth unemployment, including
expanding active labour market strategies and encouraging employers to expand quality
apprenticeships or internship programmes, while strengthening the long-term prospects of
youth through strengthening the education system, the role and effectiveness of
Vocational Education and Training and assisting the transition from school to work
(OECD, 2013c).
Helping high-school dropouts through second-chance programmes can be an effective
way to ensure they do not become stuck in low-paid jobs. For example, Second-Chance
Schools in France aim to provide youth aged 18-25 who have left school early with the
opportunity to complete a high-school diploma. It is also important that training and retraining opportunities are available to older individuals who have suffered job loss. A
number of countries, such as Australia, have opened up apprenticeship schemes to
those 25 or older, thus broadening the options for low-skilled older workers to obtain
training.
Prolonged periods of unemployment can also lead to permanent scarring and reduce
future earnings and widen inequality. For youth, different forms of hiring incentives, such
as the Youth Contract in the United Kingdom or the Zérocharges Jeunes in France and
the tax incentives introduced recently in Italy, could be considered as short-run measures
targeted to provide job opportunities, especially for the low‑skilled youth. However, they
should involve some training component and serve as a stepping stone back into regular
employment rather than becoming a low-wage trap. A good example of these schemes
was the UK Future Jobs Fund which aimed at avoiding the scarring of young people as a
result of the global economic crisis. In the medium- to longer-run, many countries need to
introduce better and more attractive options for allowing youth to combine study and
work experience and for encouraging their participation in vocational education and
training. Useful examples in this direction are provided by the German and Australian
experiences.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 42 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Often, there are also distinctive arrangements for unemployed youth including the
recently introduced European Union Youth Guarantee and its predecessor guarantees in the
Nordic countries as well as the prioritisation of education over job search for early school
leavers in Australia. In some countries, unemployed and inactive youth benefit from special
activation programmes such as intensive counselling – the Youth Guarantees – to avoid
being trapped in benefit dependency. Youth guarantees require strong co-ordination
between many service providers (health, social assistance, employment, career guidance,
education and training, etc.) to solve special barriers such as lack of housing, health
problems and devise a concrete development plan. The EU Youth Guarantee aims at
ensuring young people's successful transition into work by ensuring that, within four
months of leaving school or losing a job, young people under 25 can either find a goodquality job suited to their education, skills and experience; or acquire the education, skills
and experience required to find a job in the future through an apprenticeship, traineeship or
continued education. The EU Youth Guarantee is based on successful experience in Austria
and Finland that show that investing in school-to-work transitions for young people
pays off.
Policies that support the recognition of skills and competencies acquired during work
may be helpful, especially for youth who follow a dual path of work and study. For
instance, Slovenia has a programme in place that provides a digital tool for validation of
informally gained work experience called Nefiks. This tool produces certificates that the
young person can subsequently use to prove and demonstrate relevant skills to potential
employers, who might be concerned about credentials (OECD, 2014c).

Reforming labour market institutions
Low pay and in-work poverty are major policy challenges when addressing the
question of rising inequality. Minimum wage settings, provided they are well-designed
and embedded in appropriate tax and transfer settings can help supporting low-wage
workers and low-income families while avoiding significant job losses (OECD, 2015c).
Effective minimum-wage design requires: ensuring that decisions on the level and
adjustments of the minimum wage involve independent commissions; taking account of
differences in average income levels across regions as well as by age; ensuring
enforcement; and reducing social security contributions to lower non-wage labour costs at
the minimum wage. At the same time, minimum wages alone tend to be a rather poor
anti-poverty toll. First, while in-work poverty is associated with low wage levels, a
substantial number of minimum wage workers live in households with income above the
poverty line. Second, lack of co-ordination between minimum wage policy and other
redistribution measures, may reduce the impact of minimum wages: for instance, higher
taxes and reduced benefit entitlements could consume large parts of any minimum
wage increase.
Policies must address not just the level of wage adjustment but also its distribution. In
this respect, previous analysis has shown that declining union coverage had a
disequalising effect on the wage distribution (OECD, 2011). Improving social dialogue
and industrial relations are therefore other important elements of a more equitable and
inclusive growth. High union density and bargaining coverage, and the
centralisation/co-ordination of wage bargaining tend to go hand-in-hand with lower
overall wage inequality in both OECD countries and emerging economies, although there
is some disagreement about the size of these effects and whether they hold for women
(OECD, 2004; Golden and Londregan, 2006; Card et al., 2003; Hayter, 2015). Union
coverage, rather than union membership appears to be more important because in many
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 43

countries there is a legal or administrative extension of collective agreements within an
industrial sector.
Addressing labour market segmentation and more balanced employment protection
are also important elements of enhancing job quality and tackling inequality. Temporary
employment is associated with lower current and future long-term earnings. While
temporary jobs may provide stepping stones to more stable employment, temporary jobs
may be replacing stable jobs instead of encouraging job matching. In addition, in some
countries the likelihood of moving to a stable job is low because of high segmentation in
the labour market, highlighting the importance of having more homogenous employment
protection legislations (EPL) across workers of different types of contracts.
Since the onset of the recent economic crisis, a number of countries with relatively
strict EPL on permanent contracts undertook reforms to relax their regulations on
individual or collective dismissals, reducing the gap in the stringency of regulations
affecting permanent and temporary workers. Such reforms are expected to improve the
allocation of labour to most productive uses and also reduce labour market duality.
However, in the short-term some workers may be affected by the reallocation process; the
number of dismissed workers is likely to increase and the latter are likely to experience
income losses (OECD, 2013f). Therefore, it is important that governments have in place
adequate social protection and unemployment benefit schemes as well as effective reemployment services. Again, any success of EPL reforms will rely on a strong basis of
dialogue between the social partners and the government.
A more equitable labour market also benefits from efficient public employment
services. The responsiveness and effectiveness of activation policies have in many
countries been improved by overarching institutional reforms. Various lines of policy
actions have been considered: one option is the merging of employment assistance with
benefit administration services (United Kingdom, JobCentre Plus); another option is to
create a separate institution charged with co-ordinating the delivery of labour market and
support programmes (United States, Workforce Investment Boards); or to set up an
integrated social assistance information system (Turkey, İŞKUR). The effectiveness of
these initiatives can be further improved through careful performance management,
particularly where services are being outsourced (as in Australia).
Experience from countries such as Finland, Ireland, Switzerland and Australia,
suggests that partnership approaches between organisations and agencies (including those
in the private and not-for-profit sector) can improve the co-ordination of service delivery,
especially for disadvantaged client groups or high-unemployment areas. In Finland, the
alignment of institutional incentives – as national government and local governments
agreed to share the cost of benefit payments to the target group – accompanied the
development of jointly managed service centres for the very long-term unemployed.
The effectiveness of public and private employment services can also be improved
through performance management. Performance is often measured in terms of job
placements and, especially for harder-to-help groups, longer-term employment outcomes.
However, the targets for these outcomes are often set at the national, regional and local
office levels, by ad hoc methods such as negotiation or incremental improvements on the
previous year’s performance. Australia and Switzerland, by contrast, rate local
employment office performance in terms of gross outcomes with regression adjustments
for jobseeker and local labour market characteristics. This approach encourages the
robust operational measurement of the variables involved, helps to identify further factors

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 44 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
influencing performance and, when well-developed, generates relatively accurate and
objective ratings of local office performance.

Skills and education: Investing for opportunities
The third area for promoting equality of opportunities and tackling high inequalities is
investing in human capital through better skills and education policies. As the analysis in
Chapter 2 shows, the inability of individuals from poor socio-economic background to
access higher education and developing their human capital is at the heart of the
transmission mechanism through which income inequality lowers economic growth. The
reverse is true as well: the trend towards higher educational attainment and better skills
has been one of the most important elements to foster economic growth in the long run
and, at the same time, to partially counteract the trend toward higher earnings inequality.
Investment in human capital must start in the vital early childhood period and be
sustained through compulsory education. This will help ensure equality of opportunity for
children from disadvantaged backgrounds. Once the transition from school to work has
been accomplished successfully, there must also be sufficient incentives for workers and
employers to further invest in skills throughout the working life.

Early childhood care and education: Early years potential: use it or lose it
A focus on the early years is crucial in addressing socio-economic differences in
education. Poverty in childhood can result in gaps in the developmental trajectory for
children in the formative years; such gaps can culminate in more severe outcomes in the
long term as children are less able to take full advantage of interventions designed to
promote development. This can lead to the entrenchment of poverty and inequality in
future generations and thus to a lack of social mobility.
To combat poverty and promote child development, childcare services should provide
parents with young children the care support they need to work, as well as the safe
environments needed for pre-schoolers to learn and develop. The OECD PISA results
show that participation in quality early childhood education (as measured by the ratio of
children to staff, the programme’s duration, and spending per child) is associated with
stronger reading performance at age 15, especially for children from families with
disadvantaged socio-economic backgrounds (OECD, 2013c). Insufficient and unequal
access to childcare results from various obstacles, namely the length of parental leave, the
cost of childcare, the tax-benefit system, and the quality, accessibility and availability of
childcare.
While both OECD and emerging economy countries have achieved universal or near
universal primary education, curbing inequalities in secondary and higher education
remains a concern. In terms of educational attainment, while advanced economies are
focusing on school retention until high school completion, emerging economies need to
act on lower secondary schooling – through better learning inputs and more equality of
access – before they can turn to upper secondary education. Conditional cash transfers
can also help promoting the attendance of children from disadvantaged groups, as
experienced in several emerging economies. In addition, improving the physical
infrastructure, teachers’ qualifications and school governance are required to enhance
educational quality. South Africa, for instance, introduced a bursary programme to
encourage students entering into the teaching profession and put in place wage incentives
representing as much as 10% of a starting salary in 2007 to attract teachers in rural and
remote areas (Murtin, 2013).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 45

Addressing early school leaving – i.e. before completion of upper secondary
education – is also crucial and requires early detection of youth at risk of dropping out
early from school and remedial education for those young people who leave education too
early only to become unemployed or inactive. In New Zealand, for instance, schools send
regular reports to the Department of Education about every young person who leaves
school either with or without a qualification. These data are filtered by the department,
which identifies “at-risk” youth (based on information on qualifications and grades) and
directly refers these youth to specialised service providers (OECD, 2014c). Because
school dropout is a particularly acute problem among socio-economically disadvantaged
students, examining school dropout is intrinsically linked to issues of equity and social
mobility (OECD, 2014f).
Success at school depends not only on attendance and school resources (in terms of
teachers’ abilities or school equipment, for instance), but also on parental engagement
with the learning process and what is available in the homes. Therefore, policies that
support families with school-aged children can help reduce future poverty risks and
inequalities.

Policies for families with school-age children
These policies can be designed to encourage attendance at school (overall or at
critical stages) for low-income children, or to meet specific costs for school equipment,
uniforms and meals that might otherwise be a burden on disposable incomes in poor
families. Examples of polices designed to encourage school attendance in OECD
countries include an increase in the means-tested Family Tax Benefit in Australia
between the ages of 13 and 15. This increment is designed to increase family income, and
encourage children to stay in school, when leaving school to earn and supplement the
family income is an option considered for the child.
A broader-coverage policy designed to encourage enrolment in school in poorer
families are Mexico’s Progresa (former Oportunidades) and Brazil’s Bolsa Familia,
which conditions its cash payment on children in the families attending school between
the third grade of primary school until age 16 (see Chapter 7). Finally, in a number of
OECD countries when children get older and may consider leaving school, family cash
benefits or tax breaks are provided if children stay in further or pursue higher education.
Examples of countries applying these policies include Austria, the Czech Republic,
Germany, and Switzerland (higher education); and the United Kingdom (further
education).
Policies specifically designed to enable families to provide school equipment for
children can take the form of regular or one-off cash payments, or income tax reductions
for primary and secondary school children: an example of the former comes from France,
which provides a periodic tax allowance for families with school-aged children; examples
of the latter include the school-clothing grant in Ireland, an annual child allowances for
school-aged children for the purposes of buying school equipment in Israel, an
educational care subsidy paid as part of the Patriot’s Pension in Korea, and the
Portuguese schooling complement (paid as an element in the main Portuguese Family
Allowance). Luxembourg also pays a one-off cash “new year” school allowance per
child.
School support can also come in the form of services, and most often this means
support with food costs. Breakfast clubs, free meals, and free milk all contribute to the
nutritional needs of growing and learning children. Support for poor children in the form
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 46 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
of food supports (milk or school meals) is provided in Mexico, the United Kingdom and
the United States.
Recent analysis by the OECD has shown that countries with childcare systems that
contribute most to poverty reduction maximise access to the service through universal
provision whilst also using tapered fees structures (OECD, 2015d). Universal access
facilitates parental participation in the labour market and progressive fees help pay for the
service by the poorest parents. Denmark and France, for example, both have universal
childcare systems, but childcare provision in Denmark contributes more to poverty
reduction than in France due to a progressive fee structure. Countries in the OECD that
presently provide families with universal access to childcare (i.e. at least 15 hours a week
for at least two consecutive years, regardless of parental income or employment status),
include the Nordic countries, Australia, France, Hungary, New Zealand, Spain and the
United Kingdom. Denmark, Finland, Hungary and Sweden are examples of countries
with childcare fees structures based on differences in earnings and/or household types.

Reducing inequality in educational outcomes
Education policies focusing on equity in education can promote earnings mobility
between generations and reduce income inequality over time. Countries can work towards
this goal by giving equal opportunities to both disadvantaged and advantaged students to
achieve strong academic outcomes – laying a pathway for them to continue on to higher
levels of education and eventually secure good jobs.
Four top OECD performers on the 2012 PISA reading assessment show the potential
of this approach. Canada, Finland, Japan, and Korea all have education systems that put a
focus on equity. In each of these countries, relatively few students performed at lower
proficiency levels on the PISA reading assessment, and high proportions of students
performed better than would be expected, given their socio-economic background.
While each of these countries focuses on equity in education, their approaches differ.
In Japan and Korea, for example, teachers and principals are often reassigned to different
schools, fostering more equal distribution of the most capable teachers and school
leaders. Finnish schools assign specially-trained teachers to support struggling students
who are at risk of dropping out. The teaching profession is a highly selective occupation
in Finland, with highly-skilled, well-trained teachers spread throughout the country. In
Canada, equal or greater educational resources – such as supplementary classes – are
provided to immigrant students, compared to non-immigrant students. This is believed to
have boosted immigrant students’ performance (OECD, 2012b).
Upgrading skills to avoid obsolescence
Too many youth enter the labour market with weak cognitive skills. Low cognitive
skills in terms of numeracy and literacy are strongly associated with low long-term
earnings. But learning does not end with initial education; improving adult competencies
is an essential component to ensure the continuous employability of workers and promote
their career progression. Workers continue to learn at work, and job-related training is
essential to ensure that skills do not become obsolete and that workers can take up betterpaid jobs and thus reduce inequality.
The OECD 2013 Survey of Adult Skills (PIAAC) found large disparities in the
literacy, numeracy and problem-solving competences of adult workers. Unfortunately, the
provision of job-related training is often sub-optimal and tends to benefit workers who
need it the least – notably, those that already possess high skills. The Employer
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 47

Ownership pilot in England, for example, attempts to boost training provision by
encouraging employers to take ownership, as well responsibility, for training their
workforce. The scheme offers all employers in England direct access to up to
GBP 250 million of public investment over two years to design and deliver their own
training solutions, including apprenticeships, training courses and pre-employment
opportunities.
Training provision is particularly low in small and medium-sized enterprises (SMEs).
This is due to the challenge of finding and financing external training providers and to the
opportunity cost of providing on-the-job training by employers in very small firms. To
overcome these difficulties, Korea introduced a subsidy for private or public training
provider forming partnerships or consortia with SMEs in which the providers’ facilities or
equipment are used to develop skills. Subsidies can cover the costs of the facilities,
equipment and salaries for training personnel. Government spending on this programme
has steadily increased since 2003and in 2009 231 000 workers received such training
(Source: Korean Research Institute for Vocational Education and Training – KRIVET).
Understanding the demand for skills to ensure alignment with skills supply
Many skills are not fully used at work. This is the case, for example, of workers who
are mismatched in their job. Yet productivity depends more on actual skills use than on
skills proficiency. Hence, bringing skills use more in line with the skills proficiency of
the workforce would foster a more equal wage distribution. It is necessary to ensure that
skills’ supply is in line with demand. Skills assessment and anticipation exercises are
critical in this respect, as is the ability to effectively translate the information from such
tools into policy action.
While investments in skills are important, it is crucial to ensure that skills taught at
school are relevant for the labour market. If this is not the case, unmet demand for skills
is likely to translate in higher rewards for scarce competences resulting in even higher
income inequality. To achieve a good match between the skills possessed by the labour
force and those required by employers, countries need good information about current
and future skills demands and gaps, and to put in place effective systems and procedures
for turning this information into education, employment and migration policy and
practice. In addition, skills that are relevant to employers can only be developed by or in
co-operation with employers themselves.
Many countries have systems and tools in place for assessing and anticipating skills
needs, but approaches vary significantly in terms of: how they assess skills needs; their
time span; their methods; and their national/regional/sectoral scope. For instance, Canada
carries out analyses of current skills needs along with medium to long-run forecasts to
identify future skills needs and imbalance and tailor immediate policy intervention (e.g.
identify migration opportunities or develop short-term worker training schemes) as well
as long-term policy orientations (e.g. develop apprenticeship programmes in certain
fields).
The outputs of skills assessment and anticipation exercises are widely used to inform
a range of skills-related policies across different domains. In education, they are
commonly used to inform curriculum development and set the number of student places
in upper-secondary, post-secondary and tertiary education programmes. In many
countries skills assessment and anticipation exercises are specifically linked to the
development of Vocational Education and Training (VET) programmes. With regards to
labour market policy, information is used to update occupational standards (which
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 48 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
provide a guide for employers about the skills, training and experience needed to carry
out a job) and to design apprenticeships, re-training and on-the-job training programmes.

Tax and transfer systems for efficient redistribution
The fourth area of an effective policy strategy to curb high inequality relates to taxes
and transfers. These policies constitute the most direct and powerful instrument to
redistribute income. Most OECD countries make substantial use of income taxes and cash
transfers to reduce income gaps. On average, income inequality among the working-age
population is reduced by 26% in the OECD and by 30% in France and Germany,
compared to 19% in the United States and 4-5% in Chile and Mexico. In most emerging
economies, the effect of such redistribution is much weaker, often linked to lower levels
of per-capita social spending and less effective taxation due to a high level of informality.
At the same time, as in OECD countries, social spending on in-kind transfers – services
such as education and health – also tends to lower inequality (see Chapter 7).
In the public debate, redistribution through taxes and transfers is often claimed to
hinder economic growth. The analysis in this report suggests that, if they are welldesigned, higher taxes and transfers to reduce inequality do not necessarily harm growth.
Since taxes and transfers are so powerful in both affecting inequality and growth, getting
them right is key. Previous OECD work (Arjona et al., 2001) indicates that active social
spending, such as in-work benefits and spending on childcare, is associated with higher
growth than “passive” spending, such as on unemployment benefits, which potentially
has negative effects on growth. But the distinction between “passive” and “active”
measures, such as in the case of re-employment support, is not always possible.
Successful activation strategies, for instance, increasingly combine the two, protecting the
livelihood of job losers while encouraging a return to self-sufficiency. OECD (2012c)
discusses some of the trade-offs and complementarities between tax and transfer reforms,
inequality and growth.
Unemployment insurance mitigates the adverse effect of the risk of unemployment by
reducing earnings volatility and by alleviating concerns about not being able to find a job
once unemployed. Recent findings show that the risks of unemployment and low pay are
highly concentrated and that unemployment insurance reduces earnings volatility risk
mainly at the bottom of the distribution (OECD, 2015e). Increasing coverage of
unemployment insurance is a promising avenue for promoting worker security, provided
systems are designed to preserve incentives to work. Such security is especially important
for non-standard workers and those most excluded from the labour market, such as the
long-term unemployed, and particularly in countries where both benefit coverage and
generosity are relatively low.
Over the past decades, rising top incomes means that top earners now have a greater
capacity to pay taxes than before. Moreover, the analysis of the evolution of wealth
distribution in Chapter 6 found that the distribution of wealth is much more concentrated
than income, and is likely to become more so over time, with negative effects on
economic growth. Thus, governments should re-examine a wide range of tax provisions
to ensure that wealthier individuals contribute their share of the tax burden. This aim can
be achieved in several different ways – not only via raising marginal income tax rates on
the rich but also improving tax compliance, eliminating or scaling back tax deductions,
which tend to benefit high earners disproportionally, and reassessing the role of taxes on
all forms of property and wealth, including the transfer of assets. Some countries have
introduced base-broadening measures (Australia, Austria, Denmark, the Netherlands), or
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 49

a reduction in tax credits (France, Greece, the United Kingdom) but others have
introduced measures that reduce the personal income tax base through tax credits
(e.g. Spain).
To effectively pursue the objectives of intergenerational social mobility and equality
of opportunity, capital gains on bequeathed assets should be taxed at a standard rate and
estate taxes should be replaced with an inheritance tax. Most OECD countries tax
inheritances rather than estates, and the past several decades have seen a shift away from
estate taxes, as for example in Australia and Ireland. From the perspective of
intergenerational social mobility, taxing inheritances is preferable to taxing estates since
what matters is how much a person receives from others, not how much a person leaves
to others, and taxing inheritances would reduce wealth and income inequality if
implemented in a revenue-neutral way (see also Piketty, 2014).
Policies should also ensure that not only wealthy individuals but also multinational
firms pay their intended share of the tax burden. This includes measures for increasing
transparency and international co-operation on tax rules to minimise “treaty shopping”
(when companies structure their finances to take account of favourable tax provisions in
different countries) and tax optimisation. In that frame, the OECD-led effort on BEPS
(base erosion and profit sharing) helps developing policies to improve transparency and
tax compliance, and to ensure the automatic exchange of information between tax
authorities.
At the lower end of the income spectrum, large and persistent losses in low-income
groups underline the importance of well-targeted income-support policies. Government
transfers have an important role to play in guaranteeing that low-income households do
not fall further back in the income distribution, but they need to be paired with measures
to re-establish self-sufficiency, prevent long-term benefit dependence, and support
families’ capacities to compensate earnings losses.
Within current budgets, policies to address growing inequality could be made more
efficient, for example, by making more use of in-work benefits which encourage people
to take up paid work and give additional income support to low-income households.
During the past decade, several countries developed schemes which combine minimum
income programmes with active in-work benefits such as the Revenu de solidarité
active (RSA) in France. Also countries with much lower levels of redistribution via the
tax and transfer system increased efforts and introduced transfer programmes to increase
income protection and tackle income inequality such as Prospera, 65 y más or
SinHambre in Mexico.
More generally, while effective social protection requires a strong and sustainable
resource base, it does not necessarily mean that governments need to push up spending
levels. Ensuring that tax revenues are used efficiently means that social support measures
need to be well targeted and implemented. This requires making support accessible to
lower-income individuals, in particular. In addition, policies should be actively adapted
when incomes and labour market situations change, such as during the recent crisis.
This requires counter-cyclical social budgets, first, because the objective need for
support is greater during and after a downturn and, second, because economic upswings
alone are unlikely to undo the damage inflicted by recessions, e.g. because income losses
suffered during downturns become entrenched (OECD 2014a). For instance, countries
such as France, Portugal and the United States, have extended out-of-work benefits at the
onset of the recent crisis. Adjusting funding for active labour market policy in line with
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 50 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
unemployment is another good example. When the number of jobseekers grows during a
downturn, governments should ensure adequate resources for public employment services
and benefit and programme administration as these services act as “gateways” to
activation programmes. For example, Australia, Denmark and Switzerland automatically
adjust budgets for active labour market policies in line with labour market conditions.
Similar provisions should also be considered in other countries in order to protect this
area of social spending during times of fiscal restraint.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 51

Notes
1.

Trends in inequality followed different patterns across OECD countries: in some
countries increases in inequality were concentrated in the earlier decades of the 1980s
and 1990s (e.g. New Zealand, United Kingdom, Mexico) while in others increases
mainly occurred in later years (e.g. Nordic countries), see OECD 2011 (Figure 2).

2.

Levels of income inequality reported in India and Indonesia are closer to the OECD
average. It should be noted, however, that indicators are based on different concepts
and methods (see Chapter 7, Box 7.1).

3.

While much care has been given to make concepts and data as comparable as
possible, the assessment of wealth levels and wealth concentration depends on what is
excluded by the wealth concept used. This is, in particular the case with old-age
pensions (see Chapter 6).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 52 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS

References
Atkinson, A.B. (2015), Inequality: What Can Be Done?, Harvard University Press.
Arjona, R., M. Ladaique and M. Pearson (2001), “Growth, Inequality and Social
Protection”, OECD Labour Market and Social Policy Occasional Papers, No. 51,
OECD Publishing, Paris, http://dx.doi.org/10.1787/121403540472.
Bénabou, R. and J. Tirole (2006), “Belief in a Just World and Redistributive Politics”,
Quarterly Journal of Economics, Vol. 121, No. 2, pp. 699-746.
Card, D., Lemieux, T. and W. Riddel (2003), “Unions and Wage Inequality”, Journal of
Labor Research, Vol. 25, No. 4, pp. 519-559.
Causa, O., A. de Serres and N. Ruiz (2014), “Can Pro-growth Policies Lift All Boats? An
Analysis Based on Household Disposable Income”, Economics Department Working
Paper,
No. 1180,
OECD
Publishing,
Paris,
www.oecd.org/eco/Can_progrowth_policies_lift_all_boats_an_analysis_based_on_household_disposable_income.pdf.
Clerc, M. (2014), “Positioning on the Standards of Living Scale”, Insee Première,
No. 1515, www.insee.fr/en/themes/document.asp?ref_id=ip1515.
Cournède, B., P. Garda and V. Ziemann (2015), “Effects of Economic Policies on
Microeconomic Stability”, OECD Economics Department Working Papers, No. 1201,
OECD Publishing, Paris, http://dx.doi.org/10.1787/5js3f5cwj3jb-en.
Förster, M., A. Llena-Nozal and V. Nafilyan (2014), “Trends in Top Incomes and their
Taxation in OECD Countries”, OECD Social, Employment and Migration Working
Papers, No. 159, OECD Publishing, Paris, http://dx.doi.org/10.1787/5jz43jhlz87f-en.
Golden, M.A. and J.B. Londregan (2006), “Centralization of Bargaining and Wage
Inequality: A Correction of Wallerstein”, American Journal of Political Science,
Vol. 50, No. 1, pp. 208-213, January.
Hayter, S. (2015), “Unions and Collective Bargainingé, in J. Berg (ed.), Labour Markets,
Institutions and Inequality, Edward Elgar Publishing, ILO, Geneva, pp. 95-122.
Murtin, F. (2013), “Improving Education Quality in South Africa”, OECD Economics
Department
Working
Papers,
No. 1056,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/5k452klfn9ls-en.
Niehues, J. (2014), “Subjective Perceptions of Inequality and Redistributive Preferences:
An International Comparison”, IW-TRENDS Discussion Papers, No. 2, Cologne
Institute for Economic Research.
OECD (2015a), “Economic Policy Brief: Pro-growth Reforms: Their Impact on Income
Inequality and Household Economic Vulnerability, OECD Publishing, Paris,
www.oecd.org/economy/Pro-growth-reforms-impact-on-income-inequality-brief.pdf.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 53

OECD (2015b), Economic Policy Reforms 2015, Going for Growth, OECD Publishing,
Paris, http://dx.doi.org/10.1787/growth-2015-en.
OECD (2015c), “Focus on Minimum Wages After the Crisis: Making Them Pay, OECD,
Paris, www.oecd.org/social/Focus-on-Minimum-Wages-after-the-crisis-2015.pdf.
OECD (2015d, forthcoming), Comparing the Effectiveness and Efficiency of Family Cash
Benefits and Services, OECD Publishing, Paris.
OECD (2015e forthcoming), OECD Employment Outlook 2015, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2015-en.
OECD (2014a), Society at a Glance: OECD Social Indicators, OECD Publishing, Paris,
http://dx.doi.org/10.1787/soc_glance-2014-en.
OECD (2014b), Economic Policy Reforms 2014, Going for Growth Interim Report,
OECD Publishing, Paris, http://dx.doi.org/10.1787/growth-2014-en.
OECD (2014c), OECD Employment Outlook 2014, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2014-en.
OECD (2014d), All on Board – Making Inclusive Growth Happen, OECD Publishing,
Paris,
www.oecd.org/inclusive-growth/All-on-Board-Making-Inclusive-GrowthHappen.pdf.
OECD (2014e), The OECD Framework for Inclusive Growth, OECD Publishing, Paris,
www.oecd.org/mcm/IG_MCM_ENG.pdf.
OECD (2014f), Investing in Youth: Brazil,
http://dx.doi.org/10.1787/9789264208988-en.

OECD

Publishing,

Paris,

OECD (2013a), Economic Policy Reforms 2013, Going for Growth, OECD Publishing,
Paris, http://dx.doi.org/10.1787/growth-2013-en.
OECD (2013b), “Recommendation of the Council on Gender Equality in Education,
Employment, and Entrepreneurship”, OECD, Paris, www.oecd.org/gender/CMIN(2013)5-ENG.pdf.
OECD (2013c), “The OECD Action Plan for Youth”, www.oecd.org/els/emp/YouthAction-Plan.pdf.
OECD (2013d), PISA 2012 Results: Excellence Through Equity: Giving Every Student
the
Chance
to
Succeed,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/9789264201132-en.
OECD (2013e), OECD Skills Outlook 2013: First Results from the Survey of Adult Skills,
OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264204256-en.
OECD (2013f), OECD Employment Outlook 2013, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2013-en.
OECD (2012a), Closing the Gender Gap: Act Now, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264179370-en.
OECD (2012b), “How Pronounced Is Income Inequality Around the World – And How
Can Education Help Reduce It?”, Education Indicators in Focus, No. 04/1012, OECD,
Paris, www.oecd.org/edu/50204168.pdf

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 54 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
OECD (2012c), “Income Inequality and Growth: The Role of Taxes and Transfers”,
OECD
Economics
Department
Policy
Notes,
No.
9.
January,
www.oecd.org/eco/public-finance/49417295.pdf.
OECD (2011), Divided We Stand: Why Inequality Keeps Rising, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264119536-en.
OECD (2010), Sickness, Disability and Work: Breaking the Barriers. A Synthesis of
Findings
across
OECD
Countries,
OECD
publishing,
Paris,
http://dx.doi.org/10.1787/9789264088856-en.
OECD (2008), Growing Unequal? Income Distribution and Poverty in OECD Countries,
OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
OECD (2004), OECD Employment Outlook 2004, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2004-en.
OECD-ILO-IMF-WB (2014), “Achieving Stronger Growth by Promoting a More Gender
Balanced Economy”, Report prepared for the G20 Labour and Employment
Ministerial Meeting Melbourne, Australia, 10-11 September 2014.
Osberg, L. and T. Smeeding (2006), “‘Fair’ Inequality? Attitudes Toward Pay
Differentials: The United States in Comparative Perspective”, American Sociological
Review, Vol. 71, No. 3, pp. 450-473.
Ostry, J., A. Berg and C. Tsangarides (2014), “Redistribution, Inequality, and Growth”,
IMF Staff Discussion Note, February.
Piketty, T. (2014), Capital in the 21st Century, Harvard University Press.
Database references
OECD Income Distribution Database (IDD), www.oecd.org/social/income-distributiondatabase.htm.
OECD Wealth Distribution Database, forthcoming.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 55

Annex 1.A1
Key indicators of income inequality and poverty

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 56 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS
Table 1.A1.1. Key indicators on the distribution of household disposable income and poverty, 2007, 2011 and 2013 or most recent year
Income share in total income
Gini coefficient

Bottom
10%

Australia
Austria
Belgium
Canada
Chile
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Israel
Italy
Japan
Korea
Luxembourg
Mexico
Netherlands
New Zealand
Norway
Poland
Portugal
Slovak Republic
Slovenia
Spain
Sweden
Switzerland
Turkey
United Kingdom
United States
OECD

Poverty rate (threshold
"anchored" in 2005)

Poverty rate (relative threshold)

S90/S10 income share ratio

2007

2011

2013 or
latest
available
year

2007

2011

2013 or
latest
available
year

0.336
0.284
0.282
0.318
0.511
0.257
0.246
0.316
0.269
0.293
0.287
0.333
0.272
0.286
0.305
0.371
0.313
0.329
0.312
0.279
0.475
0.295
0.330
0.250
0.316
0.360
0.249
0.240
0.328
0.259
..
0.409
0.361
0.378

..
0.281
0.275
0.315
0.503
0.262
0.251
0.334
0.264
0.309
0.291
0.337
0.289
0.256
0.302
0.377
0.323
0.336
0.307
0.278
0.466
0.283
0.323
0.250
0.306
0.343
0.264
0.247
0.342
0.273
0.289
0.412
0.344
0.390

0.326
0.276
0.268
..
..
0.256
0.249
0.339
0.262
0.306
0.289
0.340
0.288
0.256
0.304
0.360
0.327
..
0.302
0.302
0.482
0.278
0.333
0.253
0.298
0.338
0.251
0.250
0.335
0.274
0.285
..
0.351
0.401

9.3
6.9
6.7
8.6
29.5
5.3
5.1
8.2
5.8
6.8
6.7
10.5
6.0
6.5
7.0
13.4
8.9
10.3
10.0
6.2
26.8
7.1
8.3
5.9
8.1
10.4
5.3
5.2
9.9
5.8
..
14.5
11.1
15.1

..
7.1
6.3
8.6
26.5
5.6
5.2
9.6
5.6
7.4
6.8
12.7
7.3
5.7
7.6
12.5
10.3
10.7
10.2
6.0
28.5
6.7
8.0
6.1
7.8
10.0
5.9
5.3
12.1
6.3
7.0
15.2
9.6
17.9

8.8
7.0
5.9
..
..
5.4
5.2
9.7
5.5
7.4
6.6
12.3
7.2
5.6
7.4
14.9
11.4
..
10.1
7.1
30.5
6.6
8.2
6.2
7.4
10.1
5.7
5.4
11.7
6.3
6.7
..
10.5
18.8

0.314

0.314

0.315

9.2

9.5

9.6

Bottom
20%

Bottom
40%

Top
40%

Top
20%

Top
10%

2013 or latest available year (%)

2.8
3.1
3.6
2.8
1.5
4.0
4.0
2.6
3.9
3.4
3.6
2.0
3.1
3.8
3.2
1.7
2.2
2.3
2.2
3.4
1.2
3.4
3.1
3.3
3.2
2.6
3.5
3.7
2.1
3.5
3.5
2.1
2.7
1.6
2.9

7.2
8.5
8.8
7.6
4.3
9.9
9.8
7.0
9.4
8.5
8.8
6.4
8.3
9.6
8.3
5.5
6.9
6.5
6.8
8.5
3.9
8.8
7.6
9.1
8.1
7.0
9.1
9.2
6.5
8.7
8.7
5.6
7.2
5.2
7.7

19.7
22.6
22.6
20.4
12.5
24.3
24.3
19.0
23.6
21.8
22.4
18.8
22.0
24.0
21.3
17.1
19.7
18.9
20.5
21.6
12.5
22.7
19.7
24.1
21.5
19.5
23.8
23.9
19.0
22.7
22.4
16.0
19.3
16.0
20.6

63.0
59.2
58.8
62.2
75.3
58.0
57.3
64.2
58.3
61.3
60.2
63.6
60.2
57.9
61.6
65.8
62.8
63.7
61.1
61.1
73.9
59.4
63.8
57.3
61.0
63.7
57.7
57.4
63.7
59.0
59.8
69.0
64.5
68.2
62.2

39.8
36.0
35.2
39.1
56.4
35.7
34.5
40.9
35.5
39.5
37.8
40.3
37.0
35.2
38.7
41.8
39.7
40.0
37.1
38.7
52.9
36.7
40.7
34.5
38.1
41.1
34.2
34.2
40.1
36.1
37.3
47.4
42.6
45.5
39.4

24.4
21.6
20.8
24.2
40.9
21.7
20.8
25.0
21.5
25.3
23.5
25.1
22.5
21.3
23.8
25.6
24.7
24.4
21.9
24.2
36.7
22.4
25.7
20.6
23.2
25.9
19.7
20.0
24.4
21.9
23.2
31.7
28.0
30.0
24.6

Total

By age group, latest available year

2007

2011

2013 or
latest
available
year

14.6
9.7
9.5
11.3
19.2
5.5
6.1
14.1
7.8
7.2
8.5
13.3
6.4
6.7
9.6
19.9
11.9
15.7
14.8
7.2
6.7
11.0
7.8
9.6
12.8
7.0
8.0
14.5
8.4
..
17.0
11.6
17.3

..
9.1
9.5
11.8
17.8
6.3
5.8
11.8
7.5
8.0
8.7
15.2
10.3
5.9
9.6
20.9
12.8
16.0
14.6
8.3
20.4
7.4
9.8
7.7
11.2
12.0
8.5
8.9
14.7
9.7
10.5
19.2
9.5
17.9

14.0
9.6
10.2
..
..
5.3
5.4
12.3
7.1
8.1
8.4
15.1
10.1
6.3
8.4
18.6
12.7
..
14.6
8.4
21.4
7.9
9.9
8.1
10.4
12.9
8.5
9.4
14.1
9.0
9.1
..
10.5
17.6

11.0

11.2

11.2

20.9

Total

Children
(< 18)

Youth
(18-25)

Adult
(26-65)

Elderly
(> 65)

Working
poor

12.9
10.4
11.4
14.4
23.5
8.2
2.7
11.8
4.6
11.4
7.4
21.4
11.8
8.1
8.7
24.3
17.4
15.7
8.0
12.5
25.8
10.7
12.8
5.9
12.7
17.8
14.9
8.6
21.0
8.3
8.0
28.4
10.4
19.6
13.3

7.9
10.1
11.2
13.1
15.1
5.7
21.7
13.5
15.9
13.7
12.5
21.2
11.9
11.4
10.5
16.6
14.7
18.7
9.1
8.2
15.0
21.9
10.4
30.0
10.3
15.8
7.3
7.1
16.8
17.8
6.6
16.2
10.9
20.1
13.8

11.3
8.7
9.4
11.8
15.1
4.9
3.5
12.2
6.3
7.1
7.7
14.8
9.6
5.0
8.1
14.0
12.1
13.9
9.7
8.0
18.6
6.1
8.9
5.7
10.1
12.5
7.6
8.5
13.5
7.4
6.1
14.4
9.6
15.2
9.9

33.5
11.4
10.7
6.7
20.5
2.8
4.6
12.6
7.8
3.8
9.4
6.9
8.6
3.0
6.9
24.1
9.3
19.4
49.6
3.0
31.2
2.0
8.2
4.1
8.4
8.1
4.1
15.9
6.8
9.4
23.4
18.4
13.4
21.5
12.6

4.6
7.8
6.1
8.7
14.3
4.7
3.9
9.0
4.1
7.3
3.0
15.8
7.2
6.0
5.0
13.9
12.0
12.9
..
7.9
19.0
6.6
5.8
6.7
9.2
12.2
8.0
6.6
13.6
5.8
5.7
17.8
5.3
11.9
8.7

2007

2011

2013 or
latest
available
year

..
8.2
8.4
9.8
20.6
3.3
5.0
4.4
6.4
..
9.2
11.4
..
3.8
7.2
..
10.7
..
14.4
6.9
17.6
6.1
5.5
..
5.1
10.7
3.3
5.5
8.8
..
..
..
11.4
..

..
7.3
8.1
9.3
14.7
4.2
4.7
7.1
4.9
7.1
8.6
26.7
13.0
7.2
15.3
17.8
13.5
19.6
..
8.4
19.7
7.1
6.8
5.1
4.1
12.4
2.1
6.9
12.6
5.1
7.9
..
10.6
19.5

8.2
7.9
7.8
..
..
3.6
4.6
6.0
4.6
7.5
7.9
32.3
9.1
6.9
14.8
13.2
14.9
..
..
9.6
21.1
7.0
5.7
5.0
3.8
13.6
2.3
8.2
13.7
4.8
6.1
..
11.8
19.4

7.6

9.5

9.9

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS – 57

Table 1.A1.1. Key indicators on the distribution of household disposable income and poverty, 2007, 2011
and 2013 or most recent year (cont.)

Note:
Income distribution data refer to the total population and are based on equivalised household disposable income, i.e. disposable
income adjusted for household size. The Gini coefficient takes values between 0 (where every person has the same income), and
1 (where all income goes to one person). The S90/S10 income share ratio refers to the ratio of average income of the top 10% to
the average income of the bottom 10% of the income distribution. The poverty threshold is 50% of median disposable income in
each country. Working poor are those with income below the poverty line, living in households with a working age head and at
least one worker.
The latest available data refer to 2014 for Hungary; 2013 for Finland, Hungary, Israel, Netherland and United States, 2011 for
Canada, Chile and Turkey, 2009 for Japan, and 2012 for the other countries. Data shown for 2011 refer to 2012 for Hungary,
Korea and the United States; 2010 for Mexico; and 2009 for Japan. Data shown for 2007 refer to 2008 for Australia, France,
Germany, Israel, Mexico, New Zealand, Norway, Spain, Sweden and the United States; 2006 for Chile and Japan. There is a
break in series for Israel after 2011. In the case of Japan, data are based on the Comprehensive Survey of Living Conditions,
submitted to the OECD by Japan. Other surveys, such as the National Survey of Family Income and Expenditure suggest lower
levels of income inequality and poverty in 2009. The lack of data beyond 2009 is due to the fact that no recent estimates were
provided by the Japanese authorities. The OECD average does not include Israel. The OECD average for 2011 includes 2012
data for Hungary. The OECD average for the latest year available includes 2011 data for Canada, Chile and Turkey, and 2009
data for Japan.
Source: OECD Income Distribution Database (IDD), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933208820

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 58 – 1. OVERVIEW OF INEQUALITY TRENDS, KEY FINDINGS AND POLICY DIRECTIONS

Box 1.A1.1. Towards a better understanding of people’s perceptions of income inequality
Statistics on income inequality are regularly produced by researchers and statistical offices around the
world, and easily make it to the headlines news. What remains unclear is how much inequality people perceive,
and what degree of inequality they regard as “ideal” or “acceptable”. The second issue, i.e. preferences for how
income and other valuable resources are distributed, has a long history in social sciences (Bénabou and Tirole,
2006; Osberg and Smeeding, 2006). The first issue, i.e. to what extent people have a good appreciation of
income distributions in their country and of their position in it, has received less attention although it is,
arguably, just as important.
Research based on data from the 2009 wave of the International Social Survey Programme on social
inequality shows that people systematically misperceive the level of income inequality in their countries,
although often in different directions (Niehues, 2014). For example, while people living in Hungary, Slovenia
and the Czech and Slovak Republics overestimate the income inequality in their country, those living in the
Nordic countries seem to be aware of living in a relatively equal society, and those in the United States
substantially underestimate the extent of income inequality.
The issue of people’s (mis)perception of income inequality has also attracted the interest of national
statistical agencies. In 2011, the French National Institute of Statistics and Economic Studies (INSEE) launched
a survey asking respondents to position themselves on the income scale: results showed that, while for 45% of
respondents there was little discrepancy between perceptions and realities, most poor people (three out of five)
overestimated their position, and most rich people (four out of five) underestimated it. Similarly, a survey
conducted by TNS Gallup in Finland showed that, while most people’s idea of ‘low income’ corresponds quite
closely to that used by Statistics Finland, respondents’ own income influenced their perception: the higher their
income, the higher their view of what “low” and “high” income is.
The OECD’s “Compare your income” web-tool (www.oecd.org/statistics/compare-your-income.htm)
allows users from different OECD countries to compare perceptions and realities, i.e. where in the income
distribution of their country they fit. The tool, which is similar to those available in a range of countries, is based
on
the
most
recent
data
from
the
OECD
Income
Distribution
Database
(http://stats.oecd.org/Index.aspx?DataSetCode=IDD). While the application is completely anonymous, it will
generate information on users’ perceptions and characteristics that could allow different types of analysis, once
sufficient data have been collected.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 59

Chapter 2
The impact of income inequality on economic growth

Drawing on harmonised data covering the OECD countries over the past 30 years, this
chapter first explores whether income inequality has an impact on subsequent growth. In
particular, it focuses on the growth consequences of income inequality in different parts of
the income distribution, using measures of “top” and “bottom” inequality. The chapter
then evaluates the “human capital accumulation theory”, one prominent channel through
which inequality is supposed to affect economic growth. Exploiting micro data from the
Adult Skills Survey (PIAAC), it looks at the consequences of income inequality for the skills
development of individuals with different parental education backgrounds, both in terms of
the quantity of education attained (e.g. years of schooling) and in terms of its quality
(e.g. skill proficiency).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 60 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

2.1. Introduction and key findings
As documented in Chapter 1, Income inequality has been rising over the past three
decades in the vast majority of OECD countries. Addressing these trends has moved to
the top of the policy agenda in many countries. This is partly due to worries that a
persistently unbalanced sharing of the growth dividend will result in social resentment,
fuelling populist and protectionist sentiments and leading to political instability. Recent
discussion, particularly in the United States, about whether increased inequality might be
a cause of the 2008 financial crisis has also made it more relevant to policy making. 1
Another reason for the growing interest in inequality is the concern that cumulatively
large and sometimes rapid increases in income disparity might have an effect on
economic growth and on the pace of exit from the current recession. Is inequality a
pre-requisite for growth? Or does a greater dispersion of incomes across individuals
instead undermine growth? And which are the main channels through which inequality
affects economic performance? Adding to a recent literature re-addressing these
controversial and long-standing issues (see Ostry et al., 2014; Halter et al., 2014), this
analysis will focus on the case of the OECD countries.
The chapter starts by briefly reviewing the theoretical and empirical literature on how
inequality might affect growth. Section 2.2 then presents the core of the new empirical
evidence on the links between income inequality and economic growth. Section 2.3
explores one of the main transmission mechanisms between inequality and growth: the
human capital channel. Section 2.4 draws concluding remarks.
The chapter highlights the following key findings:


When income inequality rises, economic growth falls.



The negative effect of inequality on growth is determined by the lower part of the
income distribution: not just the poorest decile but the bottom 40% of income
earners.



Redistribution through income taxes and cash benefits does not necessarily harm
growth.



Inequality has a negative impact on growth through the channel of human capital:
the wider is income inequality, the lower is the chance that low-income
households invest in education.

2.2. How inequality may affect economic growth
Over the last decades, a large body of theoretical and empirical research has
attempted to determine whether inequality is good or bad for growth. Theoretical work
has provided mechanisms supporting both possibilities, and the large empirical literature
attempting to discriminate between these mechanisms has been largely inconclusive. This
section provides a brief overview of both theoretical and empirical works, highlighting
the main methodological and measurement issues and setting the stage for the new work
on the OECD countries, described below.

Theoretical literature
Alternative theories predict that inequality can affect growth in either a positive or
negative direction. Greater inequality might reduce growth if:
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 61



Theory A: Greater inequality becomes unacceptable to voters, so they insist on
higher taxation and regulation, or no longer trust business, and pro-business
policies, all of which may reduce the incentives to invest. This is referred as the
“endogenous fiscal policy” theory (see Bertola, 1993; Alesina and Rodrik, 1994;
Persson and Tabellini, 1994; Perotti, 1996).2 In extreme cases, inequality may
lead to political instability and social unrest, with harmful effects on growth
(Alesina and Perotti, 1996; Keefer and Knack, 2000).



Theory B: The presence of financial market imperfections implies that
individuals’ ability to invest depends on their level of income or wealth. If this is
the case, poor individuals may not be able to afford worthwhile investments. For
example, lower-income households may choose to leave full-time education if
they cannot afford the fees, even though the rate of return on education (to both
the individual and society) is high. In turn, under-investment by the poor implies
that aggregate output would be lower than in the case of perfect financial
markets.3 This view, first formalised by Galor and Zeira (1993), is referred as the
“human capital accumulation” theory.4
The idea that higher inequality may result in under-investment in human capital
by the poorer segments of society has also spurred a significant amount of
research on the consequences of inequality on social mobility and the allocation
of talents across occupations (Banerjee and Newman, 1993; Fershtman et al.,
1996; Owen and Weil, 1998; Checchi et al., 1999; and Hassler et al., 2007).



Theory C: The adoption of advanced technologies depends on a minimum critical
amount of domestic demand, which might not be sufficient if the poorer sections
of society have little resources. While originating from the modelling by Murphy
et al. (1989) of the first stages of industrial take-off, and therefore initially
perceived as tangential to the case of advanced economies, the domestic demand
channel has recently been put forward again in, for example, the recent debate on
the consequences of rising US inequality for economic performance (Krueger,
2012; Bernstein, 2013).
On the other hand, greater inequality might increase growth if:



Theory D: High inequality provides the incentives to work harder and invest and
undertake risks to take advantage of high rates of return (Mirrlees, 1971; Lazear
and Rosen, 1981).5 For example, if highly educated people are much more
productive, then high differences in rates of return may encourage more people to
seek education.



Theory E: Higher inequality fosters aggregate savings, and therefore capital
accumulation, because the rich have a lower propensity to consume (Kaldor,
1955; Bourguignon, 1981).6

Empirical evidence
The large empirical literature attempting to establish the direction in which inequality
affects growth is summarised in the literature review in Cingano (2014, Annex 2). That
survey highlights that there is no consensus on the sign and strength of the relationship;
furthermore, few works seek to identify which of the possible theoretical effects is at
work. This is partly traceable to the multiple empirical challenges facing this literature.
These include:
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 62 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH


Estimation method: Most empirical works focus on reduced-form estimates of the
impact of inequality within the theoretical framework of empirical growth
models. Early studies, exploiting cross-sectional variation, yielded negative
coefficients (e.g. Alesina and Rodrik, 1994; Persson and Tabellini, 1994; Perotti,
1996). Later work using within-country variation (i.e. cross-country, time series
panel data techniques), on the other hand, found the link to be positive or not
significant (Li and Zou, 1998; Forbes, 2000; Andrews et al., 2011).7 More recent
analyses attempt to take advantage of both within-country and cross-country
variation exploiting the Generalised Method of Moments (GMM, see Box 2.1).



Data quality: The literature has also been largely constrained by the availability
and quality of income distribution data across countries, which are usually
assembled based on heterogeneous national sources (i.e. “secondary” datasets).
This implies that the inequality measures normally differ as to coverage, reference
unit, weighting and the definition of income.



Country coverage: The literature survey also highlights a role for the data country
coverage in affecting the results. The channels predicting a negative inequalitygrowth relationship (in particular, the channels of credit market imperfections and
of socio-political instability) are likely to be stronger in developing countries than
in advanced countries. Previous work suggested that that the link between
inequality and growth is negative among poor countries, but positive or
insignificant among rich countries (Barro, 2000). Studies that include both
developing and developed countries may therefore capture an average effect,
giving misleading results.



Inequality indicators: The impact of inequality on growth has been often analysed
based on a single synthetic measure of income inequality (typically, the Gini
coefficient). However, the links between inequality and growth might vary
depending on inequality in different parts of the income distribution
(Voitchovsky, 2005). For example, many of the negative mechanisms
(e.g. financial market imperfections, political instability) are associated with
inequality at the bottom end of the distribution; most of the positive mechanisms
(e.g. based on different savings propensities or on incentive considerations) are
more likely to depend on the degree of inequality in the top of the income
distribution. Hence, a single inequality statistic may end up capturing an average
effect of inequality on growth. This possibility calls for complementary indicators
of the profile of income inequality (for example, ratios of income percentiles on
either side of the median or average, or decile share ratios).

Parallel to the reduced-form inequality-growth literature, a more limited set of studies
has looked at the channels through which inequality may affect growth by focusing in
particular on the channels of endogenous fiscal policy (theory A above) and of human
capital accumulation (theory B), and, more generally, on the link between inequality and
social mobility.


Research on the endogenous fiscal policy channel provides weak evidence of a
positive association between inequality and fiscal redistribution (see Perotti, 1994,
1996; Persson and Tabellini, 1994; and De Mello and Tiongson, 2006, for a
survey); moreover, the link between redistribution (e.g. the amount of taxes) and
growth is found to be only weakly negative, or even positive (see Bergh and
Henrekson, 2011).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 63



Direct estimates of the interplay between inequality and imperfect financial
markets in shaping investment decisions produced (weak) evidence that the
negative impact of income inequality on human capital accumulation increases
according to the degree of financial imperfection (see Perotti, 1994, 1996;
Deininger and Squire, 1998). Evidence based on aggregate data, however, does
not allow inferring whether the sign and strength of the relationship varies across
individuals depending on their socio-economic background, as predicted by core
models following Galor and Zeira (1993).



The available evidence on the links between inequality and social mobility is also
largely based on cross-country correlations such as the so-called “Great Gatsby
Curve”, showing a negative relationship between inequality and intergenerational
earnings mobility in a subset of OECD countries (D’Addio, 2007; Corak, 2013).
Cross-country correlations are clearly only suggestive of the possible link
between inequality and mobility. Recent work by Chetty et al. (2014) based on
administrative data found that (upward) mobility is negatively correlated with
income inequality (and positively with school quality) across areas within the
United States.8

2.3. The impact of inequality on growth
A summary of the approach and the new evidence
The present analysis on how inequality affects growth in the OECD countries
attempts to take into account the above-mentioned issues with regard to estimation
method, data quality and coverage, and appropriate indicators. It also provides evidence
on one prominent intermediate mechanism (the financial market imperfection/human
capital accumulation theory), investigating whether inequality disproportionately affects
the investment decisions of disadvantaged individuals (see above).
Estimation method: The empirical equation estimates growth as a linear function of
initial inequality, income, and human and physical capital; the model is similar to that
used in most empirical analyses of growth determinants augmenting the Solow growth
model (see Annex 2.A1). The equation is estimated using panel data, so the baseline
regression specification takes the form:

𝑙𝑛𝑦𝑐,𝑡 − 𝑙𝑛𝑦𝑐,𝑡−1 = 𝛼𝑙𝑛𝑦𝑐,𝑡−1 + 𝑋𝑐,𝑡−1 𝛽 + 𝛾𝐼𝑛𝑒𝑞𝑐,𝑡−1 + 𝜇𝑐 + 𝜇𝑡 + 𝜖𝑐,𝑡

[1]

where c denotes a particular country and (c, t-1) is a time interval of five years. The
variable lny is the log of real GDP per capita so that the left-hand side of equation [1]
approximates five-year growth in a country. On the right hand-side, Ineq is a summary
measure of inequality; per capita GDP (yc-1) is the standard control for convergence, and
the vector X contains a minimum set of controls for human and physical capital (see
Annex 2.A1 for a detailed description of variables and sources). Using panel data allows
accounting for country and time fixed effects (c and t). The country dummies are
included to control for time-invariant omitted-variable bias, and the period dummies are
included to control for global shocks, which might affect aggregate growth in any period
but are not otherwise captured by the explanatory variables.
In the baseline specification the relevant explanatory variables are measured at the
beginning of the growth spell in order to mitigate the concerns that the GDP dynamics feed
back into inequality (reverse causality). The analysis will exploit the Generalised Method of

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 64 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Moments (GMM) as opposed to ordinary least squares or least square dummy variable
estimators (see Box 2.1 for a description). More specifically, all results are based on the
“System GMM” estimator, which exploits variation in inequality both between- and withincountry (over time). Hence, it exploits the largest source of variation in inequality
(i.e. across countries) while accounting for other potentially relevant country-specific
explanatory factors. GMM allows taking into account the estimation issues arising due to
the presence of a lagged dependent variable (lnyc,t-1), the so-called Nickell-bias.
Box 2.1. GMM (Generalised Method of Moments) estimators in growth regressions
Because most empirical growth models are based on the hypothesis of conditional convergence, growth
equations such as [1] contain some dynamics in lagged output (the independent variable lnyc,t-1) and can be
rewritten as a dynamic panel data model:
𝑙𝑛𝑦𝑐,𝑡 = (1 + 𝛼)𝑙𝑛𝑦𝑐,𝑡−1 + 𝑋𝑐,𝑡−1 𝛽 + 𝛾𝐼𝑛𝑒𝑞𝑐,𝑡−1 + 𝜇𝑐 + 𝜖𝑐,𝑡

[1a]

Standard panel data approaches to estimate [1a], like fixed effect models, are unlikely to yield unbiased
estimates of the parameters of interest ( and ). In fact, applying the within transformation, or taking first
differences, creates a correlation between lnyc,t-1 and the error term such that the fixed-effect estimator of  is
necessarily biased (Nickell, 1981). More importantly, these approaches would yield biased estimates of the
coefficients of any independent variable, including Ineqc,t-1, that is correlated with lnyc,t-1.
Specific GMM estimation techniques have been developed to deal with these problems: the first-difference
GMM estimator and the System GMM estimator. The first-difference GMM estimator, developed by Arellano
and Bond (1991), eliminates the country-specific effect by differencing model [1a], and uses lagged values of
the right-hand-side variables (e.g. lnyc,t-2, lnyc,t-3, Ineqc,t-2 etc.) as instruments for their change. Arellano and
Bond (1991) show that, in particular, consistent estimates require the absence of serial correlation in the error
term, c,t. Accordingly, they provide a test of autocorrelation in the residuals, i.e. a test that the differenced error
terms are not second-order serially correlated.
The main drawback of first-difference GMM estimates in the current context is that variables such as
inequality display notable persistence within a country. Hence, taking first differences eliminates most of the
variation in the data, and implies that the lagged levels of the explanatory variables are weak instruments for the
variables in differences giving rise to large biases and imprecision (see e.g. Blundell and Bond, 1998; Bond
et al., 2001).
Following the most recent papers on inequality and growth (Ostry et al., 2014; Halter et al., 2014), the
empirical analysis exploits the System GMM estimator proposed by Arellano and Bover (1995) and Blundell
and Bond (1998). The System GMM estimator combines first-differenced equations (as in the difference GMM)
with an additional set of equations in levels where lagged first-differences of the right-hand side variables are
used as instruments. It therefore rests on the assumption that first-differences are not correlated with the country
fixed effect. In the context of growth regressions, this implies assuming that the deviation of initial observations
(e.g. lnyc,1) from their steady states must be uncorrelated with the country-specific fixed effects (see Blundell
and Bond, 1998, p. 124). To detect possible violations of these requirements, we regularly apply difference-inHansen tests to the instruments for the level equation as a group (as suggested by Roodman, 2009).
The analysis also follows the suggestion by Roodman (2009) to account for the problem of “too many
instruments”, which requires that the number of variables in the instrumental matrix be lower than the number
of countries. Note, however, that even the System GMM has been shown to be subject to weak instruments
problems (Bazzi and Clemens, 2013).

Data quality: The analysis focuses on a sample of advanced and relatively similar
economies so as to avoid the problem that a different relationship between inequality and
growth may exist depending on the level of development (see Barro, 2000). A newlyassembled unbalanced panel was exploited, with variables measured at five-year intervals
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 65

over the period 1970-2010 for 31 OECD countries (see Annex 2.A1 for details). Data on
GDP, the working-age population and gross fixed capital formation are from the OECD
Annual National Accounts. The average years of schooling of the working-age population
are from the recently updated version of the Barro and Lee (2013) dataset.
The measures of inequality are based on the standardised data from the OECD
Income Distribution Database (IDD, www.oecd.org/social/income-distributiondatabase.htm), a high-quality data source. It contains information on income measured
both before and after income tax and cash transfers, which provides a proxy measure of
the extent of redistribution. In-kind benefits and consumption taxes, however, are
excluded, as the underlying income surveys do not provide this information.
Redistribution through public services, such as health, education, social housing and
assistance, or through services to the unemployed and active labour market policies is
therefore not taken into account.9
Inequality indicators: The OECD IDD allows a variety of measures of inequality to
be tested, including the Gini coefficient (computed on both disposable and market
income) as well as indicators that focus specifically on either the upper or the lower
ends of the distribution. The bottom inequality in a country is obtained as the ratio
between (overall) average income and average income of one bottom decile (e.g. the
second). An increase in this ratio signals a widening gap between average and poor
households, i.e. higher inequality at the bottom. The top inequality is measured as the
ratio between average income in one top decile (e.g. the eighth) and overall average
income, and therefore informs about the gap between rich and average households.
Hence the analysis can allow for the possibility that different forms of inequality have
different consequences for growth.
Sample size: While restricting the analysis to OECD countries allows using high
quality data on income distribution, it comes at the cost of focusing on a limited number
of countries and, ultimately, on a final dataset of less than 130 observations. This limits
the type of analysis one can reliably perform when using GMM techniques which,
despite its many advantages, is quite data demanding. The estimates are therefore
sensitive to reductions of the sample size stemming, for example, from focusing on subperiods or on subsets of countries. The small sample size also exacerbates the
estimation concerns arising because, in GMM, the number of instruments can easily
become “too large” relative to the number of observations (Roodman, 2009). Among
other things, this required focusing on a parsimonious set of control variables. Indeed,
extending the baseline specification (for example, to look for non-linearity or
heterogeneity of the effects) provided inconclusive results (see below). All these
caveats suggest that quantifying the inequality-growth nexus remains a daunting task
and that the coefficients estimated and discussed below are to be taken as indicative.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 66 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

Baseline results
The first part of the analysis estimates the baseline equation [1], focusing on net income
inequality (with Ineqct = Ginict); it also considers the role of redistribution as captured by the
difference between inequality in market income and in disposable income.
The empirical results show that inequality has a negative impact on economic growth.
The baseline results are reported in columns 1 to 4 of Table 2.1. The results in column 1
refer to a baseline specification in which growth depends only on initial income and
inequality. In column 2 the model is augmented with standard growth determinants such
as human and physical capital, which does not affect the above finding.10 On the other
hand, the estimated coefficients on human and physical capital are not statistically
significant, a result that is not affected by using alternative measures or specifications.11
Columns 3 and 4 explore the same model, changing the specification of the instrumental
variable matrix to address the problem of “instrument proliferation”, which has been
shown to lead to severe biases and weakened tests of instrument validity (see Roodman,
2009).12 While the p-values on the Hansen tests fall with respect to their high value when
using a high number of instruments relative to the number of countries (potentially
mirroring an instrument-proliferation problem in unrestricted regressions), the estimated
coefficients are if anything slightly larger.13
Table 2.1. The negative impact of inequality on growth in OECD countries

Net inequality (t-1)

(1)

(2)

(3)

(4)

-0.775**

-0.799**

-0.809*

-0.995***

(0.318)

(0.374)

(0.431)

(0.350)

Gross inequality (t-1)

(5)

(6)

(7)

-1.285**

-1.213**

(0.503)

(0.462)

-0.583

0.172

(1.031)

(0.612)

(Gross - Net) ineq. (t-1)
y (t-1)

(8)

0.081

-0.278

(0.686)

(1.325)

-0.140**

-0.089

-0.069

-0.081

0.047

-0.073

-0.086

0.133

(0.052)

(0.060)

(0.073)

(0.122)

(0.181)

(0.121)

(0.132)

(0.227)

-0.002

-0.005

0.004

0.009

-0.005

-0.007

0.014

(0.013)

(0.013)

(0.018)

(0.022)

(0.013)

(0.012)

(0.020)

0.216

0.521

0.187

1.606

-0.217

-0.251

2.423

(0.379)

(0.634)

(1.393)

(1.299)

(1.359)

(1.486)

(2.028)

Human capital (t-1)
Inv estment (t-1)
M2 (p-v al)

0.710

0.536

0.605

0.774

0.903

0.594

0.656

0.940

Hansen statistic (p-v al)

0.991

0.736

0.535

0.375

0.602

0.378

0.356

0.528

Observ ations

128

128

128

128

125

125

125

125

Number of countries

31

31

31

31

30

30

30

30

Number of instruments

27

31

26

16

16

18

18

16

Note: The dependent variable is Δlnyt where yt is per capita GDP, and [t-(t-1)] is a five-year period. Inequality is measured by
Gini indexes. Robust, two-step System GMM estimator with Windmeijer-corrected standard errors. All regressions include
country and period dummies. M2 are the p-values of the tests for second order serial correlation in the differenced error terms;
Hansen denotes the p-value on the Hansen test of over identifying restrictions. ***, **, * denote significance at the 1%, 5%,
10% levels, respectively.
Source: See Annex 2.A1, OECD Secretariat calculations.
12 http://dx.doi.org/10.1787/888933208837

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 67

The estimated impact of inequality on growth turns out to be sizeable. Based on the
coefficient reported in column 1, for example, lowering inequality by 1-Gini point would
translate into an increase in cumulative growth of 0.8 percentage point in the following
five years (or 0.15 point per year). Annex 2.A1 details how the estimated coefficients can
be used to infer the consequences of changes in inequality over the longer run in light of
the Solow growth model. Focusing on a 25-year horizon, the estimated coefficients imply
that a 1-Gini point reduction in inequality would raise average growth by slightly more
than 0.1 percentage point per year, with a cumulative gain in GDP at the end of the period
of around 3%.
Alternatively, one could focus on changes in inequality observed in the recent past
and infer the extent to which subsequent growth rates would have increased or decreased
had inequality not changed. Between 1985 and 2005, for example, inequality increased by
more than 2-Gini points on average across 19 OECD countries;14 the point estimates
reported in column 1 imply that this change kicked 4.7 percentage points off cumulative
growth in 1990-2010. For reference, the average of cumulated growth rates across the
same set of countries amounted to 28%. In other words, the estimated coefficients imply
that, had inequality not changed between 1985 and 2005 (and holding all other variables
constant), the average OECD country would have grown by nearly 33%.
Columns 1 to 4 in Table 2.1 are based on the inequality of disposable income. With
regard to the theoretical models referred to in Section 2.2, this measure is relevant for
those approaches which predict that inequality generates missed opportunities by the
poor (theory B) but also those models in which inequality rather represents a reward to
costly investments in human or physical capital (theory D). However, disposable
income is not the correct measure for testing the “endogenous fiscal policy” theory
(theory A). Based on this view, increased inequality in market (rather than disposable)
income would induce voters to choose a high level of (distortionary) taxation
(Milanovic, 2000). Accordingly, the results reported in column 5 of Table 2.1 replicate
the previous specification measuring inequality of income before taxes and transfers.
Though still negative, the estimated coefficient is lower in magnitude and is not
statistically significant. Hence, the analysis provides little support for this theory – at
least for the sample of OECD countries.
One prediction of some of the theories about how inequality might impact growth is
that the effect might be non-linear. Some of the political economy and socio-political
instability theories discussed above (see Benhabib, 2003) suggest that while some
inequality is unlikely to cause unrest and provides growth-enhancing incentives,
inequality can disrupt economic relations after it reaches some “tipping point” by
inviting political interference through rent-seeking behaviour and appropriation. A
similar argument might be made about investment in education, for example. However,
no such non-linearity was found in the results 15 – the effect on growth of an increase in
inequality from 20 to 21 Gini points was found to be the same as the effect of
increasing the Gini from 40 to 41. Nor was there any evidence found that the effects
varied significantly between the short term and long term.16 On the other hand, the
negative inequality-growth relationship seems to hold even at the sub-national level
(see Box 2.2).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 68 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

Box 2.2. Income inequality and economic growth in OECD regions
The OECD recently undertook an extension of its collection of standardised income distribution indicators
to the regional level. The data show that, on average, the within-country variation in inequality is large, and
comparable to that observed across countries (OECD, 2014a). Regional differences are particularly high in all
large OECD countries and in some small countries with a dominant urban centre. The data show that income
inequalities increase with city size. This has several possible explanations. First, large cities host the most
productive firms, which pay the highest rewards to observable skills and attract talented individuals (Behrens
and Robert-Nicoud, 2014). In addition, large cities may also have, on average, workers with higher unobserved
skills, since there is evidence that the increase of inequality with city size occurs also within skill groups (BaumSnow and Pavan, 2013). Finally, large cities are places of greater opportunities that attract poor people, thus
contributing to greater wage dispersion.
Importantly, the data allowed addressing the relationship between income inequalities and economic growth
at the sub-national level (Royuela et al., 2014). The analysis considers 15 OECD countries, covering three
continents, over the period 2004-12.
A clear negative correlation was observed between income inequality and economic growth in OECD
regions, especially when the patterns followed by European, and North and South American countries were
considered separately. In both continents, more unequal regions experienced a subsequent slower growth rate of
GDP per capita, on average (see figure below). Econometric results confirmed this negative relationship, which
was stronger after 2008, suggesting that higher equality might represent an element of regional resilience against
economic shocks.
The negative inequality-growth relationship was also found to be sensitive to the type of urban structure.
After having controlled for the effect of urbanisation on economic growth, higher inequality seems to be more
detrimental for growth in large cities, where inequality is already relatively higher on average.
Income inequality and growth in OECD regions, 2008-12
Europe

Canada, Chile, Mexico, United States

Average annual growth rate of GDP per capita
(2008-12)

8%

6
4

2
0

-2
-4

-6
-8

20

25

30

35

40

Inequality (Gini coefficient)

45

50

Source: Royuela, V., P. Veneri and R. Ramos (2014), “Income Inequality, Urban Size and Economic Growth in OECD
Regions”, OECD Regional Development Working Papers, Vol. 2014/10, OECD Publishing, Paris,
http://dx.doi.org/10.1787/5jxrcmg88l8r-en.
12 http://dx.doi.org/10.1787/888933207870

Redistribution
If inequality has a negative impact on long-term growth, a relevant policy question is
how to promote a win-win process to reduce inequality and boost growth. The main,

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 69

direct policy tool to reduce market income inequality is via taxes and benefits, which
however may also have a negative effect on growth (see Ostry et al., 2014; OECD, 2012).
The effect would be negative if, for example, high levels of taxes and transfers imply a
waste of resources and generate aggregate inefficiencies (as in Okun’s famous “leaky
bucket” analogy).17 If this is the case, reaching a given level of disposable income
inequality would entail a stronger drag on growth in countries featuring higher market
inequality to begin with.
The results of the analysis suggest that redistribution is, at worst, neutral to growth.
The specification in column 6 of Table 2.1 looks at the role of redistribution including
both market and disposable (“net”) income inequality. The coefficient estimated on net
inequality therefore reflects the effects of changes in inequality due to redistribution. The
coefficient remains negative, statistically significant and almost unchanged from the
previous columns. The non-significant estimate of the coefficient on market inequality
indicates that the extent of redistribution necessary to achieve a given level of net equality
has no negative direct consequences on economic growth.
This finding is further supported by alternative specifications. Column 7 shows that
after controlling for net inequality, the extent of redistribution in a country (the difference
between market and net income inequality) has no significant impact on growth. This
specification is the same as the one used by Ostry et al. (2014), who obtained similar
results for a broader set of countries. Finally, the extent of redistribution is not significant
when it is taken as the only core independent variable (see column 8). Taken together,
these results suggest that inequality in disposable incomes is bad for growth, and that
redistribution is, at worst, neutral to growth.
These results are based on a partial and relatively crude measure of redistribution and
do not therefore imply that all redistribution measures would be equally good for
growth.18 For one thing, they do not independently consider the possible contribution to
growth of other redistributive tools, such as “pre-distributive” policies that affect market
outcomes and alter income disparities before taxes and transfers. These include, for
example, education policies that allow a larger fraction of the population to benefit from
higher (skilled) wages, or labour market activation policies that favour the participation
and employment of under-represented groups.
More importantly, the impact of different redistributive measures on efficiency and
growth is in practice likely to vary, both in terms of sign and magnitude. Previous OECD
work (Arjona et al., 2001) looked at the effects of social spending on growth, dividing
such spending into “active” (social spending which attempts to change the distribution of
market income by promoting the labour market participation of part of the population
who would have lower-than-normal market incomes) and “passive” (e.g. unemployment
benefits). Active spending included active labour market policies, but also in-work
benefits and spending on childcare. This work found that active spending is associated
with higher growth, whereas more “passive” social spending is associated with lower
growth. While the approach is different from the one followed here, it suggests that not
all redistribution is necessarily equally good for growth (see also OECD, 2012).

Top and bottom inequality
To look at the growth consequences of inequality in different parts of the income
distribution (see also Voitchovsky, 2005), the Gini index of inequality is replaced with
several measures of “top” and “bottom” inequality. For example, top income inequality is
measured by the ratio of average disposable income in one top decile to average
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 70 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
disposable income in the country, and bottom inequality the ratio of average income in
the country to average income in one bottom decile.19
The results, presented in Table 2.2, suggest that lowering inequality by reducing
income disparities at the bottom of the income distribution has a greater positive impact
on economic performance than if the focus were on reducing top inequality. The
estimated coefficients imply that lowering bottom inequality by half of a standard
deviation (which is the same as changing bottom inequality in the United Kingdom to be
at the level of that in France, or that of the United States to become like that of Japan, or
Australia) would increase average annual growth by nearly 0.3-percentage point over the
subsequent 25-year period, with a cumulated gain in GDP at the end of the period in
excess of 7%.
Table 2.2. Inequality at the bottom and at the top of the income distribution
(1)

(2)

(3)

(4)

(5)

Bottom inequality
st

1 decile
Bottom inequality

2

nd

rd

decile 3 decile

(6)

(7)

(8)

(9)

Bottom and top inequality
1

th

4 decile

8

th

st

and

2

deciles 8

th

nd

and

3

deciles 8

th

rd

and

deciles 8

(10)

Top inequality
4
th

th

and

deciles

th

th

9 decile 10 decile

-0.031**

-0.071*

-0.121*

-0.196*

-0.032*

-0.084***

-0.133***

-0.198**

(0.012)

(0.037)

(0.067)

(0.111)

(0.018)

(0.029)

(0.047)

(0.083)

-0.038

-0.367

-0.220

-0.066

-0.571

-0.065

(0.750)

(0.469)

(0.403)

(0.448)

(0.451)

(0.050)

Top inequality
M2 (p-v al)

0.318

0.305

0.333

0.537

0.266

0.193

0.248

0.338

0.311

0.378

Hansen statistic (p-v al)

0.436

0.513

0.615

0.120

0.703

0.807

0.823

0.753

0.449

0.309

Observ ations

94

94

94

94

94

94

94

94

94

94

Number of countries

30

30

30

30

30

30

30

30

30

30

Number of instruments

11

11

11

11

13

13

13

13

11

11

Note: The dependent variable is lnyt where [t-(t-1)] is a five-year period. Bottom inequality is measured by the ratio between
mean disposable income in the economy (𝑌̅) and mean income of one bottom decile specified in the column heading (𝑦̅𝑛 , with
n=1,.., 4). An increase in the indicator in column 1, for example, implies a widening disparity between average overall income
and the average income of the bottom 10% of the population. Top inequality is measured as the ratio between the average
income of one top decile, specified in the column heading, and overall average income in the economy (𝑌̅). All regressions
include country and period dummies and a control for beginning of period GDP per capita (see Annex 2.A1 for a detailed
description of variables and sources). A robust, two-step System GMM estimator with corrected (Windmeijer, 2005) standard
errors is used. M2 are the p-values of the tests for second-order serial correlation in the differenced error terms. Hansen denotes
the p-value on the Hansen test of over-identifying restrictions. ***, **, * denote significance at the 1, 5 and 10% levels,
respectively.
Source: See Annex 2.A1; OECD Secretariat calculations.
12 http://dx.doi.org/10.1787/888933208845

The negative effect of bottom income inequality on growth proves robust. The basic
approach is to focus on the poorest households in the population (i.e. the gap in incomes
between the poorest decile and the average, see column 1). But it also holds – and is
remarkably similar in magnitude – when focusing on the second, third or fourth income
decile, which rather captures the relative income conditions of the lower-middle class
(columns 2 to 4).20 Moreover, it holds even when inequality on the upper end of the
distribution is simultaneously accounted for in the specification (columns 5 to 8). These
findings imply that countering the negative effect of inequality on growth is not (just)
about tackling poverty and the least well-off in society, it needs to be about addressing
low incomes more generally.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 71

Changes in top income inequality, however, are found to have no statistically
significant impact on economic growth (see columns 9 and 10, which confirm the results
obtained when top inequality is considered together with bottom inequality, in columns 5
to 8). It is important to note that the income data used do not provide information on
trends in concentration at the very top of the income distribution, e.g. movements of the
top percentile and above.21
These findings shed further light on the relative importance of the alternative avenues
through which income inequality is supposed to affect subsequent growth across OECD
countries. As first pointed out by Voitchovsky (2005), most of the mechanisms predicting
a negative effect of income inequality on growth emphasise the role of income disparities
at the bottom end of the distribution. For example, the human capital accumulation theory
(theory B above) predicts that inequality would be harmful because it raises the relative
costs of education of an increasing fraction of families in the bottom half of the
distribution. Higher inequality at the top is rather a signal of the existence of high rewards
to risky investments, and therefore more directly linked to the theories implying a
positive effect of inequality on growth (theory D above, for example). However, the
present findings differ from those of Voitchovsky (2005), who found support for both
bottom and top inequality having negative growth consequences.
In terms of the theoretical mechanisms highlighted in Section 2.2, the findings in this
section seem to indicate that one important way in which income inequality affects
growth is by lowering the investment and/or occupational opportunities of disadvantaged
individuals, as in the financial market imperfection/human capital accumulation theory
(theory B). Accordingly, the next section attempts to test this theory more directly,
looking at the links between inequality and investment in human capital by individuals
with different socio-economic backgrounds.

2.4. Inequality, social mobility and human capital accumulation
Across OECD countries, income inequality is negatively associated with average
educational attainment. Panel A of Figure 2.1 shows this by plotting a simple correlation
between the population share enrolled in upper secondary education and the Gini
coefficients of disposable income inequality (a similar picture emerges when looking at
tertiary enrolment). While consistent with the results from early cross-country regressions
linking inequality to human capital investment (e.g. Perotti, 1996, Deininger and Squire,
1998), this simple correlation is not in itself confirmation of the human capital
accumulation/financial market imperfection theory (theory B above). To test this, it
would be necessary to see whether the sign and strength of the relationship between
inequality and education varies across individuals in accordance with their socioeconomic background.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 72 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Figure 2.1. Inequality, enrolment and mobility across OECD countries
Panel A. Secondary enrollment

Panel B. Inequality and mobility

0.80
0.70

0.60
0.50
0.40

0.90
SVN
CZE HUN ITA
SVK
BELISL
FINAUT FRA POL EST
NOR
NZL
SWE
PRT
KOR
GRC ISR
CHE LUX
GBR
DNK NLD
JPNIRL
USA
DEU AUS
ESP

CHL
TUR

MEX

0.30

Intergenerational earnings elasticity

Enrolled in secondary educa tion

0.90

DNK

FIN

0.85
0.80

NOR

CAN
KOR JPN

SWE

0.75

NZL

AUS

0.70

DEU
CHE

0.65

ESP

0.60
FRA

0.55

GBR

ITA

0.50

USA

0.45
20

30

40

Inequality (Gini coefficient)

50

20

25

30

35

40

Inequality (Gini coefficient)

Note: Panel A: The graph reports the ratio between the number of students enrolled in upper secondary education and the population
aged 15-19. The two variables are sourced from the Education and Training Statistics in OECD.Stats (“Education and Skills”) and the
OECD Demography Statistics http://dx.doi.org/10.1787/5f958f71-en, and relative to 2010. Inequality (Gini coefficient) is measured
when individuals were aged 10-14, that is in 2005. Panel B: Compiled from different sources as in D’Addio, A.C. (2007),
“Intergenerational Transmission of Disadvantage: Mobility or Immobility Across Generations?”, OECD Social, Employment and
Migration Working Papers, No. 52, OECD Publishing, Paris, http://dx.doi.org/10.1787/217730505550; and OECD (2008), Growing
Unequal? Income Distribution in OECD Countries, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
Source: See Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207806

The negative cross-country correlation between inequality and intergenerational earnings
mobility, the so-called “Great Gatsby Curve” (see D’Addio, 2007; Corak, 2013), is also
linked to the same theory. This relationship (reproduced in Panel B of Figure 2.1) suggests
that in societies experiencing higher inequality, individuals might become less and less able
to move outside the earnings class in which they were born. It is therefore compatible, in
particular, with the idea that inequality lowers the investment and occupational opportunities
of disadvantaged individuals (upward mobility). As for any other result based on pure crosscountry variation, however, this finding is likely to suffer from biases induced by observed
and unobserved time-invariant country-specific confounding factors.
The analysis in this section uses individual-level survey data (from the OECD Adult
Skills Survey, PIAAC) to estimate whether the link between educational attainments and
inequality varies with parent educational background (PEB, a proxy for socio-economic
background). In doing this, it exploits within-country variation to account for timeinvariant observed and unobserved country characteristics that might affect both
attainment and the level of inequality. Studies of social mobility across OECD countries
(e.g. Causa and Johansson, 2009; and OECD, 2010b) provide broad support for the idea
that students’ achievements strongly depend on parents’ education. The current analysis
extends these results, exploring whether the degree of educational mobility declines as
inequality increases. Because human capital differences across individuals explain the
largest fraction of income differences, confirmation of this hypothesis would contribute to
understanding why inequality and intergenerational economic mobility are negatively
correlated across countries or areas within a country.22
To exploit such variation in a cross-sectional survey like the PIAAC, the exercise
uses differences in human capital attainment across age cohorts (within a country). More
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 73

specifically, individuals are pooled by five-year age groups (indexed with t), and each
group is assigned the measure of inequality in their country at the time they were aged
between 10 and 14.23 The baseline empirical equation is:
𝐻𝐶𝑖,𝑡,𝑐 = 𝜷𝟏 PEB𝑖,𝑡,𝑐 ∗ 𝐼𝑛𝑒𝑞𝑡,𝑐 + 𝜷𝟐 PEB𝑖,𝑡,𝑐 + 𝜃X 𝑖,𝑡,𝑐 + 𝜇𝑡 + 𝜇𝑐 + ϵ𝑖,𝑡,𝑐

[2]

where HC is a measure of human capital for individual i in country c, PEB is a set of
three indicators for her parent educational background (low, medium or high), and Ineq is
an index of inequality in the country (see Box 2.2 for details). In this specification, the
three parameters in 2 measure average educational outcomes of individuals with
different parental backgrounds, while those in 1 capture whether such averages vary with
the extent of income inequality in the country. This procedure allows running panel
regressions (country c, period t) accounting for country fixed effects (𝜇𝑐 ) and common
shocks (𝜇𝑡 ). Hence, the parameters  can be estimated accounting for time-invariant
country determinants that might bias cross-country estimates. This would be the case, for
example, if inequality is correlated with the quality of the educational system, or with
other policies and institutions that affect educational outcomes. (For a detailed description
of the other variables considered and of the estimation approach, see Box 2.3.)24
Box 2.3. The Programme for the International Assessment of Adult
Competencies (PIAAC) survey
The data used in this section are sourced from the OECD Programme for the International Assessment of
Adult Competencies (PIAAC), a survey administered to representative samples of the working-age (15-64)
population in 24 OECD countries between 2010 and 2011 (for more details, see OECD, 2013). It includes a rich
battery of questions covering demographic characteristics (age, gender, place of residence, religion), working
history and educational attainment as well as a direct measure of skill proficiency in three domains: literacy,
numeracy and problem-solving in technology-rich environments. The PIAAC measures each of the three skill
domains on a 500-point scale.
For each individual, the data allow measuring educational outcomes both in terms of formal achievements
(e.g. highest degree obtained) and in terms of actual skills (e.g. numeracy score). Recent studies suggest that
skills represent a potentially better measure of human capital than standard indicators of the highest level of
formal education attained. For example, using indicators of skill proficiency (e.g. based on international test
scores in literacy, science and math) dramatically improves the explanatory power of human capital in crosscountry growth regressions (see e.g. Hanushek and Woessmann, 2012).
The PIAAC data also report the level of education and main occupation of parents, which are used to
construct a measure of parent educational background (PEB). The variable is defined as follows. An individual
is assigned a low PEB if neither parent has attained upper secondary education; a medium PEB if at least one
parent has attained secondary and post-secondary, non-tertiary education; and a high PEB if at least one parent
has attained tertiary education.
Other individual characteristics that are likely to be relevant for educational decisions that are made when
the child is at an early age are used as controls (in matrix X). These include gender, parents’ immigration status,
whether the individual speaks a native or foreign language and the region of residence. Proficiency scores in
problem-solving will sometimes be used as a proxy for individual ability. Because the exercise assumes a role
for measured inequality in the country at earlier ages, individuals born abroad are excluded from the analysis.

This approach measures the degree of inter-generational educational persistence in terms
of average outcome differences among individuals in the three groups (and to anticipate the
results reported below, unsurprisingly it finds the persistence to be strong). However,
crucially, it also allows (changes in) inequality to affect individuals with different PEBs
differently. If the most important effect of higher income inequality is to increase the
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 74 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
incentives to invest in education, higher disparities should be associated with increased
achievement by at least part of the population. On the other hand, evidence that achievement
declines with rising inequality and that this effect is stronger among the poor would support
the idea that, interacting with financial market imperfections, inequality significantly lowers
the opportunities for education and upward social mobility for disadvantaged individuals.
This section tests these alternatives using three different sets of outcomes:


The first is a set of measures of the quantity of human capital accumulated by the
individual, including the probability of attaining tertiary education, and the
number of completed years of formal education.



The second is a set of indexes of skill proficiency, capturing cognitive ability and
therefore also accounting for the quality of the education completed.



The third measure is an index of the probability of employment, so moving beyond
education to explore the impact of inequality on labour market opportunities.

The results of all three approaches indicate that widening income disparities lowers
the outcomes of individuals from low socio-economic backgrounds, but does not affect
those of individuals from medium and high backgrounds. As in the case of growth
regressions, therefore, the results strongly support the idea that higher inequality lowers
opportunities for education (and social mobility) for disadvantaged individuals in the
society, an effect that dominates the potentially positive impacts through incentives.

Inequality and the quantity of human capital
Highest educational level attained
The first evidence supporting the negative effects of inequality on opportunities refers
to the highest level of formal education achieved. This is obtained by estimating an
ordered probit where the dependent variable (𝐻𝐶𝑖,𝑡,𝑐 = 𝐸𝐷𝑈𝑖,𝑡,𝑐 ) takes three ranked
values: low if the individual reports having attained less than lower secondary education,
medium in case of upper secondary education, and high in case of tertiary education (see
Annex 2.A1 for details). In this framework, the estimated parameters 1 and 2 allow
predicting the average probability of achieving each educational level by parental
education background and inequality level (i.e., at different levels of inequality). In
Panel A of Figure 2.2, each line indicates the average predicted probability of graduating
from university by PEB as a function of inequality (measured in Gini points).
Consistent with the substantial evidence for significant inter-generational educational
persistence, the estimated probability of graduating from university is highest for
individuals who have highly educated parents – just over 40% of them receive tertiary
education, compared to an average of around 30% of those who have parents with
medium levels of education.
However, the graph also shows that the probability of tertiary education decreases
with inequality, but only in the case of low PEB individuals. Based on the underlying
estimates, an increase in inequality of around 6 Gini points (corresponding to the income
inequality differential between the United States and Canada in 2010) would lower the
probability of individuals with parents of low educational background being in tertiary
education by around 4 percentage points.25 On the other hand, inequality does not have
any impact on the probability of graduating from tertiary education in the case of
individuals with a medium or high family background.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 75

Inequality is also associated with a significant increase in the probability that low
PEB individuals attain at most lower secondary education (see Panel B of Figure 2.2).
This probability is predicted to increase, on average, by nearly 5 percentage points
following an increase in inequality of 6 Gini points. However, there does not seem to be
an association between inequality and the attainment probability of richer individuals.
Figure 2.2. Probability of educational attainment by parent educational background (PEB) and inequality
Panel A. Probability of tertiary education
Low PEB

Medium PEB

High PEB

0.50

Probability of tertiary education

0.45
0.40

0.35
0.30
0.25
0.20

0.15
0.10
18

20

22

24

26

28

30

32

34

36

38

36

38

Inequality (Gini coefficient)

Panel B. Probability of lower secondary education
Low PEB

Medium PEB

High PEB

Probability of low er secondary education

0.50
0.45
0.40

0.35
0.30
0.25
0.20

0.15
0.10
18

20

22

24

26

28

30

32

34

Inequality (Gini coefficient)

Note: The two panels report the average predicted probability that individuals from poor, medium and high parental
(educational) backgrounds attain tertiary education (Panel A) or at most lower secondary education (Panel B), as a function of
the degree of inequality (Gini points) in the country at the time they were around 14 years old. Low PEB: neither parent has
attained upper secondary education; medium PEB: at least one parent has attained secondary and post-secondary, non-tertiary
education; high PEB: at least one parent has attained tertiary education. The bars indicate 95% confidence intervals. The vertical
dashed lines indicate the 25th, the median and the 75th percentiles of the underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207811

Average years of schooling
The same results qualitatively hold if focusing on a summary measure of the abovementioned outcomes, such as the number of years of completed schooling (YS).
Table 2.A1.1 reports the results obtained for estimating [2] when 𝐻𝐶𝑖,𝑡,𝑐 = 𝑌𝑆𝑖,𝑡,𝑐 . The
results in column 1 report the baseline estimates for the vector 2, indicating that higher
income inequality is negatively (and significantly) related to average schooling for low
background individuals, while the link is not significant in the case of medium or high
background individuals (as in the case of attainment probabilities). The remaining
columns correspond to alternative specifications of the control set X, which are detailed in
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 76 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
the table note. Importantly, the results are robust to accounting for the country’s level of
per-capita GDP (column 3), for country-specific trends in inequality and educational
achievements (column 4) and for the interaction between time dummies (i.e. age cohort)
and country dummies (column 5).26
To get a better understanding of the relevance of the estimated coefficients,
Figure 2.A1.1 reports the average predicted number of years of schooling by educational
background as a function of inequality, using the results from the baseline specification
(column 1). According to these estimates, increasing inequality by around 6 Gini points
lowers human capital among low background individuals by almost 0.5 years of
schooling. This represents more than 50% of the predicted schooling differential with
individuals from medium family backgrounds.27

Inequality and skill proficiency
Numeracy score
The second set of evidence about the effects of inequality on human capital is obtained
by looking at test scores (the PIAAC provides measures of proficiency in numeracy and
literacy). Figure 2.3 reports the average predicted numeracy score by the father’s educational
background as a function of inequality. As in the previous case, Figure 2.3 shows that
numeracy scores decrease with inequality in the case of individuals from low backgrounds.
In contrast, the average scores of more advantaged individuals are unaffected by widening
income inequality. According to these estimates, an increase in inequality of around 6 Gini
points is associated with a 6-point fall in the numeracy score of low background individuals.
This is a significant amount – it accounts for nearly 40% of the gap between their average
predicted numeracy score (261) and that of individuals with medium parental backgrounds.
One obvious concern in the context of the present analysis is about the extent to which
these measures actually reflect skills acquired while in education, given that skills are likely
to depreciate with age, and might be complemented by those accumulated at work. These
concerns are addressed empirically in Table 2.A1.2 and discussed in detail in Annex 2.A1.
In particular, the core results are unaffected by controlling for country-specific trends in
human capital achievements and inequality, and for occupation-specific and countryspecific rates of depreciation. Moreover, previous results suggest that skills measured in the
PIAAC largely reflect those accumulated while in education (OECD, 2014b).
In principle, the results reported in Figure 2.3 might simply be a consequence of the
previous observation, whereby a lower amount of skills just reflects a lower quantity of
education. However, the results are robust, albeit slightly lower in magnitude, even if
conditioning the estimates on the level of formal education, therefore insulating the
estimates from the negative consequences of inequality on the quantity of education.
Figure 2.4 plots the patterns of numeracy scores by level of inequality as predicted by
such exercise (see column 7 of Table 2.A1.2 for the underlying parameter estimates). The
figure shows that low background individuals see their skills decrease as inequality rises
even when they are compared with higher background individuals who attained the same
amount of formal education. Relative to the previous, unconditional case (represented by
the dotted blue line), the decreasing pattern is only slightly less steep. This suggests that a
large part of the lower proficiency of low PEB individuals can be traced to a worse
“quality” of the educational track (e.g. they attended worse quality schools/universities)
or to a lower amount of effort (e.g. hours) while studying.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 77

Figure 2.3. Average numeracy score by parent educational background (PEB) and inequality
Low PEB

Medium PEB

High PEB

300

Numeracy score

290
280

270
260
250
240

18

20

22

24

26

28

30

32

34

36

38

Inequality (Gini coefficient)

Note: The graph plots the average predicted numeracy score for individuals from low, medium and high parental (educational)
backgrounds, as a function of the degree of inequality (Gini points) in the country at the time they were around 14 years old.
Marginal effects obtained using estimates are shown in Table 2.A1.2, column 1. Low PEB: neither parent has attained upper
secondary education; medium PEB: at least one parent has attained secondary and post-secondary, non-tertiary education; high
PEB: at least one parent has attained tertiary education. The bars indicate 95% confidence intervals. The vertical dashed lines
indicate the 25th, the median and the 75th percentiles of the underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207829

Figure 2.4. Average numeracy score conditional on education by parent educational background (PEB)
and inequality
Low PEB

Medium PEB

High PEB

300

Numeracy Score

290

280

270

260

250
18

20

22

24

26

28

30

32

34

36

38

Inequality (Gini coefficient)

Note: The graph plots the average predicted numeracy score for individuals from low, medium and high parental (educational)
backgrounds, as a function of the degree of inequality (Gini points) in the country at the time they were around 14 years old.
Marginal effects obtained using estimates are shown in Table 2.A1.2, column 7 (i.e. conditioning the degree of formal
education). The dotted blue line replicates the results reported in Figure 2.4 in the case of low PEB individuals. Low PEB:
neither parent has attained upper secondary education; medium PEB: at least one parent has attained secondary and postsecondary, non-tertiary education; high PEB: at least one parent has attained tertiary education. Dotted lines represent baseline
probabilities for each group. The bars indicate 95% confidence intervals. The vertical dashed lines indicate the 25th, the median
and the 75th percentiles of the underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207839

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 78 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

Literacy score
Using literacy scores as a measure of the quality of human capital delivers very
similar findings. Figure 2.A1.2 plots the average predicted literacy score by educational
background as a function of inequality (using the specification in column 1 of
Table 2.A1.3). An interquartile (25th-75th) increase in inequality is associated with a
lower literacy score of low background individuals by slightly less than seven points. For
comparison, the average predicted differential in literacy score with the medium PEB
group amounts to 15 points.

Inequality and labour market outcomes
The third set of evidence suggesting that higher inequality lowers the amount of
opportunities available to disadvantaged individuals emerges from looking at their labour
market outcomes. The PIAAC allows an analysis of the probability of not being
employed, on average, over the working life.28 As in the case of educational outcomes,
this probability significantly increases with inequality in the case of low background
individuals. The predicted probabilities plotted in Figure 2.5 imply that, for example, the
probability of low background individuals not being employed rises by around
3 percentage points (or 20% of their baseline probability of non-employment) as
inequality widens by 6 Gini points. Again, the corresponding probabilities for richer
individuals are unaffected by inequality. A similar picture emerges when looking at the
probability of not being employed at the date of interview, which increases with
inequality for low PEB individuals only.
Figure 2.5. Probability of not being employed over the working life by parent educational background (PEB)
and inequality
Low PEB

Medium PEB

High PEB

Probability of not being employed

0.2

0.15

0.1

0.05
18

20

22

24

26

28

30

32

34

36

38

Inequality (Gini coefficient)

Note: The graph plots the average predicted probability of not being employed over the working life for individuals from low,
medium and high parental (educational) background, as a function of the degree of inequality (Gini points) in the country at the
time they were around 14 years old. Low PEB: neither parent has attained upper secondary education; medium PEB: at least one
parent has attained secondary and post-secondary, non-tertiary education; high PEB: at least one parent has attained tertiary
education. Dotted lines represent baseline probabilities for each group. The bars indicate 95% confidence intervals. The vertical
dashed lines indicate the 25th, the median and the 75th percentiles of the underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207843

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 79

Discussion and extensions
As in the case of growth regressions, these findings seem more supportive of the
“opportunity” arguments raised to explain the sources of the inequality-growth nexus
rather than the “incentive” arguments. If higher inequality (e.g. a higher skill wage
premium) increased the incentives to invest in education, this should be reflected in
increased attainment by at least part of the population. The finding that attainment is in
fact lowered, and that this happens only to the poorer segments, signals that income
availability significantly determines the opportunities of education and social mobility.
Interesting extensions of this exercise include checking whether the strength of the
link between inequality and educational attainment varies with countries’ institutional
characteristics or policy settings. This would help to shed light on the potential role of
policy in offsetting the adverse long-run consequences of widening inequality. The
characteristics of education systems and the effectiveness of educational expenditure in
levelling the playing field are examples of ways to capture such policies. Preliminary
attempts to explore these issues proved inconclusive, however.

2.5. Conclusion
This chapter contributes to a large empirical literature estimating the impact of
inequality on growth. Drawing on harmonised data covering the OECD countries over the
past thirty years, the econometric analysis suggests that income inequality has a sizeable
and statistically significant negative impact on growth, and that achieving greater equality
in disposable income through redistributive policies has no adverse impact on growth.
Moreover, it suggests that it is inequality at the bottom of the income distribution that
hampers growth. Further analysis based on OECD PIAAC data suggests that one key
channel through which inequality negatively affects economic performance is through
lowering investment opportunities (particularly in education) of the poorer segments of
the population.
These findings have relevant implications for policy makers concerned about slow
growth and rising inequality. On one hand it points to the importance of carefully
assessing the potential consequences of pro-growth policies on inequality (OECD, 2015):
focusing exclusively on growth and assuming that its benefits will automatically trickle
down to the different segments of the population may undermine growth in the long run
inasmuch as inequality actually increases. On the other hand it indicates that policies that
help limiting or – ideally – reversing the long-run rise in inequality would not only make
societies less unfair, but also richer. In particular, the present analysis highlights the
importance of two pillars of a policy strategy for tackling rising inequalities and
promoting equality of opportunities and plead for opting for “win-win” policies that aim
both at reducing inequality and promoting economic growth. This approach is also
consistent with the OECD Strategy for Inclusive Growth (OECD, 2014c).
One policy avenue for reducing inequality involves reforms to tax and benefit policies
(see also OECD, 2012). Recent OECD work that has focused on top incomes (Förster
et al., 2014) suggests that, as top earners now have a greater capacity to pay taxes than
before, governments may consider re-examining their tax systems to ensure that wealthier
individuals uphold their fair share of the tax burden. This aim can be achieved in several
different ways – not only via raising marginal tax rates on the rich but also by improving
tax compliance, eliminating or scaling back tax deductions that tend to benefit high
earners disproportionally, and by reassessing the role of taxes on all forms of property
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 80 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
and wealth, including the transfer of assets. Broadening the tax base by closing loopholes
in the current tax code has the potential to promote both efficiency and equity. This is
particularly the case for the taxation of capital income, which is highly concentrated
among wealthy households and represents a significant fraction of their total income. The
unequal tax treatment of income from different asset classes increases inequality in some
cases and distorts the allocation of capital.
However, the present chapter suggests that it is even more important to focus on
inequality at the bottom of the income distribution. Government transfers have an
important role to play in guaranteeing that low-income households do not fall further in
the income distribution. This implies paying careful attention to the design of income
support, for both those in work and out of work. This is not only restricted to cash
transfers. Other important elements of this pillar are policies to promote and increase
access to public services. This concerns services such as high-quality childcare and
education or access to health or housing. Such measures immediately smooth inequality
stemming from cash incomes, but they also constitute a longer-term social investment to
foster upward mobility and create greater equality of opportunity in the long run. A final
important element is the active promotion of inclusive employment, an area where the
“double dividend” of reducing inequality and increasing growth has always been most
apparent. Increasing employment can contribute to sustainable cuts in income inequality,
provided that employment gains occur in jobs that offer career prospects. A key challenge
for policy therefore is to facilitate and encourage access to employment for underrepresented groups, such as youths, older workers, women and migrants.29
In terms of social policy, social assistance and minimum income policies are aimed at
poverty alleviation. The analysis in this chapter suggests, however, that it is not just
poverty (i.e. the incomes of the lowest 10% of the population) that inhibits growth. What
is suggested instead is that policy makers need to be concerned about the bottom 40%
more generally – including the vulnerable lower-middle classes at risk of failing to
benefit from the recovery and future growth. Anti-poverty programmes will not be
enough.
The other major set of policy insights from this chapter concerns the links between
inequality and human capital. The evidence strongly indicates that high inequality hinders
the ability of individuals from low economic backgrounds to invest in their human
capital, in terms of both the level of education but even more importantly the quality of
education. This would imply that education policy should focus on improving access by
low-income groups, whose educational outcomes are not only worse on average from
those of middle and top income groups, but also more sensitive to increases in inequality.
However, the performance of disadvantaged individuals might not respond significantly
to policies aimed at lowering the direct private costs of, in particular, tertiary education
(e.g. tuition costs, or the availability of grants). The adverse impact of inequality may, in
fact, still operate via the differential effects of foregone earnings on schooling decisions
in different segments of the income distribution, via its effect on the allocation of parental
inputs in children’s human capital production, or via the ability of parents to select
optimal schooling environments (e.g. neighbourhoods). Policy needs therefore to take
account of the fact that in unequal societies low socio-economic groups are likely to have
underinvested in formal education. Accordingly, strategies to foster skills development
should include improved job-related training and education for the low-skilled (on-thejob training) and better access to formal education over their working lives.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 81

Notes
1.

Rajan (2010) argued that rising inequality in the United States induced low-income
individuals to borrow beyond their means to sustain consumption, and that this
overleveraging sowed the seeds of the crisis. Stiglitz (2012) and Acemoglu (2011)
claimed that increasing political influence of the rich and the financial industry
contributed to the financial excesses that generated the crisis. Fitoussi and Saraceno
(2010) argued that the roots of the crisis lie in a structural change in income
distribution that has been going on for the past three decades.

2.

According to the “endogenous fiscal policy” theory, the negative link between
inequality and growth rests on the combination of two basic mechanisms: an
economic mechanism positing that redistributive tools (e.g. capital income taxes)
lower the private returns to investment, and a political mechanism predicting that
higher inequality would induce more redistribution as poor individuals would prefer
larger tax rates than the rich.

3.

With perfect financial markets and decreasing returns to individual investment,
efficiency is maximised as poor individuals would borrow from the rich to realise the
optimal amount of investment. If financial markets are not available, underinvestment by the poor implies that aggregate output would be less than optimal, a
loss that would in general increase the degree of wealth heterogeneity (see e.g.
Bénabou, 1996; Aghion et al., 1999).

4.

Aghion and Bolton (1997) and Piketty (1997) explicitly modelled the supply side of
the credit market, explaining imperfections based on moral hazard (e.g. problems of
input verifiability) or enforcement problems stemming from contract incompleteness
(e.g. due to output verifiability).

5.

Mirrlees (1971) focused on the principal-agent setting where an (observable) output
depends on an unobservable effort. In that context, rewarding the agent independently
of output performance will discourage her from making any effort while allowing for
wage dispersion would encourage the effort. More broadly, Rebelo (1991) showed
that in a variety of growth models high investment or income tax rates would
discourage capital accumulation and imply lower growth rates.

6.

Kaldor (1955) suggested that, because the savings propensity out of labour income is
lower than that out profits, richer individuals (i.e. those earning more income from
capital) will tend to save more than the poor. This hypothesis was formalised in the
context of a Solow model by Bourguignon (1981), who showed that when savings are
a convex function of income, there may exist multiple steady states characterised by
different degrees of inequality. In this case, output is shown to be larger in the
unequal steady states not only at the aggregate level, but also for all individuals (i.e.
the unequal Pareto equilibrium dominates the egalitarian one).

7.

One interpretation of these differences is that panel data approaches are successful in
controlling for country-specific effects. Another possibility, however, is that they end
up eliminating most of the variation in the data, exacerbating measurement error
biases and reflecting in practice only the short-run effects of inequality. But many of

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 82 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
the theoretical effects of inequality on growth may take a significant amount of time
to materialise (changes in education, or in political stability, for example).
8.

More specifically, upward mobility is negatively related to inequality when this is
measured by Gini coefficients, which is consistent with the “Great Gatsby Curve”
documented across countries. Top 1% income shares are not highly correlated with
intergenerational mobility, however. By contrast, Bloome (2015) found that the US
states in which income inequality has increased the most have not been more likely to
suffer a decline in intergenerational income mobility.

9.

See Verbist et al. (2012), who estimate that the combined effect of in-kind transfers
for education, health and care reduces net income inequality by around one-fifth in
OECD countries, on average.

10.

Physical capital is proxied by the ratio of real non-residential fixed capital formation
to GDP, while human capital is measured by average years of schooling of the
working-age (25-64) population.

11.

This result is not completely surprising, as several other GMM studies that focus on
advanced economies estimated non-significant coefficients for one or more of the
standard growth determinants. This issue is discussed in detail in Annex 2.A1,
Section A2.

12.

With GMM estimators, the set of available instruments (i.e. the lagged values of the
independent variables) is potentially large, and using too many instruments may
weaken their effectiveness (Roodman, 2009). It is therefore important to check the
robustness of the results when reducing the instrumental variable matrix. Specifically,
for the inequality variable two lags are used as instruments in columns 1 and 2, and
one lag is used in column 3. In column 4, one lag is used and the instrument matrix
has been collapsed into one column (i.e., column 5 has the minimum possible set of
instruments). For all other variables, only one lag is used as instrument, and the
instrument matrix has been collapsed.

13.

The Arellano-Bond test indicates that serial correlation in the residuals, which
potentially undermines the use of lagged variables as instruments, should not be a
concern. The Hansen test of over-identifying restrictions does not suggest that any
instruments might be invalid.

14 .

The subsample of countries for which a time series of inequality and growth data is
available include Austria, Belgium, Canada, Denmark, Finland, France, Germany,
Ireland, Italy, Japan, Mexico, Netherlands, New Zealand, Norway, Spain, Sweden,
Turkey, the United Kingdom and the United States.

15.

This was tested by adding in a quadratic term (Gini2).

16.

This was tested by adding further lags of the inequality variable (e.g. Ginit-2). Using
data on a larger sample of countries, Halter et al. (2014) find that higher inequality
helps economic performance in the short term but reduces the growth rate in the
longer run.

17.

Okun’s (1975) prominent “leaky bucket experiment” refers to the fact that, when
government attempts to transfer income from rich to poor individuals, “(…) money
must be carried (…) in a leaky bucket. Some of it will simply disappear in transit, so
the poor will not receive all the money that is taken from the rich” (Okun, 1975,
p. 91). Okun attributed these losses to the administrative costs of taxing and
transferring, and to disincentive effects, mainly in the labour supply.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 83

18.

A detailed report investigating how tax structures can best be designed to support
GDP per capita growth is OECD (2010a)

19 .

More specifically, denote average disposable income in the country as 𝑌̅, and the
mean disposable income of the nth decile as 𝑦̅𝑛 . Bottom inequality is measured as the
ratio between 𝑌̅ and average income in the lower deciles of the distribution (the focus
is on the first to the fourth decile): 𝐵𝐼 = 𝑌̅⁄𝑦̅𝑛 (for n<5). Vice versa, top inequality is
measured as the ratio between average income in the upper decile and average
disposable income in the country: 𝑇𝐼 = 𝑦̅𝑛 ⁄𝑌̅ (for n>7). In both cases, an increase in
the indicator suggests widening inequalities.

20.

This is because the increase in the estimated coefficient on bottom inequality as one
moves from specification in column 1 to column 4 is almost entirely offset by a fall in
the standard deviation of the corresponding variable.

21.

Unfortunately, the available sources of top income data (e.g. the World Top Income
Database) only include pre-tax income shares of top deciles and percentiles. Further,
only 18 OECD countries are included in this dataset (see Atkinson et al., 2011;
OECD, 2014b). As a result, it is not possible to extend the analysis to consider the
role of top inequality based, for example, on the top 1% pre-tax income shares.

22.

Works focusing on within-country variation include Chetty et al. (2014), who find
that (upward) mobility is robustly negatively correlated with income inequality across
US counties, and Bloome (2015), who finds that the US states in which income
inequality has increased the most have not experienced a decline in intergenerational
income mobility.

23.

For example, the exercise assumes that the educational outcomes of individuals born
in 1966-70 informs about schooling decisions taken around 1980. Hence, in the
statistical analysis the outcomes of those individuals can be related to Ineq1980,c.
Following this reasoning, the outcomes of the cohorts (1966-70, 1971-75, …, 199195) were related to the inequality measured in (1980, 1985, …, 2005, which
correspond to t in equation [2]).

24.

The specification allows accounting also for unobserved drivers of both individual
attainments across subsequent cohorts and changes in inequality. This is obtained by
controlling for country-specific trends and for country-cohort dummies, absorbing
shocks that are specific to different country-cohort pairs (for example, the
introduction of country-specific educational policies affecting some but not all
cohorts, or the existence of country-specific rates of skills depreciation – when HC is
measured in terms of test scores).

25.

This amounts to more than one-fifth of the baseline probability of tertiary education
attainment of low PEB individuals (18%), and more than one-third of the probability
differential relative to medium PEB individuals.

26.

Because inequality varies at the country-year level, this last specification only allows
estimating the differential effect of inequality on average schooling by PEB (i.e. whether
higher inequality lowers schooling by low relative to high PEB individuals). It does not
identify, however, the consequences of inequality on the number of years of schooling
for each group. However, comparing the results with the baseline regression is
informative as to the relevance of potential biases from country-specific depreciation
rates.

27.

Estimating specification [2] allowing for nonlinear effects of inequality (e.g.
interacting family background with a quadratic in inequality) reveals a similar overall

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 84 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
picture, but suggests that the negative effects of increased income dispersion on
education attainment are slightly stronger when inequality is below the median.
28.

Each individual is asked to report the number of years spent in paid employment
(experience), and the number of years since she left the educational system (potential
experience). This information allows computing the fraction of time spent out of
employment (a measure of the probability of not being employed) over the
working life.

29.

Recent OECD work shows that several pro-growth policies helping to narrow
(bottom) income inequality (Causa et al., 2014). This is the case of reducing
regulatory barriers to domestic competition, trade and FDI, of stepping-up job-search
support and activation programmes and of tightening unemployment benefits for all
categories of jobseekers.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 85

References
Acemoglu, D. (2011), “Thoughts on Inequality in Financial Crisis”, Presentation at the
American Economic Association Annual Meeting, Denver, Colorado, January.
Aghion, P. and P. Bolton (1997), “A Theory of Trickle-down Growth and Development”,
Review of Economic Studies, Vol. 64, No. 2, pp. 151-172.
Aghion, P., E. Caroli and C. Garcia-Peñalosa (1999), “Inequality and Economic Growth:
The Perspective of the New Growth Theories”, Journal of Economic Literature,
Vol. 37, No. 4, pp. 1615-1660.
Alesina, A. and D. Rodrik (1994), “Distributive Politics and Economic Growth”,
Quarterly Journal of Economics, Vol. CIX, No. 2, pp. 465-490.
Alesina, A. and R. Perotti (1996), “Income Distribution, Political Instability, and
Investment”, European Economic Review, Vol. 40, No. 6, pp. 1203-1228.
Andrews, D., C. Jencks and A. Leigh (2011), “Do Rising Top Incomes Lift All Boats?”,
B.E. Journal of Economic Analysis and Policy, Vol. 11, No. 1, pp. 1-45, January.
Arellano, M. and S. Bond (1991), “Some Tests of Specification for Panel Data: Monte
Carlo Evidence and an Application to Employment Equations”, Review of Economic
Studies, Vol. 58, No. 2, pp. 277-297.
Arellano, M. and O. Bover (1995), “Another Look at the Instrumental Variable
Estimation of Error-Components Models”, Journal of Econometrics, Vol. 68, No. 1,
pp. 29-51.
Arjona, R., M. Ladaique and M. Pearson (2001), “Growth, Inequality and Social
Protection”, OECD Labour Market and Social Policy Occasional Papers, No. 51,
OECD Publishing, Paris, http://dx.doi.org/10.1787/121403540472.
Arnold, J., A. Bassanini and S. Scarpetta (2011), “Solow or Lucas? Testing Speed of
Convergence on a Panel of OECD Countries”, Research in Economics, Vol. 65, No. 2,
pp. 110-123.
Atkinson, A., T. Piketty and E. Saez (2011), “Top Incomes in the Long Run of History”,
Journal of Economic Literature, Vol. 49, No. 1, pp. 3-71.
Behrens, K. and F. Robert-Nicoud (2014), “Survival of the Fittest in Cities. Urbanisation
and Inequality”, Economic Journal, Vol. 124, No. 581, pp. 1371-1400.
Banerjee, A. and A. Newman (1993), “Occupational Choice and the Process of
Development”, Journal of Political Economy, Vol. 101, No. 2, pp. 274-298.
Barro, R.J. (2000), “Inequality and Growth in a Panel of Countries”, Journal of Economic
Growth, Vol. 5, No. 1, pp. 5-32.
Barro, R. and J. Lee (2013), “A New Data Set of Educational Attainment in the World,
1950–2010”, Journal of Development Economics, Vol. 104, pp. 184-198.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 86 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Bassanini, A. and S. Scarpetta (2002), “Does Human Capital Matter for Growth in OECD
Countries? A Pooled Mean Group Approach”, Economics Letters, Vol. 74, No. 3,
pp. 399-405.
Baum-Snow, N. and R. Pavan (2013), “Inequality and City Size”, Review of Economics
and Statistics, Vol. 95, No. 5, pp. 1535-1548.
Bazzi, S. and M. Clemens (2013), “Blunt Instruments: Avoiding Common Pitfalls in
Identifying the Causes of Economic Growth”, American Economic Journal:
Macroeconomics, , Vol. 5, No. 2, American Economic Association, pp. 152-186,
April.
Behrens, K. and F. Robert-Nicoud (2014), “Survival of the Fittest in Cities, Urbanisation
and Inequality”, Economic Journal, Vol. 124, No. 581, pp. 1371-1400.
Bénabou, R. (1996), “Inequality and Growth”, NBER Working Paper No. 5658,
Cambridge, United States, July.
Benhabib, J. (2003), “The Trade-off Between Inequality and Growth”, Annals of
Economics and Finance, Vol. 4, No. 2, pp. 491-507.
Bergh, A. and M. Henrekson (2011), “Government Size And Growth: A Survey And
Interpretation Of The Evidence”, Journal of Economic Surveys, Vol. 25, No. 5, Wiley
Blackwell, pp. 872-897, December.
Bernstein, J. (2013), “The Impact of Inequality on Growth”, Center for American
Progress Paper.
Bertola, G. (1993), “Factor Shares and Savings in Endogenous Growth”, American
Economic Review, Vol. 83, No. 5, pp. 1184-1198.
Bloome, D. (2015), “Income Inequality and Intergenerational Income Mobility in the
United States”, Social Forces, forthcoming.
Blundell, R. and S. Bond (1998), “Initial Conditions and Moment Restrictions in
Dynamic Panel Data Models”, Journal of Econometrics, Vol. 87, No. 1, pp. 115-143.
Bond, S., A. Hoeffler and J. Temple (2001); “GMM Estimation of Empirical Growth
Models”, CEPR Discussion Papers, No. 3048.
Bourguignon, F. (1981), “Pareto Superiority of Unegalitarian Equilibria in Stiglitz’
Model of Wealth Distribution with Convex Saving Function”, Econometrica, Vol. 49,
No. 6, pp. 1469-1475.
Caselli, F., G. Esquivel and F. Lefort (1996), “Reopening the Convergence Debate: A
New Look at Cross-Country Growth Empirics”, Journal of Economic Growth, Vol. 1,
No. 3, pp. 363-389.
Castelló-Climent, A. (2010), “Inequality and Growth in Advanced Economies: An
Empirical Investigation”, Journal of Economic Inequality, Vol. 8, No. 3, pp. 293-321,
September.
Castelló-Climent, A. and R. Doménech (2002), “Human Capital Inequality and Economic
Growth: Some New Evidence”, Economic Journal, Vol. 112, No. 478, pp. C187C200.
Causa, O. and A. Johansson (2009), “Intergenerational Social Mobility”, OECD
Economics Department Working Papers, No. 707, OECD Publishing, Paris,
tp://dx.doi.org/10.1787/223106258208.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 87

Causa, O., A. de Serres and N. Ruiz (2014), “Can Growth-enhancing Policies Lift All
Boats? An Analysis Based on Household Disposable Income”, OECD Economics
Department Working Papers, No. 1180, OECD Publishing, Paris,
www.oecd.org/eco/Can_pro-growth_policies_lift_all_boats_an_analysis_based_on_
household_disposable_income.pdf.
Checchi, D., A. Ichino and A. Rustichini (1999), “More Equal But Less Mobile?
Education Financing and Intergenerational Mobility in Italy and in the US”, Journal of
Public Economics, Vol. 74, No. 3, pp. 351-393.
Chetty, R. (2014), “Where Is the Land of Opportunity? The Geography of
Intergenerational Mobility in the United States”, with N. Hendren, P. Kline and
E. Saez, Quarterly Journal of Economics, Vol. 129, No. 4, pp. 1553-1623.
Cingano, F. (2014), “Trends in Income Inequality and its Impact on Economic Growth”,
OECD Social, Employment and Migration Working Papers, No. 163, OECD
Publishing, Paris, http://dx.doi.org/10.1787/5jxrjncwxv6j-en.
Cohen, D. and M. Soto (2007), “Growth and Human Capital: Good Data, Good Results”,
Journal of Economic Growth, Vol. 12, pp. 51-76.
Corak, M. (2013), “Income Inequality, Equality of Opportunity, and Intergenerational
Mobility”, Journal of Economic Perspectives, Vol. 27, No. 3, pp. 79-102.
D’Addio, A.C. (2007), “Intergenerational Transmission of Disadvantage: Mobility or
Immobility Across Generations?”, OECD Social, Employment and Migration Working
Papers, No. 52, OECD Publishing, Paris, http://dx.doi.org/10.1787/217730505550.
de la Fuente, A. and R. Doménech (2013), “Cross-Country Data on the Quantity of
Schooling: A Selective Survey and Some Quality Measures”, Working Paper,
No. 720, Barcelona Graduate School of Economics.
de la Fuente, A. and R. Doménech (2006), “Human Capital in Growth Regressions: How
Much Difference Does Data Quality Make?”, Journal of the European Economic
Association, Vol. 4, No. 1, pp. 1-36.
De Mello, L. and E. Tiongson (2006), “Income Inequality and Redistributive Government
Spending”, Public Finance Review, Vol. 34, No. 3, pp. 282-305.
Deininger, K. and L. Squire (1998), “New Ways of Looking at Old Issues: Inequality and
Growth”, Journal of Development Economics, Vol. 57, No. 2, pp. 259-287.
Fershtman, C., K. Murphy and M. Weiss (1996), “Social Status, Education, and Growth”,
Journal of Political Economy, Vol. 104, No. 1, pp. 108-132.
Fitoussi, J.P. and F. Saraceno (2010), “Inequality and Macroeconomic Performance”,
OFCE /POLHIA Working Papers, No. 2010-13, Paris.
Forbes, K.J. (2000), “A Reassessment of the Relationship Between Inequality and
Growth”, American Economic Review, Vol. 90, No. 4, pp. 869-887.
Förster, M., A. Llena-Nozal and V. Nafilyan (2014), “Trends in Top Incomes and their
Taxation in OECD Countries”, OECD Social, Employment and Migration Working
Papers, No. 159, OECD Publishing, Paris, http://dx.doi.org/10.1787/5jz43jhlz87f-en.
Galor, O. and J. Zeira (1993), “Income Distribution and Macroeconomics”, Review of
Economic Studies, Vol. 60, No. 1, pp. 35-52.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 88 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Halter, D., M. Oechslin and J. Zweimüller (2014), “Inequality and Growth: The
Neglected Time Dimension”, Journal of Economic Growth, Vol. 19, No. 1,
pp. 81-104.
Hanushek, E. and L. Woessmann (2012), “Do Better Schools Lead to More Growth?
Cognitive Skills, Economic Outcomes, and Causation”, Journal of Economic Growth,
Vol. 17, No. 4, pp. 267-321.
Hassler, J., J. Rodríguez Mora and J. Zeira (2007), “Inequality and Mobility”, Journal of
Economic Growth, Vol. 12, No. 3, pp. 235-259.
Islam, N. (1995), “Growth Empirics: A Panel Data Approach”, Quarterly Journal of
Economics, Vol. 110, No. 4, pp. 1127-1170.
Kaldor, N. (1955), “Alternative Theories of Distribution”, Review of Economic Studies,
Vol. 23, pp. 83-100.
Keefer, P. and S. Knack (2000), “Polarization, Politics and Property Rights”, World Bank
Policy Research Working Paper, No. 2418, August.
Krueger, A. (2012), “The Rise and Consequences of Inequality”, Presentation made to the
Center for American Progress in Washington, DC, January
Lazear, E.P. and S. Rosen (1981), “Rank-Order Tournaments as Optimum Labor
Contracts”, Journal of Political Economy, Vol. 89, No. 5, pp. 841-864.
Lee, K., M. Pesaran and R. Smith (1997), “Growth and Convergence in Multi-country
Empirical Stochastic Solow Model”, Journal of Applied Econometrics, Vol. 12, No. 4,
pp. 357-392.
Li, H. and H. Zou (1998), “Income Inequality Is Not Harmful for Growth: Theory and
Evidence”, Review of Development Economics, Vol. 2, No. 3, pp. 318-334.
Mankiw, G., D. Roemer and D. Weil (1992), “A Contribution to the Empirics of
Economic Growth”, Quarterly Journal of Economics, Vol. 107, No. 2, pp. 407-437.
Milanovic, B. (2000), “The Median-voter Hypothesis, Income Inequality, and Income
Redistribution: An Empirical Test with the Required Data”, European Journal of
Political Economy, Vol. 16, No. 3, pp. 367-410.
Mirrlees, J. (1971), “An Exploration in the Theory of Optimum Income Taxation”,
Review of Economic Studies, Vol. 38, No. 114, Wiley Blackwell, pp. 175-208, April.
Murphy, K.M., A. Shleifer and R. Vishny (1989), “Income Distribution, Market Size, and
Industrialisation”, Quarterly Journal of Economics, Vol. CIV, No. 3, pp. 537-564.
Nickell, S. (1981), “Biases in Dynamic Models with Fixed Effects”, Econometrica,
Vol. 49, No. 6, pp. 1417-1426, November.
OECD (2015), “Structural Reforms and Income Distribution”, OECD Economic Policy
Paper, No. 13, OECD Publishing, Paris.
OECD (2014a), How’s Life in Your Region? Measuring Regional and Local Well-being
for
Policy
Making,
OECD
Publishing,
Paris,
http://dx.doi.org/
10.1787/9789264217416-en.
OECD (2014b), “Non-regular Employment, Job Security and the Labour Market Divide”,
Chapter 4 in the OECD Employment Outlook, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2014-en.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 89

OECD (2014c), All on Board – Making Inclusive Growth Happen, OECD Publishing,
Paris,
www.oecd.org/inclusive-growth/All-on-Board-Making-Inclusive-GrowthHappen.pdf.
OECD (2013), Skills Outlook: First Results from the Survey of Adult Skills, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264204256-en.
OECD (2012), “Income inequality and growth: The role of taxes and transfers”, OECD
Economics Department Policy Notes, No. 9. January 2012.
OECD (2010a), “Responding to the Crisis while Protecting Long-term Growth”, in
OECD, Economic Policy Reforms 2010: Going for Growth, OECD Publishing, Paris,
http://dx.doi.org/10.1787/growth-2010-3-en.
OECD (2010b), “Tax Policy Reform and Economic Growth”, OECD Tax Policy Studies,
No. 20, OECD Publishing, Paris. http://dx.doi.org/10.1787/9789264091085-en.
OECD (2008), Growing Unequal? Income Distribution in OECD Countries, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
Okun, A. (1975), Equality and Efficiency: The Big Trade-off, Brookings Institution,
Washington DC.
Ostry, J., A. Berg and C. Tsangarides (2014), “Redistribution, Inequality, and Growth”,
IMF Staff Discussion Note, February.
Owen, A.L. and D.N. Weil (1998), “Intergenerational Earnings Mobility, Inequality and
Growth”, Journal of Monetary Economics, Vol. 41, No. 1, Elsevier,
Perotti, R. (1996), “Growth, Income Distribution and Democracy: What the Data Say”,
Journal of Economic Growth, Vol. 1, No. 2, pp. 149-187.
Perotti, R. (1994), “Income Distribution and Investment”, European Economic Review,
Vol. 38, No. 3-4, pp. 827 835.
Persson, T. and G. Tabellini (1994), “Is Inequality Harmful for Growth?”, American
Economic Review, Vol. 84, No. 3, pp. 600-621.
Pesaran, M.H. and R. Smith (1995), “Estimating Long-run Relationships from Dynamic
Heterogeneous Panels”, Journal of Econometrics, Vol. 68, No. 1, pp. 79-113.
Pesaran, M.H., Y. Shin and R. Smith (1999). “Pooled Mean Group Estimation of
Dynamic Heterogeneous Panels”, Journal of the American Statistical Association,
Vol. 94, No. 446, pp. 621-634.
Piketty, T. (1997) “The Dynamics of the Wealth Distribution and the Interest Rate with
Credit-rationing”, Review of Economic Studies, Vol. 64, No. 2, pp. 173-189.
Pritchett, L. (2000a), “Understanding Patterns of Economic Growth: Searching for Hills
among Plateaus, Mountains, and Plains”, World Bank Economic Review, Vol. 14,
No. 2, pp. 221-250.
Rajan, R. (2010), Fault Lines: How Hidden Fractures Still Threaten the World Economy,
Princeton University Press.
Rebelo, S. (1991), “Long-Run Policy Analysis and Long-Run Growth”, Journal of
Political Economy, Vol. 99, No. 3, University of Chicago Press, pp. 500-521, June.
Roodman, D. (2009), “A Note on the Theme of Too Many Instruments”, Oxford Bulletin
of Economics and Statistics, Vol. 71, No. 1, pp. 135-158.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 90 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Royuela, V., P. Veneri and R. Ramos (2014), “Income Inequality, Urban Size and
Economic Growth in OECD Regions”, OECD Regional Development Working
Papers, Vol. 2014/10, OECD Publishing, Paris, http://dx.doi.org/10.1787/
5jxrcmg88l8r-en.
Solow, R. (1956), “A Contribution to the Theory of Economic Growth”, Quarterly
Journal of Economics, Vol. 70, No. 1, pp. 65-94
Stiglitz, J. (2012), The Price of Inequality, W.W. Norton and Company.
Verbist, G., M.F. Förster and M. Vaalavuo (2012), “The Impact of Publicly Provided
Services on the Distribution of Resources: Review of New Results and Methods”,
OECD Social, Employment and Migration Working Papers, No. 130, OECD
Publishing, Paris, http://dx.doi.org/10.1787/5k9h363c5szq-en.
Voitchovsky, S. (2005), “Does the Profile of Income Inequality Matter for Economic
Growth?”, Journal of Economic Growth, Vol. 10, No. 3, pp. 273-296.
Windmeijer, F. (2005), “A Finite Sample Correction for the Variance of Linear Efficient
Two-step GMM Estimators”, Journal of Econometrics, Vol. 126, No. 1, pp. 25-51.

Database references
Luxembourg Income Study (LIS) Database, www.lisdatacenter.org, microdata runs
completed between 22 April 2013 and 7 November 2014.
OECD Income Distribution Database (IDD), www.oecd.org/social/income-distributiondatabase.htm.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 91

Annex 2.A1
Estimating the inequality, social mobility and growth nexus

A1.

Introduction
This annex provides background information and estimation details on the empirical
exercises summarised in the main text. Section A2 below describes the methodology and
the data used in cross-country growth regressions (Section 2.2 in the main text).
Section A3 focuses on the approach taken to investigate the link between inequality and
educational mobility (Section 2.2 in the main text), presenting additional results.

A2.

The impact of inequality on growth
The growth equation
Mankiw et al. (1992) showed how empirical growth equations similar to the one
analysed here can be derived from a neoclassical growth model (Solow, 1956) augmented
in order to take into account human capital as a factor of production. These equations
start from a constant return-to-scale production function as:
𝑌(𝑡) = 𝐾(𝑡)𝜃 𝐻(𝑡)𝛽 (𝐴(𝑡)𝐿(𝑡))1−𝜃−𝛽

[A.1]

where Y, K and H are output, physical and human capital, respectively, L is labour, A is
labour-augmenting technology and 𝜃 and β are the partial elasticities of output with
respect to physical and human capital. As in the Solow model, L and A grow exogenously
at rates n and g, respectively: L(t) = L(0)ent and A(t) = A(0)egt. The number of effective
units of labour A(t)L(t) then grows at rate n+g. Physical capital depreciates at rate .
Let sk and sh be the fraction of income invested in physical and human capital,
respectively. Defining quantities in [A.1] in terms of units of effective labour input
A(t)L(t) (y = Y/AL, k = K/AL, and h = H/AL), the evolution of the economy is then
determined by:
𝑘̇ (𝑡) = 𝑠𝑘 𝑦(𝑡) − (𝑛 + 𝑔 + 𝛿) 𝑘(𝑡)

[A.2]

ℎ̇(𝑡) = 𝑠ℎ 𝑦(𝑡) − (𝑛 + 𝑔 + 𝛿) ℎ(𝑡)

[A.3]

Under the assumption that +<1 (i.e. of decreasing returns to reproducible factors), this
system of equations can be solved to obtain steady-state values of k* and h* defined by:
1−𝛽 𝛽

1⁄(1−𝜃−𝛽)

𝑠
𝑠ℎ
𝑘∗ = ( 𝑘
)
𝑛+𝑔+𝛿

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

[A.4]

 92 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
𝑠𝑘𝜃 𝑠ℎ1−𝜃
ℎ =(
)
𝑛+𝑔+𝛿

1⁄(1−𝜃−𝛽)

[A.5]

∗

Substituting [A.4] and [A.5] into the production function and taking logs yields the
expression for the steady-state output in intensive form. The latter can be expressed either
as a function of sh (investment in human capital) and the other variables or as a function
of h* (the steady-state stock of human capital) and the other variables (Mankiw et al.,
1992). From an empirical point of view, the choice between the two depends on the
nature of the available data. In this chapter human capital is measured by the average
years of education of the working-age population, and thus the expression is in terms of
human capital stock (h). The steady-state path of output y* can therefore be written as:
∗

𝑌(𝑡)
𝜃
𝛽
𝑙𝑛 (
ln 𝑠𝑘 +
ln ℎ∗
) = 𝑙𝑛𝐴(0) + 𝑔𝑡 +
𝐿(𝑡)
1−𝜃−𝛽
1−𝜃−𝛽
−

[A.6]

𝜃+𝛽
𝑙𝑛 (𝑛 + 𝑔 + 𝛿)
1−𝜃−𝛽

Let y* be the steady state level of output in efficiency units and y(t) its value at time t,
then the transitional dynamics to the steady state can be expressed as:
𝜕𝑙𝑛 𝑦
= 𝜆 [ln 𝑦 ∗ − ln 𝑦]
𝜕𝑡
where 𝜆 = (𝑛 + 𝑔 + 𝛿) (1 − 𝜃 − 𝛽) is the rate of convergence. For example,
if 𝜃 =  = 1⁄3, and 𝑛 + 𝑔 + 𝛿 = 0.06 then the convergence rate would equal 0.02.
This implies that the economy moves halfway to the steady state in about 35 years. Under
the assumption that 𝜃+β < 1 (i.e. decreasing returns to reproducible factors), this equation
implies that lny approaches lny* exponentially,
ln 𝑦(𝑡) − ln 𝑦 ∗ = 𝑒 −𝜆𝑡 [ln 𝑦(𝑡 − 𝑠) − ln 𝑦 ∗ ]
which can be rewritten to have an expression for the growth of income:
𝑙𝑛𝑦(𝑡) − 𝑙𝑛𝑦(𝑡 − 𝑠) = (1 − 𝑒)−𝜆𝑠 (𝑙𝑛𝑦 ∗ − 𝑙𝑛𝑦(𝑡 − 𝑠) )

[A.7]

Substituting y* from [A.6] (with 𝜙(𝜆) = (1 − 𝑒)−𝜆𝑠 ):
𝜃

𝑙𝑛 𝑦(𝑡) − 𝑙𝑛 𝑦(𝑡 − 𝑠) = −𝜙(𝜆)𝑙𝑛𝑦(𝑡 − 𝑠) + 𝜙(𝜆) 1−𝜃−𝛽 ln 𝑠𝑘 +
𝛽

[A.8]

𝜃+𝛽

𝜙(𝜆) 1−𝜃−𝛽 ln ℎ∗ − 𝜙(𝜆) 1−𝜃−𝛽 𝑙𝑛 (𝑛 + 𝑔 + 𝛿)
Hence, in a Solow model, output growth is a function of the initial level of income
and of the ultimate determinants of the steady state. This implies that estimating an
equation as [A.8] would allow inferring the impact of each growth determinant on the
subsequent pattern of growth.1 This is because the coefficient ̂𝛼 , estimated on lagged
output in equation [1], allows recovering the speed of convergence: 𝜆̂ = − ln(1 − 𝛼̂)⁄𝑠,
with s = 5. Moreover, the coefficient estimated on a given growth determinant X (call this
coefficient 𝛾, as in the case of inequality in [1]) allows computing the impact of this
̂∗ = −(𝛾̂⁄𝛼̂ ) ∗ ∆𝑋). Exploiting these
determinant on the steady state level of output (∆𝑙𝑛𝑦
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 93

two estimates and equation [A.7] one can, for example, calculate the implied effect of a
change in inequality on long-run (i.e. 25-year) growth.
The figures discussed in Section 2.2, for example, are obtained from the coefficients
estimated in the first column of Table 2.1. Based on those estimates, a 1-Gini point
reduction in inequality would increase the steady state level of per capita GDP by 5.7%
(∆ ln 𝑦 ∗ = −(𝛾̂ ⁄𝛼̂ ) ∗ (−1) = 0.0569). Differentiating [A.7] yields an expression for the
percentage change in GDP at year t (which is s years ahead of the current period) as a
̂
function of ∆ ln 𝑦 ∗ : ∆ ln 𝑦(𝑡) = (1 − 𝑒)−𝜆𝑠 (∆ ln 𝑦 ∗ ). Finally, the estimated speed of
convergence is 𝜆̂ = − ln(1 − 𝛼̂)⁄5 = 0.029. These estimates imply that a 1-Gini point
reduction in inequality would increase GDP per capita by 3% after 25 years (with a gain
in average growth of nearly 0.115% per year). These values also imply that GDP would
cover slightly more than half of the distance from the new steady state over the same
horizon.2
Since the beginnings of the empirical growth literature in the early 1990s,
equation [A.8] has been extended to account for a variety of long-run growth
determinants (such as public and social capital, trade openness, financial development,
quality of institutions, etc.). Early works focussing on the role of inequality include
Persson and Tabellini (1994) and Alesina and Rodrik (1994).
Moreover, equation [A.8] can be estimated for any time interval. Because inequality
indicators have not been measured at high frequency in the past across countries, the
present application will exploit five-year intervals (i.e. s = 5). This allows using a
dynamic fixed-effect (DFE) specification estimated with GMM methods, accounting for a
country-specific component in the error term, which is a likely source of bias in early
cross-country regressions of long-run per capita GDP growth on inequality.
However, DFE specifications typically impose homogeneity of all slope coefficients,
and homogeneity of the rate of convergence appears to be at odds with data for OECD
countries (Bassanini and Scarpetta, 2002). Pesaran and Smith (1995) show that, under
slope heterogeneity, GMM (and simple least square dummy variable) dynamic fixedeffect estimates of the speed of convergence are usually affected by a downward
heterogeneity bias. Accordingly, Arnold et al. (2011) rather looked at an error
correction (ECM) version of equation [A.8], using pooled mean group (PMG) estimators,
which allow the speed of convergence to the steady state to differ across countries. This is
a realistic approach, as both exogenous (i.e. Solow) and endogenous (i.e. Uzawa-Lucas)
growth models imply that the speed of convergence to the steady state differs across
countries because of cross-country heterogeneity in population growth, technical change
and the progressiveness of income tax. Moreover, the approach permits to discriminate
between growth theories by just glancing at the estimated parameters. In fact, for
plausible values of the parameters, the Solow model implies a much slower speed of
convergence to the steady state than that implied by the Lucas model (Arnold et al., 2011,
concludes that the estimated speed of convergence is compatible with endogenous growth
theories). As mentioned, this empirical approach could not be taken in the case of the
present analysis due to the lack of time series variation in inequality data.

The empirical model and data
The baseline regression considered in the analysis augments the above estimating
equation by including inequality among the determinants of steady-state income. It is
estimated empirically exploiting a newly assembled (unbalanced) panel of data covering

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 94 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
OECD countries over the period 1970-2010. More specifically, the baseline estimating
equation is:
𝑙𝑛𝑦𝑐,𝑡 − 𝑙𝑛𝑦𝑐,𝑡−𝑠

[1]

= 𝛼𝑙𝑛𝑦𝑐,𝑡−𝑠 + 𝛾𝐼𝑛𝑒𝑞𝑐,𝑡−𝑠 + 𝛽1 𝐻𝐶𝑐,𝑡−𝑠 + 𝛽2 𝐼𝑛𝑣𝑐,𝑡−𝑠 + 𝜇𝑐
+ 𝜇𝑡 + 𝜖𝑐,𝑡
In the baseline specification:


Inequality is measured by the Gini coefficient. The analysis will also focus on measures
capturing income disparities at the top/bottom of the distribution. More specifically,
denote average disposable income in the country as ̅
Y, and the mean disposable income
of the nth decile as y̅n , then bottom inequality is measured as: BI = ̅
Y⁄y̅n (for n<5),
̅ (for n>7).
and vice versa, top inequality is measured as TI = y̅n ⁄Y



The main source of the inequality data is the OECD-IDD dataset.3 The dataset
contains a number of standardised indicators based on the central concept of
“equivalised household income”, i.e. the total income received by households
adjusted for household size with an equivalence scale. Income data refer to cash
income. This includes earnings; self-employment income; capital income (rent,
dividends and interest). The figures for public transfers and household taxes are also
included, which allows distinguishing “market” and “disposable” income (measured
after taxes and transfers). The analysis assumes the household is the unit within
which income sources are pooled and equally shared. The income attributed to each
person is adjusted for household size based on a common equivalence elasticity (the
square root of household size). While satisfactorily covering the last part of the
period 1970-2010, the IDD presents more missing values in the early sub-periods
and has therefore been integrated with information from the key figures from the LIS
(Luxembourg Income Study) database.4



Output is measured by the log of real GDP relative to the population aged 25-64 in
country c and year t (lnyc,t ) expressed in 2005 USD at purchasing power parity.
The analysis exploit five-year intervals (i.e. s = 5), so that the left-hand side
variable measures five-year growth rates of per capita GDP (Source: OECD
Annual National Accounts5).



Physical capital is proxied by the ratio of real non-residential fixed capital
formation to real GDP expressed in 2005 USD at purchasing power parity (Source:
OECD Annual National Accounts).



Human capital is measured by average years of schooling of the workingage (25-64) population. The baseline specification focuses on the level of this
variable (as in recent works on inequality and growth, see e.g. Halter et al., 2014),
but the consideration of a log transformation, which would be consistent with the
derivation in Section A2, did not change the results. The data are sourced from the
latest (2013) version of the widely used Barro and Lee dataset.6 In general, the
quality of available cross-country data on human capital has been shown to be
relatively poor (De la Fuente and Doménech, 2013). The high quality education
data re-constructed by Arnold et al. (2011) are however available only for a subset
of OECD countries and could not be used in the present analysis, as they would
imply a substantial reduction in the sample size.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 95

Panel data allow estimating the empirical link between inequality and growth while
accounting for country and period fixed effects (𝜇𝑐 , 𝜇𝑡 ). The baseline specification does
not, on the other hand, account for the last term in [A.7], cumulating population growth,
capital depreciation and technological progress (n+g+). The chapter focuses on a
simplified specification for several reasons. First, since sample size is already limited by
the availability of inequality statistics, and especially since panel estimation requires a
large number of observations, this simple specification helps maximise the degrees of
freedom. Second, within-country variation in population growth is unlikely to differ a lot
for OECD countries (capital depreciation is assumed constant and technological growth is
unobserved). Third, the adopted model is the one typically used to estimate the effect of
inequality on growth (see e.g. Perotti, 1996; Forbes, 2000; Halter et al., 2014).
It is important to discuss the possible reasons why the results shown in Table 2.1 do not
point to a positive direct effect of human capital on growth. Those findings are in fact hard
to reconcile with the large amount of evidence on the positive consequences of education
on individual productivity (from the labour literature) and on the significant contribution of
human capital to aggregate growth (from growth accounting). Yet these findings parallel
those obtained in several other growth studies exploiting panel data (Islam, 1995; Prichett,
2000) and, in particular, those adopting GMM estimation techniques (Caselli et al., 1996;
Bond et al., 2001; Castello-Climent, 2010).
One explanation is that, while eliminating one source of bias, exploiting withincountry variation dramatically lowers the precision of estimates when variables either
display high stability over time or trend in one direction (such as the stock of human
capital). This concern is even more serious given the high volatility of growth rates
measured at short horizons (e.g. five years) and the likelihood of substantial errors in the
measurement of human capital (De la Fuente and Doménech, 2006; Cohen and Soto,
2007). When variables are highly persistent, lagged levels can be weak instruments for
first differences, so that the (first difference) GMM panel data estimator is likely to be
severely biased in short panels. With System GMM, identification therefore relies on
lagged first differences having some explanatory power for levels, which might not be the
case of available measures of human capital.
One further source of bias in GMM estimates arises from (cross-country) parameter
heterogeneity (Lee et al., 1997). To address these issues, previous OECD research
(Bassanini and Scarpetta, 2002; Arnold et al., 2011) carefully re-constructed high quality,
yearly education data and looked at an error correction (ECM) version of the underlying
growth model, estimated using pooled mean group (PMG) techniques developed by
Pesaran, Shin and Smith (1999). This approach allows to deal with parameter
heterogeneity and to estimate short- and long-run coefficients separately for each growth
determinant. The results suggest that across 21 OECD countries human capital has a
robust, positive and significant impact on long-run growth. Those data and the approach
could not be used in the present analysis, however, due to the lack of yearly data on
inequality for a sufficiently large number of OECD economies.

A3.

Inequality, social mobility and human capital investment
This section describes the methodology used to test the relevance of the human
capital accumulation channel, presenting additional results.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 96 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

Inequality, background and educational attainments
The link between inequality, family background and the highest level of education
attained is estimated through an ordered probit model. An ordered probit can be derived
form a latent variable model 𝑦 ∗ = 𝑋𝛽 + 𝜖 where y* is unobserved and the error term is
normally distributed. The latent continuous variable would in this application measure the
individual’s desired level of education. While this is unobserved, we assume that there exist
two threshold levels in the support of y* that determine observable changes in education
attainment. The three available educational alternatives can be ranked and are defined as
follows: low if the individual reports having attained less than lower secondary education,
medium in case of upper secondary education and high in case of tertiary education. These
cut points 1 and 2 are therefore such that y=low if y*<1, y=med if 1<y*<2, and y=high
if y*>2. The parameters  and the cut points  can be estimated by maximum likelihood.
In this application, the estimated parameters 1 and 2 of equation [2] in the text allow
predicting the average probability of achieving each educational level by parental
education background and inequality level (i.e., at different levels of inequality).
Figure 2.4 in the main text shows that higher inequality lowers the probability of tertiary
education for low background individuals, and increases their probability of lower
secondary (or no) education.
Table 2.A1.1 reports further results obtained estimating [2] in the text when 𝐻𝐶𝑖,𝑡,𝑐 =
𝑌𝑆𝑖,𝑡,𝑐 , and where YS = years of schooling. The baseline estimate of the vector 2
(column 1) indicates that higher inequality is negatively (and significantly) related to
average schooling by low background individuals, while the link is not significant in the
case of medium or high background individuals (as in the case of attainment
probabilities). The remaining columns present alternative specifications of the control
set X, which are detailed in the table note.
Table 2.A1.1. Years of schooling, family background and inequality
(1)

Inequality X low PEB
Inequality X medium PEB
Inequality X high PEB

(2)

(3)

(5)

Lev el of

(6)

Baseline

Indiv idual controls

-0.076***

-0.061***

-0.095***

-0.061*

-0.043*

(0.024)

(0.021)

(0.022)

(0.034)

(0.022)

-0.013

0.004

-0.024

-0.010

0.010

(0.025)

(0.023)

(0.024)

(0.036)

(0.013)

dev elopment

Country specific trend Country X y ear dummy

-0.019

-0.002

-0.024

-0.019

(0.026)

(0.024)

(0.025)

(0.038)

Observ ations

64 562

64 562

62 315

64 562

64 562

R-squared

0.343

0.390

0.352

0.351

0.360

Note: The dependent variable is the number of years of schooling. All regressions control for family background
(high/med./low), country and year (age-cohort) dummies. Column 2: Adds individual controls (age, gender, region, language,
and parents’ birth place). Column 3: Adds the interaction between PEB and average GDP per capita. Column 4: Adds a countryspecific trend. Column 5: Controls for the interaction country*year (age cohort) dummies. Cluster-adjusted (country*age cohort)
standard errors are in parentheses. ***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933208850

Figure 2.A1.1 plots the average predicted number of years of schooling by educational
background as a function of inequality, using the results from the first column.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 97

Figure 2.A1.1. Years of schooling by parent educational background (PEB) and inequality
Low PEB

Medium PEB

High PEB

Years of schooling

14

13

12

11

18

20

22

24

26

28

30

32

34

36

38

Inequality (Gini coefficient)

Note: The graph plots the average predicted years of schooling for individuals from low, medium and high parental (educational)
backgrounds, as a function of the degree of inequality (Gini points) in the country. Marginal effects obtained using estimates in
column 1 of Table 2.3. Low PEB: neither parent has attained upper secondary education; medium PEB: at least one parent has
attained secondary and post-secondary, non-tertiary education; high PEB: at least one parent has attained tertiary education. The
bars indicate 95% confidence intervals. The vertical dashed lines indicate the 25th, the median and the 75th percentiles of the
underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207856

Inequality, background and skills
One key element of the PIAAC is a skill assessment exercise, consisting of a set of test
questions organised into three domains: numeracy, literacy and problem-solving. The test
results are used to impute to each participant an indicator of skill proficiency, which is
transformed into a scale ranging from 0 to 500. Hence, the survey offers the possibility to
estimate model (1) using as a dependent variable one of the available measures of skill
proficiency.
One obvious concern, however, is the extent to which the tests actually reflect skills
acquired while in education. On the one hand, proficiency in literacy and numeracy as
accumulated at school is likely to depreciate with age. On the other, skills measured later
in life might not just reflect those accumulated at school. These concerns are somehow
mitigated by the following considerations. First, if skill depreciation occurs at the same
rate for individuals in different countries, its effect would be captured by the time (age
cohort) dummies. These are always included in the specifications showed in
Table 2.A1.2, e.g. columns 1 to 3). To capture this pattern more flexibly, column 4 allows
for occupation-specific depreciation rates, measured by cohort*2-digit occupation
dummies using current information. More importantly, the specification allows
controlling for country-specific time trends and even for time- (i.e. cohort-) country
dummies, thus accounting for country-specific rates of depreciation. Finally, previous
work focusing on the consequences of job-specific training on skills as measured by the
PIAAC did not find any significant relationship, suggesting that they largely reflect skills
accumulated while studying (OECD, 2014b).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 98 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH
Table 2.A1.2. Numeracy scores, family background and inequality
(1)

(2)

(3)

(4)

(5)

(6)

Indiv idual

Lev el of

Skills

Country

Country X y ear

controls

dev elopment

depreciation

specific trend

dummy

-1.077***

-1.004***

-1.034***

-1.051***

-1.006***

(0.304)

(0.293)

(0.258)

(0.284)

(0.283)

-0.244

-0.148

-0.141

-0.310

(0.267)

(0.260)

(0.228)

(0.246)

Baseline
Inequality X low PEB
Inequality X medium PEB
Inequality X high PEB

(7)

(8)

Formal education

Ability

-0.997***

-0.773***

-0.485**

(0.259)

(0.263)

(0.195)

-0.287

-0.271*

-0.307

-0.076

(0.251)

(0.142)

(0.250)

(0.163)

-0.008

0.057

0.147

-0.005

-0.010

-0.024

0.088

(0.275)

(0.269)

(0.244)

(0.260)

(0.274)

(0.260)

(0.179)

Observ ations

65 500

65 500

63 253

65 500

65 500

65 500

65 485

51 560

R-squared

0.177

0.195

0.184

0.250

0.182

0.185

0.285

0.679

Note: The dependent variable is the PIAAC numeracy score. All regressions control for Family Background (high/med./low),
country and year (age-cohort) dummies. Column 2: Adds individual controls (age, gender, region, language, and parents’ birth
place). Column 3: Adds the interaction between PEB and average GDP per capita. Column 4: Skills depreciation columns
include interactions of age cohort*occupation (2-digit classification). Column 5: Adds a country-specific trend. Column 6:
Controls for the interaction country*year (age cohort) dummies. Column 7: Accounts for education (3-group dummy).
Column 8: Includes problem-solving score (when >0) as a proxy for ability. Cluster-adjusted (country*age cohort) standard
errors are in parentheses. ***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933208869

Literacy score
Figure 2.A1.2 reports the average predicted literacy score by educational background
as a function of inequality, using the baseline specification (column 1) of the subsequent
Table 2.A1.3.
Figure 2.A1.2. Literacy scores by parent educational background (PEB) and inequality
Low PEB

Medium PEB

High PEB

Probability of not being employed

310
300

290
280
270
260
250

18

20

22

24

26

28

30

32

34

36

38

Inequality (Gini coefficient)

Note: The graph plots the average predicted literacy score for individuals from poor, medium and high family (educational)
background, as a function of the degree of inequality (Gini points) at the time they were around 14 years old. Low PEB: neither
parent has attained upper secondary education; medium PEB: at least one parent has attained secondary and post-secondary,
non-tertiary education; high PEB: at least one parent has attained tertiary education. The bars indicate 95% confidence intervals.
The vertical dashed lines indicate the 25th, the median and the 75th percentiles of the underlying distribution of inequality.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933207862
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH – 99

Table 2.A1.3. Literacy scores, family background and inequality
(1)

(2)

(3)

Indiv idual

Lev el of

controls

dev elopment

-1.110***

-1.013***

-0.996***

-1.073***

-0.795***

-0.780***

-0.843***

-0.510***

(0.300)

(0.294)

(0.248)

(0.270)

(0.231)

(0.216)

(0.265)

(0.169)

-0.306

-0.210

-0.138

-0.384

-0.081

-0.059

-0.366

-0.125

(0.277)

(0.272)

(0.229)

(0.253)

(0.208)

(0.116)

(0.266)

(0.159)

-0.312

-0.227

-0.084

-0.305

-0.032

-0.326

-0.190

(0.275)

(0.269)

(0.240)

(0.255)

(0.227)

(0.269)

(0.159)

Observ ations

65 500

65 500

63 253

65 500

65 500

65 500

65 485

51 560

R-squared

0.174

0.180

0.181

0.252

0.181

0.184

0.287

0.718

Baseline
Inequality X low PEB
Inequality X medium PEB
Inequality X high PEB

(4)
Skills depr.

(5)

(6)

(7)

Country

Country X

Formal

specific trend y ear dummy

education

(8)
Ability

Note: The dependent variable is the PIAAC Numeracy score. All regressions control for PEB (high/medium/low), country and
year (age-cohort) dummies. Column 2: Adds individual controls (age, gender, region, language, and parents’ birthplace).
Column 3: Adds the interaction between PEB and Average GDP per capita. Column 4: Skills depreciation columns include
interactions of age-cohort*occupation (2-digit classification). Column 5: Adds a country-specific trend. Column 6: Controls for
the interaction country*year (age-cohort) dummies. Column 7: Accounts for education (3-group dummy). Column 8: Includes
problem-solving score (when >0) as a proxy for ability. Cluster-adjusted (country*age cohort) standard errors are in parentheses.
***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Source: OECD Secretariat calculations based on PIAAC data. See Box 2.3 and Annex 2.A1.
12 http://dx.doi.org/10.1787/888933208877

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 100 – 2. THE IMPACT OF INCOME INEQUALITY ON ECONOMIC GROWTH

Notes

1.

Empirically, such an equation has been adapted to assess the relevance of a variety of
growth determinants, allowing for different specifications of the functional form; see
Section A2 in Annex 2.A1 for a precise definition of the variables and specification
considered here.

2.

The country-specific implied effects shown in Figure 2.3 are obtained using the same
estimated coefficients and equation. The only difference is that, for each country, the
impact on growth was obtained cumulating the effects of each of the four five-year
changes in inequality observed between 1985 and 2005. Hence, for example,
∆𝐼𝑛𝑒𝑞1985−90 induces a shift in 𝑙𝑛𝑦 ∗ which affects the growth of GDP during 20 years
̂
(∆ ln 𝑦(2010) = (1 − 𝑒)−20𝜆 ∆ ln 𝑦 ∗ (∆𝐼𝑛𝑒𝑞1985−90 ) while ∆𝐼𝑛𝑒𝑞2000−05 only affects
̂
GDP during five years (∆ ln 𝑦(2010) = (1 − 𝑒)−5𝜆 ∆ ln 𝑦 ∗ (∆𝐼𝑛𝑒𝑞2000−05 ). Hence
earlier shifts would have a larger impact on GDP at the end of the period than would
subsequent shifts of the same magnitude.

3.

See www.oecd.org/social/income-distribution-database.htm.

4.

See www.lisdatacenter.org/data-access/key-figures/.

5.

The data can be downloaded at http://dotstat.oecd.org/Index.aspx.

6.

The data can be downloaded at www.barrolee.com/data/dataexp.htm.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 101

Chapter 3
Income inequalities during the crisis and fiscal consolidation

This chapter looks at the distribution of income during the recent global financial and
economic crisis and subsequent period of fiscal consolidation. In particular, it describes
trends in market and disposable income inequality. It analyses drivers of earnings
inequality using a decomposition of the employment and wage effects. It examines the
redistribution of income via taxes and transfers, their role as automatic stabilisers and
impact on income inequality. The chapter also explores trends in relative poverty rates
and anchored poverty rates, as well as poverty rates by age groups. Finally, it
summarises detailed analyses of tax and benefit reforms in ten OECD countries
implemented as part of fiscal stimulus and fiscal consolidation programmes and their
impact and incidence on household income.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of
such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements
in the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 102 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION

3.1. Introduction and key findings
Most OECD countries entered the global economic crisis with historically high levels
of income inequality (OECD, 2011a). In fact, some scholars even suggest that high
income inequality was a direct or indirect cause of the crisis, at the least in some
countries, such as the United States (Rajan, 2010; Stiglitz, 2012; and Fitoussi and
Saraceno, 2010).
The impact of the crisis itself on inequality, over and above the long-term trend is
uncertain, a priori. On the one hand, lower (and especially negative) returns from capital
reduce incomes at the top of the distribution, where capital income tends to be
concentrated, thereby narrowing the gap between rich and poor.1 On the other hand,
higher unemployment rates will increase inequality, particularly when job losses are
concentrated among lower-income groups.
In many OECD countries hard hit by the crisis, household income inequality and
poverty did in fact increase. However, the final impact was also affected to a significant
extent by the role played by the tax-benefit systems as well as the specific measures
implemented during the period. Although this suggests a continuation of the long-term
trend of rising inequality, a closer look reveals that the crisis significantly changed the
factors underlying the rise of income inequality and poverty in most OECD countries.
Throughout this chapter, “the crisis” refers to the period between 2007 and 2011.
This chapter is divided into six sections. Section 3.2 analyses the extent to which
income inequality has increased since 2007, what drove the rise, how sensitive the results
are to different measures, and how taxes and benefits contributed to the trends in
inequality. Section 3.3 examines how much poverty increased, using relative and
“anchored” poverty thresholds, and it explains why poverty among the elderly fell in
many countries during the crisis. Section 3.4 discusses how taxes and benefits cushioned
the impact of the crisis on household income and why this effect has become weaker in
recent years. Section 3.5 summarises ten in-depth country reviews analysing the
tax-benefit measures implemented in these countries during the crisis and their impact on
income distribution. Section 3.6 summarises and concludes.
The following key findings emerge from this analysis:


Income inequality, before accounting for taxes and benefits, rapidly increased
during the crisis in most countries. Income inequality after taxes and benefits
was already at an all-time high before the crisis, and it continued to rise. The rise
was particularly large when focusing on inequality at the lower end of the
income distribution.



Income poverty also increased during the crisis, especially when compared to
pre-crisis real income benchmarks (“anchored poverty”). In many countries
average household incomes declined in real terms. Poverty rates increased
among children and youth while they fell among the elderly.



There have been two distinct phases since the beginning of the crisis. In the
initial years, automatic stabilisers and fiscal stimulus measures cushioned the
household sector better than other sectors of the economy. As the economic
difficulties continued, benefits started to expire and governments introduced
fiscal consolidation programmes; the cushioning diminished, especially in the
countries hit hardest (e.g. Greece, Ireland and Spain).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 103



While tax-benefit reforms varied considerably across countries and across time,
in the initial years of the crisis many countries reduced taxes and/or raised
benefits, thus reinforcing the impact of automatic stabilisers on household
incomes. Many of these countries later reversed these policies and raised taxes
and cut benefits. Families with children (especially lone parents) tended to lose
more, partly due to cuts in family benefits. Broadly, measures had a progressive
effect among in-work households, with gains or smaller losses on low-earning
families and greater losses on higher-earning families.

3.2. Income inequality continued to increase through the crisis
The dispersion of market income (i.e. gross income from labour and capital)
increased strongly during the crisis. On average across the OECD, the Gini coefficient of
market income inequality increased by 1.5 percentage points between 2007 and 2011
(Figure 3.1, Panel A). This rise in market income inequality was continuous throughout
the crisis period (Figure 3.1, Panel B) and represented a shift in the drivers of inequality
of disposable income. Since the mid- to late 1990s, falling redistribution had been the
main factor driving the rise of inequality of disposable income (OECD, 2011a). The
recent rise in market income inequality was widespread: out of the 30 OECD countries
where trend data is available, it increased by more than 0.3 point in 21 countries and
decreased by more than 0.3 point in four, while in the other five there was “no significant
change”. Market income inequality rose particularly strongly in Greece, Ireland and
Spain, while it fell considerably in Poland and Israel. The drivers of the changes are
analysed below.
Higher tax-benefit redistribution cushioned the steep rise in market income inequality
during the crisis. After taking account of income taxes and cash benefits – a proxy for
“redistribution” – the rise in inequality was considerably less for disposable income than
for market income. Part of the rise in redistribution was “automatic”. With progressive
tax-benefit systems in place, greater inequality automatically leads to more redistribution,
even if no policy action is taken (Immervoll and Richardson, 2011). The rise in
redistribution was reinforced by tax-benefit reforms implemented in the early years of the
crisis, which were reversed in later years. The trend in disposable income inequality
describes two distinct phases. In the first years of the crisis, disposable income inequality
stalled or even fell slightly. In later years, it increased once again (Figure 3.1, Panel B).
These two phases in the trend are linked to the role of taxes and benefits and the fiscal
reforms implemented in the period, which are discussed in Sections 3.4 and 3.5.
Between 2007 and 2011 disposable income inequality increased in a majority of
countries. Out of the 33 OECD countries where trend data is available for disposable
income inequality, it rose by more than 0.3 point in 15 countries and fell by more than
0.3 point in nine, and in the other nine there was “no change”. The differences in
inequality between market income and disposable income varied considerably across
countries, thus revealing significant differences in the ability of tax-benefit systems to
cushion the rise of inequality. In Iceland, Portugal and Belgium, income inequality fell
after accounting for taxes and benefits, despite rises in market income inequality.2
Tax-benefit redistribution also alleviated higher market income inequality in Ireland,
Greece, Luxembourg, Slovenia, the United Kingdom, Norway and Finland, and, to a
lesser extent, in Spain, France, the United States and Estonia. In Poland, market income
inequality fell to a larger extent than disposable income inequality, as tax-benefit reforms
generally favoured middle and top incomes.3 In Israel and the Slovak Republic,
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 104 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
disposable income inequality increased due to a fall in redistribution. Similarly, in
Sweden a rise in market income inequality was accompanied by a fall in redistribution. In
Spain, disposable income inequality increased so sharply that, while in 2007 it was below
the OECD average, by 2011 it was among the ten most unequal countries in the OECD
(OECD, 2014a).
Figure 3.1. Income inequality during the crisis
1
Panel A. Percentage point change in Gini coefficient, 2007-11, total population
1

Panel A. Percentage point change in Gini coefficient, 2007-2011, total population
Gini at disposable income

Gini at market income (↗)
8
6
4
2
0

-2
-4

Panel B. Percentage change in Gini coefficient, 2007=100, OECD, total population2

x 0.01

Gini at market income

Gini at disposable income

104

103

102

101

100

99
2007

2008

2009

2010

2011

1. 2007 refers to 2006 for Chile and Japan: 2008 for Australia, Germany, Finland, France, Israel, Mexico, Norway and Sweden.
2011 refers to 2009 for Japan; 2010 for Austria, Belgium, Ireland and the United Kingdom and 2012 for Australia, Hungary,
Korea and United States; Switzerland is not available. OECD refers to the unweighted average (Turkey, Mexico and Hungary
are not included.
2. OECD refers to the unweighted average of 26 countries. Austria, Belgium, Switzerland, Hungary, Japan, Mexico,
New Zealand and Turkey are not included.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207881

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 105

The fall in employment drove the rise in market income inequality
Labour income is, by far, the largest source of income of working-age households
(OECD, 2011a), and thus drives market income inequality. Labour income can become
more unequally distributed either as a result of employment changes (e.g. higher
unemployment increasing the number of working-age individuals with low or non-labour
income) or changes in the distribution of wages (i.e. changes in the pay gap between
lower- and higher-paid workers); it is important to disentangle these two effects.
In most countries, it was the fall in employment that drove the rise in labour income
inequality during the crisis years. Figure 3.2 presents the employment and wage effects
on overall labour income inequality of working-age individuals (including employees and
self-employed workers). Estimates are based on a theoretical model proposed by
Atkinson and Brandolini (2006) and applied in OECD (2011a). During the crisis, the
employment effect (i.e. a rise in unemployment or inactivity) was the main driver in most
countries where labour income inequality rose. Before the crisis the wage effect was the
main driving force in most countries (OECD, 2011a). In Spain, out of the 8 point rise in
labour income inequality, six were due to the fall in employment and two to the wage
effect. In Greece, Ireland and Portugal, the rise in labour income inequality was also due
to the employment effect. However, in these three countries wage inequality fell perhaps
in part due to reductions in public sector wages, which had larger effects in the upper part
of the income distribution of these countries (Callan et al., 2011; Avram et al., 2013).
Only in Australia, Austria, France, Denmark and Italy, higher wage dispersion among
workers was the main – and in most cases the sole – driver of rising labour income
inequality. In the few countries that registered a fall in labour income inequality, the wage
effect was generally the main driving force.
Figure 3.2. Decomposition of changes in the Gini coefficient of labour income
Percentage point change in Gini coefficient, 2007-11, working-age individuals1
Residual

Wage effect

Employment effect

Change in Gini (↗)

8
6
4
2
0
-2

-4

1. Gini coefficient of labour income among the entire working-age population estimated by assigning zero earnings to
non-workers.
Source: OECD Secretariat calculations from the European Union Statistics on Income and Living Conditions (EU-SILC, 2008
and 2012), the Survey of Labour and Income Dynamics (SLID, 2007 and 2010) for Canada, Household, Income and Labour
Dynamics in Australia (HILDA, 2008 and 2012) for Australia and the Current Population Survey (CPS, 2008 and 2012) for the
United States.
12 http://dx.doi.org/10.1787/888933207898

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 106 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Capital and business incomes also play a role in market income inequality,
particularly at the very top of the income distribution, where they are concentrated.
According to OECD (2014c), the richer the income group, the higher its share of capital
and business income. For the five countries analysed (Canada, France, Italy, Spain and
the United States), capital income and business incomes are the largest income sources of
individuals in the top 0.01%, except in Canada. The United States was the only country
where capital and business income also drove the income growth of the top 0.1% as well.
But it must be kept in mind that capital income may be underreported in some countries;
some types of capital income are not subject to income tax and consequently not reported
in the tax files from which top income data is extracted (Förster et al., 2014).

The bottom 10% fared worse than the rest of the population
Lower-income households either lost more during the crisis or benefited less from the
recovery than other income groups. Figure 3.3 shows the annual change in household
disposable income for the median, bottom and top 10%, between 2007 and 2011. On
average, the income of the bottom 10% fell at a much higher rate than that of the top 10%
and the median. Out of the 33 countries for which data are available, the bottom 10% fared
worse than the top 10% in 21 countries and worse than the median in 27 countries.
The income fall at the bottom was particularly striking in some of the countries hit
hardest by the crisis. In Spain, Greece, Ireland, Iceland, Hungary, Mexico and Estonia,
the average income of the bottom 10% fell 5% or more per year. Except for Iceland, in all
of these countries the bottom fared considerably worse than the median and the top 10%.
In countries where median household incomes were less affected by the crisis, the
patterns were more mixed. In Austria, France, the Slovak Republic, Sweden and the
United States, the bottom 10% fared worse than the median and the top, whereas in
Belgium, Chile and Poland the bottom fared better than the top.
Figure 3.3. Changes in household real disposable income by income group
Annual percentage change between 2007 and 2011,1 total population
Median income (↗)

Bottom 10%

Top 10%

%
6
4

2
0
-2
-4

-6
-8
-10
-12
-14

1. 2007 refers to 2006 for Chile and Japan: 2008 for Australia, Germany, Finland, France, Israel, Mexico, Norway and Sweden.
2011 refers to 2009 for Japan; 2010 for Austria, Belgium, Ireland and the United Kingdom and 2012 for Australia, Hungary,
Korea and United States; Switzerland is not available. OECD33 refers to the unweighted average.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207907

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 107

On average across the OECD, the fall in income at the bottom started earlier and was
more intense than for other income groups. While median and top incomes were still
rising in 2008, bottom incomes were already falling. Since then, in each year incomes at
the bottom fell at least as much as the worst-performing income group (see Figure 3.4).
The trends were not the same across all countries. Some countries (e.g. Spain) followed
the OECD average trend, with incomes at the bottom faring considerably worse than for
other income groups since the beginning of the crisis. In other countries, incomes at the
bottom fared either better or similarly to other income groups in the early years of the
crisis, but performed considerably worse in later years. Greece is the most dramatic
example of this, with bottom incomes rising until 2009 and then plummeting afterwards.
On the other hand, in Iceland top incomes increased in 2008 and then fell strongly
afterwards.
Figure 3.4. Income trends at the bottom, top and median of the distribution
Percentage change in household disposable income by income group, 2007=100%, total population 1

95

125

2009

2011

Spain

100
75

50
2007

50
2007
125

2011

2009

2011

Portugal

100

50
2007

Iceland

100
75

75

2009

125

x 0.01

100
75

x 0.01

x 0.01

90
2007

Greece

Top 10%

50
2007
125

2009

2011

Estonia

100
75

2009

2011

50
2007

125

Ireland

100
75

x 0.01

100

125

x 0.01

OECD292

Median income

x 0.01

105

x 0.01

x 0.01

Bottom 10%

50
2007
125

2009

2011

United States

100
75

2009

2011

50
2007

2009

2011

1. Countries are ranked by average annual percentage change in median income between 2007 and 2011 (cf. Figure 3.3.)
2. OECD refers to the unweighted average of 29 OECD countries. Austria, Belgium, Japan, New Zealand and Switzerland are
not included.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207914

The fact that the bottom 10% fared worse than other income groups during the later
years of the crisis is particularly worrying as it exacerbates a long-term trend. Focusing
on 11 countries for which long-term data are available, Figure 3.5 shows that, in the last
25 years, incomes for the bottom 10% increased much less than the rest as they grew less
during expansions and fell more during recessions.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 108 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Figure 3.5. Trends in household disposable income by income group over the last 25 years
Percentage change, index 1985=1, OECD17,1 total population2
Bottom 10%

Median income

Mean income

Top 10%

1.60

1.50
1.40
1.30

1.20
1.10
1.00
1985

1990

1995

2000

2005

2010

1. OECD is the unweighted average of 17 countries: Canada, Germany, Denmark, Finland, France, United Kingdom, Greece,
Israel, Italy, Japan, Luxembourg, Mexico, Netherlands, Norway, New Zealand, Sweden and United States.
2. 1985 refers to 1983 for Sweden, 1984 for France, Italy, Mexico and United States and 1986 for Finland, Greece, Luxembourg
and Norway; 1990 refers to 1989 for France and United States and 1991 for Italy and Sweden, and is corrected to include
Greece, Japan, Luxembourg, Mexico and Norway; 1995 refers to 1994 for Greece, Mexico and United Kingdom and 1996 for
France and Luxembourg; 2005 refers to, 2003 for Japan and New Zealand, 2004 for Germany, Finland, Mexico, Norway and
Sweden and 2006 for Italy; 2007 refers to 2006 Japan and 2008 for Germany, Finland, France, Israel, Mexico, Norway, New
Zealand, Sweden and United States: 2011 refers to 2009 for Japan 2012 for Mexico, Netherlands and United States.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207920

Box 3.1. Do the results differ when using other inequality measures?
The results presented above focus on estimates based on the Gini coefficient, an indicator that is particularly
sensitive to changes around the middle part of the distribution (Lambert, 2001). The trends in income inequality
during the crisis and particularly the levels differ depending on the indicator used to measure it. Indicators that
are less sensitive to changes at the bottom part of the distribution suggest relatively small changes, including a
slight fall in 2009. However, indicators that are more sensitive to changes at the bottom indicate a continuous
increase in income inequality which intensified in more recent years.
According to the S90/S10 indicator (which compares the income share of the top 10% with the
bottom 10%), between 2007 and 2011 income inequality increased of almost 5% on average across OECD
countries. The rise in inequality was also significant when looking at the income share of the top and bottom
20% (S80/S20), as well as the ratio of the highest income of the bottom 90% and bottom 10% (P90/P10) and the
ratio of the median to the highest income at the bottom 10% (P50/P10). Income inequality increased less when
comparing the top and middle of the distribution (P90/P50). The Palma index (the ratio between the income
share of the top 10% and the bottom 40%) describes a similar pattern as the Gini coefficient.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 109

Box 3.1. Do the results differ when using other inequality measures? (cont)
Trends of different income inequality indicators through the crisis
Percentage change in income inequality,1 2007=100%, OECD,2 total population

x 0.01

Gini

Palma

S80/S20

P90/P50

S90/S10

P90/P10

P50/P10

105
104

103
102
101
100
99
2007

2008

2009

2010

2011

1. See text above for explanation of indicators.
2. OECD is the unweighted average of 29 countries; Austria, Belgium, Japan, New Zealand and Switzerland are not
included.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933208012

3.3. Income poverty increased, no matter how it is measured
The higher income losses at the bottom of the income distribution raise concerns
about poverty. Indeed, income poverty increased during the crisis. However, the intensity
of the change and in some countries even its direction differ depending on how poverty is
measured. Income poverty increased less when measured against a relative threshold than
against an anchored poverty threshold. While the relative poverty threshold is based on
the current median income (hence variable), the anchored poverty threshold is based on
the median income of a previous year (hence fixed). Since median incomes fell as a result
of the crisis (see Figure 3.3), so did the relative poverty threshold.
Indicators of relative income poverty, such as the ones most commonly used by the
OECD and the European Union (OECD, 2008; European Commission, 2014), are based on
the notion that poverty is defined in relation to a country’s general level of prosperity at a
given point in time (Atkinson et al., 2002). Furthermore, unlike measures based on an
absolute poverty threshold, relative measures allow international comparisons
independently of specific national definitions of basic needs (Förster and Mira d’Ercole,
2012). However, in examining changes over time, a relative approach, based on percentages
of current mean or median income, may overemphasise the comparative dimension of
poverty at the expense of real absolute changes in living standards. Methodologically, this
issue is often illustrated by the fact that a strictly relative indicator would suggest no change
in poverty if the incomes of all households were to double or halve.
The analysis in this section shows evidence of a similar type of issue during the
economic crisis. In a number of countries, relative poverty thresholds fell following a
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 110 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
decrease in real median incomes. As relative poverty thresholds fall, households
previously considered “in poverty” may now be classified differently, even though their
income remains the same. This situation may even apply to households whose income
falls, if the fall is lower than the median.
Although this type of result is consistent with the relative approach (i.e. inter-personal
comparisons), it fails to recognise people’s comparisons to their own circumstances in a
(recent) previous period. Anchored poverty lines provide an alternative that addresses this
issue. Similarly to the relative approach, anchored poverty lines are based on relative
living standards (e.g. median income). However, instead of using current living standards
(as in the relative approach), anchored lines are based on the living standards of a
previous year, adjusted for inflation. The anchored poverty threshold used in this analysis
is based on the median income of 2005.
Between 2007 and 2011, the OECD anchored poverty rate increased by about one
percentage point (see Figure 3.6). Relative poverty grew by just a third of a percentage
point, from 11.2% to 11.5%. In Greece, anchored poverty more than doubled (rising from
12% to 27%), reflecting the dramatic effect of the crisis on household income. The
relative poverty rate also went up in Greece, but just by 1 percentage point. Anchored
poverty rates also increased considerably in Spain (8 points), Ireland, Iceland, Mexico
and Hungary. In some countries, most noticeably Estonia, Portugal, New Zealand, Ireland
and Iceland, relative poverty fell while anchored poverty increased, indicating that the
income of some low-income groups (particularly the elderly, see below) did not fall as
much as the median. On the other hand, anchored poverty rates fell while relative poverty
increased in Israel, the Slovak Republic and Poland, as those at the bottom benefited from
income growth but not to the same extent as the median. Both anchored and relative
poverty fell in Chile, Finland and the United Kingdom.
Figure 3.6. Changes in relative and anchored poverty rates during the crisis
Percentage point change, 2007-11,1 total population
Changes in anchored poverty rate (↗)

8

Changes in relative poverty rate

15

6
4
2
0
-2
-4
-6
-8

1. 2007 refers to 2006 for Chile and Japan: 2008 for Australia, Germany, Finland, France, Israel, Mexico, Norway and Sweden.
2011 refers to 2009 for Japan; 2010 for Austria, Belgium, Ireland and the United Kingdom and 2012 for Australia, Hungary,
Korea and United States; Switzerland is not available. OECD33 refers to the unweighted average.
2. Hungary is the only country were relative poverty rate increased more than anchored poverty rate. This is due to a
combination of very low but different anchored and relative poverty rates in 2007 (2% and 6%, respectively) and a strong fall of
low incomes between 2007 and 2011 (see Figure 3.3).
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207934
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 111

Poverty increased for all age groups except the elderly
The age profile of relative poverty changed considerably during the crisis. Relative
poverty increased among all age groups except the elderly. Between 2007 and 2011, on
average across the OECD, poverty increased by 1 point among children, 2 points among
youths and 1 point among working-age adults (see Figure 3.7). In 2007, the elderly
(particularly aged 75 or more) were the age group with the highest poverty incidence. By
2011, the young and children had taken this place. OECD (2008) had already identified a
long-term trend of increasing poverty rates among the young and falling poverty rates
among the elderly. The crisis clearly accelerated this process.
Figure 3.7. Relative poverty rates by age group in 2007 and 2011 1, 2
2007

75+
75+

51 - 65

65 - 75
65 - 75

51 - 65

75+

65 - 75

51 - 65

41 - 50

26 - 40

0 - 17

18 - 25

75+

51 - 65

65 - 75

0%

41 - 50

0%
26 - 40

0%
0 - 17

0%

18 - 25

10%

75+

20%

10%

51 - 65

20%

10%

65 - 75

20%

10%

41 - 50

20%

26 - 40

30%

0 - 17

30%

18 - 25

30%

41 - 50

United States

30%

26 - 40

Estonia

41 - 50

0 - 17

75+

51 - 65

65 - 75

41 - 50

26 - 40

0 - 17

Portugal

18 - 25

0 - 17

Spain

75+

0%

51 - 65

0%

65 - 75

0%

41 - 50

0%

26 - 40

10%

18 - 25

10%

75+

10%

51 - 65

10%

65 - 75

20%

41 - 50

20%

26 - 40

20%

0 - 17

20%

18 - 25

30%

26 - 40

Ireland

30%

18 - 25

Iceland

30%

0 - 17

Greece

30%

18 - 25

OECD32

2011

1. 2011 refers to 2012 for United States. OECD refers to the simple (unweighted) average of 32 OECD countries, Switzerland
and Korea are not included
2. Countries are ranked by average annual percentage change in median income between 2007 and 2011 (cf. Figure 3.3.)
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207942

For the first time since the OECD started collecting this data, in 2011 the poverty rate
of people aged 66 to 75 was lower than the population average (OECD, 2014a). Between
2007 and 2011, the OECD average relative poverty rate fell by 2.6 points among people
aged 66 to 75 and by 4 points among people over age 75. The fall in elderly poverty was
widespread: poverty among people aged 66 to 75 fell by 1 point or more in 18 countries
and among people over age 75 in 21 countries (Figure 3.8). In Estonia, relative poverty
among people over age 75 fell by 27 points. Substantial falls in poverty also took place in
New Zealand, Spain, Portugal Ireland and Greece. On the other hand, in Poland, Korea,
Turkey, Austria, Mexico and Canada elderly poverty increased noticeably between 2007
and 2011.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 112 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Figure 3.8. Changes in relative poverty among the elderly during the crisis
Percentage point change in relative elderly poverty rates, 2007 to 2011,1 by age group of elderly
People aged 66-75 (↗)

People aged 76+

15
%
10
5
0
-5
-10
-15
-20
-25
-30

1. 2007 refers to 2006 for Chile and Japan: 2008 for Australia, Germany, Finland, France, Israel, Mexico, Norway and Sweden.
2011 refers to 2009 for Japan; 2010 for Austria, Belgium, Ireland and the United Kingdom and 2012 for Australia, Hungary,
Korea and United States; data for Switzerland and for people aged 76+ in Korea are not available. OECD33 refers to the
unweighted average.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207952

Relative poverty rates among the elderly fell due to a combination of falling poverty
thresholds and rising incomes. Elderly poverty rates can be quite responsive to slight
changes in the poverty line, as in many countries a considerable part of low-income
pensioners are clustered around the poverty threshold. On the other hand, elderly incomes
were not hit as hard by the crisis, as these depend more on public pension transfers and
are less sensitive to changes in the labour market. Even in some of the countries hit
hardest by the crisis, such as Estonia, Portugal, Spain and Italy, the average incomes of
the elderly either rose or at worst stagnated since 2007. In Greece, Iceland and Ireland
elderly average income fell, but considerably less than for other age groups. Elderly
income also outperformed that of the rest of the population in some countries that
experienced overall income growth, such as Israel, Norway, the Slovak Republic and
Sweden. Only in Turkey was elderly income considerably outperformed by that of the
rest of the population (see Figure 3.9).
This protection of the elderly from the crisis can be considered a positive outcome of
the social protection systems. Not only are the elderly less able to adapt to changing
circumstances in the labour market and find new sources of income but their household
income is also on average lower than for the rest of the population. Across OECD countries,
elderly income is 13% lower than the population average (see Figure 3.9). On the other
hand, the rise in child poverty is particularly worrying. Poverty in childhood can have a
damaging and lasting effect on people’s future outcomes, such as cognitive and behavioural
development or health outcomes. In the long run, early childhood poverty is associated with
reduced adult working hours (and so earnings) and higher poverty risks and welfare
dependency later in life (OECD, 2011b; Duncan et al., 2010).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 113

Figure 3.9. Changes and relative differences in household disposable income by age group
Percentage difference to population average, 2011 5

Annual percentage change in household disposable income, 2007 to 2011 1, 2, 3, 4
Children
-10

0

Adults

Youths
10

-10

0

10

-10

0

Elderly
10

-10

0

Children

Total (↗)
10

-10

0

10

-40

0

Youths
40 -40

0

Adults
40 -40

0

Elderly
40 -40

0

40

GRC
ISL
IRL
ESP
MEX
PRT
EST
ITA
NLD
HUN
JPN
LUX
GBR
NZL
SVN
CZE
USA
FRA
DNK
BEL
DEU
ISR
AUS
CAN
NOR
TUR
FIN
KOR
AUT
SWE
CHL
POL
SVK
-

-

-

-

CHE
OECD336

-10

0

10

-10

0

10

-10

0

10

-10

0

10

-10

0

10

-40

0

40

-40

0

40

-40

0

40

-40

0

40

1. Annual percentage changes in household disposable income between 2007 and 2011, by age group and total population.
2. 2007 refers to 2006 for Chile and Japan: 2008 for Australia, Germany, Finland, France, Israel, Mexico, Norway and Sweden.
2011 refers to 2009 for Japan; 2010 for Austria, Belgium, Ireland and the United Kingdom and 2012 for Australia, Hungary,
Korea and United States; Switzerland is not available.
3. Children refer to 0-17 years old; youth refers to 18-25 years old; adults refer to 26-65 years old and elderly refers to over
65 years old.
4. Household incomes are equivalised by household size (www.oecd.org/els/soc/OECD-Note-EquivalenceScales.pdf).
5. Difference between average household disposable income of each age group and that of total population in 2011.
6. OECD33 refers to the unweighted average of 33 countries. Trend data for Switzerland is missing.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207962

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 114 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION

3.4. Taxes and benefits cushioned the impact of the crisis on household income, but
mostly during the first years
Significant increases in public expenditure on social benefits and lower personal
income tax burdens prevented inequality in disposable income from rising as much as
disparities in market income during the early phase of the crisis. This cushioning effect
resulted from the stabilising properties of the tax-benefit system, in some cases reinforced
by fiscal stimulus measures. However, in more recent years the effect diminished as
entitlements to social benefits expired and most governments switched from fiscal stimulus
to fiscal consolidation to tackle increases in the public deficits and high debt-to-GDP ratios.
Taxes and benefits acted as automatic stabilisers, offsetting part of the income losses
resulting from the economic downturn. Public spending on social benefits typically
increases as more people claim unemployment or other safety-net benefits. At the same
time, tax revenues fall as the tax base (e.g. income or consumption) shrinks.
In the early stages of the crisis several governments reinforced the stabilising effect of
their tax-benefit systems through fiscal stimulus packages that raised public expenditure
and reduced taxes (as for example in the United States between 2007 and 2010, see
Section 3.5). As the crisis continued, however, governments, facing large and increasing
budget deficits, reversed these policies and began implementing fiscal consolidation
programmes. In many countries, the consolidation measures included rises in personal
income taxes (Spain in 2011 and 2012) and reductions in social cash benefits (Spain in
2013, Ireland in 2011 and 2012, Portugal in 2012), thus directly affecting household
income.4
Figure 3.10 clearly illustrates these two phases of fiscal policies during the crisis, with
2010 as a transition year. Between 2007 and 2009, personal income taxes decreased and
cash social benefits increased in response to the fall in market income. The changes in
both types of policies were considerable: benefit spending rose by 10% and tax revenues
fell by 4%. Between 2010 and 2011, market income continued to fall, but now the
cushioning effect was reversed, with taxes rising and benefits falling.
In general, similar patterns took place in most OECD countries at the onset of the
crisis, with benefits rising and taxes falling. In later years, trends in taxes and benefits
across countries started to diverge. Figure 3.11 shows developments in market income,
taxes and benefits in eight selected countries for which annual income data are available.
In Greece, benefits increased significantly up to 2009 but started to fall in 2010. By 2011,
benefits were below pre-crisis levels. Following a dramatic decline in market income,
taxes plunged in 2010, but recovered slightly in 2011, as reforms were introduced. In
Iceland and Estonia, benefits rose steeply in 2008, but started to fall in 2009, while taxes
picked up in 2010 and 2011, as market income began to recover. In Ireland, the
substantial tax hikes started in 2009 raised taxes well above their pre-crisis levels. On the
other hand, benefits continued to increase up to 2010 but fell sharply in 2011. In Spain,
despite the fiscal consolidation measures introduced since 2010, benefits have not fallen
but tax levels have – suggesting that the rises in tax rates were not sufficient to
compensate the fall in taxable income. In Portugal, as market income worsened, benefits
continued to rise. On the other hand, following reforms implemented in 2010 and 2011,
taxes increased back to their pre-crisis levels. In the United States, the tax-benefit
cushioning effect started with a rise in benefits in 2008 and 2009, followed by a reduction
of pension contributions in 2011. In Finland, as in other Nordic countries, benefits
increased significantly as market incomes stagnated or fell.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 115

Figure 3.10. Trends in taxes, benefits and market income during the crisis
Percentage change, 2007=100%, OECD,1 total population

x 0.01

Household market income

Cash social benefits

Personal income taxes

112
110
108
106
104
102
100
98
96

94
2007

2008

2009

2010

2011

1. Household incomes are equivalised by household size and adjusted for inflation using changes in consumer price indices and
differences in purchasing power using 2010 PPP indices for private consumption. OECD average refers to the unweighted
average of 26 countries. Austria, Belgium, Switzerland, Hungary, Japan, Mexico, Switzerland and Turkey are not included.
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207976

Figure 3.11. Trends in taxes, benefits and market income across countries
Percentage change, 2007=100%, total population1
Household market income

80

80
2009

2011

60
2007

120
100
80
60
2007

2009

2011

60
2007

120
100

2011

60
2007

100
80

2009

2011

60
2007

120
100

2011

60
2007

2009

2011

Finland

80
2009

120

United States

80
2009

100

Estonia

x 0.01

x 0.01

Portugal

120

80

x 0.01

60
2007

100

x 0.01

100

120

Spain

x 0.01

120

Personal income taxes

Ireland

x 0.01

Iceland
x 0.01

x 0.01

Greece

Cash social benefits

120
100
80

2009

2011

60
2007

2009

2011

1. Countries are ranked by average annual percentage change in median income between 2007 and 2011 (cf. Figure 3.3).
Source: OECD Income Distribution Database (2014), www.oecd.org/social/income-distribution-database.htm.
12 http://dx.doi.org/10.1787/888933207982

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 116 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION

3.5. Tax-benefit measures during the crisis: A summary of detailed analyses of
ten countries
The tax-benefit measures implemented during the crisis varied considerably not only
across time and countries but also with respect to how they were designed and their
impact on the incomes of households in different demographic, labour market and
earnings circumstances. The previous sections showed that the tax-benefit system played
an important role in cushioning the impact of the crisis in its initial years. Part of the
cushioning effect was automatic – with taxes falling and benefits rising as earnings and
employment tumbled. However, several new measures introduced by governments during
the period also had an important impact shaping the role of taxes and benefits,
particularly with the introduction of fiscal consolidation programmes.
This section summarises policy changes applied to taxes and benefits between 2008
and 2013 for a selection of OECD countries (Estonia, France, Germany, Greece, Iceland,
Ireland, Portugal, Spain, the United Kingdom and the United States) including those hit
hardest by the crisis.5 Simulations with the OECD tax-benefit model are used to make an
in-depth analysis of policy changes by policy, type of measure and impact on household
income according to demographic and labour market characteristics and earning levels.

Context
The economic crisis had a major impact on public finances. As public revenues fell
and expenditure rose, many OECD governments ran large budget deficits (see Table 3.1).
Part of the mounting pressure on public finances had to do with the cushioning of the
household sector that resulted from the stabilising properties of the tax-benefit system and
fiscal stimulus programmes. As the ratios of public budget deficits-to-GDP rose (in many
countries reaching double figures), most OECD governments introduced fiscal
consolidation measures to reduce public borrowing and debt levels.
Table 3.1. Public budget deficits during the crisis
Government net lending, as a percentage of GDP, 2007-13
2007

2008

2009

2010

2011

2012

2013

Estonia

1.7

-3.6

-2.3

0.2

1.0

-0.3

-0.5

France

-2.5

-3.2

-7.2

-6.8

-5.1

-4.9

-4.1

Germany

0.3

0.0

-3.0

-4.1

-0.9

0.1

0.1

Greece

-6.7

-9.9

-15.2

-11.0

-10.1

-8.6

-12.2

Iceland

5.1

-12.9

-9.4

-9.5

-5.3

-3.7

-2.0

Ireland

0.2

-7.0

-13.9

-32.4

-12.6

-8.1

-5.7

Portugal

-3.0

-3.8

-9.8

-11.2

-7.4

-5.5

-4.9

Spain

2.0

-4.4

-11.0

-9.4

-9.4

-10.3

-6.8

United Kingdom

-2.9

-4.9

-10.7

-9.5

-7.5

-5.9

-5.6

United States

-3.7

-7.2

-12.8

-12.2

-10.7

-9.0

-5.7

OECD10

-0.9

-5.7

-9.5

-10.6

-6.8

-5.6

-4.7

Note: OECD10 is the average of the ten countries analysed in this section.
Source: OECD (2014), OECD Economic Outlook, No. 96, November, http://dx.doi.org/10.1787/data-00717-en.
12 http://dx.doi.org/10.1787/888933208889

OECD (2009) assessed the fiscal position of OECD countries in the early years of the
crisis and also documented and examined the fiscal stimulus programmes introduced in
the period 2008-10. The study suggested that most of the increase in budget deficits in
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 117

this period was a cyclical effect due to the operation of automatic stabilisers. On average,
the size of the impact of the automatic stabilisers over the period 2008-10 was about three
times the size of the fiscal stimulus programmes. While most countries announced fiscal
stimulus measures, they varied substantially in size, with the United States announcing
the largest programme. On average, their cumulative impact on fiscal balances amounted
to about 3.5% of GDP. Most packages adopted a broad range of measures, adjusting
various taxes and spending programmes simultaneously. A majority of countries gave
priority to tax cuts over spending rises. In the United States, in 2008, the stimulus focused
entirely on tax cuts, whereas in 2009 about two-thirds was on spending measures. Tax
cuts were concentrated on personal income taxes and to a lesser extent on business taxes,
with the United Kingdom and Portugal cutting VAT. On the spending side, most
countries increased public investment. Transfers to households were often made more
generous, in particular for those on low incomes. A few countries also announced larger
subsidies to the business sector.
OECD (2011c) and OECD (2012) analysed and compared the fiscal consolidation
plans announced for the period 2009-15. OECD (2012) categorised countries into four
groups (see Table 3.2) based on the different pressures and pace of fiscal consolidation.
The analysis indicated that while most countries announced large deficit reductions, the
consolidation plans varied significantly in size, across time and by composition.
Table 3.2. Countries by category of fiscal consolidation
Consolidation category

Countries

A. Countries w ith IMF/EU/ECB programmes

Greece, Ireland and Portugal

B. Countries under distinct market pressure

Belgium, Hungary , Italy , Poland, the Slov ak Republic, Slov enia and Spain
Austria, Canada, the Czech Republic, Denmark, Finland, France,

C. Countries w ith substantial deficits and/or debt, but less market pressure

Germany, Iceland, Israel, Japan, Mex ico, the Netherlands, New Zealand,
the United Kingdom and the United States

D. Countries w ith no or marginal consolidation needs

Australia, Chile, Estonia, Korea, Lux embourg, Norw ay , Sw eden,
Sw itzerland and Turkey

Note: Countries analysed in this section are in bold.
Source: OECD (2012), Restoring Public Finances: 2012 Update, OECD Publishing, Paris, http://dx.doi.org/10.1787/
9789264179455-en.

According to OECD (2012), the size of the announced consolidation plans varied
significantly depending on the country’s fiscal position and the current status and time
frame of the consolidation plan. Countries with the largest economic and public finance
imbalances announced larger consolidation plans. The three countries with programmes
with the IMF/EU/ECB (Greece, Ireland and Portugal) announced the largest packages, all
above 10% of GDP (Figure 3.12, Panel A). Iceland, Spain and the United Kingdom also
adopted fiscal consolidation plans amounting to over 5% of GDP, whereas France,
Estonia and Germany announced plans with a cumulative impact of between 3% and 5%
of GDP. The plans also changed considerably as circumstances evolved. For example,
between 2011 and 2012 the volume of fiscal consolidation was revised by at least 1% of
GDP in 17 countries and as much as 5% of GDP in Portugal. Similarly, the time span of
the consolidation plans also varied and tended to be occasionally revised.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 118 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
The fiscal consolidation plans tended to focus on reducing expenditure, partly scaling
back the spending rises implemented during the fiscal stimulus. On average, about twothirds of the consolidation packages announced by 2012 relied on reducing expenditure
and one-third on increasing revenue (Figure 3.12, Panel B). There was nevertheless a
significant variation in the composition of consolidation measures. Countries with smaller
consolidation plans focused mostly on expenditure measures. Some countries with large
consolidation plans (including Germany, Iceland, Ireland, Spain and the United
Kingdom) focused on reductions in expenditure that exceeded the OECD average;
whereas others (including France, Greece and Portugal) took the middle ground. Only a
few countries, particularly those that were already withdrawing from consolidation
(including Estonia), relied on tax increases for the majority of their consolidation.
Figure 3.12. Fiscal consolidation plans, 2009-15
Panel A. Cumulative fiscal consolidation plans as a percentage of GDP
Consolidation plans in 2012-15

Implemented consolidation in 2009-11

20
%

18
16

14
12
10
8
6

4
2

0
-2

Panel B. Share of consolidation plans on expenditure reductions and revenue enhancements
Revenue enhancements

Expenditure reductions

100

80

60

40

20
0

Note: The data are the sum of annual incremental consolidation from 2009/10 until 2015 as reported by the national authorities.
Korea, Japan and the United States have not reported an announced concrete consolidation plan and are not included in the
figure. Australia reports consolidation but applies a broader definition of the term consolidation than this report. Norway is not
applying a consolidation plan.
Source: OECD (2012), Restoring Public Finances: 2012 Update, OECD Publishing, Paris, http://dx.doi.org/10.1787/
9789264179455-en.
12 http://dx.doi.org/10.1787/888933207991

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 119

Governments can use a number of instruments to achieve their fiscal consolidation
objectives. Table 3.3 lists the instruments most frequently reported by OECD
governments for fiscal consolidation plans between 2009 and 2015. Rawdanowicz et al.
(2013) examined how policy instruments may influence equity as well as their impact on
long-term growth, and then assessed the equity implications of the consolidation plans
announced in OECD countries. Similarly, Cournède et al. (2013) evaluated the potential
short- and long-term impact of policy instrument on growth, equity and current accounts
(see Table 3.4). Their results suggest that half of OECD countries could reduce excess
debt mainly through moderate adjustments to instruments that have at most limited sideeffects on growth and equity. They also found that a smaller number of countries faced
more difficult choices, as they would need either to make bigger adjustments in areas
where spending cuts or tax hikes are least harmful or to rely significantly on
consolidation instruments with substantial adverse side-effects. These trade-offs could be
alleviated through structural reforms in the delivery of public services and in taxation.
Table 3.3. Fiscal consolidation instruments reported by governments
Expenditure
Welfare
Wage cuts
Health
Pensions
Infrastructure
Staff reduction
Other
Revenue
Consumption taxes
Personal income tax
Corporate income tax
Property tax
Tax expenditures
Non-tax rev enue
Social security contributions
Tax on financial sector
Improv ing tax compliance

Note: Instruments are sorted in decreasing order by reporting frequency. Policies fully analysed in this section are in bold, and
policies partially analysed in bold and italic.
Source:
OECD
(2012),
Restoring
http://dx.doi.org/10.1787/9789264179455-en.

Public

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

Finances:

2012

Update,

OECD

Publishing,

Paris,

 120 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Table 3.4. Summary assessment of growth and equity effects of fiscal consolidation instruments
Growth

Equity

Short-term

Long-term

Short-term

Long-term

Education

--

--

-

--

Health serv ices prov ided in kind

--

-

-

-

Other gov ernment consumption (ex cluding family policy )

--

+

-

Spending cuts

++

Pensions
Sickness and disability pay ments

-

+

--

Unemploy ment benefits

-

+

-

-

Family

-

-

--

--

Subsidies

-

++

+

+

Public inv estment

--

--

Personal income tax es

-

--

+

+

Social security contributions

-

--

-

-

Corporate income tax es

-

--

+

+

Env ironmental tax es

-

+

*

-

Consumption tax es (other than env ironmental)

-

-

-

Recurrent tax es on immov able property

-

Other property tax es

-

Sales of goods and serv ices

-

Revenue increases

+

++

+

-

-

Note:
+ sign reflects positive welfare effects. – sign reflects negative welfare effects.
* This + sign reflects positive welfare effects as the long-term impact on output narrowly defined as GDP may be ambiguous.
Source: Cournède, B., A. Goujard and Á. Pina (2013), “How to Achieve Growth- and Equity-friendly Fiscal Consolidation?
A Proposed Methodology for Instrument Choice with an Illustrative Application to OECD Countries”, OECD Economics
Department Working Papers, No. 1088, OECD Publishing, Paris, http://dx.doi.org/10.1787/5k407lwvzkkh-en.

Tax and benefit measures implemented during the crisis
Countries used their tax-benefit toolkits in different ways. Table 3.5 summarises the
aggregate effects of measures introduced between 2008 and 2013, by policy type, based
on simulations with the OECD tax-benefit model (see Box 3.2). Households gained from
personal income tax cuts in Germany, the United Kingdom and the United States, and
from benefit rises in Estonia, France, Germany, Iceland and the United States.
Households lost mainly from benefit cuts particularly in Ireland and Portugal but also in
Greece, Spain and the United Kingdom. Tax rises led to significant losses to households
in Greece, Iceland, Ireland and Portugal, and to a lesser extent in France and Spain.
Value-added tax rates (VAT) increased in most countries, particularly Greece and Spain.
Measures on pensions were less homogeneous across countries.
Different mechanisms were used both to cut benefits and to raise taxes. Benefits were
cut by lowering levels, thresholds or rates, restricting eligibility, reducing durations and
diminishing real-term amounts by failing to update them in line with inflation (fiscal
drag). Taxes were raised by implementing new taxes, increasing the number and/or level
of tax rates, and reducing tax deductions in nominal or real terms by fiscal drag.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 121

Box 3.2. Assessing policy changes with the OECD tax-benefit model
The OECD tax-benefit model
The OECD tax-benefit model (TBM) computes the effects of tax and benefit systems on household
disposable income. The model contains information for 33 OECD countries, and six EU non-OECD countries,
currently for years 2001 to 2013. The benefit model covers unemployment, social assistance, housing, family
and employment-conditional benefits. The tax system covers personal income taxes and social security
contributions. The results capture the effects of taxes and benefits on the incomes of working-age individuals
and their families both in and out of work (see www.oecd.org/social/benefits-and-wages.htm).
By modelling taxes and benefits in detail and in an integrated way, the TBM can assess structural policy
reforms such as changes in tax exemptions or benefit eligibility, as well as account for the complexity of the
knock-on effects and interactions between policies and policy elements. Furthermore, as the results are
computed at the household level, the analyses based on this model account for differences in household and
personal circumstances, such as household composition, labour market status and earning levels.
The scope of the modelled policies focuses on those with a direct and immediate effect on the household
income of working-age individuals in employment or unemployment. Pensions, sickness and disability payments,
public services, in-kind benefits and indirect, capital and self-employment taxes and contributions are not included.

Simulations
Simulations with the TBM compute changes in the amount of taxes and benefits for a group of family types.
The characteristics of these family types vary in terms of family composition, labour circumstances and earning
levels. Six family types are considered: single individuals, single parents with children, one-earner couples with
and without children and two-earner couples with and without children. In all the families there is at least one
active working-age individual (either employed or unemployed). If present, other working-age individuals in the
household are assumed to be either inactive (in one-earner households) or working full-time and earning the
average wage in the country (two-earner households). The labour market circumstances of active working-age
individuals include unemployment (from 1 to 60 months out of work) and employment (with earnings ranging
up to 200% of the average wage in the country).

Results
The results are computed as unweighted averages by groups of family types. The size and composition of
the groups depend on the level of analysis. Full aggregates (see, for example, Table 3.5) are based on the
unweighted average of all analysed family types (i.e. household composition, labour market circumstances and
earning levels, amounting to 1 560 households). While these results are not representative at the population
level, as they do not account for the frequency of each of these household circumstances,* they are sensitive to
and illustrative of the effects of policy changes under different household circumstances. Furthermore, these
results are broadly consistent with those drawn from actual and population-representative data (see below).
Thus, though TBM simulations may not accurately estimate magnitudes, findings suggest that they provide
appropriate indicators for a qualitative assessment of the overall effect of tax-benefit reforms.
Also, due to the same caveats, the results can only partially assess the distributive impact of fiscal reforms
through different breakdowns, including by earnings levels. A full distributive incidence analysis would require
that simulations were computed on representative household survey data, as it is done with tax-benefit
microsimulation models such as the EU tax-benefit microsimulation model (EUROMOD). On the other hand,
analysis based on “typical” families allows many of the determinants of tax and benefit amounts to be held
constant while changing one household characteristic at a time, thus facilitating the understanding of existing
policy instruments as well as the differences between them across countries and different points in time.
Although pensions and value-added tax (VAT) are not simulated by the OECD tax-benefit model, statutory
changes are shown, based on additional national sources. Only changes that affect current pension levels
(e.g. indexation rules) are indicated. Other changes (e.g. retirement age) are not considered. Similarly, only
changes in VAT standard rates are indicated. Other changes (e.g. tax classification of goods) are not considered.
* See Immervoll, H. et al. (2004) and Levy (forthcoming) for an assessment of the scope and incidence of the policies
simulated by the TBM and the frequency of the standard family types in the actual population of each country.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 122 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Unemployment benefits became more generous in the United States via both an
automatic mechanism that is triggered once state-level unemployment rates exceed a predefined threshold but also by the decision of the Obama administration to extend the
maximum duration to 99 weeks (Immervoll and Richardson, 2013). In Iceland, the
duration of unemployment benefits was also extended in the early years of the crisis, but
was subsequently brought back to its original level. In Spain, Ireland and Greece, the
unemployment insurance level was reduced, but in Spain a new benefit for the long-term
unemployed was introduced, and in Greece the means-tested threshold for unemployment
assistance was raised. In Portugal, the benefit amount and duration were reduced but
eligibility conditions relaxed. In Germany, long-term unemployment benefits were
reduced as transition measures implemented in the Hartz labour market reform
(OECD, 2007) were phased out.
Family benefits increased in Germany as the levels were increased above inflation. In
Ireland and Portugal, family benefit levels were lowered in nominal terms. In the United
Kingdom, Spain, Greece and Iceland, benefit levels fell in real terms, as they were kept
frozen or increased below inflation. In the United Kingdom, eligibility for child benefit,
previously a universal benefit, was effectively withdrawn from high-income taxpayers.
In-work benefits are widely used in the United States, the United Kingdom, Ireland
and France as policies to protect low-earning families and to increase working incentives
(Pearson and Scarpetta, 2000; Immervoll and Pearson, 2009). In the United States these
were increased and temporarily supplemented in 2008 and 2009. In Ireland, the levels
were increased considerably in 2008 to 2010. In the United Kingdom, some elements of
the in-work benefits were cut and others increased. The levels of working tax credit were
frozen, except for the child care tax credit, which increased above inflation. The child tax
credit had the family element frozen, the baby element removed and the tapering
increased, but the child element increased above inflation. In France, several amounts of
the in-work benefit, Prime pour l’emploi, were kept frozen and the lump-sum in-work
benefit (Prime de retour à l’emploi) was abolished and replaced as part of the new social
assistance benefit (Revenu de solidarité active).
Housing benefits were cut by keeping levels frozen in Germany, and by a reform in
the United Kingdom. In France, Ireland and Iceland, despite being kept frozen in some
years, housing benefit levels in 2013 were higher in real terms than in 2007. Housing
benefits are not simulated in Portugal, Spain and the United States.
Social assistance levels fell in real terms due to fiscal drag in Germany and to
nominal cuts in Portugal. On the other hand, social assistance benefits increased both in
Estonia, Iceland and the United States, as levels rose above inflation, and in France with
the introduction of the new social assistance benefit (Revenu de solidarité active), which
is more generous than the previous system for working families with low earnings.
Real-term benefit levels remained more or less constant in Ireland and the United
Kingdom, and no social assistance benefit is simulated in Spain and Greece.
Employee social security contributions increased in Iceland and the United Kingdom,
as rates were raised, and in Greece, as the contribution ceiling was lifted. Contributions
fell both in Germany, as rates were reduced and contribution floors failed to increase with
inflation, and in Ireland as health contributions were replaced by a new income tax (see
below). In the United States pension insurance contributions were temporarily reduced
in 2011 and 2012, but brought back to their previous level in 2013. In real terms,
employee social security contributions remained about the same in Estonia, France,
Iceland, Portugal and Spain.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 123

Personal income tax rates were raised in France, Greece, Iceland, Portugal, Spain and
the United Kingdom. Tax deductions were reduced via nominal cuts in Ireland and
Portugal and by fiscal drag in Estonia, Iceland, Ireland, Portugal and Spain. New income
taxes were introduced in Ireland (“Universal Social Charge”) and Greece (“Special
solidarity contribution”). On the other hand, income tax expenditures were increased in
the United Kingdom (exemption limit), the United States (tax credits) and Germany
(tax allowances).
Public pension levels fell due to indexation below inflation in Germany, Iceland,
Portugal and Spain. Pension levels were nominally reduced in Greece (the number of
payments per year was reduced from 14 to 12) and Portugal (due to an extraordinary
pension contribution). On the other hand, pension levels increased above inflation in
Estonia, France, Ireland, the United Kingdom and the United States.
Value-added tax (VAT) rates were increased in virtually all countries. In absolute
terms, Spain and Greece saw higher increases. In France, the standard rate did not
change, but the reduced rate effectively rose as an additional rate was added.6 In
Germany, rates did not change in the period, but increased considerably in 2007. In the
United States, there is no federal value-added tax on goods or services. Instead, sales tax
is common in most states.
Table 3.5. Simulated overall effect of tax-benefit measures, 2008-13, by policy type

Total
+

UB

France
Germany
Greece
Iceland
Ireland
Portugal
Spain

+
+
+
---

+
-

United Kingdom
United States
OECD10

+
-

Estonia

+
-

Working-age cash benefits
FB
IWB
HB
+
-

-

+

+

SA
+

+
-

+

+
+

+
+
-

+
+

Personal income taxes
Total
SC
PIT
+
----+
+
-

+
-+
+
-

Other policies*
Pensions
VAT
++
-

+
-----

+
--+
-

+
+
-

+
+
-

---

Note:
+ sign indicates a measure that has a positive effect on household income (i.e. a tax cut or benefit rise).
– sign indicates a measure that has a negative effect on household income (i.e. a tax rise or benefit cut).
* Results for other policies (pensions and VAT) are not simulated, and simply indicate statutory changes.
FB = family benefits; HB = housing benefits; IWB = in-work benefits; PIT = personal income taxes; SA = social assistance
benefits; SC = social security contributions; UB = unemployment benefits; VAT = value-added tax.
Underlying detailed analyses are available at www.oecd.org/social/inequality-and-poverty.htm.
Source: OECD tax-benefit model (www.oecd.org/social/benefits-and-wages.htm); OECD Tax Database
(www.oecd.org/tax/tax-policy/tax-database.htm); Adiego, M., M. Burgos, M. Paniagua and T. Pérez (2014), “Spain 20102013”, EUROMOD Country Report; Adiego, M., O. Cantó, H. Levy, M. Paniagua and T. Pérez (2012), “Spain 2007-2010”,
EUROMOD Country Report; Bardens, J. and R. Cracknell (2014), “2014 Benefit Uprating”, House of Commons, Social and
General Statistics; DPS (2015), “Rates of Payment”, Department of Social Protection, Ireland; Leventi, C., A. Karakitsios,
M. Matsaganis
and
P. Tsakloglou
(2014),
“Greece
2009-2013”,
EUROMOD
Country
Report,
www.iser.essex.ac.uk/euromod/resources-for-euromod-users/country-reports; MISSOC (2015), “MISSOC Comparative
Tables Database”, retrieved on 02/01/2015 from http://missoc.org; SSA (2015), “Cost-of-Living Adjustment (COLA)
Information for 2015”, US Social Security Administration, www.ssa.gov/news/cola/.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 124 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION

The effect of tax-benefit measures on household income
In line with findings in the sections above (see Figures 3.10 and 3.11), simulation
results presented in Table 3.6 shows that in most countries households tended to gain
from the policy changes implemented in 2008 and 2009 (i.e. fiscal stimulus) and to lose
from those implemented from 2010 to 2012 (i.e. fiscal consolidation). The effects of
policy changes in 2013 were less homogeneous across countries. Taking the changes
introduced in the period 2008-13 as a whole, the countries can be organised in three
groups. In four countries (Estonia, France, Germany and the United States), the
cumulated effect of tax-benefit changes on household income was positive. In two
countries (Spain and the United Kingdom), the impact on household income was
moderately negative. Finally, in four countries (Greece, Iceland, Ireland and Portugal),
the impact was negative.
Table 3.6. Simulated overall effect of tax-benefit measures by year
2008

2010

2011

2012

2013

2007-2013

-

-

-

+

+

+

-

-

-

+

+

+

+

-

+

+

+

+

+

-

--

-

--

--

-

-

-

-

-

+

--

Ireland

+

-

-

--

-

-

--

Portugal

+

+

-

--

-

--

--

Spain

+

+

-

-

-

-

-

United Kingdom

+

+

-

-

-

-

-

United States

+

+

-

-

+

OECD10

+

+

-

-

-

-

-

Estonia

+

France

-

Germany

-

Greece
Iceland

2009

Note:
+ sign indicates a measure that has a positive effect on household income (i.e. a tax cut or benefit rise).
– sign indicates a measure that has a negative effect on household income (i.e. a tax rise or benefit cut).
Underlying detailed analyses are available at www.oecd.org/social/inequality-and-poverty.htm.
Source: OECD tax-benefit model (www.oecd.org/social/benefits-and-wages.htm).

Table 3.7 summarises the impact of the tax and benefit measures by family
composition. Generally, the impact did not vary substantially by family type. However,
there were some differences in intensity and also a few exceptions. Families with children
tended to lose more than those without children, partly due to cuts in family benefits.
Lone-parent families experienced greater losses across countries, except in Germany and
the United States where the measures had a positive impact. Two-earner couples with or
without children also tended to lose more than other family types. This was the case
especially in countries that increased income taxation, such as France, Greece, Ireland,
Iceland and Portugal, but also Estonia. On the other hand, two-earner families fared better
than other family types in the United Kingdom as they gained from an increase in the
income tax personal allowance (a standard deduction).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 125

Table 3.7. Simulated overall effect of tax-benefit measures by family composition, 2008-13
No children
Single person

Two children

One-earner

Two-earner

couple

couple

Lone parent

One-earner

Two-earner

couple

couple
-

Estonia

+

+

-

-

+

France

+

+

-

-

+

-

Germany

+

+

+

+

+

++

Greece

--

--

---

--

--

---

Iceland

-

-

--

--

-

--

Ireland

--

-

--

---

--

---

Portugal

--

--

---

--

---

---

Spain

-

-

-

-

-

-

United Kingdom

-

-

+

-

-

+

United States

+

+

+

+

+

+

OECD10

-

-

-

-

-

--

Note:
+ sign indicates a measure that has a positive effect on household income (i.e. a tax cut or benefit rise).
– sign indicates a measure that has a negative effect on household income (i.e. a tax rise or benefit cut).
Underlying detailed analyses are available at www.oecd.org/social/inequality-and-poverty.htm.
Source: OECD tax-benefit model (www.oecd.org/social/benefits-and-wages.htm).

Figure 3.13 shows the average impact of the tax and benefit measures taken between
2008 and 2013 on families with different labour market circumstances and earning levels.
The results are expressed as a proportion of the average wage in each country in 2007,
and thus represent absolute changes, despite being presented as percentages. Furthermore,
the results are computed as an unweighted average from four family types (single person,
one-earner couple, lone parent with two children, and one-earner couple with two
children). Each chart is divided into two parts. The left-hand side presents the results for
households in which the earner has been unemployed and out-of-work for 1 to 60 months.
The right-hand side presents the results for households in which the earner has been inwork and has earnings that range between 1% and 200% of the national average wage.
The impact of the tax-benefit measures varies considerably across households with
different labour market circumstances. In most countries measures affecting personal
direct taxes and cash benefits had a progressive effect on households in work. Generally,
low-earning families fared better than higher-earning families. The main exception is
Germany, where higher-earning families gained from the tax cuts, and to some extent
Portugal and the United Kingdom, where benefits for low-earning families fell
considerably. Results are less clear-cut for working-age households in unemployment, as
in several countries their losses are considerable.
Similar results were found using micro-simulation techniques. Looking at
12 EU countries, including all eight analysed here, De Agostini et al. (2014) found that
the distributional effects of changes in personal direct taxes, public pensions and cash
benefits were broadly progressive, except in Germany and Estonia. The distributional
effects became more regressive once VAT changes were introduced.7 For the United
States, Larrimore, Burkhauser and Armour (2013) found that tax and benefit reforms
played an important role to cushion the impact of the crisis, particularly among the
bottom quintile of the distribution.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 126 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Figure 3.13. Simulated overall effect of tax-benefit measures implemented in 2008-13, by labour market
circumstance and earning level
Change in household disposable income as a proportion of national average wage

10

Estonia
,20
,10
,0
-,10
-,20
-,30

20

%
%

- 20

- 20

- 30

- 30
In-work (% of AW)
Germany

Out-of-work
(no. months)
x 0.01

20

20

10

0

0

- 10

- 10

- 20

- 20

- 30

- 30

Iceland

%

Out-of-work
(no. months)
x 0.01

20

In-work (% of AW)
20
%

10

10

0

0

- 10

- 10

- 20

- 20

- 30

- 30
Out-of-work
(no. months)

In-work (% of AW)

In-work (% of AW)
Greece

%

10

Out-of-work
(no. months)

France

0

- 10

%%
%

Housing benefits
Total

In-work (% of AW)

- 10

Out-of-work
(no. months)
x 0.01

In-work benefits
Personal income taxes

10
Out-of-work
(no.
months)

0

x 0.01

Family benefits
Social security contributions
x 0.01

20
%
x 0.01

x 0.01

Unemployment benefits
Social assistance benefits

In-work (% of AW)
Ireland

Out-of-work
(no. months)

In-work (% of AW)

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 127

Figure 3.13. Simulated overall effect of tax-benefit measures implemented in 2008-13, by labour market
circumstances and earning levels (cont.)
Change in household disposable income as a proportion of national average wage

10

Portugal
,20
,10
,0
-,10
-,20
-,30

In-work benefits
Personal income taxes
20

In-work (% of AW)
0

- 10

- 20

- 20

- 30

- 30

In-work (% of AW)

United Kingdom

Out-of-work
(no. months)
x 0.01

20

%

10

0

0

- 10

- 10

- 20

- 20

- 30

- 30

In-work (% of AW)

In-work (% of AW)

20
%

10

Out-of-work
(no. months)

Spain

%

- 10

Out-of-work
(no. months)

Housing benefits
Total

10
Out-of-work
(no.
months)

0

x 0.01

Family benefits
Social security contributions
x 0.01

20
%
x 0.01

x 0.01

Unemployment benefits
Social assistance benefits

United States

Out-of-work
(no. months)

In-work (% of AW)

Note:
Differences in the amount taxes and benefits between 2007 and 2013, due to the policy changes implemented between 2008 and
2013. Amounts expressed as a percentage of the average wage in each country. See main text for further details.
Underlying detailed analyses are available at www.oecd.org/social/inequality-and-poverty.htm.
Source: OECD tax-benefit model (www.oecd.org/social/benefits-and-wages.htm).
12 http://dx.doi.org/10.1787/888933208002

Households in unemployment for less than one year gained in Germany, but lost in
Ireland, Portugal, Greece and Spain as well as to some extent in the United States and
Iceland. The policy changes for the longer-term unemployed were rather irregular. This
group lost in Portugal, due to a cut in the duration of unemployment benefits. When these
expire, households fall into social assistance benefits, and their levels were also cut. In
Ireland and Spain, the long-term unemployed lost due to real-term cuts in unemployment
assistance, and in Germany due to the phase-out of a transition measure. The long-term
unemployed benefited from the extension of benefit duration in the United States and
from a real-term increase in the level of housing benefit in Estonia.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 128 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Low-earning families (earning less than half the average wage) faced considerable
losses in Portugal due to cuts in social assistance. Low-earning families also faced some
losses in the United Kingdom, due to housing benefit reform, and in Ireland, due to cuts
in family benefits. On the other hand, lower-earning families gained in France, the United
States and Estonia.
Families earning between half and one time the average wage gained from higher
family benefits in Germany and higher tax deductions in the United Kingdom and the
United States. On the other hand, higher taxes reduced the incomes of these families in
Greece, Ireland and Iceland.
In Germany, family earnings between once and twice the average wage gained from
the higher tax deductions implemented in 2010 (see OECD, 2010; OECD, 2011d). In all
other countries, families either lost or remained more or less the same. The largest losses
happened in Greece, Iceland, Ireland and Portugal. More moderate losses also took place
in Spain and the United Kingdom.

3.6. Conclusion
Household income inequality and poverty increased during the crisis. The impact
during the different phases of the downturn and recovery was largely affected by the taxbenefit systems and measures introduced. The largest impact of the crisis was not so
much on the level of income inequality as on the nature of its drivers. In the decade prior
to the crisis, income inequality was driven upwards by weakening tax-benefit
redistribution, while inequality before taxes and benefits often stalled (OECD, 2011a). In
the first years of the crisis, income inequality before taxes and benefits increased strongly
but taxes and benefits cushioned the rise. In more recent years, while income inequality
before taxes and benefits has continued to rise, the tax-benefit cushion weakened, thus
accelerating the overall upwards trend in disposable income inequality.
During the crisis, labour income inequality rose due to declines in employment (i.e.
rising unemployment and inactivity), rather than to higher inequality in wages (i.e. rising
pay gap), as had been the case in the pre-crisis period (OECD, 2011a). The persistent
increase in unemployment in many OECD countries has exerted considerable downward
pressure on real wage growth, however the wage adjustment costs have been shared quite
evenly across workforce groups (OECD 2014c). Lower employment pushed labour
income inequality up in most OECD countries, particularly in Spain, Estonia, Ireland,
Greece, the United Kingdom and Portugal. Australia, Austria, France, Denmark and Italy
were the only countries where the pay gap was the main driver of higher labour income
inequality. In most of the few countries where labour income inequality fell, the lower
pay gap was also the main driver.
Income inequality increased considerably more when looking at indicators that are
more sensitive to changes at the bottom part of the income distribution, as lower-income
households either lost more during the crisis or benefited less from the recovery than
other income groups. In most countries, the bottom 10% fared worse than the top 10%
and the median. These results suggest the continuation of a long-term trend, in which the
growth of bottom incomes has been persistently outperformed by middle and, especially
top incomes.
Independently of the indicator used, income poverty increased in most countries
during the crisis. Poverty increased sharply when measured using an anchored poverty
threshold, i.e. fixing the poverty line at a pre-crisis level in real terms. Poverty increased
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 129

less when measured using relative thresholds based on current income levels, as these
have fallen with the crisis.
Trends in household disposable income and poverty were uneven across age groups.
Income rose and poverty fell among the elderly, while all other age groups saw their
incomes fall and poverty rise. Income losses were largest among young people, followed
by adults and children. In the first years of the crisis, incomes of the elderly were less
exposed than of other age groups. A priori, this can be expected given that the main effect
of the crisis was through the labour market and hence on the working-age population,
while the elderly depend more on social transfers (particularly pensions) than on labour
income. In more recent years, pensioners too were affected by fiscal consolidation
measures to reduce public pension expenditure. Protecting the income of the elderly is an
explicit goal of most social protection systems. Not only are the elderly less likely to be
able to adapt to changing circumstances in the labour market and to find new sources of
income but their household income is also on average lower than that of the rest of the
population.
Taxes and benefits cushioned the fall in household income and the rise in inequality.
Significant increases in public expenditure on social benefits and lower personal income
tax burdens prevented household disposable income levels from falling and inequality
from rising as much as levels of market income. This cushioning effect resulted from the
stabilising properties of the tax-benefit system, in some cases reinforced by fiscal
stimulus reforms. However, as the crisis continued, the effect diminished as entitlement to
social benefits expired and most governments implemented fiscal consolidation
programmes to tackle rising public deficits and debt-to-GDP ratios.
The size, composition and effect of the fiscal stimulus and fiscal consolidation
programmes varied substantially across time and countries, reflecting both different fiscal
positions and policy strategies. In many countries, households tended to gain from the
policy changes implemented in 2008 and 2009 and to lose from those implemented
between 2010 and 2012. The effects of the policy changes in 2013 were less homogenous
across countries.
While overall fiscal stimulus and consolidation programmes concentrated more on
public expenditure, tax-benefit measures targeted at working-age households were more
balanced, with hikes in personal income tax playing an important role in several
countries. The direction and intensity of the effect of these measures on household
income depended on household circumstances. Families with children (especially lone
parents) tended to lose more, partly due to cuts in family benefits. Broadly, measures had
a progressive effect among households in-work, producing either gains or smaller losses
on low-earning families and greater losses on higher-earning families. Results are less
clear-cut for working-age households in unemployment, as in several countries their
losses are considerable.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 130 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION

Notes
1.

Using data from the World Top Incomes Database, Förster et al. (2014) find that over
the years 2008 and 2009 the income share of the top 1% decreased significantly in
countries for which data are available. The top earners indeed experienced a higher
income shock than the rest of the population, thereby reinforcing the idea that top
incomes are more cyclically sensitive than others.

2.

According to Decoster et al. (2014), tax-benefit reforms implemented in Belgium
between 2007 and 2012 boosted average household disposable income in a
progressive way, ranging from an extra 6% in the bottom 10% to 0.1% in the top
10%.

3.

According to Domitrz et al. (2013), the reforms introduced between 2006 and 2011
increased the value of the Gini coefficient by 0.46 percentage points. Morawski and
Myck (2010) estimate that due to the 2007 child tax credit reform, households with
children in the bottom decile gained on average about PLN 7.60 per month, while
those in the top 40% gained over PLN 100 per month on average
(PLN 100 = approximately USD 26). Myck et al. (2013) suggest that tax-benefit
reforms weakened labour market incentives for families who have children and are
eligible for safety net benefits, and that work incentives improved over the years
despite the negative effect of these reforms only due to significant real wage growth.

4.

In 2012, more than two-thirds of OECD countries planned to reduce spending on
working-age social benefits and about 40% of countries planned to do the same on
pensions (OECD, 2014b).

5.

The
underlying
detailed
country
analyses
www.oecd.org/social/inequality-and-poverty.htm.

6.

In 2012, an additional rate (“taux intermédiaire”) of 7%, higher than the reduced rate
(5.5%), was introduced and applied to most goods and services previously taxed at
the reduced rate. In 2014, the taux intermédiaire rose to 10% and the standard rate
from 19.6% to 20%.

7.

OECD/KIPF (2014) showed that VAT systems are regressive when measured as a
percentage of income, but are generally either proportional or slightly progressive
when measured as a percentage of expenditure.

are

available

at

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 131

References
Adiego, M., M. Burgos, M. Paniagua and T. Pérez (2014), “Spain 2010-2013”,
EUROMOD Country Report, www.iser.essex.ac.uk/euromod/resources-for-euromodusers/country-reports.
Adiego, M., O. Cantó, H. Levy, M. Paniagua and T. Pérez (2012), “Spain 2007-2010”,
EUROMOD Country Report, www.iser.essex.ac.uk/euromod/resources-for-euromodusers/country-reports
Atkinson, A.B. and A. Brandolini (2006), “From Earnings Dispersion to Income
Inequality,” in F. Farina and E. Savaglio (eds.), Inequality and Economic Integration,
Routledge, London.
Atkinson, A.B., Cantillon, B., Marlier, E. and Nolan, B. (2002), Social Indicators, The
EU and Social Inclusion, Oxford University Press.
Avram, S., F. Figari, C. Leventi, H. Levy, J. Navicke, M. Matsaganis, E. Militaru, A.
Paulus, O. Rastrigina and H. Sutherland (2013), “The Distributional Effects of Fiscal
Consolidation in Nine Countries”, Working Paper EM 2/13, Microsimulation Unit,
University of Essex, Colchester.
Bardens, J. and R. Cracknell (2014), “2014 Benefit Uprating”, House of Commons,
Social and General Statistics, www.parliament.uk/briefing-papers/sn06774.pdf
Callan T., C. Leventi, H. Levy, M. Matsaganis, A. Paulus and H. Sutherland (2011), “The
Distributional Effects of Austerity Measures: A Comparison of Six EU Countries”,
Research Note 2/2011, European Observatory on the Social Situation and
Demography, European Commission.
Cournède, B., A. Goujard and Á. Pina (2013), “How to Achieve Growth- and Equityfriendly Fiscal Consolidation? A Proposed Methodology for Instrument Choice with
an Illustrative Application to OECD Countries”, OECD Economics Department
Working
Papers,
No.
1088,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/5k407lwvzkkh-en.
De Agostini, P., A. Paulus, H. Sutherland and I. Tasseva (2014), “The Effect of Taxbenefit Changes on the Income Distribution in EU Countries since the Beginning of
the Economic Crisis”, EUROMOD Working Papers EM9/14, EUROMOD at the
Institute for Social and Economic Research.
Decoster, A., S. Perelman, D. Vandelannoote, T. Vanheukelom and G. Verbist (2014),
“Efficiency and Equity Aspects of 20 Years of Tax and Benefit Reforms in Belgium:
Microsimulation Results”, Flemosi Working Paper.
Domitrz, A., Morawski, L., Myck, M. and Semeniuk, A. (2013), “Dystrybutywny wpływ
reform podatkowo-świadczeniowych z lat 2006-2011”, Bank i Kredyt, Vol. 44, No. 3,
pp. 261-286.
DPS – Department of Social Protection (2015), “Rates of Payment”, Department of
Social Protection, Dublin, Ireland.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 132 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Duncan, G., K. M. Ziol-Guest and A. Kalil (2010), “Early Childhood Poverty and Adult
Attainment, Behaviour and Health”, Child Development, Vol. 81, No. 1, pp. 306-325.
European Commission (2014), Employment and Social Developments in Europe 2013.
Farinha, C. and V. Junqueira (2012), “Portugal 2007-2010”, EUROMOD Country Report,
www.iser.essex.ac.uk/euromod/resources-for-euromod-users/country-reports.
Farinha, C., V. Junqueira and R. Figueuras (2013), “Portugal 2009-2013”, EUROMOD
Country Report, www.iser.essex.ac.uk/euromod/resources-for-euromod-users/countryreports.
Fitoussi, J.P. and F. Saraceno (2010), “Inequality and Macroeconomic Performance”,
OFCE /POLHIA Working Papers, No. 2010-13, Paris.
Förster, M. and M. Mira d’Ercole (2012), “The OECD Approach to Measuring Income
Distribution and Poverty: Strengths, Limits and Statistical Issues”,
http://umdcipe.org/conferences/oecdumd/conf_papers/Papers/The%20OECD%20App
roach%20to%20Measuring%20Income%20Distribution%20and%20Poverty.pdf.
Förster, M., A. Llena-Nozal and V. Nafilyan (2014), “Trends in Top Incomes and their
Taxation in OECD Countries”, OECD Social, Employment and Migration Working
Papers, No. 159, OECD Publishing, Paris.
Immervoll, H. and M. Pearson (2009), “A Good Time for Making Work Pay? Taking
Stock of In-Work Benefits and Related Measures across the OECD”, OECD Social,
Employment and Migration Working Papers, No. 81, OECD, Paris,
http://dx.doi.org/10.1787/225442803245.
Immervoll, H. and L. Richardson (2011), “Redistribution Policy and Inequality Reduction
in OECD Countries: What Has Changed in Two Decades?”, OECD Social,
Employment and Migration Working Papers, No. 122, OECD Publishing, Paris,
http://dx.doi.org/10.1787/5kg5dlkhjq0x-en.
Immervoll, H. and L. Richardson (2013), “Redistribution Policy in Europe and the United
States: Is the Great Recession a ‘Game Changer’ for Working-age Families?”, OECD
Social, Employment and Migration Working Papers, No. 150, OECD Publishing,
Paris, http://dx.doi.org/10.1787/5k44xwtc0txp-en.
Immervoll, H., M. Mira-D’Ercole and P. Marianna (2004), “Benefit Coverage Rates and
Household Typologies. Scope and Limitations of Tax-Benefit Indicators”, OECD
Social
Employment
and
Migration
Working
Paper,
No. 20,
www.oecd.org/els/workingpapers.
Lambert, P. (2001), The Distribution and Redistribution of Income, Third Edition,
Manchester University Press, Manchester.
Larrimore, J., R. Burkhauser and P. Armour (2013), “Accounting for Income Changes
over the Great Recession (2007-2010) Relative to Previous Recessions, the
Importance of Taxes and Transfers”, NBER Working Paper Series, No. 19699,
www.nber.org/papers/w19699.
Leventi, C., A. Karakitsios, M. Matsaganis and P. Tsakloglou (2014), “Greece 20092013”, EUROMOD Country Report, www.iser.essex.ac.uk/euromod/resources-foreuromod-users/country-reports.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION – 133

Levy, H. (forthcoming), “Characteristics of Benefit Recipients: Scope and Limitations of
Tax-Benefit Indicators”, OECD Social, Employment and Migration Working Papers,
OECD Publishing, Paris.
MISSOC – Mutual Information System on Social Protection (2015), “MISSOC
Comparative Tables Database”, retrieved on 02/01/2015 from www.missoc.org/.
Morawski, L. and M. Myck (2010), “Distributional Effects of the Polish Child Tax Credit
and its Potential Reforms”, CenEA Microsimulation Report MR01/10.
Myck, M., A. Domitrz, L. Morawski and A. Semeniuk (2013), “Financial Incentives to
Work in the Context of a Complex Reform Package and Growing Wages: The Polish
Experience 2005-2011”, CenEA Working Paper Series WP01/13.
OECD (2014a), “Rising Inequality: Youth and Poor Fall Further Behind”, Insights from
the OECD Income Distribution Database, Directorate for Employment, Labour and
Social
Affairs
and
Statistics
Directorate,
OECD,
Paris,
www.oecd.org/social/OECD2014-Income-Inequality-Update.pdf.
OECD (2014b), Society at a Glance 2014: OECD Social Indicators, OECD Publishing,
Paris, http://dx.doi.org/10.1787/soc_glance-2014-en.
OECD (2014c), “Sharing the Pain Equally? Wage Adjustments during the Crisis and
Recovery”, OECD Employment Outlook 2014, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2014-5-en.
OECD (2012), Restoring Public Finances: 2012 Update, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264179455-en.
OECD (2011a), Divided We Stand: Why Inequality Keeps Rising, OECD Publishing,
Paris, http://dx.doi.org/10.1787/9789264119536-en.
OECD (2011b), Doing Better for Families,
http://dx.doi.org/10.1787/9789264098732-en.

OECD

Publishing,

Paris,

OECD (2011c), “Restoring Public Finances: Fiscal Consolidation in OECD Countries”,
Special issue of the OECD Journal on Budgeting, Vol. 2011/2,
http://dx.doi.org/10.1787/budget-v11-2-en.
OECD
(2011d),
Taxing
Wages
2010,
http://dx.doi.org/10.1787/tax_wages-2010-en.
OECD
(2010),
Taxing
Wages
2008-2009,
http://dx.doi.org/10.1787/tax_wages-2009-en.

OECD
OECD

Publishing,

Paris,

Publishing,

Paris,

OECD (2009), “The Effectiveness and Scope of Fiscal Stimulus”, OECD Economic
Outlook
Interim
Report,
March
2009,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/eco_outlook-v2008-sup2-16-en.
OECD (2008), Growing Unequal? Income Distribution and Poverty in OECD Countries,
OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
OECD (2007), Benefits and Wages 2007: OECD Indicators, OECD Publishing, Paris,
http://dx.doi.org/10.1787/ben_wages-2007-en.
OECD/Korea Institute of Public Finance (2014), “The Distributional Effects of
Consumption Taxes in OECD Countries”, OECD Tax Policy Studies, No. 22, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264224520-en.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 134 – 3. INCOME INEQUALITIES DURING THE CRISIS AND FISCAL CONSOLIDATION
Pearson, M. and S. Scarpetta (2000), “An Overview: What Do We Know About Policies
to Make Work Pay?”, OECD Economic Studies, No. 31, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_studies-v2000-2-en.
Rajan, R. G. (2010), “Fault Lines: How Hidden Fractures Still Threaten the World
Economy”, Princeton University Press, Princeton, United States.
Rawdanowicz, L., E. Wurzel and A. Christensen (2013), “The Equity Implications of
Fiscal Consolidation”, OECD Economics Department Working Papers, No. 1013,
OECD Publishing, Paris, http//dx.doi.org/10.1787/5k4dlvx2wjq0-en.
SSA – US Social Security Administration (2015), “Cost-of-Living Adjustment (COLA)
Information for 2015”, US Social Security Administration, www.ssa.gov/news/cola/.
Stiglitz, J.E. (2012), The Price of Inequality: How Today’s Divided Society Endangers
Our Future, W.W. Norton & Company, New York.

Database references
OECD Income Distribution Database (IDD), www.oecd.org/social/income-distributiondatabase.htm.
OECD Tax Database, www.oecd.org/tax/tax-policy/tax-database.htm.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 135

Chapter 4
Non-standard work, job polarisation and inequality

This chapter provides evidence for the implication of trends in non-standard work for
individual and household earnings and income inequality. It first presents the sociodemographic characteristics of non-standard workers before discussing the contribution
of non-standard work to overall changes in employment. It shows that, in a majority of
OECD countries, standard jobs have disappeared in the middle of the distribution in
terms of wages and skill, while non-standard jobs have contributed to an increase in jobs
at both ends of the distribution. Non-standard jobs tend to pay lower wages than
standard jobs, especially at the bottom of the earnings distribution, thereby raising
earnings inequality. The chapter then looks at the impact of non-standard work on
household incomes and shows that non-standard workers living alone or with other nonstandard workers suffer from higher chances of low income and poverty. Finally, the
chapter examines the work incentives and adequacy effects of tax and benefit rules. It
finds that some non-standard workers, such as the self-employed, usually face different
statutory rules and shows that taxes and benefits reduce poverty gaps for non-standard
workers but create work disincentives for moving from inactivity to work.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of
such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in
the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 136 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

4.1.

Introduction and key findings
Changes in earnings – which constitute three-quarters of household income – and in
labour market conditions have been identified as the most important direct driver of rising
income inequality. This concerns, in particular, changes in the distribution of gross wages
and salaries, which have become more dispersed in most OECD countries in the past
25 years. But this is also linked to changes in employment patterns, working conditions
and labour market structures. For instance, growing levels of non-standard work, such as
part-time work, casual work and work on temporary contracts, may help to explain the
puzzle of increasing inequality despite aggregate employment growth prior to the global
economic crisis.
The effects of the rising share of employment in non-standard work (NSW)
arrangements have gained centre stage in policy debates in recent decades. Since the
1980s, labour markets in OECD countries have been subject to major structural changes.
The employment protection legislation (EPL) became less strict in countries where
protection had been relatively strong to start with, while countries where the strictness of
the EPL was below average in 1985 tended to stick with a similar policy in the late 2000s
(OECD, 2011). Alongside these institutional changes, demographic and societal
developments – ageing and higher female labour market participation – have also
profoundly modified the labour force. Finally, structural changes in employment due to
growth in services and knowledge jobs, a greater use of ICTs and just-in-time delivery
have all had implications for the demand and supply drivers of atypical forms of work. As
NSW is often portrayed as being associated with lower earnings and with job insecurity,
this has drawn attention to its potentially adverse impact on the distribution of individual
earnings as well as of household income more generally.
Evidence from OECD (2011) has shown the impact of non-standard work on the level
of overall earnings inequality: adding the earnings of part-time workers to the distribution
of full-time employees increased earnings inequality by almost 20%, and adding selfemployed workers increased inequality by a further 5%. In addition, policy reforms such
as weaker employment protection for temporary contracts have tended to increase
employment opportunities but were associated with wider wage inequality.
There is however a lack of empirical evidence on the detailed channels through which
non-standard work may affect the distribution of individual and household income.
Non-standard employment might be associated with poorer labour conditions (wages,
working time, job security, leave entitlements, etc.), particularly in the case of dual or
segmented labour markets, if firms use such arrangements for cost or flexibility reasons
or as a probationary device. On the other hand, part-time, temporary and self-employment
arrangements may be attractive to certain workers, and workers might choose this type of
employment to achieve a better work-family life balance, higher life satisfaction or, in the
case of self-employment, a greater sense of control. The degree of mobility between both
segments is also likely to influence whether there are persistent wage differentials
between both sectors.
The chapter is organised as follows. Section 4.2 defines different forms of
non-standard work and the demographic composition of these workers. Section 4.3
analyses the extent to which employment growth stems from non-standard work and how
NSW contributes to job polarisation. Section 4.4 looks at the question of whether nonstandard jobs pay less and whether such jobs improve employment prospects. It also
discusses the implications for the distribution of earnings. The contribution of NSW to
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 137

household income inequality and poverty is discussed in Section 4.5. Finally, Section 4.6
presents the impact of tax-benefit policies on income adequacy and on work incentives
for non-standard workers.
The key findings from this chapter are:


Non-standard work (temporary, part-time and self-employment taken together)
represents one-third of total employment in the OECD, ranging from a low of
under 20% in the eastern European countries (except Poland) to 46% or more in
the Netherlands and Switzerland. Women (especially part-time), youth
(especially temporary jobs) and workers with lower level of education are overrepresented in NSW, as are workers in small firms.



Close to half of employment growth since the 1990s and up to the global
economic crisis has been in the form of non-standard work; the share reaches
almost 60% of if the crisis years are included.



Non-standard work contributes to job polarisation, i.e. to jobs disappearing in the
middle of the distribution relative to those at the bottom and at the top: nearly all
employment losses in middle-skill occupations were in standard work contracts,
while job gains in high- and low-skill jobs were mainly in NSW.



Non-standard work is not always a stepping stone to stable employment.
Temporary contracts increase the chances of acquiring a standard job compared
with remaining unemployed, but a part-time job or self-employment does not
increase the chances of a transition to a standard job.



Non-standard workers are worse off in terms of many aspects of job quality.
They tend to receive less training and, in addition, those on temporary contracts
have more job strain and have less job security than workers in standard jobs.
Earnings levels are also lower in terms of annual and hourly wages but, for parttimers, once other demographic and job characteristics are taken into account, the
differences in hourly wages tend to disappear. On the other hand, compared with
permanent workers, temporary workers face substantial wage penalties, earnings
instability and slower wage growth.



Non-standard work tends to lower wages at the bottom of the earnings
distribution, while the effect is often neutral at the top, thereby contributing to
increased individual earnings inequality.



Adding earnings from non-standard work to households where standard work is
the norm increases household earnings inequality by three Gini points on average
and help explain about 20% of household income inequality.



Slightly more than half of non-standard workers are the main breadwinners in
their household, and the great majority of them (80% or more) live in a
household with two persons or more, including children.



While not all low-wage non-standard workers live in low-income households,
households with non-standard work arrangements are overrepresented at the lower
end of the household income distribution. But the household constellation matters:
low-income and poverty risks are five and ten times higher respectively if NSW is
the main source of earnings rather than if NSW live with a standard worker.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 138 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

4.2.



About 60% of working poor households are households where the main source of
earnings is NSW.



Non-standard workers face different statutory and effective entitlements to taxes
and benefits in comparison to workers in standard jobs. For the self-employed,
this is due to structurally different policy rules, while for part-timers it is the
particular circumstances of these jobs that lead to different outcomes in terms of
adequacy and incentives. In most countries, taxes and benefits significantly
reduce in-work poverty gaps for NSW, though they are more effective for parttime than for self-employed workers.

A snapshot of non-standard work
There is no universally accepted definition of non-standard work arrangements. In its
broadest sense, NSW may be defined as all employment relationships that do not conform
to the “norm” of full-time, regular, open-ended employment with a single employer (as
opposed to multiple employers) over a long time span. Such a broad definition of nonstandard employment includes three partly overlapping types: a) self-employment (ownaccount workers1); b) temporary or fixed-term contracts; and c) part-time work.2 It is
clear that such a definition comprises very different groups of workers: for some (e.g.
involuntary part-timers), this employment may have job characteristics associated with
precariousness (low pay, instability); for others (e.g. voluntary part-timers with long
tenure), such a job may actually be a desired outcome. Furthermore, transforming this
definition into comparable cross-country statistics is not without problems, and the
process is constrained by data availability (Box 4.1).
Box 4.1. Defining non-standard forms of employment

Figures on non-standard employment are not easily comparable across countries because of national
differences in definition and measurement. The difficulties in defining non-standard work on a comparable basis
are accentuated if attempts are made to link non-standard forms of employment with wages and household
earnings, as few data sources contain information on both employment and wages over time. Labour force
surveys or household surveys typically ask respondents first, to classify themselves as employees or selfemployed according to their status in their main job, and then ask employees to report on their type of contract
and their working hours. Self-reporting errors may be present in such information, and figures should be used to
indicate broad levels and trends across countries.
In its broadest sense, NSW arrangements are defined by what they are not: full-time dependent employment
with a contract of indefinite duration, or what is generally considered the “standard” work arrangement. This
definition generally implies that self-employed own-account workers and all part-time workers fall under
“non-standard workers”. While problematic – as this lumps together precarious and non-precarious forms of
work – this convention is followed by a large part of academic international and national research (e.g. Houseman
and Osawa, 2003; Wenger, 2003; Görg et al., 1998; Kalleberg et al., 1997; Kalleberg, 2000; Leschke, 2011), as
well as by international organisations (e.g. International Labour Organisation, World Bank, Eurofound).
As noted above, this chapter breaks down non-standard employment into three separate categories: 1) selfemployed (own-account), 2) temporary full-time employees and 3) part-time employees (including permanent
and temporary contracts). Unpaid family workers are excluded from the analysis. Where possible, a distinction
is made to break down the category of part-time employees into voluntary and non-voluntary part-timers, as
well as part-timers on temporary and permanent contracts.
The distinction between different forms of employment has become increasingly blurred. There is a
growing grey area, for instance between self-employment and wage employment (OECD, 2000). The growth in
the numbers of self-employed contractors working for just one company or franchisees constitute groups on the
borders of dependent and self-employment.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 139

Box 4.1. Defining non-standard forms of employment (cont.)
Temporary jobs for the purpose of this analysis are defined as dependent employment of limited duration,
including temporary work agency, casual, seasonal or on-call work. Definitions across countries outside the
European Union are not harmonised and are based on different approaches. For Korea, workers in temporary
jobs include fixed-term jobs or jobs of a limited duration, which is close to so-called contingent workers, as well
as other atypical workers, i.e. temporary agency workers, individual contract workers, at-home workers, on-call
workers and others. In the case of Australia, a broad definition of temporary work includes jobs of fixed-term
duration, those employed through a labour hire or a temporary work agency as well as casual workers. Casual
workers may lack entitlements to key fringe benefits such as paid vacation or sick leave or may not be protected
by legislation against unfair dismissal, but might otherwise have continuous and stable employment, and are
therefore one form of atypical or NSW. In this respect, this definition follows the work undertaken by the
Australia Productivity Commission (2006) in classifying casual work as one form (and the most sizeable one)
of non-standard work.
Part-time employees are defined based on their weekly working hours, namely working less than 30 hours
per week. This may differ from national definitions which use different hours thresholds. Part-time work is also
further disaggregated into part-time temporary and part-time permanent jobs when the data is available.

Employment in NSW arrangements in the OECD today is sizeable, comprising on
average one-third of total employment (Figure 4.1). Permanent full-time employment
remains nonetheless the norm in a majority of OECD countries, although there is
substantial diversity across countries. In the Netherlands, more than one job in two is
non-standard (though more than half of these are permanent part-time jobs), while in
some eastern European countries the share is less than one in four jobs.

Different forms of non-standard work and their prevalence across the OECD
The three main forms of non-standard work, i.e. self-employment, temporary
employment and part-time work, account for fairly similar shares on average in the
OECD, but they differ greatly by country (Figure 4.1, Panel A). For instance, selfemployment is the most prevalent form of non-standard work in Greece, Turkey and the
Czech Republic. On the other hand, part-time employment represents close to or over
60% of total non-standard employment in the Netherlands, the Nordic countries (except
Finland), Belgium, Luxembourg and Switzerland, while it is only 12% in Korea and
Poland. In Australia, where a broad definition of temporary employment also includes
casual workers (Box 4.1), this type of work accounts for 85% (43%) of part-time
(full-time) workers with a temporary employment contract.
Part-time workers are a very heterogeneous group with very different labour supply
patterns. Some people work part-time because they wish to do so and would not take on
full-time employment, while others do so because there is no full-time employment
available. On average, involuntary part-time accounts for close to 30% of total part-time
employment, with just under half of this associated with a temporary contract (Figure 4.1,
Panel B). There are, however, large variations across countries. In Greece, Spain and
Italy, over 60% of part-timers want to work more hours but could not find full-time jobs.
In contrast, in Austria, Luxembourg, the Netherlands, Belgium and Switzerland, part-time
work is predominantly voluntary and is associated with a permanent contract.
The characteristics and preferences of workers, as well as institutional factors and the
sectoral composition of employment, all play a role in explaining cross-country differences
in the share of non-standard workers. The tax wedge, product market regulations,
employment protection legislation and the size of the public sector have been found to have
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 140 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
an impact on the incidence of different categories of non-standard work. For instance, there
is a well-established negative relationship between the level of GDP and the selfemployment rate (Acs et al., 1994). In addition, self-employment rates tend to be high in
countries where the public sector is small, taxation levels are high, product market
regulation (PMR) is tight3 and the rule of law is weakly enforced (OECD, 1999; Schuetze,
2000; Torrini, 2005). Temporary employment tends to be higher in countries with stricter
employment protection legislation for regular workers (OECD, 2014; Chen et al., 2015,
forthcoming). One explanation put forward is that the employment protection of permanent
jobs has a minor impact on total employment, but leads to a stronger substitution of
temporary jobs for permanent jobs (Cahuc et al., 2012).
Figure 4.1. Share of non-standard employment by type, 2013
Panel A. Non-standard forms of employment as a percentage of total employment
Part-time temporary employees

Full-time temporary contract

Part-time permanent employees

Self-employment

%
70
60
50
40
30

20
10

0

Panel B. Part-time employment by type
Involuntary temporary

Involuntary permanent

Voluntary temporary

Voluntary permanent

%
100
90
80
70
60
50
40
30
20
10
0

Note: Sample restricted to paid and self-employed (own account) workers aged 15-64, excluding employers, student workers and
apprentices. Breakdown of part-time employment by voluntary/involuntary is not possible for non-European countries. Panel A. For
Australia, 42.6% of full-time temporary contract are casual; and 85.2% of part-time temporary employees are casual.
Source: European Union Labour Force Survey (EU-LFS, 2013), Household, Income and Labour Dynamics in Australia
(HILDA, 2012), Japan Labour Force Survey “Basic Tabulation” (2012), Korean Labor & Income Panel Study (KLIPS, 2009)
and Labour Force Survey (LFS, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208028

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 141

Typical profiles of non-standard employment by worker characteristics
Given the predominance of part-time work in total non-standard employment in many
countries, women are disproportionately represented among non-standard workers in
about half of the OECD countries (Figure 4.2). They represent close to 70% of nonstandard workers in Luxembourg, Austria and Switzerland, and more than 60% in most
Nordic countries, the Netherlands, Germany, France, Belgium and Japan. If part-timers
are excluded, women account for roughly 38% of non-standard employment (i.e. fulltime temporary employment and self-employment), with higher shares (close to 50%) in
Finland, Luxembourg and Portugal.
Figure 4.2. Share of women in non-standard employment, 2013
All non-standard work (↘)

Excluding part-time employment

%
80
70
60

50
40
30
20
10

Note: Sample restricted to paid and self-employed (own account) workers aged 15-64, excluding employers, student workers
and apprentices.
Source: European Union Labour Force Survey (EU-LFS, 2013), Household, Income and Labour Dynamics in Australia
(HILDA, 2012), Japan Labour Force Survey “Basic Tabulation” (2012), Korean Labor & Income Panel Study (KLIPS, 2009)
and Labour Force Survey (LFS, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208033

While youth represent a small fraction of total non-standard workers (less than 25%),
they are the group with the highest incidence of NSW (Figure 4.3). The incidence of
NSW is 40% among younger workers (versus 30% for prime-age), and it is over 50% in
Australia, the Netherlands, southern Europe and Poland. This mostly involves younger
workers on temporary contracts. On average in the OECD, 43% of temporary workers are
aged 15 to 29, and this share is over 60% in Austria, Germany and Switzerland,
suggesting that these jobs are often entry ports for young workers (figures not shown).
Education and skill levels also matter. The incidence of non-standard employment is
highest among workers with a lower level of education (around 44%) and lowest among
the higher educated (Figure 4.4). In particular, the incidence of non-standard work among
those with a lower level of education is over 60% for workers in Greece, Korea, Poland,
Switzerland, Germany and the Netherlands.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 142 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.3. Incidence of non-standard employment by age group, 2013
15-29 (↘)

30-49

50-64

%
80

70
60
50
40

30
20
10
0

Note: Sample restricted to paid and self-employed (own account) workers aged 15-64, excluding employers, student workers
and apprentices.
Source: European Union Labour Force Survey (EU-LFS, 2013), Household, Income and Labour Dynamics in Australia (HILDA,
2012), Korean Labor & Income Panel Study (KLIPS, 2009) for Korea and Labour Force Survey (LFS, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208044

Figure 4.4. Incidence of non-standard employment by educational attainment, 2013
Lower educated (↘)

Middle educated

Higher educated

%
70
60
50

40
30
20
10
0

Note: Sample restricted to paid and self-employed (own account) workers aged 15-64, excluding employers, student workers
and apprentices. Lower educated corresponds to basic education to levels 0 to 2 of the International Standard Classification of
Education (ISCED), middle educated to ISCED 3-4 and higher educated to ISCED 5-6.
Source: European Union Labour Force Survey (EU-LFS, 2013), Household, Income and Labour Dynamics in Australia (HILDA,
2012), Korean Labor & Income Panel Study (KLIPS, 2009) for Korea and Labour Force Survey (LFS, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208059

Non-standard workers are also more likely to be found in small firms (Figure 4.5).
Small firms might be more likely to make use of NSW arrangements as a screening
process, or they may prefer more flexible work arrangements to cope with fluctuations in
demand when lay-off costs for permanent workers are high (Bentolila and Saint-Paul,
1994). Indeed, a little less than half of all non-standard workers (excluding the selfemployed) are working in small enterprises. Incidence of non-standard work is highest
among small firms in the Netherlands with over 60%, followed by Switzerland, Germany
and Australia, where close to 50% of workers in small firms are in NSW.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 143

Figure 4.5. Incidence of non-standard employment by firm size, 2013
%

Small (<19) (↘)

Medium (20-49)

Large (50+)

70

60

50

40

30

20

10

0

Note: Sample restricted to paid workers aged 15-64, excluding employers, self-employment, student workers and apprentices.
For Australia and Canada, medium size refers to 20-99 workers and large to 100+ workers.
Source: European Union Labour Force Survey (EU-LFS, 2013), Household, Income and Labour Dynamics in Australia
(HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009) for Korea and Labour Force Survey (LFS, 2013) for
Canada.
12 http://dx.doi.org/10.1787/888933208061

4.3. The role of non-standard employment in overall employment growth and job
polarisation
How did non-standard work contribute to employment growth?
Non-standard employment has increased in OECD countries, but only at a moderate
level. On average, the share of non-standard employment increased by two percentage
points between 1985 and 1995 in 12 countries for which data is available (Figure 4.6),
and by another two points between 1995 and 2013 for a larger sample of countries.
Non-standard employment has grown significantly in the Netherlands, where it increased
by almost 30 percentage points, as well as in Austria, France, Germany and Luxembourg.
In some Nordic countries (Iceland, Norway, Denmark), Greece and Korea, on the other
hand, it declined by 20% or more. In Spain, non-standard employment increased until
1995 but declined in the subsequent decade.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 144 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.6. Trends in shares of non-standard employment, shares in total employment, 1985-2013
%
70

2013 (↘)

1985

1995

60
50
40
30
20
10
0

Note: Sample restricted to paid and self-employed (own account) workers aged 15-64, excluding employers, student workers
and apprentices.
1. Indicates 1998 instead of 1995 for Czech Republic, Estonia, Hungary, Luxembourg, Poland, Slovak Republic, Slovenia, and
Switzerland.
Source: European Union Labour Force Survey (EU-LFS, 1985, 1995, 2013), Household, Income and Labour Dynamics in
Australia (HILDA, 2001, 2012) for Australia, Korean Labor & Income Panel Study (KLIPS, 1999, 2009) for Korea and Labour
Force Survey (LFS, 1997, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208079

While the overall increase was moderate, NSW represents a substantial fraction of the
overall employment created since 1995: on average across countries, more than 40% of the
growth in employment in the period up to the global economic crisis is attributable to nonstandard jobs (Figure 4.7, Panel A). However, the trends vary largely by country. In
Austria, Germany, the Netherlands and the Czech and Slovak Republics all employment
growth over this period can be attributed to non-standard employment. During this same
period, the total number of workers in standard employment in these countries has fallen,
especially in Germany. Nevertheless, standard work was an important source of
employment growth prior to the crisis in many other OECD countries, such as Norway and
Greece but also Iceland and Hungary (where non-standard jobs declined).
The share of part-time permanent employment increased in more than half of the
countries up to the global economic crisis and has contributed in itself to more than onethird of non-standard employment growth and 15% of overall employment growth. The rise
in part-time employment was very prominent in Ireland, Luxembourg and Belgium, with
growth of at least 10%. On the other hand, part-time employment declined sharply in some
of the Nordic countries. Evidence shows that higher female labour force participation
accounts for more than half of the growth in part-time employment in Europe and the
United States during the 1980s and 1990s (OECD, 2010). Further evidence shows that parttime work has not developed at the expense of full-time employment (e.g. Jaumotte, 2003;
Genre et al., 2005). In countries where part-time employment is widespread, inactivity rates
are much lower. The relationship between part-time work, full-time work and inactivity
varies significantly across different demographic groups. Higher part-time shares are
unambiguously associated with a larger labour supply for youth and older workers, while
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 145

prime-age women may be substituting part-time work for full-time work in order to balance
work and family life (OECD, 2010).
Temporary employment has increased in more than three-quarters of the countries and
constitutes the bulk of growth in non-standard employment. The increases during the precrisis period were particularly large in Poland, Portugal and Spain, with growth of over
10%. In Poland, all employment growth during this period was in the form of temporary
employment, while other types of jobs declined. There is also some evidence that
strict EPL for permanent workers together with the weakening of regulations for
temporary employment have contributed to the growth in the share of temporary jobs in
some European countries, such as Spain.
Trends in self-employment are more mixed, with most countries showing stability and
even a small decline, although a few countries experienced larger changes. Hungary and
Poland have the largest relative declines in self-employment rates. This downward trend is
strongly correlated with a reduction in the agricultural sector in OECD countries. At the
same time, there has been growth in the numbers of own-account self-employed working
for just one company. For some of this group, self-employment may be linked to tax
incentives or employment protection legislation, i.e. the phenomenon of so-called “false”
self-employment, especially in the sectors of construction, real estate and business
activities. For instance, as a response to this phenomenon, a tax reform was introduced in
the Czech Republic in 2004 to halt the spread of “false” self-employment, although the ban
was overturned in 2007. While it remains difficult to isolate the effect of policy reforms
from other factors, the incidence of own-account work increased less in the Czech Republic
than in the Slovak Republic during this period (OECD, 2008a). In Italy, the legislation
introduced in 1997 and 2003 to legalise temporary work agencies (while reforming
collaboration agreements) may have led to an increase in self-employed workers who are in
fact working for the same company, but as own-account workers.
The pattern of employment dynamics evolved differently during the recent global
crisis (2007-13). Foremost, instead of the approximate 17% growth in employment
recorded in the pre-crisis period (1995-2007), Figure 4.7 (Panel B) reports, on average, a
2% drop in the total number of persons employed in the latter period. There is, however,
large cross-country variation. In about half of the countries, the loss in employment is
mainly associated with standard jobs. In Greece and Ireland, for instance, the decline in
the number of standard workers is responsible in itself for 15% drop in total employment
over this period.
Full-time temporary workers were also hit hard during the global economic crisis. In
Spain, for example, the losses of such jobs accounted for the biggest part of the drop in
total employment, while in Portugal and Slovenia this represented 30% and 40%,
respectively. The start of a recovery is underway in some countries and in a third of those
(including Germany, Luxembourg, Norway, Belgium, Switzerland and Sweden, positive
employment growth occurred in standard work during this period. It is also noteworthy
that in Germany the number of workers of this type shrank between 1995 and 2007, but
then increased again slightly. Changes in the relative share of standard and non-standard
workers during the economic recession have led to a discussion about whether the crisis
led to a “deskilling” of the workforce, with a destruction of full-time permanent jobs and
a rise of more atypical jobs. However, the opposite might have occurred as in some
countries, a large share of temporary jobs were eliminated, thus the economic crisis could
have led to an up-skilling. So far, the evidence on this topic is inconclusive
(Gallie, 2013).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 146 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.7. Employment growth by type of employment
Percentage
Panel A. 1995-2007
Standard work

Self-employment (own account)

Full-time temporary

Part-time permanent

Part-time temporary

Part-time permanent

Part-time temporary

80
70
60
50

40
30
20
10
0
-10
-20

-30

Panel B. 2007-2013
Standard work

Self-employment (own account)

Full-time temporary

%
20
15

10
5
0
-5
-10
-15
-20

-25

Note: Working-age (15-64) workers, excluding employers as well as students working part-time. Countries are ranked from left
to right in decreasing order for total employment growth. Temporary for Australia includes both casual and fixed-term work.
Source: European Union Labour Force Survey (EU-LFS), Labour Force Survey for Canada, Household, Income and Labour
Dynamics in Australia (HILDA) for Australia.
12 http://dx.doi.org/10.1787/888933208089

Interestingly, except in Canada and the Nordic countries, the number of part-timers
continued to grow, albeit moderately, during the crisis and consolidation/recovery phase.
Because of this growth in part-time employment and the important drop in standard
employment in many countries, 56% of employment growth can be attributed to
non-standard employment in the period from 1995 to 2013 as a whole.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 147

Is job polarisation linked to the growth of non-standard employment?
When employment is divided into three broadly defined tasks – abstract, routine and
non-routine manual4 – a clear and robust pattern of polarisation in job tasks can be seen in
most EU countries between the mid-1990s and 2010 (Figure 4.8). The employment shares
of routine-task jobs, which are traditionally composed of middle-skill standard job
workers, have declined significantly in all OECD countries. At the same time, there has
been an increase in non-routine manual jobs (9%), which were more often non-standard
jobs and a large increase in the employment share of abstract jobs (21%). The growth of
non-routine manual jobs (e.g. drivers or care workers) is more visible in Switzerland,
Ireland and Portugal, where the employment share of such jobs grew by 20% or more. In
Portugal and Switzerland, for instance, this corresponds to more than two-thirds of the
total change in employment over the period. The observed polarisation by task in OECD
countries since the mid-1990s has been driven primarily by within-sector movements in
employment rather than by changes in the use of tasks between sectors. The withincomponent alone can explain about 80% of the increase in the share of abstract jobs and
63% of the reduction of routine jobs (see Table 4.A1.1).
While the literature on job polarisation often emphasises that the adoption of
technology significantly alters the tasks performed by workers at their jobs, the analysis
below adds another dimension to this phenomenon: the role of non-standard employment.
In most countries nearly all the growth in low-skill/non-routine manual jobs (Figure 4.8,
Panel C) was in non-standard employment, while losses in middle-skill/routine jobs were
primarily associated with standard employment (Panel B), and both non-standard and
standard work on average contributed roughly equally to the increase in abstract jobs
(Panel A). The graphs also demonstrate a certain degree of “substitution” whereby some
standard workers performing low-skilled tasks were replaced by workers with the same
skill but on non-standard contracts (e.g. Austria, Germany, Italy, Luxembourg, the
Netherlands and Slovenia). In a few countries like Austria, Germany, Italy, the Netherlands
and Switzerland, even growth in high-skilled occupations (abstract jobs) was entirely driven
by non-standard employment. Since nearly all job losses, regardless of the type of task,
were associated with regular work, while growth in employment took place mainly in the
form of non-standard employment, technological advancement alone cannot be the only
explanation for job polarisation. Labour market institutions and policies have also probably
played a role in the patterns of substitution observed in certain countries.
In reality, there is considerable variation in the level of skills used to perform the three
broad tasks defined above (and thus wages), and as a result, different patterns of job
polarisation could emerge depending on how the skill/value of a job is measured. Another
way to look at employment polarisation is to use the “jobs-based” approach to analyse
employment shifts (see Annex 4.A1 for a detailed description).5 In general, the analysis
based on this approach and presented in Figure 4.9 shows a clear trend towards job
polarisation in more than half of the countries (i.e. 11 out of 19) for which linked job-wage
data is available (Panels A and B).6 In eight countries the pattern of employment shifts is
strongly U-shaped across job deciles, with contractions of employment shares in the middle
of the distribution and expansions at both ends. In Germany, for instance, the employment
share of the least-paid and highest-paid jobs increased by about 2.7 and 2.3 percentage
points, respectively, between 1995 and 2010, whereas the share of employment has fallen in
most other job deciles. Polarisation varies across countries, however. For instance, jobs
vanished mostly in the lower-middle (i.e. 2nd-5th) deciles of the skill distribution in Belgium,
Canada, France and Norway, but in the more central (i.e. 3rd-7th) deciles in Germany,
Finland, the Netherlands and the United Kingdom.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 148 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.8. Contribution of non-standard work/standard work to changes in employment share by task,
1995/98-2010
Panel A. Abstract task
Standard work

Non-standard work

%
80

70
60
50
40

30
20
10
0
-10

-20

Panel B. Routine task
Standard work

Non-standard work

%
20

10
0

-10
-20
-30
-40
-50

Panel C. Non-routine manual task
Standard work

Non-standard work

%
40
30
20
10
0
-10
-20

-30

Note: Standard and non-standard workers, respectively, as defined in the text, with occupations classified as follows: Abstract
(ISCO88: 12-34); Routine (ISCO88: 41-42, 52, 71-74, 81-82 and 93); and Non-routine manual (ISCO88: 51, 83 and 91). The
overall sample is restricted to workers aged 15-64, excluding employers as well as students working part-time.
Source: European Union Labour Force Survey (EU-LFS).
12 http://dx.doi.org/10.1787/888933208091

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 149

Using the “job-based” approach to classify employment confirms that a shift from
standard towards non-standard employment (i.e. de-standardisation of the employment
contract) played a crucial role in the trend in job polarisation between 1995 and 2010.
Among countries exhibiting a polarisation in jobs over this period (Panels A and B),
some degree of polarisation in standard employment alone is observed, with a
significant decline in this type of employment in the middle of the job spectrum. The
pattern of polarisation becomes more visible once non-standard is factored in, as net
expansion in such work has been concentrated mostly in both the lowest-paid and
highest-paid occupations.7
Similarly, non-standard employment also intensifies the patterns towards job
upgrading in Luxembourg, Italy and Sweden where there was an obvious expansion of
atypical jobs in the top of the job distribution (Panel C). The growth in self-employed
professionals may be part of the story of this development. Interestingly, in Poland and
the Czech Republic changes in non-standard work tend to reduce the degree of
upgrading in employment, as expansion of non-standard jobs was concentrated mostly
in the middle to lower-end of the job deciles.
It is also noteworthy that in a few countries, such as Germany, the Netherlands and
Poland, a decline in mid-range standard jobs was accompanied by similar growth in
non-standard employment within the same job deciles, resulting in a moderate overall
loss of jobs in the middle.
These findings suggest that the “routinisation” story cannot be the only explanation
for the hollowing out of the middle, since the vanishing standard jobs in the middle, if
driven by technology, cannot be easily replaced by workers with the same skill but in
non-standard forms of employment.
Other mechanisms, in particular institutional changes such as those favouring more
flexible labour force, are also at work in reshaping the pattern of employment
polarisation in these countries. Some of the countries that experienced polarisation were
characterised by high levels of employment protection legislation for regular jobs (fulltime or part-time permanent) and low levels of protection for temporary jobs (full-time
or part time), while others were not. Supply side factors, such as an increase in
educational attainment or migration,8 could also accelerate/decelerate the pace of job
polarisation.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 150 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.9. Percentage-point change in employment share due to non-standard work/standard work
by job decile, mid-1990s to 2010
Panel A. Strong polarisation
Belgium

Germany

ppt ΔNon-standard work
ppt ΔStandard work

Finland
ppt ΔNon-standard work
ppt ΔStandard work

ppt ΔNon-standard work
ppt ΔStandard work

4

4

4

2

2

2

0

0

0

-2

-2

-2

-4

- -4.7

1

2

3

4

5

6

7

8

9 10

-4

1

2

3

France

4

5

6

7

8

9 10

-4

ppt ΔNon-standard work
ppt ΔStandard work

ppt ΔStandard work

2

2

0

0

0

-2

-2

-2

4

5

6

7

8

9

10

-4

- 4.4

1

2

3

4

Canada

4

2

2

0

0

-2

-2

3

4

5

6

8

9 10

8

9

10

9

10

-4

1

2

3

4

5

6

7

8

9 10

ppt ΔStandard work

4

2

7

7

ppt ΔNon-standard work

ppt ΔStandard work

1

6

6

Norway

ppt ΔNon-standard work

-4

5

5

ppt ΔStandard work

2

3

4

ppt ΔNon-standard work

4

2

3

United Kingdom

4

1

2

Netherlands
ppt ΔNon-standard work

4

-4

1

7

8

9

10

-4

1

2

3

4

5

6

7

8

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 151

Figure 4.9. Percentage-point change in employment share due to non-standard work/standard work
by job decile, mid-1990s to 2010 (cont.)
Panel B. Moderate polarisation
Hungary

Spain

Greece

ppt ΔNon-standard work

ppt ΔNon-standard work

ppt ΔStandard work

ppt ΔStandard work

ppt ΔNon-standard work
ppt ΔStandard work

4

4

4

2

2

2

0

0

0

-2

-2

-2

-4

1

2

3

4

5

6

7

8

9 10

-4

1

2

3

4

5

6

7

8

9 10

-4

1

2

3

4

5

6

7

8

9 10

Panel C. Upgrading
Luxembourg

Italy

Poland

ppt ΔNon-standard work

ppt ΔNon-standard work

ppt ΔNon-standard work

ppt ΔStandard work

ppt ΔStandard work

ppt ΔStandard work

10

8

4

5

4

2

0

0

0

-5

-4

-2

-10

1

2

3

4

5

6

7

8

9 10

-8

1

2

3

4

Czech Republic

5

6

7

8

9 10

-4

-4.8

1

2

3

4

Sweden
ppt ΔNon-standard work

ppt ΔStandard work

ppt ΔStandard work
4

2

2

6

7

8

9 10

Estonia

ppt ΔNon-standard work
4

5

ppt ΔNon-standard work
ppt ΔStandard work
6
4

2
0

0

-2

-2

0

-2
-4

-4

1

2

3

4

5

6

7

8

9

10

-4

1

2

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

3

4

5

6

7

8

9

10

-6

1

2

3

4

5

6

7

8

9

10

 152 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.9. Percentage-point change in employment share due to non-standard work/standard work
by job decile, mid-1990s to 2010 (cont.)
Panel D. Other
Slovak Republic

Portugal
ppt ΔNon-standard work
ppt ΔStandard work

ppt ΔNon-standard work
ppt ΔStandard work
4

4

2

2

0

0

-2

-2

-4

1

2

3

4

5

6

7

8

9 10

-4

1

2

3

4

5

6

7

8

9 10

Note: Working-age (15-64) workers, excluding employers as well as students working part-time.
Source: European Union Labour Force Survey (EU-LFS).
12 http://dx.doi.org/10.1787/888933208102

4.4. Wage gaps between standard and non-standard workers and their impact on
the distribution of earnings
All types of NSW arrangements pay lower hourly wages and lower annual earnings
than do standard jobs (Figure 4.10).9 The median annual earnings of all non-standard
workers are almost half the level of those for standard workers across the OECD. Median
annual earnings for part-timers are less than half those of standard workers and even 70%
lower for part-time temporary workers, reflecting a lower take-home pay due to fewer
working hours. Earnings for temporary workers and for own-account self-employed
workers – who have similar weekly hours as standard workers – are still some 40%
lower. Workers on temporary contracts have particularly low annual earnings in Estonia,
Canada and Luxembourg. The annual median earnings of the self-employed are closer to
the level of standard workers in the Czech and Slovak Republics, Hungary and,
especially, Korea, but they are considerably lower in Canada and Estonia.
For the selection of countries for which data are available, the hourly wages for
temporary and part-time employees are around 70 to 80% of the median hourly wages for
standard workers (Figure 4.10, Panel B), which is smaller than the difference for annual
wages. The gap in hourly wages is nevertheless high for temporary workers in
Luxembourg, Korea and Canada and for part-time workers in Canada and Japan, who
earn less than 60% of the hourly wage of a standard worker. In terms of hourly wages,
part-time temporary jobs pay less than part-time permanent jobs, though the differences
are small or insignificant in most countries, with notable exceptions (Belgium,
Luxembourg, but also France, Greece and Poland). Part-time jobs pay higher hourly
wages than temporary jobs in some countries, while they tend to pay less in the
English-speaking countries and the Slovak Republic.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 153

Figure 4.10. Earnings ratio between standard and non-standard workers (standard workers = 1), 2012
Panel A. Median annual earnings
All non-standard workers (↘)

Self-employed (own account)

FT Temp. contracts

PTPE employees

PT Temp. employees

1.2
1.1

1
0.9
0.8
0.7
0.6
0.5

0.4
0.3
0.2
0.1
0

Panel B. Median hourly wages, employees only
Full-time temporary (↘)

Part-time permanent

Part-time temporary

1.2
1.1
1
0.9

0.8
0.7
0.6
0.5
0.4
0.3

0.2
0.1
0

Note: Sample restricted to paid workers aged 15-64, excluding employers, self-employment, student workers and apprentices.
Temporary contracts for Australia include both casual and fixed-term work. FT: Full-time, PT: Part-time, PTPE: Part-time
permanent employment.
1. For seven EU-SILC (European Union Statistics on Income and Living Conditions) countries for which information on
monthly wages is not available in the cross-sectional files, hourly wages are imputed from the 2012 longitudinal EU-SILC files,
except Estonia and the Slovak Republic which are from 2010. Specifically, hourly wages are calculated as annual earnings
divided by annual hours worked. Annual employee earnings are available from the survey, while annual hours worked (total
weeks work*hours worked per week) are derived using monthly vectors of labour force activity (PL211A-PL211L) and as well
as weekly hours worked variable (PL060). All wages are expressed in national currency units and are CPI adjusted. Hourly
wages are computed as monthly earnings divided by the total number of hours worked per week (x4).
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Japan Labour Force Survey “Basic Tabulation” (2012), Korean Labor & Income Panel
Study (KLIPS, 2009) and Survey of Labour and Income Dynamics (SLID, 2010) for Canada.
12 http://dx.doi.org/10.1787/888933208111

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 154 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

Low pay: a result of different characteristics of workers or a penalty on
non-standard work?
The observed wage penalties may be partly the result of different worker
characteristics. First, standard and non-standard workers may have different levels of
human capital and be concentrated in particular occupations and at different stages in
their life-cycles. There is evidence that women overall face occupational segregation,
particularly part-time female workers, reflecting both demand and supply factors (Bardasi
and Gornick, 2008). Temporary or fixed-term contracts may be more prevalent in certain
sectors such as construction, hotels and retail, where wages may be lower. Second,
employers may also pay lower hourly wages mainly to part-timers as a way to
compensate for higher fixed labour costs. Third, even in cases where hourly wages are the
same for standard and non-standard work, part-time and temporary workers may receive
lower earnings because of lower additional pay compensation such as bonuses: evidence
suggests that fixed-term, temporary agency workers and part-time workers are less likely
to benefit from profit-sharing (but are as likely to receive paid overtime and individual
performance pay) and that part-timers are also less likely to receive overtime pay and
team-based bonuses (Venn, 2011).
At the same time, standard and non-standard workers may receive different salaries
because they have different unobserved characteristics or because of asymmetric information
with respect to their ability between employers and job applicants or workers.10 Part-time
workers, for instance, may be more productive because there is an inverted U-shaped
relationship between hourly efficiency and the number of hours, and part-timers might be
found on the rising part of the efficiency hours (Booth and Wood, 2006). On the other hand,
temporary contracts may be used as a screening device. Firms might also use temporary
contracts as a probation device if they cannot screen potential employees directly at a
suitable cost to assess whether they are of high ability. In particular, firms can hire temporary
workers or workers though temporary help agencies so as to obtain economies of scale in
screening and training temporary workers (Autor, 2001; Houseman and Polivka, 1999).
Prospective employers may use the type of employment or their earnings history as a sign of
low ability or lower productivity and offer them lower wages. Controlling for individual
unobserved characteristics is crucial to disentangle the reasons behind wage differentials (see
Box 4.2. for the econometric approach).
Previous studies suggest that, once adjustments are made for personal and job
characteristics, the wage gap between non-standard workers and standard workers
narrows significantly, although an unexplained portion remains. Part-time wage
differentials are related to the extent to which part-time workers are concentrated in lowwage occupations. Evidence suggests that much of the wage difference between part-time
and full-time workers is explained by differences in workers and jobs, with differences in
job characteristics such as occupation or sector being by far more important, and that the
wage penalty might be small but rises over the working life, as a result of lower
experience levels and accumulated human capital (Hirsch, 2005; Bardasi and Gornick,
2008; OECD, 2010). The under-investment in human capital associated with short-term
contracts can give rise to lower wages for temporary workers. Another possibility is that
workers accept lower wages with the expectation that this would be followed by more
stable careers. In the case of temporary workers, occupational segregation within firms is
also responsible for an important portion of the unadjusted wage gap (50%) while higher
observed skills for those on indefinite contracts working in the same occupation accounts
for 30% of the wage differential (De la Rica Goiricelaya, 2004).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 155

Box 4.2. Estimating the wage penalty between standard and non-standard workers
Individual wages may be considered as a function of individual and job characteristics, including the type of
contract:

𝑙𝑛𝑤𝑖𝑡 = 𝑋𝑖𝑡′ 𝛽 + 𝛼𝑁𝑆𝐷𝑖𝑡 + 𝜇𝑖 + 𝜀𝑖𝑡
where i=1,…, N represents the number of individuals at each wave and t=1,…, T is the number of waves,
w is the hourly wage rate in period t for individual i, X is a vector of characteristics that influence wages
including individual and job characteristics, NSD denotes non-standard employment status (either part-time or
temporary), 𝜇 is an unobserved individual effect and 𝜀 is a random error term. Wages are estimated separately
for men and women, as the coefficients of certain covariates may vary by gender.
The inability to measure the unobserved individual effects leads to biased estimates of 𝛼 if individual fixed
effects are correlated with non-standard employment status. Panel-data techniques can be used to focus on wage
changes as a result of changes in non-standard employment status, conditional on the values of the individual
fixed effects. With the differencing, the permanent component 𝜇 is purged, and the resulting estimates of the
equation yield consistent estimates of the coefficients on the assumption that 𝜇 is constant within the relevant
time frame. This also relies on the assumption that X and NSD are orthogonal to the error term 𝜀. In addition,
first-differencing resolves endogenous selection and non-random attrition problems as long as they are related to
the time-invariant individual components.
Three specifications from both cross-sectional and panel data are estimated in the empirical analyses. The
baseline specification includes ordinary least squares (OLS) estimates from pooled cross-sectional data to
paint a general picture about the wage differences between non-standard and standard workers, controlling
for observable characteristics. Specifications (2) and (3) further add interaction terms to examine whether the
extent of the wage gap varies by age or skill groups, respectively. To take into account unobserved individual
heterogeneity, estimates of the fixed effects model are also provided for four countries (Australia, Germany,
Korea and the United Kingdom) for which panel data on hourly wages is available. The fixed effects panel
analysis cannot be used for EU-SILC countries, as information on the hourly wage is not available. The
samples are restricted to paid employees aged 15-64, excluding self-employed workers. In all specifications,
the dependent variable is the logarithm of hourly wages expressed in 2010 constant currency. The main
parameters of interest are dummy variables representing different types of non-standard contracts – full-time
temporary work (TE), part-time permanent (PTPE) and part-time temporary work (PTTE) – as their
coefficients capture the (log) wage differential with reference to standard work. A negative (positive)
coefficient therefore indicates a wage penalty (premium) for non-standard workers. To facilitate
interpretation, we translate coefficients into percentage difference in hourly wages between various groups of
interest and the reference group in Tables 4.1 and 4.2 for men and women, respectively. Since hourly wages
are log-transformed, the percentage difference in hourly wages between temporary and standard (reference)
employment, for instance, can be obtained by [exp(bTE) – 1]*100, where bTE is the estimated coefficient on
the temporary contract dummy.

Temporary workers in all countries face a wage penalty, even after controlling for
observable individual, family and work characteristics (Tables 4.1 and 4.2).11 On average,
a temporary contract worker receives an hourly wage that is 11% lower for men than their
counterparts in standard jobs (13% lower for women). The wage penalty ranges between
almost zero in Australia to 19% in Greece. Similar magnitudes are found in the wage
penalties for men and women in most countries. Some noticeable exceptions include
Greece, Ireland, Portugal and Korea where the wage penalty tends to be some four to five
points higher for women than for men on temporary contracts.
Similarly, in most OECD countries part-time workers also tend to earn lower hourly
wages than their standard full-time counterparts. However, the degree of the penalty
varies depending on the type of contract. In general, the wage penalty is smaller for those
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 156 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
with a permanent employment contract, compared with part-timers with a temporary
contract, and for women: it represents 13% for part-time temporary male workers and 9%
for their permanent counterpart, while it is 12% for part-time temporary women and 4%
for women in part-time permanent employment. The wage penalty for working part-time
is especially pronounced in Germany, Ireland and Poland (the latter for men only). In
Germany, for instance, the hourly wage for part-time men in temporary jobs is 33% (24%
for women) lower than that for full-time standard workers. This may partly reflect the rise
of mini-jobs in Germany. Interestingly, part-time work is not associated with any wage
penalty in Portugal (for those with permanent jobs). Australia stands out as the only
country where a small part-time wage premium is found for female workers.
The analysis reveals that the young and the low-skilled face additional wage penalties
in the case of temporary workers in almost all countries [Tables 4.1 and 4.2,
specifications (2) and (3)].12 This suggests that the pay levels of young and low-skilled
individuals may take longer to converge with the levels of standard jobs if workers start a
career with a temporary job. While the wages of temporary workers increase with age and
skill level, they grow more slowly than those of standard workers. As a result, the wage
differences between temporary and standard workers tend to widen with age or skill. This
implies that years of labour market experience may not be valued in the same way for
temporary workers as for standard workers. Having a higher education level does not
eliminate the wage disadvantage faced by temporary workers. In most countries, those
with a university degree are still at a considerable wage disadvantage compared with
peers in standard work.
For part-time workers, mixed cross-national results are found for wage penalties by
age or skill levels. In a number of countries, such as Austria, Belgium, Greece and Korea,
a wage penalty for male part-timers compared with their full-time standard equivalents is
more pronounced among young workers but tends to be less significant or even
disappears for older workers. The opposite, however, is found in Ireland, Italy and
Portugal. Similarly, the part-time wage penalty for temporary contract workers is largely
driven by less-educated cohorts in some countries (Austria, Belgium) but not in others,
such as Germany, where the wage penalty for part-time temporary workers tends to be
fairly high and equal across all skill groups.
There are significant gender differences in the wage penalty with part-time work. In
general, among those with a permanent contract, the part-time penalty is rather moderate
or negligible for women regardless of age or skill groups, but is still apparent for men. In
most countries, with the exception of Germany, Ireland and the United Kingdom, a wage
premium is even found among young part-time women with a permanent contract.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 157

Table 4.1. Estimates of (log) hourly wage gap between standard and non-standard workers, pooled OLS: Men
AUS
(HILDA)
Specification (1)
Temporary employment (TE)
Part-time permanent employment (PTPE)
Part-time temporary employment (PTTE)
Other controls
Constant
Number of observations
Specification (2)
TE
PTPE
PTTE
Age1529
Age5064
TE·age1529
TE·age5064
PTPE·age1529
PTPE·age5064
PTTE·age1529
PTTE·age5064
Other controls

AUT

BEL

DEU
(GSOEP)

ESP

GBR

GBR
(BHPS)

GRC

HUN

IRL

ITA

KOR
(KLIPS)

POL

PRT

-0.026*** -0.148*** -0.116*** -0.189*** -0.160*** -0.102*** -0.146***' -0.138*** -0.135*** -0.151*** -0.165*** -0.104*** -0.187*** -0.125***
-0.088*** -0.055*** -0.051** -0.239*** -0.100*** -0.174*** -0.116*** -0.074*** -0.075*** -0.194*** -0.063*** -0.070** -0.146*** -0.140***
-0.01
-0.051
-0.170*** -0.507*** -0.191*** -0.143*** -0.134*** -0.107*** -0.128*** -0.291*** -0.171*** -0.177*** -0.244***
-0.037
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
3.108***
25 449

2.808***
22 817

-0.012
-0.185***
-0.098***
-0.246***
0.008
-0.013
-0.044**
0.169***
0.133***
0.132***
0.093***
Yes

-0.129*** -0.133*** -0.190*** -0.167*** -0.144*** -0.157*** -0.142*** -0.156*** -0.201*** -0.170*** -0.099*** -0.206*** -0.131***
-0.097*** -0.075** -0.185*** -0.067*** -0.251*** -0.088*** -0.127*** -0.135*** -0.171*** -0.077***
-0.058
-0.105***
-0.067
-0.233*** -0.166** -0.560*** -0.184***
-0.023
-0.199*** -0.106*** -0.118** -0.302*** -0.192***
0.053
-0.291***
-0.024
-0.181*** -0.173*** -0.207*** -0.159*** -0.254*** -0.244*** -0.240*** -0.135*** -0.266*** -0.205*** -0.209*** -0.191*** -0.215***
0.097*** 0.141*** 0.072*** 0.128***
0.008
-0.017*** 0.179***
0.015*
0.113*** 0.128*** -0.081***
0.009
0.111***
-0.042
0.044
-0.040*** 0.071***
0.024
0.019
0.091*** 0.068***
0.089*
0.046*** -0.102*** 0.100***
0.038**
-0.043
-0.066
0.045*** -0.091*** 0.170***
0.014
-0.108***
0.001
0.097
-0.047*** 0.067*** -0.045*** -0.057**
-0.027
-0.147* -0.137***
-0.056
0.069
-0.019**
0.054
0.141** 0.159*** 0.136*** -0.159**
-0.068
-0.023
0.158***
0.074
-0.062***
-0.059
0.141*** -0.063** 0.168***
0.08
-0.134*** -0.045**
0.160**
-0.069*
-0.158
0.122
-0.028
0.152***
0.033
-0.098
0.038
0.079*
0.038
0.237*** 0.127*** -0.452*** 0.177***
0.027
0.419***
0.048
-0.123*** -0.086**
-0.210*
0.201**
-0.106**
-0.059
-0.105* -0.096*** -0.146**
0.011
-0.111
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes

3.109***
25 449

2.809***
22 817

Other controls

-0.036***
-0.064**
-0.002
-0.130***
0.144***
0.006
0.029*
-0.158***
0.019
0.018
-0.050**
Yes

Constant
Number of observations

3.110***
25 449

Constant
Number of observations
Specification (3)
TE
PTPE
PTTE
Less than high school (Less HS)
University (Univ.)
TE·Less HS
TE·Univ.
PTPE·Less HS
PTPE·Univ.
PTTE·Less HS
PTTE·Univ.

2.861***
3 799

2.863***
3 799

3.028***
62 434

2.835***
12 838

2.683***
67 237

9.053***
21 572

-0.142*** -0.126*** -0.196*** -0.194*** -0.076** -0.123*** -0.140***
-0.067*** -0.110*** -0.236*** -0.157*** -0.151*** -0.184*** -0.113***
-0.041
-0.186** -0.479*** -0.189*** -0.185*** -0.361*** -0.129***
-0.139*** -0.094*** -0.075*** -0.131*** -0.106*** -0.127*** -0.085***
0.124*** 0.139*** 0.154*** 0.080*** 0.162*** -0.139*** 0.146***
-0.03
0.02
0.007
0.083***
-0.092
-0.031
0.041**
-0.012
0.014
0.017
-0.054***
-0.045
-0.037
-0.084***
-0.103***
0.024
-0.110*** 0.104***
-0.071
0.044
0.043
0.108*** 0.154***
0.028
-0.027 0.116***
0.037
0.081
-0.131
-0.048
-0.131***
0.029
0.052
0.304***
0.018
0.04
0.152
-0.042
-0.044
0.08
0.334***
0.071
Yes
Yes
Yes
Yes
Yes
Yes
Yes

-0.136***
-0.105***
-0.141***
-0.132***
0.339***
-0.007
0.023
0.114*
0.089
0.063
-0.046
Yes

-0.171***
-0.121***
-0.183***
-0.101***
0.213***
0.085
-0.001
-0.124***
-0.074
-0.104
-0.264***
Yes

-0.164***
-0.069***
-0.136***
-0.102***
0.199***
0.024***
-0.093***
0.029*
-0.050*
-0.033
-0.194***

-0.131***
-0.032
-0.173***
-0.241***
0.250***
0.225***
-0.307***
0.133
-0.151**
0.259***
-0.162***

Yes

Yes

2.810***
22 817

1.407***
24 866

2.830***
12 838

2.682***
67 237

9.053***
21 572

2.488***
42 703

2.072***
16 557

1.407***
24 866

9.055***
21 572

2.948***
19 051

2.487***
42 703

2.072***
16 557

2.684***
67 237

2.418***
46 520

2.951***
19 051

2.486***
42 703

2.835***
12 838

3.027***
62 434

2.412***
46 520

2.949***
19 051

1.410***
24 866

2.866***
3 799

3.026***
62 434

2.416***
46 520

2.073***
16 557

1.147***
42 964

1.146***
42 964

1.886***
16 478

1.886***
16 478

-0.184*** -0.260***
-0.194*** -0.232*
-0.278***
-0.058
-0.125*** -0.240***
0.198*** 0.346***
0.008
0.200***
-0.039*** -0.076**
0.176***
0.146
-0.166
0.132***
0.056
0.01
0.217***
0.028
Yes
Yes
1.147***
42 964

1.906***
16 478

Note: All regressions control for age groups, levels of education, marital status, the presence of children, limited health condition, dummies for region of residence, occupation and year
effects. Temporary for Australia includes both casual and fixed-term work. Robust standard errors are calculated. ***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2004-2012, cross-sectional files), British Household Panel Survey (BHPS, 1992-2009), German SocioEconomic Panel (GSOEP, 1999-2012), Household, Income and Labour Dynamics in Australia (HILDA, 2001-2012), Korean Labor & Income Panel Study (KLIPS 1999-2009).
12 http://dx.doi.org/10.1787/888933208953
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 158 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Table 4.2. Estimates of (log) hourly wage gap between standard and non-standard workers, pooled OLS: Women
AUS
(HILDA)

AUT

BEL

DEU
(GSOEP)

ESP

GBR

GBR
(BHPS)

GRC

HUN

IRL

ITA

KOR
(KLIPS)

POL

PRT

Specification (1)
Temporary employment (TE)
-0.033*** -0.137*** -0.106*** -0.220*** -0.152*** -0.121*** -0.151*** -0.191*** -0.141*** -0.171*** -0.152*** -0.152*** -0.167*** -0.156***
Part-time permanent employment (PTPE) 0.050*** -0.024*** 0.041*** -0.130*** -0.052*** -0.074*** -0.122*** -0.052*** -0.019 -0.107*** -0.040*** 0.009 -0.051*** 0.006
Part-time temporary employment (PTTE) 0.014** -0.116*** -0.188*** -0.332*** -0.106*** -0.152*** -0.122*** -0.149*** -0.063*** -0.174*** -0.150*** -0.171*** -0.101*** -0.002
Other controls
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Constant
Number of observations
Specification (2)
TE
PTPE
PTTE
Age1529
Age5064
TE·age1529
TE·age5064
PTPE·age1529
PTPE·age5064
PTTE·age1529
PTTE·age5064
Other controls
Constant
Number of observations
Specification (3)
TE
PTPE
PTTE
Less than high school (Less HS)
University (Univ.)

3.168*** 2.679*** 2.834*** 2.907*** 2.225*** 2.838*** 2.275*** 1.980*** 1.338*** 2.699*** 2.491*** 8.897*** 1.096*** 1.720***
25 184
19 876
3 229
58 159
37 378
19 656
45 292
13 023
24 766
13 179
54 202
14 260
37 748
15 855
-0.050***
0.048***
-0.008
-0.182***
-0.001
0.039**
0.029
-0.003
0.003
0.067***
0.005
Yes

-0.160***
-0.015**
-0.118***
-0.221***
0.098***
0.067**
0.009
0.031
-0.048***
0.129***
-0.087**
Yes

-0.100***
0.041***
-0.182***
-0.212***
0.167***
0.018
0.007
0.062*
-0.060*
0.250***
-0.340***
Yes

-0.196***
-0.115***
-0.329***
-0.200***
0.056***
-0.078***
0.042**
-0.01
-0.042***
0.055***
-0.054***
Yes

-0.151***
-0.054***
-0.114***
-0.188***
0.135***
0.059***
-0.093***
0.086***
-0.037***
0.098***
-0.080***
Yes

-0.139***
-0.085***
-0.244***
-0.213***
-0.015
0.063
-0.018
0.044**
0.015
0.134**
0.196***
Yes

-0.143***
-0.149***
-0.119***
-0.199***
-0.058***
-0.012
0.009
0.109***
0.038***
-0.003
-0.02
Yes

-0.207***
-0.074***
-0.172***
-0.262***
0.166***
0.119***
-0.126***
0.099***
0.008
0.134***
-0.068**
Yes

-0.146***
-0.014
-0.029
-0.145***
0.052***
0.031
-0.004
0.188***
-0.075***
-0.024
-0.078*
Yes

-0.179***
-0.107***
-0.143***
-0.302***
0.096***
0.061
-0.095
0.097***
-0.036*
0.071
-0.126***
Yes

-0.136***
-0.033***
-0.171***
-0.233***
0.158***
0.029***
-0.113***
0.131***
-0.102***
0.183***
-0.126***
Yes

-0.160***
-0.016
-0.171***
-0.133***
-0.056***
-0.022
0.072**
0.055
0.124*
-0.098***
0.183***
Yes

-0.179***
-0.025*
-0.132***
-0.239***
0.079***
0.094***
-0.061***
0.107***
-0.112***
0.150***
-0.014
Yes

-0.143***
0.053**
-0.037
-0.186***
0.175***
0
-0.090***
-0.004
-0.112***
0.103**
0.002
Yes

3.168*** 2.678*** 2.822*** 2.900*** 2.222*** 2.842*** 2.287*** 1.982*** 1.337*** 2.675*** 2.485*** 8.905*** 1.094*** 1.716***
25 184
19 876
3 229
58 159
37 378
19 656
45 292
13 023
24 766
13 179
54 202
14 260
37 748
15 855

Other controls

-0.101*** -0.080* -0.205*** -0.158***
-0.028***
0.03
-0.141*** -0.084***
-0.109*** -0.163*** -0.350*** -0.103***
-0.153*** -0.120*** -0.067*** -0.120***
0.195*** 0.176*** 0.092*** 0.104***
0.014
-0.036
0.051
-0.025 0.043***
0.006 -0.101***
-0.06
-0.030*
-0.017
-0.014
0.004
0.05
0.003
0.084***
0.028*
0.022
0.005
0.044***
0.002
-0.039** -0.016
-0.108*
0.028
0.057***
0.012
-0.017
0.056
0.062*** -0.088***
Yes
Yes
Yes
Yes
Yes

Constant
Number of observations

3.166*** 2.681*** 2.834*** 2.913*** 2.224*** 2.842*** 2.273*** 1.972*** 1.339*** 2.700*** 2.488*** 8.902*** 1.084*** 1.717***
25 184
19 876
3 229
58 159
37 378
19 656
45 292
13 023
24 766
13 179
54 202
14 260
37 748
15 855

TE·Less HS
TE·Univ.
PTPE·Less HS
PTPE·Univ.
PTTE·Less HS
PTTE·Univ.

-0.040***
0.042***
0.020*
-0.052***
0.143***

-0.065
-0.088***
-0.202***
-0.109***
0.181***

-0.163***
-0.118***
-0.15***
-0.107***
0.159***

-0.181***
-0.073***
-0.098***
-0.071***
0.191***

0.004
0.031
0.043**
-0.095*
0.012 -0.054***
0.074*** 0.036***
0.049
0.012 -0.031*** 0.068*
0.195*
0.019
0.011
0.076
0.051* -0.156***
Yes
Yes
Yes

-0.146***
-0.029*
-0.093***
-0.092***
0.325***
0.054**
-0.031
-0.006
0.075**
0.061
0.111*
Yes

-0.109***
-0.123***
-0.218***
-0.154***
0.200***

-0.159***
-0.036***
-0.150***
-0.127***
0.155***
-0.140* 0.029***
-0.074
-0.003
0.053** 0.021**
0.01
-0.087***
0.094** 0.042***
0.057 -0.096***
Yes
Yes

-0.197***
-0.037
-0.284***
-0.203***
0.272***
0.151***
-0.096***
0.059
0.158***
0.264***
0.165***
Yes

-0.129*** -0.148***
-0.036*** -0.005
-0.089*** -0.009
-0.063*** -0.190***
0.299*** 0.402***
-0.025 0.081***
-0.156*** -0.216***
-0.001
0.039
-0.055** -0.227***
-0.029
0.013
-0.024
-0.011
Yes
Yes

Note: All regressions control for age groups, levels of education, marital status, the presence of children, limited health condition, dummies for region of residence, occupation and year
effects. Temporary for Australia includes both casual and fixed-term work. Robust standard errors are calculated. ***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2004-2012, cross-sectional files), British Household Panel Survey (BHPS, 1992-2009), German SocioEconomic Panel (GSOEP, 1999-2012), Household, Income and Labour Dynamics in Australia (HILDA, 2001-2012), Korean Labor & Income Panel Study (KLIPS, 1999-2009).
12 http://dx.doi.org/10.1787/888933208964
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 159

Wage differences between standard and non-standard workers could also be driven by
some unobserved individual effects like ambition or preferences. For instance, non-standard
workers may receive lower wages simply because they are less productive or because they
have preferences for more flexible work arrangements. Taking such effects into account
tends to reduce the hourly pay gap. This analysis is undertaken for four OECD countries:
Australia, Germany, Korea and the United Kingdom (Table 4.3). Overall, being a
temporary contract worker is still associated with a wage penalty in three of the
four countries, but the difference in hourly wages is more moderate: about 5% lower for
men and 8% for women compared with their standard work counterparts, holding
everything else equal. In Australia, there is no longer a difference in hourly wages between
standard workers and temporary contract workers with similar characteristics.
Once individual fixed effects are taken into account, the part-time penalty is no longer
present, especially for those with a permanent contract, except for German men. If anything,
a wage premium is found in Australia and, to a lesser extent, for women in Korea. By
contrast, part-time temporary workers still face some wage disadvantage in Germany and
Korea (for women). In Australia, a substantial part-time premium (10-12% for women,
18-24% for men) was found once unobserved individual heterogeneity was taken into
account: one hypothesis is that firms may have to pay more to attract part-time workers
because of the high effective marginal tax rates for second-earners (Booth and Wood, 2006).
Distinct wage effects of NSW by age and skill group hold once unobserved effects are
taken into account. In three panel countries where a wage penalty is observed for temporary
workers, this penalty is higher for younger workers. In Germany, for instance, the hourly
wages of young male (female) temporary workers are 11% (16%) lower than those of their
age-similar equivalents in standard jobs. The comparable wage differences for other age
groups, however, tend to be rather modest – about 5% or smaller.
With respect to those working in permanent jobs, there is no obvious wage penalty for
part-time women in all age or skill groups, while for men the observed small or insignificant
wage gap masks two contrasting effects across the population. In Korea, for instance, a wage
penalty (13%) is found for young male workers – compared with standard full-time
workers – whereas a wage premium (18%) is estimated for older workers. As for individuals
with a temporary contract, the wage penalty also tends to be much higher among the young
for both men and women. In the United Kingdom, on the contrary, older part-time workers
(men) are at a higher risk of receiving a wage penalty, as they earn about 10% less than their
age-similar counterparts in standard jobs. Moreover, in Australia wage premiums are found
for all part-time workers, regardless of age or skill level.
In sum, four main messages can be drawn from the findings of this section. First, in most
countries being a full-time temporary worker is associated with a wage penalty. This result is
robust in both cross-sectional and panel analyses regardless of whether or not controlling for
unobserved individual heterogeneity. Second, part-timers – in particular those with a
temporary contract – also tend to have lower hourly wages than their regular full-time
equivalents. However, part-time wage penalties diminish or even disappear when
unobserved fixed effects are taken into account, suggesting that unobservable characteristics,
such as individual preferences or ability, play an influential role in determining the wage gap
for part-timers. Third, the extent of the wage penalty is not homogeneous across all nonstandard workers. In many countries, the penalty primarily affects young workers, especially
those with a temporary employment contract. Finally, there is a considerable cross-national
variation in wage penalties associated with non-standard work. In general, full-time
temporary contract workers tend to fare worse in hourly wages in Austria, while having a
part-time job is harshly penalised in Germany but rewarded in Australia.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 160 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Table 4.3. Fixed effects estimate of wage gaps
AUS
(HILDA)
Specification (1)
Temporary employment (TE)
-0.001
Part-time permanent employment (PTPE) 0.172***
Part-time temporary employment (PTTE) 0.172***
Other controls
Yes
Constant
Specification (2)
TE
PTPE
PTTE
Age1529
Age5064
TE·age1529
TE·age5064
PTPE·age1529
PTPE·age5064
PTTE·age1529
PTTE·age5064
Other controls

Men
DEU
KOR
(GSOEP)
(KLIPS)

GBR
(BHPS)

AUS
(HILDA)

Women
DEU
KOR
(GSOEP)
(KLIPS)

GBR
(BHPS)

-0.063***
-0.047**
-0.242***
Yes

-0.083***
-0.008
-0.012
Yes

-0.068***
0.028
0.004
Yes

0.008
0.125***
0.139***
Yes

-0.092***
-0.009
-0.048***
Yes

-0.101***
0.060*
-0.074***
Yes

-0.082***
-0.001
-0.011
Yes

2.872***

2.678***

8.852***

2.053***

2.775***

2.552***

8.487***

1.826***

0.007
0.178***
0.174***
-0.001
-0.053***
-0.024
0.005
-0.023
0.019
-0.02
0.038
Yes

-0.052***
-0.003
-0.156***
-0.059***
-0.025***
-0.067***
0.016
-0.092*
-0.078*
-0.091
-0.235***
Yes

-0.046**
0.013
0.136***
-0.073***
-0.039**
-0.090***
-0.038
-0.157*
0.152
-0.296***
-0.212**
Yes

-0.029
0.112***
0.047
-0.092***
-0.094***
-0.074***
-0.019
-0.063
-0.218***
-0.047
-0.11
Yes

0.004
0.124***
0.145***
-0.011
-0.037***
-0.003
0.029
-0.011
0.02
-0.02
0.004
Yes

-0.051***
-0.003
-0.036**
-0.055***
-0.014*
-0.130***
0.017
0.002
-0.01
-0.064**
0.007
Yes

-0.079***
0.053
-0.056*
-0.062***
-0.069***
-0.078**
0.022
0.027
0.037
-0.121**
0.133**
Yes

-0.035*
-0.014
0.015
-0.056***
-0.079***
-0.102***
-0.027
0.043***
0.035**
-0.069
0.002
Yes

Constant
Specification (3)
TE
PTPE
PTTE
Less than high school (Less HS)
University (Univ.)
TE·Less HS
TE·Univ.
PTPE·Less HS
PTPE·Univ.
PTTE·Less HS
PTTE·Univ.
Other controls
Additional 1

2.872***

2.677***

8.842***

2.051***

2.778***

2.538***

8.487***

1.833***

0.008
0.155***
0.183***
-0.105***
0.081***
-0.019
-0.015
0.038
0.024
-0.005
-0.036
Yes
Yes

-0.067***
-0.053*
-0.229***
-0.012
0.111***
0.031
0.004
-0.07
0.046
-0.082
-0.02
Yes
Yes

-0.077***
-0.052
-0.044
-0.077
0.092**
0.058
-0.136***
0.159
0.079
0.121
0.028
Yes
Yes

-0.064***
-0.028
-0.137**
-0.001
0.055***
-0.04
8
0.08
0.074
0.187**
0.215***
Yes
Yes

-0.001
0.127***
0.119***
-0.050*
0.106***
0.022
0.008
-0.019
0.004
0.003
0.055**
Yes
Yes

-0.071***
-0.018*
-0.052***
-0.028
0.108***
-0.017
-0.055***
0.056**
0.013
0.072**
-0.023
Yes
Yes

-0.148***
0.031
-0.172***
-0.037
0.04
0.147***
-0.019
0.016
0.132
0.219***
0.171**
Yes
Yes

-0.118***
-0.013
-0.043
0.004
0.057***
0.064
0.049*
0.055***
0.001
0.063
0.063
Yes
Yes

Constant
Number of observations
Numbers of groups

2.869***
25 449
5 015

2.676***
62 434
12 340

8.852***
26 952
5 557

2.055***
42 703
5 889

2.782***
25 184
5 156

2.551***
58 159
12 061

8.503***
14 260
4 382

1.83***
45 292
6 294

Note: All regressions control for age groups, level of education, marital status, the presence of children, limited health condition,
dummies for region of residence, occupation and year effects.
1. Additional controls in national panels include dummies for industry, firm size, and job tenure (for Australia and Germany).
Temporary for Australia includes both casual and fixed-term work. Robust standard errors are calculated. ***, **, * denote
significance at the 1%, 5%, 10% levels, respectively.
Source: British Household Panel Survey (BHPS, 1992-2009), German Socio-Economic Panel (GSOEP, 1999-2012), Household,
Income and Labour Dynamics in Australia (HILDA, 2001-2012), Korean Labor & Income Panel Study (KLIPS, 1999-2009).
12 http://dx.doi.org/10.1787/888933208974

Having identified a wage penalty for temporary workers, it is important to investigate
whether they have an equal or better chance of upward earnings mobility compared with
workers in standard employment. If temporary jobs are a form of probation, a low wage
during the probationary period will be followed by higher future wages and there should
be little loss in terms of long-term wages (Booth et al., 2002). A conversion from a
temporary to a permanent job (with the same employer) should thus be associated with
upward earnings mobility. Also, because temporary workers are more likely to switch
jobs voluntarily given their short-term contracts, they may be more likely to take
advantage of higher-paying job opportunities than less mobile permanent workers
(Amuedo-Dorantes and Serrano-Padial, 2007). Table 4.4 looks at contract and earnings
mobility for temporary workers.13
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 161

Table 4.4. Change of employment contract and earnings mobility, average over panel periods
Australia (HILDA)
Transition Downward Stay
Upward

Total

Transition Downward

SW-SW
SW-TE
TE-SW
TE-TE

17.6
24.1
24.7
20.7

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

Upward

Austria
Stay

Upward

Total
100
100
100
100

Total

27.0
54.4
18.6
28.3
43.3
28.3
27.7
45.5
26.8
28.4
44.4
27.2
Czech Republic
Transition Downward Stay
Upward

Total

18.6
27.7
31.2
23.9

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

17.3
24.5
22.5
20.3

100
100
100
100

Upward

Total

63.6
54.2
58.5
58.5
France
Transition Downward Stay

Upward

Total

22.7
59.0
18.3
35.6
36.6
27.7
23.5
43.5
33.1
41.2
45.6
13.2
Germany (GSOEP)
Transition Downward Stay
Upward

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

11.5
23.0
31.5
19.9

100
100
100
100

Total

79.2
60.6
55.0
66.2
Greece
Transition Downward Stay

Upward

Total

SW-SW
SW-TE
TE-SW
TE-TE

14.8
17.1
20.1
18.1

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

13.6
18.7
17.2
17.3

100
100
100
100

Upward

Total

Transition Downward

Upward

Total

64.3
16.6
55.6
21.4
58.5
20.1
53.1
25.7
Korea (KLIPS)
Transition Downward Stay
Upward

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

100
100
100
100

Total

71.8
15.1
56.1
23.3
60.3
25.1
57.1
25.9
Luxembourg
Transition Downward Stay
Upward

Total

SW-SW
SW-TE
TE-SW
TE-TE

19.8
19.5
37.3
21.1

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

12.9
23.8
23.8
19.5

100
100
100
100

Upward

Total

75.8
52.1
63.1
63.6
Portugal
Transition Downward Stay

Upward

Total

16.2
21.4
24.7
23.7

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

100
100
100
100

Upward

Total

13.8
72.6
13.6
18.1
57.1
24.8
13.8
66.2
20.0
15.4
61.9
22.8
Slovak Republic
Transition Downward Stay
Upward

Total

17.3
67.4
15.3
23.7
49.7
26.6
23.1
54.3
22.6
24.6
51.8
23.6
United Kingdom (BHPS)
Transition Downward Stay
Upward

100
100
100
100

SW-SW
SW-TE
TE-SW
TE-TE

100
100
100
100

Total

SW-SW
SW-TE
TE-SW
TE-TE

100
100
100
100

66.8
55.4
57.5
61.2
Belgium
Transition Downward Stay
SW-SW
SW-TE
TE-SW
TE-TE

15.6
20.6
17.8
18.2

16.8
24.8
14.6
17.3

64.6
47.5
54.2
58.9
Estonia
Transition Downward Stay
SW-SW
SW-TE
TE-SW
TE-TE

73.9
67.8
68.7
69.8
Hungary
Transition Downward Stay
SW-SW
SW-TE
TE-SW
TE-TE

11.3
15.1
11.2
12.2

19.1
23.0
21.4
21.1

63.0
45.5
48.3
57.2
Poland
Transition Downward Stay
SW-SW
SW-TE
TE-SW
TE-TE

17.2
35.1
14.5
21.7

16.7
24.4
20.1
20.0

67.1
54.2
55,15
56.4
Spain
Transition Downward Stay
SW-SW
SW-TE
TE-SW
TE-TE

12.0
23.6
19.4
16.0

72.1
54.6
54.6
60.4

16.0
21.8
26.2
23.6

19.2
21.3
18.9
21.2

9.4
16.4
13.5
13.9

11.8
14.8
16.2
15.5

74.6
66.6
66.7
67.2
Italy
Stay

13.1
20.7
14.6
16.9

11.4
24.1
13.1
16.9

20.6
23.3
22.9
23.8

59.7
50.5
52.4
54.0

19.7
26.2
24.8
22.3

Note: SW: standard full-time permanent employment; TE: full-time temporary contract. A worker is said to have upward mobility if
he/she moved up at least one earnings quintile from year t-1 to t; similarly downward mobility refers to a move to a lower quintile. Stay
refers to workers who remained in the same earnings category. Temporary for Australia includes both casual and fixed-term work.
Source: British Household Panel Survey (BHPS, 1992-2009) for the United Kingdom, German Socio-Economic Panel (GSOEP,
1999-2012) for Germany, European Union Statistics on Income and Living Conditions (EU-SILC, 2004-2012) for other
European countries, Household, Income and Labour Dynamics in Australia (HILDA, 2001-2012) for Australia, Korean Labor &
Income Panel Study (KLIPS ,1999-2009) for Korea.
12 http://dx.doi.org/10.1787/888933208988

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 162 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
On average, about one-third to one-half of full-time employees made a transition into
another earnings category within one year (Table 4.4). A great majority of workers,
however, remained in the same earnings quintile, regardless of changes in the type of
contract. This is confirmed by evidence from Australia and several European countries
that shows, in particular, strong state dependence among low-wage workers
(Buddelmeyer et al., 2010; Cappellari and Jenkins, 2008; Mosthaf, 2011; Stewart, 2007,
Uhlendorff, 2006). In general, upward earnings mobility is higher for those experiencing
a change from temporary to permanent work (TE-SW). In Belgium, Estonia, France and
Korea, for instance, about one in three workers who moved from TE to SW increased
their earnings. Because of data limitations, it is not possible to distinguish whether or not
a change of contract also involves a change of employer. It is therefore difficult to infer
whether these upgrades in earnings reflect workers’ gains in productivity, as people
initially on temporary contracts who display high ability are later offered permanent
positions at a firm, or whether this is due to a (more intensive) job search in the pursuit of
a better career match. Similarly, switching from standard employment to a temporary
contract (SW-TE) often results in earnings losses (especially in Estonia and Korea).
Although entering a standard job often leads to a wage rise, there are exceptions. In
Hungary, Spain and the Slovak Republic, for instance, nearly one in two workers changed
earnings category when moving from a temporary contact to standard employment; about onehalf of them experienced upward mobility while the other half ended up with lower earnings.
Finally, in several countries (Austria, the Czech Republic and Korea) temporary
workers who remained in the same contract type tended to be relatively mobile in
earnings in both directions (upward and downward), suggesting greater earnings
instability among such workers. A fall in earnings is more common in Estonia where
more than 41% of persistent temporary workers experienced a fall in earnings, as shown
by moving to a lower quintile.

Are non-standard jobs stepping stones?
The analysis has shown that significant wage penalties are associated with temporary
work, and in some countries with part-time jobs for men, compared with standard permanent
employment, even when controlling for the observed and unobserved characteristics of
workers. A related concern is whether this wage differential has a long-lasting impact on
wages over the career, or whether temporary workers can catch up with their counterparts
who started permanent jobs earlier. Spells of low-paid jobs may lead to depreciation in
human capital, which may compromise the possibility to find better-paid jobs and thus
generate persistence in low-paid employment. On the other hand, non-employment may lead
to a larger loss of human capital and often has a scarring effect on subsequent employment;
getting any job, even a low-paid, less stable job or one involving limited hours, is therefore
often put forward as a way to improve future employment and wage prospects. In this view,
non-standard jobs may offer unemployed individuals a transition to more stable jobs. This
sub-section investigates whether such “stepping stone” effects exist in the short run.
Past findings on whether non-standard forms of employment improve or hinder
labour market prospects vary across countries and the type of non-standard employment
considered. Part of the literature suggests that temporary jobs are often stepping stones to
permanent employment (e.g. Gagliarducci, 2005; Icchino et al., 2008). However, findings
differ with respect to the type of temporary contract and other workers characteristics. For
instance, in the United Kingdom, a large fraction of people on fixed-term contracts
mainly move into permanent jobs while transition rates are much lower for workers on
seasonal or casual jobs, especially for part-time workers (Booth et al., 2002). Weaker
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 163

labour market prospects for temporary and contingency workers have also been found in
the United States (Autor and Houseman, 2005), Spain (Amuedo-Dorantes, 2000; EstebanPretel et al., 2009) and Japan (Yu, 2011).
Box 4.3. Estimating the probability of labour market transitions
Because it is difficult to differentiate the effects of holding a given job (i.e. whether holding a non-standard job
leads to human capital depreciation) from the characteristics and motivations that lead individuals to choose such jobs
(i.e. whether individuals choose non-standard jobs), it is challenging empirically to test the “stepping-stone”
hypothesis.
Testing the “stepping stone” hypothesis requires that the causal effect of holding a non-standard job (state
dependency*) be isolated from the impact of confounding factors, such as differences in educational levels and
motivations between standard and non-standard workers (unobserved heterogeneity). The analysis in this
sub-section is performed for 17 OECD countries.
To analyse transitions between different states and address the issue of state dependence, a dynamic probit
model is used. This model estimates the conditional probability of being in standard work at t, conditional on
previous employment status (Lit-1) and demographic characteristics (Xit), while also controlling for unobservable
individual heterogeneity (  i ). The general form is:

Pr(Lit  1   Lit 1 , X it ,  i )  ( L'it 1  X it'    i ).

(1)

In estimating the dynamic model, the problem of initial conditions needs to be taken into account: an individual’s
labour market status at the start of the panel is not randomly distributed and will be influenced by unobservable
individual heterogeneity (  i ). Failing to take into account the initial condition problem will lead to overstating the

level of state dependence. Indeed, controlling for initial labour market status, demographic characteristics and
household income leads to large drops in the coefficients of the lagged labour market status. This indicates that not
controlling for initial conditions would seriously bias the estimates. Following Wooldridge (2002), the distribution of
the individual effects is parameterised as a linear function of the initial employment status at the first wave of the
panel and of the time means of the regressors, assuming that this has a conditional normal distribution:
_

 i  c0  L'i 0   X 'i    i .

(2)

Substituting (2) into (1) yields:
_

Pr(Lit  1   Lit 1 , X it ,  i )  ( L'it 1  X it'   c0  L'i 0   X 'i    i ).

(3)

For the probability of being in standard work, past labour market status includes dummies for lagged standard
work, lagged NSW (temporary, part-time and self-employed) and lagged inactivity. The coefficients for NSW at
time t-1 should be interpreted as the difference in the likelihood of being in standard work at wave t as compared
with being unemployed at t-1. In addition to the previous labour market state, the estimation controls for individual
_

characteristics (X) and for initial conditions ( L'i 0 and

X ). The former include age, education, marital status, an

indicator for self-reported health, household incomes and geographic location (as well as year effects).
Alternatively, a single equation can be estimated (a dynamics multinomial logit) to capture all labour market
transitions simultaneously (e.g. Buddelmeyer and Wooden, 2011). This would allow individual heterogeneity (  i ) to
be correlated across the different labour market choices. For the sake of simplicity, this study assumes independent
choices in labour market status and estimates equation (3) in a univariate framework. That is, the probability of
transition into standard work and transition into non-employment is estimated by two separate probit models.
* State dependency arises when individuals who have experienced an event in the past are more likely to experience the
event in the future than are individuals with identical characteristics who have not experienced the event (Heckman, 1981).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 164 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
The analysis undertaken for 17 OECD countries supports the stepping-stone hypothesis
in most cases, at least in the short-run – but only for some types of non-standard jobs, in
particular temporary jobs. The estimations presented in Figure 4.11 capture the marginal
effects of the labour market and job status in the previous year on the probability of having
a standard job. The reference group is that of the unemployed, i.e. the numbers for standard
work, temporary, part-time, self-employed and inactive should be interpreted as how much
more or less likely such individuals are to move to a standard job in the next year compared
with unemployed people with similar characteristics.
In most countries, full-time temporary workers have a higher probability of entering into
standard employment than the unemployed. In Belgium, for instance, the likelihood of
having a standard job in the current period for men is 32 percentage points. The steppingstone effect for temporary jobs is also noticeable in the Czech Republic, Germany, Portugal
and the United Kingdom and in the Slovak Republic for women. France is an exception
where female workers on temporary jobs have the same likelihood as the unemployed to
obtain a standard job, and in the case of men are slightly less likely. Korea also stands out
since both men and women in temporary employment have lower probabilities of moving
into standard employment than the unemployed.14 Other research for OECD countries shows
however that transition rates remain low when considering a longer time span and that
inequalities are likely to persist: less than 50% of workers who were on temporary contracts
in a given year were employed with full-time permanent contracts three years later (OECD,
2014). In many countries, prime age and older workers in temporary jobs have a better
chance of using such jobs as “stepping stones” than younger workers (Chen et al., 2015).
On the other hand, having a part-time job or self-employed work does not necessarily
improve the chances of getting a permanent full-time job compared to being unemployed.
An observable increased probability for part-timers compared with the unemployed is
found only in about one-third of the countries studied, noticeably for men in Belgium and
Spain, and for women in Hungary. In several countries part-time work reduces the
likelihood of finding a standard job compared to being unemployed: in Australia, Korea
and Luxembourg for both men and women (in Germany and the Slovak Republic for
men, and in the United Kingdom for women). This confirms the findings from other
studies showing that part-time jobs are used as stepping stones in very few cases and that
retention in part-time employment is high: about two-thirds of part-timers stay in parttime employment for more than one year (OECD, 2010).
With regard to another dimension of job quality,15 job security, non-standard workers
are also worse off than full-time permanent employees. Job security, proxied by the
probability of job loss within the next six months, is markedly different across types of
jobs: non-standard workers are more than twice as likely as standard workers to lose their
job within six months (Chen et al., 2015). Previous temporary employment increases the
likelihood of unemployment in almost every country, especially for men, compared with
being in standard work (Annex 4.A3, Figure 4.A3.1). On the other hand, part-time work
or self-employment increases the risk of dropping out of the labour market. Indeed,
working part-time is associated with a higher risk of inactivity compared with standard
workers in about three-quarters of countries, on average by a margin of 4.5 percentage
points. Moreover, in some countries, both temporary workers and part-timers face a
double income and security penalty that can increase their risk of poverty: not only do
their contracts tend to be more precarious, but they also have less coverage by
unemployment insurance systems, since shorter periods of work (and shorter working
hours for part-timers) make them less likely to meet the eligibility conditions for these
insurance schemes (OECD, 2010; OECD, 2014).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 165

Figure 4.11. Influence of previous labour market status on the probability of having a standard employment
Women

Men

Australia

Austria
***
***

***

Standard work t-1

***

***
***

***

***
***

Temporary t-1

Belgium

***
***

**

**
***

Part-time t-1

***

***

Self-employed t-1

***
**

***
***

Inactive t-1

***

***

Unemployed t-1 (ref.)
-40% -20%

0%

20%

40%

60%

80%

-40% -20%

0%

Czech Republic

20%

***

***

***
***

Temporary t-1

80%

-40% -20%

0%

20%

40%

60%

80%

France
***
***

***

***
***

*

Part-time t-1
Self-employed t-1

60%

Estonia
***

Standard work t-1

40%

**
***

***

Inactive t-1

***

***
***

***

***

***

***

***

***

***

Unemployed t-1 (ref.)
-40% -20%

0%

20%

40%

60%

80%

-40% -20%

0%

Germany

20%

60%

80%

-40% -20%

0%

***
***

***

***
***

20%

***

***

Self-employed t-1

Inactive t-1

***
***

***

***

80%

60%

80%

***

***

***
***
***

60%

***
***

***
***

***

Part-time t-1

40%

Hungary

Greece

Standard work t-1
Temporary t-1

40%

*

***
***

***

***

Unemployed t-1 (ref.)
-40% -20%

0%

20% 40% 60% 80%

-40% -20%

Italy

***

Temporary t-1

***
***

Part-time t-1

***
***

Inactive t-1

***

20%

40%

60%

80%

-40% -20%

Korea

Standard work t-1

Self-employed t-1

0%

0%

20%

Luxembourg

***
***

***

***
***

***
***
***

**

**
**

***

***
***

***
***

40%

***

***

***
***

***
***

***

***

Unemployed t-1 (ref.)
-40% -20%

0%

20%

40%

60%

80%

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

-40% -20%

0%

20%

40%

60%

80%

-40% -20%

0%

20%

40%

60%

80%

 166 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.11. Influence of previous labour market status on the probability of having a standard employment
(cont.)
Women

Men

Poland

Slovak Republic

Portugal
***
***

Standard work t-1

***

***
***

Temporary t-1

***

***

***
***

***

***
***

Part-time t-1
Self-employed t-1

*
***

Inactive t-1

**

***

***

**

**
***
**

***

*

***

***

Unemployed t-1 (ref.)
-40% -20%

0%

20%

40%

60%

80%

-40% -20%

Spain

20%

40%

***
***

Part-time t-1

***
***

80%

-40% -20%

0%

20%

40%

60%

80%

***
***

Temporary t-1

60%

United Kingdom

***

Standard work t-1

***
***
***
**

Self-employed t-1
Inactive t-1

0%

***
***
***

***

******

Unemployed t-1 (ref.)
-40% -20%

0%

20%

40%

60%

80%

-40% -20%

0%

20%

40%

60%

80%

Note:
Marginal effects from lagged employment status on the probability of standard employment based on a random-effects dynamic
probit, controlling for initial conditions. ***, **, * denote significance at the 1%, 5%, 10% levels, respectively.
Samples include persons present in at least three consecutive waves and aged 15-64, and excluding retired individuals.
Temporary for Australia includes both casual and fixed-term work.
All regressions include regional dummies, year dummies, age groups, a dummy for the presence of children, a dummy for
whether the individual suffers from a health problem, a dummy for whether the individual is married, dummies for low-skilled
and high-skilled (corresponding to ISCED 0-2 and ISCED 5-6) and equivalised household net income. To control for initial
conditions, initial labour market status dummies are included as well as averages for time-varying regressors based on
Wooldridge (2002). To test whether including additional controls alter the transition probabilities, separate regressions were
estimated for countries for which additional variables are available (e.g. including job tenure, total time in employment since
full-time education and its square, and shares of the foreign-born for Australia; an indicator of ethnicity for the United Kingdom;
and an indicator of foreign-born and work experience for Germany). The results are very similar to the baseline specifications.
Source: British Household Panel Survey (BHPS, 2004-2009) for the United Kingdom, German Socio-Economic Panel (GSOEP,
2004-2012) for Germany, European Union Statistics on Income and Living Conditions (EU-SILC, 2004-2012) for other
European countries, Household, Income and Labour Dynamics in Australia (HILDA, 2004-2012) for Australia, Korean Labor &
Income Panel Study (KLIPS, 2004-2009) for Korea.
12 http://dx.doi.org/10.1787/888933208120

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 167

In terms of the work environment, differences between standard and non-standard
workers depend largely on the type of contract, except with respect to training, which
tends to be lower for all non-standard workers. Full-time temporary and part-time
workers are 20% and 40% less likely respectively to receive training, while training for
part-time permanent workers is marginally lower than for standard workers (Chen et al.,
2015). From the perspective of financial efficiency, if jobs are temporary or workers are
loosely attached to the labour market, it can be inefficient for workers to invest in specific
human capital or for firms to provide firm-specific training. Across European countries,
temporary workers in 40% of the countries report a slightly higher incidence of working
on tight deadlines (Chen et al., 2015). Temporary workers report a higher incidence of job
strain, which tends to be driven both by higher job demands and lower job resources: they
report higher exposure to both physical health risk factors at work and workplace
intimidation, while enjoying less autonomy, fewer learning opportunities and less support
from their colleagues (OECD, 2014). However, the incidence of job strain tends to be
lower among part-time contracts as they have lower job demands.

How does non-standard work affect the overall distribution of earnings?
The previous sub-sections have discussed the wage penalty for non-standard workers
at the median, but the evidence of job polarisation has shown that NSW appears to have
increased both at the bottom and at the top of the job distribution. This section looks at
the earnings gap across the distribution to ascertain whether wage differentials exist and
are similar at different parts of the distribution and then make the link with wage
inequality. The following analysis investigates whether an increase in non-standard work
implies an increase in workers in the lowest earnings segments, thereby making earnings
more unequal.
The impact of NSW on the overall wage distribution is likely to depend on the
concentration of non-standard workers in particular parts of the wage distribution.
Figure 4.12, Panel A confirms that non-standard workers (excluding the self-employed)
are more likely to be found in the lower part of the earnings distribution, particularly in
the lowest three deciles.16 More than half of employees in the lowest decile of earnings
are non-standard workers, and this figure reaches more than two-thirds in Germany and
Canada. In contrast, the share of NSW is below 15% in the top decile. The presence of
non-standard workers at the top of the distribution is particularly low in Hungary (5%),
Spain and Korea (8-9%).
In almost all countries, there is an earnings gap between standard and non-standard
workers that is significantly larger at the bottom of the wage profile: the so-called stickyfloor effect. Using the unconditional quantile regression models (see Box 4.4),
Figure 4.12, Panel B shows how the wage penalty associated with a marginal increase in
NSW varies for the different points of the wage distribution. On average, a rise in the
share of NSW leads to lower log hourly wages of around 18% to 24% for the lowest 40%
of the distribution. The earnings gap for non-standard workers decreases for each decile
between the middle of the distribution and the top, virtually disappearing for the top 10%.
The shape and magnitude of the gap are in line with previous work on temporary
contracts (Bosio, 2014; Mertens et al., 2007; Santangelo, 2011). As a result, an increase
in the share of NSW should contribute to widening overall wage inequality, since it
increases inequality at the bottom end of the distribution and has a neutral effect on wage
inequality at the top end.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 168 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

Box 4.4. Assessing the impact of non-standard work along the earnings distribution
For investigating the impact of non-standard work on different deciles of the earnings distribution, and the
resultant impact on inequality, a methodology using unconditional quantile regressions (UQR) is used, which
was introduced by Firpo et al. (2007) and further developed in Fortin et al. (2010).
This method is based on regressions in which the dependent variable is a transformation, the Recentered
Influence Function (RIF) of the outcome variable, i.e. the unconditional quantile. Each quantile of earnings is
regressed against NSW, age (and its square), gender, education, industry and occupation, as well as regional
controls.
While conditional quantile regression allows the estimated return to a given characteristic to vary according
to the conditional quantile of an individual, which can be thought of as the individual’s position in a virtual
distribution in which everybody else has the same observed characteristics, unconditional quantile regressions
allow estimating the impact of a small locational shift in the distribution of a variable of interest on the entire
(unconditional) distribution of the dependent variable.
The coefficient on the dummy of NSW from an unconditional quantile regression (UQR) gives an
estimation of the impact of NSW on inequality by showing whether the effect of NSW on wages is different at
different points of the distribution. One caveat of the UQR, however, is that it does not allow for a control of
endogeneity in the selection into NSW.

Although there is a higher earnings gap for NSW at the bottom, different country
profiles emerge (see Annex 4.A3, Figure 4.A3.2). In one set of countries including
Australia, Greece and Luxembourg, NSW results in lower wages below the middle of
the distribution, and this wage penalty disappears in the upper deciles, turning into a
wage premium in the top decile. In other countries, including Germany, Hungary, Italy
and Spain, NSW tends to decrease wages in almost all deciles of the distribution, but
the difference decreases monotonically with the upper deciles. In Canada, Ireland and
the United Kingdom, the earnings gap is more pronounced in the bottom 20% to 40%
of the distribution rather than the bottom 10%. Portugal is an exception in that the wage
penalty is fairly small at the bottom of the distribution and more substantial in the upper
middle part.
In general, the earnings gap is more marked in Germany, with NSW lowering log
hourly wages by more than 40% in the lowest decile. On the other hand, in Australia the
earnings gap at the bottom of the distribution is smaller, while there is a significant wage
premium of 20% to NSW at the top.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 169

Figure 4.12. Non-standard work and earnings by decile of hourly wages, OECD-14 average
Panel A. Share of non-standard work in earnings deciles
100%

Interval of the impact of non-standard
work on log hourly wages ranging
between 25% and 75% of values

80%

60%

Country with the highest and
lowest impact on wage
associated with non-standard
work for each decile

Median impact

40%

20%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Earning percentiles

Panel B. Effect of non-standard work on wages by decile
Log hourly wage change

0.25
0.15
0.05
-0.05
-0.15

-0.25
-0.35
-0.45

-0.55

10%

20%

30%

40%

50%

60%

70%

80%

90%

Earning percentiles

Note: The box for each quantile represents the interval of the impact of non-standard work on log hourly wages ranging between
25% and 75% of values, with the black line representing the median impact. The circles represent the country with the highest
and lowest impact on wage associated with NSW for each decile. OECD-14 refer to Australia, Austria, Canada, Germany,
Greece, Hungary, Ireland, Italy, Korea, Luxembourg, Poland, Portugal, Spain, United Kingdom.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour Dynamics
in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Canada Labour Force Survey (LFS, 2013).
12 http://dx.doi.org/10.1787/888933208137

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 170 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

4.5.

The impact of non-standard work on household income and poverty
This section investigates how non-standard work contributes to household earnings
and income inequality. This aspect has remained unexplored in the literature. Previous
work has attempted to make a link between the polarisation of jobs and earnings by
arguing that part of rising household earnings inequality is related to the polarisation
between jobless and job-rich households (see Gregg and Wadsworth, 1996). However,
this approach fails to account for the large part of inequality that is explained by
inequality within working households. Part-time workers and temporary workers tend to
have lower hourly wages and/or annual earnings (see Section 4.4 above), while the
earnings of self-employed individuals are more dispersed, both at the top and at the
bottom of the distribution. In addition, household earnings and income are also influenced
by working hours and months worked during the years for the different workers in the
household. How this translates into household earnings depends on the household
composition of workers. Are non-standard workers the main or even the only earners, or
are they typically the “secondary” earners within a household? If temporary or part-time
workers earn less but supplement the earnings of a main earner with a standard job, an
increased share of them will lead to higher household earnings in households where
previously there was only one earner and smooth the overall distribution. On the other
hand, if non-standard workers are concentrated mainly in households where all earners
are in non-standard jobs, the impact may be different and can increase inequality.

Are low-wage non-standard workers in low-income households?
Do non-standard workers change their relative distributional position when total
household incomes, rather than individual earnings, are considered? In particular, what
proportion of non-standard workers in the bottom part of the individual earnings
distribution remain in the bottom when all other income sources are pooled within the
households?
While the share of non-standard workers is sizeable in many countries, their
contribution to household earnings can be very different. The extent to which nonstandard workers are main or secondary earners in a household with multiple workers has
distributional implications. Figure 4.13 presents the share of non-standard workers as the
main or secondary income earner, with a breakdown by household type (number of adults
and children).17 On average across the countries, just under 50% of all non-standard
workers are the main earners (right panel). The shares are higher for Korea and Greece
(over 62%), but lower for Japan, Luxembourg and Switzerland (35%, 38% and 39%). It is
striking that almost half of these workers (47%) have dependent children. Given that
some non-standard workers, in particular temporary contract workers, tend to earn less
and suffer greater earnings instability than standard workers, individuals living with such
non-standard/main earners are at a greater risk of falling into the bottom part of the
earnings distribution.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 171

Figure 4.13. Share of non-standard workers as the main/secondary income earner in a household
Main earner (%)

% 70

Single
Couple with children
Two adults+
60
50
40

Secondary earner (%)
Single parent
Couple no children

30

20

10

Couple with children

0

0

10

20

Couple no children

30

40

50

Two adults+

60

70 %

AUS
AUT
BEL
CHE
CAN
CZE
DEU
ESP
EST
FRA
GBR
GRC
HUN
ITA
IRL
JPN
KOR
LUX
POL
PRT
SVK
Total

Note: “Children” refers to persons aged 17 or less or young adults (18-24) who were economically inactive and living with at
least one parent. Data for Japan refer to respondent and spouses aged 20-64 and there is no information on earnings for other
household members.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada, Japan Household Panel Survey (JHPS, 2012).
12 http://dx.doi.org/10.1787/888933208147

In addition to the earnings of other household members, non-labour incomes,
including social transfers, need to be considered. Pooling data over 15 EU countries,
Table 4.5 shows in the first column that about 45% of non-standard workers are in the
bottom quintile of individual earnings. The rest of the table presents a matrix showing the
relative distributional positions of non-standard workers within the household context.
The columns refer to quintiles of individual earnings (among workers), and the rows
correspond to quintiles of household equivalised income (for the whole working-age
population). More than one-third of non-standard workers who were in the bottom
quintile of individual earnings remained in the bottom quintile when household
equivalised income was considered; another 24% moved to the following quintile, 20% to
the middle, and the remaining 22% advanced to the top two quintiles. Strong movements
are also found for non-standard workers in other earnings quintiles, as two-thirds to
three-quarters of them are positioned in different quintiles in terms of the household
equivalised income. The exceptions are the richest non-standard workers, those in the top
quintile of the earnings distribution: 90% of them remain in the top two quintiles of the
household income distribution.18

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 172 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Table 4.5. Distributional position of non-standard workers in quintiles of household equivalent income,
by quintile of individual earnings, pooled 15 EU countries, 2012
Quintiles of household equivalent income

Quintiles of
individual earnings

Share of nonstandard workers

1

2

3

4

5

Total

1
2
3
4
5
Total

45.4
24.8
12.9
8.8
8.1
100.0

34.2
17.3
7.9
2.7
0.6
21.1

24.2
22.2
17.8
10.4
2.9
19.9

19.6
24.6
26.3
19.7
8.2
20.8

14.3
22.9
28.4
32.9
19.4
20.3

7.8
13.1
19.6
34.4
68.9
19.7

100
100
100
100
100
100

Note: Quintiles of individual earnings are defined based on all workers, while quintiles of household equivalent income are
defined based on the whole working-age population.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012).
12 http://dx.doi.org/10.1787/888933208999

These average patterns, however, differ markedly by household type: low-earning
non-standard workers are more likely to remain at the bottom of the income distribution if
they live with another non-standard worker rather than with a standard worker. To
illustrate this, Figure 4.14 presents a graphical form of Table 4.5 for two-earner
households with a breakdown by two employment types: households with two nonstandard workers, and mixed households with a standard and a non-standard worker. It
comes as no surprise that non-standard workers who live in a household with other nonstandard workers tend to remain in the same position, especially those in the bottom
earnings quintile: about 42% of them remain in the bottom quintile of household
equivalised income (first bar in Figure 4.14). The comparable figure is much lower (12%)
for those living with a standard worker (i.e. mixed households) (second bar in
Figure 4.14). Similar patterns are also found for non-standard workers in other lowearnings quintiles, that is, the chances of remaining in the second/third quintile (or falling
below) are higher for those in NSW households than those in mixed households. This
suggests that the income inequality impact of non-standard employment, if any, happens
mainly through the increase in NSW households, not through the growth in mixed
households.
Figure 4.14 (Panel B) presents the results by country, including the European countries
from Table 4.5 as well as Australia, Canada and Korea. Non-standard workers in the lowest
earnings quintile seem to fare better in the household income distribution in Japan, Ireland
and Australia, where slightly under one-quarter of these workers remain at the bottom
quintile when all income sources from other household members were pooled.
Interestingly, low-earnings, non-standard workers in those three countries seem to get more
financial support from their household members or have received noticeable non-labour
incomes, as about half of non-standard workers in the bottom earnings quintile in these
countries found themselves in the upper three quintiles of the household income. On the
other hand, low-earnings, non-standard workers in Estonia, Luxembourg and Greece face a
higher risk of low income, as more than 40% of them remain in the lowest quintile of
household income.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 173

Figure 4.14. Distributional position of non-standard workers in household income quintiles,
by quintile of individual earnings, 2012
Panel A. For two-earner non-standard workers and mixed standard and non-standard worker households,
pooled 15 EU-SILC countries
Quintile of household equivalised income
Bottom

2nd

3rd

4th

Top

Two-earner mixed SW/NSW households

Two-earner NSW only households
100%
80%
60%
40%
20%
0%

2nd

Bottom

3rd
Quintile of individual earnings

4th

Top

Panel B. By country
Quintile of household equivalised income

Bottom

2nd

3rd

4th

Top

100%

80%

60%

40%

20%

0%

EST LUX GRC FRA ESP POL ITA CZE HUN CAN PRT SVK KOR AUT GBR DEU CHE BEL AUS IRL JPN
Non-standard workers at the bottom quintile of individual earnings

Total

Note: Quintiles of individual earnings are defined based on all workers, while quintiles of household equivalent income are
defined based on the whole working-age population. “NSW only households” are households with only non-standard workers;
“Mixed households” are households with both standard and non-standard workers. Data for Japan refer to respondent and
spouses aged 20-64 and there is no information on earnings for other household members.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada, Japan Household Panel Survey (JHPS, 2012).
12 http://dx.doi.org/10.1787/888933208158

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 174 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
In spite of changes in the distributional position, households with NSW arrangements
are still more likely to be found at the lower end of the household income distribution than
other working households. Figure 4.15 reports that on average 36% of NSW households are
situated in the bottom income quintile of household equivalised income. While this share is
lower than that of jobless households (54%), it represents more than three times the share of
households with only standard workers. The risk of being in the bottom quintile for
NSW households is highest in Estonia, Luxembourg and the Slovak Republic where over
45% of NSW households fall into the lower part of the income distribution. Working
poverty (which will be discussed below) is a concern when having a working household
member (in a non-standard job) does not improve the position in the income distribution. It
is remarkable that in Greece and Luxembourg the share of NSW households falling into the
bottom quintile is very close to that of jobless households.
Figure 4.15. Household employment patterns and household equivalised income, 2012 or most recent year
Percentage of households in the bottom quintile of the household equivalised income distribution, by household
and employment pattern
%

Non-standard worker households (↘)

Standard worker households

Jobless households

Mixed SW/NSW households

80

70
60
50
40

30
20
10
0

Note: “Standard worker households” refers to households where all adult members (16-64) are in standard work or to
households with the presence of both standard worker(s) and jobless adult member(s). Similarly, “Non-standard worker
households” refers to households where either all adult members are in non-standard work or there are only non-standard
workers and non-working adult members. “Mixed SW/NSW households” refers to households with both standard and nonstandard workers. Jobless households refer to households without any worker during the year. Household incomes are
equivalised by family size. The cut-offs of income quintiles are calculated based on the entire population. Data for Japan refer to
respondent and spouses aged 20-64 and there is no information on earnings for other household members.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada, Japan Household Panel Survey (JHPS, 2012).
12 http://dx.doi.org/10.1787/888933208160

On the other hand, mixed households with both standard and non-standard workers
are better placed in the income distribution. On average their risk of falling into the
bottom of the income distribution is very similar to that of standard worker households:
only about 7% of mixed households are found in the lowest income quintile. There are,
however, some country differences, with the shares ranging from below 2% in Australia
to around 15% in Luxembourg. Mixed SW/NSW households fare slightly better in the
distributional position than SW households in all countries with the exception of Portugal.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 175

How does non-standard work affect household earnings and income inequality?
Do NSW arrangements widen the distribution of household earnings? Since almost half
of all non-standard workers fall in the lowest earnings, a larger (and in many countries
growing) share of non-standard workers may increase overall household earnings
inequality. On the other hand, nearly half of non-standard workers live in a household with
a standard worker, and these are less likely to be at the lower end of the distribution.
The analysis below presents earnings inequality by successively introducing
households with different employment patterns in the calculation of inequality measured
by the Gini coefficient of equivalised household earnings19 (Figure 4.16). The first bar
shows the level of earnings inequality among households that comprise only standard
workers (including those living with non-workers). The second and third bars then show
the level of inequality by successively adding households with mixed standard/nonstandard workers and households with only non-standard workers, respectively. Finally,
the triangle represents the estimates of household earnings inequality with the inclusion
of households where no-one works.
Figure 4.16. Gini coefficient of equivalised household earnings for households with different employment
patterns, 2012 or most recent year
Standard worker households (↗)

+ mixed SW/NSW households

+ Non-standard worker households

+ jobless households

0.55
0.50
0.45
0.40
0.35
0.30
0.25

0.20
0.15
0.10

Note: The Gini coefficient takes values between 0 for a perfectly equal income distribution where every person has the same
income, and 1 which refers to a situation of maximum inequality where all income goes to one person. “Standard worker
households” refers to households where all adult members (16-64) are in standard work or to households with the presence of
both standard worker(s) and jobless adult member(s). Similarly, “Non-standard worker households” refers to households where
either all adult members are in non-standard work or there are only non-standard workers and non-working adult members.
“Mixed SW/NSW households” refers to households with both standard and non-standard workers. Jobless households refer to
households without any worker during the year.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010).
12 http://dx.doi.org/10.1787/888933208173

Figure 4.16 reveals that including earnings from households with non-standard
workers has two opposing effects. On the one hand, when mixed SW/NSW households
were added, household earnings inequality declines by one percentage point, from 0.32
to 0.31 on average (more than two points in Austria and Luxembourg), compared with
inequality among households with only standard workers. The lower level of inequality
for mixed worker households may be related to the fact that such households have at least
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 176 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
two workers, while standard work households include both single-earner households
(households with SW and jobless members) and multiple earners in standard work. On
the other hand, when households with only non-standard workers were added, inequality
increases significantly across the board, by about four percentage points, to a Gini
coefficient of 0.35, on average. The dis-equalising effect of non-standard work at the
household level is particularly pronounced in Ireland, Greece and Spain, where adding
households with non-standard workers increases household earnings inequality by
7-8 percentage points. Finally, inequality increases more markedly when jobless
households are included (average Gini coefficient of 0.41). The largest increases (more
than 10 points in Gini) were found in Belgium and Ireland.
The extent to which non-standard employment affects the distribution of household
income depends not only on the earnings level but also on the non-labour incomes
received in these households. Inequality would widen if households with NSW
arrangements also received less income from other sources, including social transfers,
compared to standard employment households. This, however, does not seem to be the
case, as NSW households receive a considerable share of their income in the form of
transfers (see Annex 4.A3, Table 4.A3.1), which in general have an equalising effect on
the income distribution (OECD, 2011, Chapters 6 and 7). To identify the impact of nonstandard employment on household income inequality, a decomposition is performed by
income sources (Figure 4.17), breaking down household incomes into four main sources
(i.e. earnings, capital, public transfers and taxes).20
Figure 4.17. Breakdown of household equivalent income inequality (Gini) by income source, 2012
Standard work

Non-standard work

Unknown earnings

Investment

Net transfers

Total Gini (↘)

0.6
0.5
0.4
0.3
0.2
0.1

0
-0.1
-0.2
-0.3

Note: Non-standard work sources refer to earnings from full-time temporary contracts, part-time jobs as well as self-employment.
Unknown earnings are labour incomes for which the source (i.e. from SW or NSW work) cannot be identified. Benefits include all
transfers from government. All income sources are equivalised by family size. The sample refers to working-age households. Note
that for France, Hungary and the United Kingdom there are some differences between the current results and the Gini coefficients
reported in the OECD Income Distribution Database, for which information is collected through questionnaires.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Survey of Labour and Income Dynamics (SLID, 2010).
12 http://dx.doi.org/10.1787/888933208182

On average across countries, earnings from non-standard work contributed about 20%
to cross-sectional household income inequality. The contribution is higher for Australia
and Belgium, reaching close to 30%, possibly due to their relatively higher share of
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 177

NSW earnings in income (28% and 26%, respectively) and the high correlation of
NSW earnings with overall income. Earnings from non-standard work tend to be more
dispersed than earnings from standard jobs, as non-standard jobs are more heterogenous
(Annex 4.A3, Table 4.A3.2): the factor dispersion (in terms of Gini coefficient) for NSW
earnings is about 0.8, compared with 0.55 for SW earnings. In addition to being more
unequally distributed, earnings from NSW are more concentrated in the households at the
lower part of the income distribution, as the factor correlation between NSW earnings and
total household income is low (on average about 0.33).

Non-standard work and poverty
Non-standard employment may have a strong impact on poverty, as many
NSW households are situated in the lower part of the income distribution. Previous
OECD work has shown that access to a job is a major factor limiting the risk of poverty,
but being employed per se is often not sufficient to escape poverty. While the poverty
rate among jobless households is more than double the rate observed among working
households, there are also significant in-work poverty risks in many countries. Indeed,
7% of individuals living in households with at least one worker are poor in the OECD,
and the working poor comprise more than 60% of all the poor individuals of working age
(OECD, 2008a).
Are non-standard workers and their households at higher risk of poverty than standard
workers? To address this issue, a conventional poverty threshold is used with 50% of the
median equivalised household disposable income.21 The analysis presents the proportion of
households falling below the poverty line, for various household work types (Figure 4.18).
This first confirms the importance of employment for protecting against poverty and second
highlights the role of employment types. When averaged over the OECD countries for
which data are available, jobless households have the highest poverty rate, at slightly over
40%, with over 50% in Germany and Australia, and 60% in Canada. At the same time,
households with only non-standard workers also face a high risk of poverty, at around 22%
on average across countries, while mixed households with both standard and non-standard
workers have lower poverty rates (2-4%). This means that the risk of poverty depends on
the combination of the type of employment with household composition, i.e. whether nonstandard workers live with other non-standard workers (or jobless household members), or
with standard workers. NSW poverty rates range from close to or above 30% in Canada,
Greece, Portugal, Estonia and Spain to 12% or below in Belgium and Ireland.22
One of the striking findings of Figure 4.18 is that in some countries (e.g. Greece,
Luxembourg) the poverty rate for NSW households (based on their net income) is very
similar to that for jobless households. A possible reason for this is that NSW households
on average pay more taxes or have less access to benefits than their jobless counterparts
(see Annex 4.A3, Table 4.A3.1). In other words, low-paid NSW households have high net
effective tax rates. This may create a disincentive to work, especially when non-standard
jobs are the only options to find work.
Working-poor households account for around half of all poor households of working
age (Figure 4.19), with most of them in NSW households. Indeed, among the workingpoor households, about 60% are concentrated in households with non-standard
employment. NSW households represent an important fraction of the working poor
particularly in Australia, Ireland and the southern European countries (except Portugal).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 178 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.18. Poverty rates of households for various household employment patterns,
2012 or most recent year
Non-standard work (↘)

Standard work

Mixed SW/NSW

Jobless

%
70
60
50
40
30
20

10
0

Note: The poverty line is calculated based on half of the median equivalised household income calculated for the entire population.
“Standard work” refers to households where all adult members (16-64) are in standard work or to households with the presence of both
standard worker(s) and jobless adult member(s). Similarly, “Non-standard work” refers to households where either all adult members
are in non-standard work or there are only non-standard workers and non-working adult members. “Mixed SW/NSW” refers to
households with both standard and non-standard workers. Jobless households refer to households without any worker during the year.
Data for Japan refer to respondent and spouses aged 20-64 and there is no information on earnings for other household members.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada, Japan Household Panel Survey (JHPS, 2012).
12 http://dx.doi.org/10.1787/888933208192

Figure 4.19. Distribution of poor households by household employment pattern, 2012 or most recent year
Non-standard work (↘)

Standard work

Mixed SW/NSW

Jobless

100%
80%
60%
40%
20%

0%

Note: The poverty line is calculated based on half of the median equivalised household income calculated for the entire population.
“Standard work” refers to households where all adult members (16-64) are in standard work or to households with the presence of both
standard worker(s) and jobless adult member(s). Similarly, “Non-standard work” refers to households where either all adult members
are in non-standard work or there are only non-standard workers and non-working adult members. “Mixed SW/NSW” refers to
households with both standard and non-standard workers. Jobless households refer to households without any worker during the year.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada.
12 http://dx.doi.org/10.1787/888933208204

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 179

To what extent does redistribution through taxes and benefits reduce the poverty risk
for non-standard worker households? Figure 4.20 presents the impact of taxes and
transfers in reducing poverty, comparing NSW with jobless households. Households with
non-standard workers see their poverty rates reduced by a third, from 34% to 22% on
average. In comparison, taxes and transfers have a much larger impact on jobless
households, halving their poverty rate from 76% to just 39% after taxes and transfers. The
poverty-reducing impact of taxes and transfers for NSW households tends to be
particularly large in Belgium, Ireland and the United Kingdom. For instance,
NSW households in the United Kingdom have one of the highest incidences of
poverty (50%) in terms of market income. Redistribution lowers their poverty risk to
20%, below the OECD average of 22%. The strong poverty-reducing effect in the United
Kingdom (and other countries) reflects certain tax/benefit policies that are especially
helpful to non-standard workers and their families.
By contrast, taxes and transfers have no impact on poverty reduction for
NSW households in Korea and Italy. In Greece, poverty among NSW households even
increases after redistribution. In other southern European countries, Portugal and Spain,
poverty reduction for NSW households was also modest.
Figure 4.20. The impact of taxes and transfers on poverty reduction
Percentage reduction in poverty due to tax/transfer
Panel A. Non-standard worker households

Panel B. Jobless households

%
20

%
20

0

0

-20

-20

-40

-40

-60

-60

-80

-80

-100

-100

Note: The poverty line is half of the median equivalised household income calculated for the entire population. Figures represent
the difference between the poverty rate for disposable income and for market income in percentage points. For Korea market
income refers to after tax before public social and government tranfers and is not comparable with the other countries.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010) for Canada.
12 http://dx.doi.org/10.1787/888933208213

4.6.

The role of taxes and benefits with regard to non-standard work
Personal income taxes and cash social transfers are not neutral to the form of work.
Differences in treatment result directly or indirectly from how policies are designed.
Some policies are directly targeted to particular types of work by differentiating the

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 180 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
“eligibility” of taxes and benefits for different types of work. For example, in many
countries self-employed workers are not eligible for unemployment insurance benefits.
Similarly, the rules for social insurance contributions for self-employed workers are
different from those for employees. Some differences in taxes and benefits between
different types of workers are indirect, in the sense that they result from how different
policies interact in the context of non-standard work. For example, means-tested benefits
may be more generous for part-time workers as a result of their lower earnings, which is
in turn due to working fewer hours.
This section analyses how the role of taxes and benefits of workers in non-standard
jobs differs with respect to those in standard jobs. The analysis looks at qualitative factors
(statutory differences) and quantitative indicators (net benefits, adequacy and work
incentives). The quantitative analysis is carried out for part-time and self-employed
workers using the results of simulations with tax-benefit models.23 The analysis of
temporary workers is limited to statutory differences.

Statutory differences between standard and non-standard work
People in NSW may be subject to different tax and benefit rules. In general, selfemployed workers are more likely to experience different statutory treatment than people
in other forms of NSW (Table 4.6). In most cases, the benefit rules for part-time and
temporary workers are the same as for standard workers. In most countries,
unemployment and work injury benefits for the self-employed are different than for
standard workers. Sickness and maternity, old-age, disability and survivors benefits are
also different in some countries. Even family benefits are different for self-employed
workers in Belgium and Italy (where benefit rules are different) and Chile, Greece and
Mexico (where some family benefits are not available for self-employed workers).
The most common difference with standard workers is the exclusion of workers in
non-standard work from benefits related to unemployment and work injury. In 19 out of
the 34 OECD countries self-employed workers are not eligible for unemployment
benefits. In three countries, some part-time workers are not eligible for unemployment
benefits.24 In ten countries, self-employed workers are not eligible for work injury
benefits.
The second most common difference concerns variations in the content of the benefits
(e.g. the coverage or payment level). For example, in the United Kingdom, self-employed
workers are not eligible for statutory sick pay (which is paid by the employer) but for
employment and support allowance, which is less generous. Benefit content differences
are common for self-employed workers, particularly regarding old-age, disability and
survivor, and sickness and maternity benefits. In a few countries, the content of benefits
also differs for part-time and temporary workers. For example, some part-time workers in
Denmark (working less than nine hours per week) and temporary workers in Canada
(casual and seasonal agricultural workers) are not eligible for the earnings-related
pension. Finally, in some countries, the enrolment of the self-employed in some benefits
is optional. These optional schemes are particularly common for insurance benefits
related to work injury, sickness/maternity, unemployment and old-age/disability/survivor.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 181

Table 4.6. Statutory benefit differences between non-standard and non-standard work, by benefit, 20101, 2
Part-time 3
Australia
Austria
Belgium
Canada
Chile
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Israel
Italy
Japan
Korea
Luxembourg
Mexico
Netherlands
New Zealand
Norway
Poland
Portugal
Slovak Republic
Slovenia
Spain
Sweden
Switzerland
Turkey
United Kingdom
United States

ODS
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

SM
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

WI
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Temporary worker
UB
0
0
0
0
0
0
1
0
0
0
0
0
0
0
-2
0
0
-2
-2
0
4

0
0
0
0
0
0
0
0
0
0
0
0
0

FB
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

ODS
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

SM
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0

WI
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-2

UB
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
1
0
0
0
0
4

0
0
0
0
0
0
0
0
0
0
0
0
-2

Self-employed
FB
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

ODS
-2
1
1
0
-1
0
-1
0
0
1
1
-2
0
0
1
0
1
-2
0
0
-1
0
0
0
0
1
1
0
1
0
-1
1
1
1

SM
0
1
1
-1
1
1
0
0
0
1
-2
-2
0
0
1
0
1
1
0
1
-1
0
0
1
-1
-1
0
0
1
0
0
0
1
0

WI
-1
0
-2
-2
0
-2
-1
0
-1
1
-2
-2
0
0
-2
0
0
-2
-1
0
-1
0
0
-1
0
0
-2
0
-1
0
-1
1
-2
-2

UB
0
-2
-2
-2
-2
0
0
0
1
-2
-1
-2
0
0
-2
-2
-2
-2
-1
1
4

-2
0
-2
-2
-2
-1
-1
-2
-2
-2
-2
1
-2

FB
0
0
1
0
-2
0
0
0
0
0
0
-2
0
0
0
0
1
0
0
0
-2
0
0
0
0
0
0
0
0
0
0
0
0
0

1. FB: Family benefits, ODS: Old age, disability and survivors, SM: Sickness and maternity, UB: Unemployment, WI: Work
injury.
2. Colour code: “dark grey”: no benefit, “light grey”: optional enrolment, “blue”: different rules from standard workers, “white”:
same rules as the general scheme.
3. Part-time workers are excluded if working less than nine hours a week.
4. There is no unemployment benefit in Mexico. Labour law requires employers to pay dismissed employees a lump sum.
Source: Social Security Administration (2010), Social Security Programs Throughout the World: Asia and the Pacific,
Government Printing Office; Social Security Administration. (2010), Social Security Programs Throughout the World: Europe,
Government Printing Office; Social Security Administration (2011), Social Security Programs Throughout the World: The
Americas, Government Printing Office.

Figure 4.21 shows that, in general, the amounts of taxes/benefits paid/received by
part-time workers are similar, if not identical, to those paid/received by standard workers.
In contrast, the amounts of taxes/benefits paid/received by self-employed workers are
usually substantially different. Whether self-employed workers do better or worse than
standard workers depends on how employer social insurance contributions are accounted.
Legally, employer social insurance contributions are paid by the employer. However, the
actual incidence of such contributions may effectively fall on the employer (lower profits)
or be transferred to the employee (lower wages) or to consumers (higher prices).25 Here,
the scenarios in which employer contributions fall either on the employer or the employee
are assessed. Generally, self-employed contributions are larger than employee
contributions but smaller than the sum of employee and employer contributions. Hence,

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 182 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
self-employed workers tend to do worse than employees if the actual incidence of
employer contributions falls on employers. Conversely, self-employed workers tend to do
better than employees if the actual incidence of employer contributions falls on
employees. However, there are some significant exceptions. Independently of the
incidence of employer contributions, the self-employed fare worse than employees in
Hungary and Luxembourg, and better than employees in Portugal, Germany, Austria, the
Netherlands and the United Kingdom.
Figure 4.21. Differences in tax-benefit amounts between part-time/self-employed workers and workers
in standard jobs, 2010
Part-time employee

Self-employment (employers' social contributions included)

Self-employment (employers' social contributions excluded)

CZE

FIN

AUS
AUT
BEL
CAN
CHE

ESP
FRA

HUN

LUX

POL
SVK
SWE
-40%

GBR
IRL

ITA

-50%

DEU
DNK

-30%

ISL
ISR

JPN
KOR

NLD
NOR
NZL
PRT
SVN
USA

-20%

-10%

0%

10%

20%

30%

40%

Note:
Differences expressed as percentage of average wages (AW) in the country.
Part-time results are for employees working between 16 and 30 hours per week and paid the average hourly wage in the country,
Self-employed results are for people working full-time (40 hours per week) and earning between 40% and 160% of the average
wage in the country (self-employment estimates are available only for EU countries). Four sets of typical families are
considered: single adult living alone, single parent with children, single-earner couples with and without children.
Source: OECD tax-benefit models and EUROMOD.
12 http://dx.doi.org/10.1787/888933208228

Adequacy of net benefits for low-earning workers in non-standard jobs
As discussed in Section 4.5, workers in non-standard jobs face a higher risk of being
in poverty as their earnings tend to be lower. Taxes and benefits that are designed
appropriately can reduce this risk by increasing the income that families effectively take
home. The analysis below measures the adequacy of tax-benefit systems in protecting
families relying on low earnings from part-time or self-employment. In either case,
family earnings before taxes and benefits are equivalent to those received by a person
working 20 hours per week at the country’s average hourly wage. Adequacy is measured
by comparing equivalised family income after taxes and benefits to the national poverty
line, defined as 50% of median disposable income.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 183

Comparisons between employment and self-employment incomes are not
straightforward in practice. Self-employed income is often unpredictable as it tends to
fluctuate across time considerably more than wages and salaries. Also, evidence shows
that self-employed workers underreport income to the tax authorities more than
employees do (Feldman and Slemrod, 2007). Bearing these differences in mind, the
analysis measures the amount of taxes and benefits while assuming a scenario in which
self-employed incomes are constant across the year and accurately reported to the tax and
benefit authorities.
In all the countries analysed, workers working half-time (20 hours per week) and
earning the average hourly wage would earn an income (before taxes and benefits) above
the poverty line, if living alone (Figure 4.22, Panel A). If they live in families and are the
single earner, in several countries the family income would be below the poverty line.
The proportion of countries with families in poverty increases with the number of family
members. Only in Germany, the Netherlands and the United Kingdom would the income
of a couple with two children earning half the monthly average wage lift them above the
poverty line.
Taxes and benefits considerably alter these results for families with part-time
workers, redistributing income from smaller to larger families. Bearing in mind that the
results are illustrative of a specific wage level, single part-timers pay more in taxes than
they receive in benefits – in Slovenia, the fall in disposable income brings singles below
the poverty line. In some countries, including Ireland, New Zealand and Japan, couples
without children pay less in taxes than they receive in benefits, but in most countries the
opposite is the case, and in some countries (Austria, Italy, Sweden and Switzerland),
disposable income falls below the poverty line. In most countries, single parents who
have two children and work part-time receive more in benefits than they pay in taxes, and
in many of these countries the resulting increase in disposable income brings the
household above the poverty line. With the exception of Korea, Spain and Switzerland,
couples with children are net beneficiaries of taxes and benefits.
In several OECD countries, taxes and benefits are not sufficient to prevent the income
of low-earning families headed by self-employed workers from falling below the poverty
line. In comparison to part-time employees, a larger number of families are left in
poverty. In the case of families composed by a single individual, the amount of taxes
exceed the benefits in all the countries analysed, and in several of them (Luxembourg,
Poland, the Slovak Republic, Slovenia and Sweden) family disposable income falls below
the poverty line.
In most countries, single-earner couples without children also pay more in taxes than
they receive in benefits. In 16 countries, the income of single-earner couples without
children is below the poverty line, and in ten of these this is a direct effect of negative net
benefits. Only in Ireland the income of single-earner couples rises above the poverty line
after accounting for taxes and benefits.
As with part-time workers, net benefits tend to be more generous for families with
children. In 16 countries, net benefits are positive for single parents with children. In six
of these countries, this results in family disposable income rising above the poverty line.
Among couples with children, net benefits are positive in 17 countries, lifting income
above the poverty line in six of these. Yet in 14 countries, family income is below the
poverty line, in the case of the Netherlands as the direct result of negative net benefits.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 184 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.22. Tax-benefit adequacy
Part-time workers (black), self-employed workers (blue), earning half the average wage, before (bars)
and after (arrows) taxes and benefits, 2010
Panel A. Single

Panel B. Single parent

105%

105%

95%

95%

85%

85%

75%

75%

65%

65%

55%

Poverty line

55%

45%

45%

35%

35%

25%

25%

Panel C. Single-earner couple

105%

Poverty line

Panel D. Single-earner couple with two children

105%

95%

95%

85%

85%

75%

75%

65%

65%

55%

Poverty line

55%
Poverty line

45%

45%

35%

35%

25%

25%

Note:
The results are expressed as a percentage of the average equivalised household disposable income in the country.
Source: OECD tax-benefit models and EUROMOD.
12 http://dx.doi.org/10.1787/888933208238

Work incentives
How do taxes and benefits affect the work incentives of workers in non-standard
jobs? Do they encourage or deter workers in non-standard jobs from increasing their
working hours or moving into standard jobs? Do they make work pay or do they reduce
the financial incentives to work harder or even to work at all?
This sub-section addresses these questions by using the results of simulations
computed with the OECD/EC tax-benefit models (for part-time workers) and
EUROMOD (for self-employed workers).26 Following the literature on labour supply
(Heckman, 1974; Blundell and Macurdy, 1999; Brewer et al., 2010; and Blundell et al.,
2011), a distinction is made between the intensive margin of labour supply, which
measures incentives to increase the intensity of work by those at work (i.e. variations in
the number of hours of work) and the extensive margin, which measures “qualitative
shifts” from out-of-work to in-work or from NSW to standard work.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 185

Incentives from inactivity to part-time work
Do taxes and benefits provide incentives for individuals out of work to move into
NSW? This section assesses these incentives by measuring participation tax rates (PTR).
Simulations assume that out-of-work people are “inactive”, in the sense that they are not
eligible for unemployment benefit but may receive social assistance and other benefits, if
they are entitled. In the situation of NSW, part-time and self-employed workers are
assumed to work half-time at the average hourly wage.
In many OECD countries the income incentives to move from inactivity to part-time
employment are rather small (Figure 4.23). This result is in line with findings from
previous studies (OECD, 2007; OECD, 2009; OECD, 2010). On average, the
participation tax rate (PTR) is 70%. Put another way, 70% of the earnings obtained in
moving from inactivity (in receipt of social assistance) to part-time employment are
“taken away” due to higher taxes and lower benefits. Across countries, PTRs range from
less than 30% in Italy27 and the United States to more than 90% in Switzerland and
Denmark. In 15 countries the PTRs exceed 80%.
The main cause of such high rates is the reduction or removal of social assistance
benefits. On average, 45 percentage points of PTR are due to social assistance benefits. In
some countries this effect is partly compensated by in-work benefits (or other
employment-related instruments) that increase the financial returns of work (Pearson and
Scarpetta, 2000; Immervoll and Pearson, 2009). In Ireland and the United States, in-work
benefits reduce the PTRs by 35% and 20%, respectively. Housing benefits may also play
a role in reducing the financial rewards of part-time work. With a few exceptions, social
contributions play a larger role than personal income tax in increasing PTRs. Through
progressive tax rates, exemption limits and other deductions, income taxes have more
scope than social contributions, which usually rely solely on contribution bases and fixed
rates, to modulate the tax burden on low-earning workers.
Income incentives to move from inactivity to self-employment on a part-time basis
are also limited in most countries analysed. In Germany, Hungary, Portugal, Denmark
and Luxembourg, people in inactivity have little income incentive to take up selfemployed work, as 90% or more of their earnings are “taken away” by lower benefits or
higher taxes or contributions. Participation tax rates for self-employment are higher than
for employees, especially in Hungary, Ireland and Sweden, and on average by onequarter. This is the case in all but four of the analysed countries.
Social benefits are the main driver of PTRs among self-employed workers; on
average more than half of the rate is due to social benefits. About one-third of PTRs are
due to social contributions. Also, predictably, social contributions are the main source of
differences between the PTRs of the self-employed and employees.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 186 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.23. Participation tax rates for part-time workers
Participation tax rates (PTRs) for inactive persons who are receiving social assistance and move into part-time work,
decomposed by taxes and benefits, 2010
Housing benefits

Family benefits

In-work benefits

Social assistance

Social contributions

Income taxes

PTRs for inactive persons (0-20) (↗)
140%
120%
100%
80%

60%
40%
20%
0%
-20%
-40%
-60%

Note:
PTRs for inactive persons (0-20): PTRs for inactive persons (working 0 hour) receiving social assistance and moving into parttime work (working 20 hours).
The results are computed as the average of four sets of typical families: single adult living alone, single-earner couple without
children, single parent with two children, single-earner couple with two children.
Source: OECD tax-benefit models.
12 http://dx.doi.org/10.1787/888933208240

Figure 4.24. Participation tax rates for self-employed workers
Participation tax rates (PTRs) for inactive persons receiving social assistance and moving into self-employed work,
decomposed by taxes and benefits, 2010
Benefits

Social contributions

Income taxes

PTR of self-employed workers (↗)

PTR of equivalent standard workers

120%
100%
80%

60%
40%
20%
0%
-20%

Note:
The results are computed as the average of four sets of typical families: single adult living alone, single-earner couple without
children, single parent with two children, single-earner couple with two children.
Source: EUROMOD, Tax-benefit microsimulation model for the European Union.
12 http://dx.doi.org/10.1787/888933208259

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 187

Incentives for moving from part-time to full-time work
This sub-section assesses the incentives that non-standard workers would have to
move into standard jobs. The indicator used to measure such incentives is referred to here
as the transition tax rate (TTR) and calculates how much of the earnings increase is
“taken away” through higher taxes and lower benefits.28
The income incentives to move from part-time to full-time employment tend to be
higher than those for moving from inactivity to part-time employment. On average, the
transition tax rate (TTR) for moving from working 20 hours per week to working
40 hours per week is 48%. This ranges from less than one-third in Korea, Spain and
Portugal to two-thirds or more in Denmark, the United Kingdom, Ireland, Japan and the
Netherlands.
Personal income taxes are the main driver, and social contributions the second largest,
of TTRs, accounting for 19 and 12 percentage points, respectively, on average. The role
of social benefits is scattered. Housing benefits produce rates that are well above average
in Japan and the United Kingdom. Family benefits considerably reduce the work
incentives of families with children in English-speaking countries (except New Zealand
and the United States). Social assistance benefits, overall, play a smaller role than in the
transition from inactivity to part-time employment. In-work benefits have different effects
across countries. In Ireland and the United States, in-work benefits increase the TTRs, as
they are withdrawn from workers with higher earnings. On the other hand, in New
Zealand and Sweden, the TTRs are negative, thus rewarding transitions from part-time to
full-time work.
Figure 4.25. Transition tax rates from part-time to full-time employment
Transition tax rates (TTRs) for part-time employee to move to full-time employment, decomposed by taxes and benefits, 2010
Housing benefits
In-work benefits
Social contributions
TTRs for part-time employee (20-40 hours) (↗)

Family benefits
Social assistance
Income tax

80%
70%
60%
50%
40%
30%
20%
10%
0%

-10%
-20%

Note:
TTRs for part-time employee (working 20 hours) moving into full-time work (working 40 hours).
The results are computed as the average of four sets of typical families: single adult living alone, single-earner couple without
children, single parent with two children, single-earner couple with two children.
Source: OECD tax-benefit models.
12 http://dx.doi.org/10.1787/888933208262

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 188 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Self-employed workers face lower income disincentives to move from part-time work
to full-time work than from inactivity to part-time work. On average, less than half of the
earnings increase due to working full-time instead of half-time is “taken away” with
lower benefits or higher taxes and contributions. In 13 out of 19 countries, families keep
at least half of the increase in earnings, and in virtually all countries they keep at least
40%. The rates for self-employed workers are, on average, similar to those faced by
employees, although there are significant differences across countries.
Personal income taxes, social contributions and social benefits have a similar impact
on transition tax rates when all analysed countries are taken as a whole. Nevertheless,
there are important differences in the impact of social contributions across countries, in
particular between countries with high rates of self-employed contributions and those
where self-employed contributions are set at a fixed amount.
Figure 4.26. Transition tax rates from part-time to full-time self-employment
Transition tax rates (TTRs) for self-employed workers to move to full-time work, decomposed by taxes and benefits, 2010
Benefits

Social contributions

Income taxes

20-40: TTR of self-employed worker (↗)

SW: TTR of equivalent standard workers

90%
80%
70%
60%
50%

40%
30%
20%
10%

0%

Note:
The results are computed as the average of four sets of typical families: single adult living alone, single-earner couple without
children, single parent with two children, single-earner couple with two children.
Source: EUROMOD, Tax-benefit microsimulation model for the European Union.
12 http://dx.doi.org/10.1787/888933208274

4.7.

Conclusion
The concept of NSW is a “fuzzy” one. As a benchmark, this chapter uses a broad
definition of atypical or NSW arrangements that, following international convention,
includes any temporary, part-time and own-account self-employment work: taken
together, this kind of employment comprises a sizeable share (33%) of total employment
in the OECD, shared roughly equally between temporary jobs, permanent part-time jobs
and self-employment. The shares of non-standard employment range from a low of under
20% in some eastern European countries to 46% or more in the Netherlands and
Switzerland. Part-time employment is particularly frequent in the Netherlands,
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 189

Switzerland and the Nordic countries. Self-employment is more important in Greece, the
Czech and Slovak Republics and Turkey. Temporary employment is largest in relative
size in Poland, Portugal, Korea and Spain.
One-quarter of men but 40% of women are in non-standard employment, mainly due
to strong gender imbalances in part-time work. Youth and workers with a lower level of
education are over-represented in NSW – close to half of temporary workers are under
30 years of age, and the incidence of temporary employment is 30% higher for those with
lower level of education than for those with medium education. Non-standard workers are
also more likely to be found in small firms. Half of all non-standard workers are the main
breadwinners in a household, and a large majority of them live in two-or-more-person
households, often with children.
More than 40% of the 17% increase in the number of employed people between 1995
and 2007 was attributable to non-standard jobs. In the six years since the global economic
crisis, standard jobs were destroyed while part-time employment continued to increase.
There has been strong growth in part-time permanent jobs over the long-term in most
OECD countries, and some saw a doubling in their share. Temporary employment also
increased in more than three-quarters of the OECD, especially in Poland and
southern Europe. On the other hand, self-employment either declined or was stable in
most countries.
Jobs are becoming polarised. The tasks-based approach, for instance, shows that the
employment shares of routine-task jobs, traditionally composed of middle-skill workers,
have declined from 53% to 41% between 1995 and 2010. At the same time, there has
been a large increase in the employment share of abstract (from 28% to 38%), and to a
lesser extent, non-routine manual jobs (from 18% to 21%). Similarly, the jobs-based
approach, which that takes into consideration a particular occupation within a particular
industry, finds evidence for job polarisation by showing a U-shaped pattern of
employment shifts across job deciles in more than half of the countries. The shift towards
job polarisation is closely related to the development of non-standard employment: much
of the decline in middle-skill employment is due to a decrease in standard work contracts,
while growth in high-skill and low-skill jobs is mainly associated with non-standard
employment.
Having a non-standard job improves the chances of advancing into a permanent job
for only a few. In particular, in most countries those on temporary contracts are more
likely to move into standard employment than those who were unemployed, while being
in part-time or self-employment in general does not improve the prospects of standard
work compared with the unemployed. This suggests some evidence of stepping-stone
effects for temporary workers, at least in the short run. At the same time, non-standard
workers face higher labour market insecurity and are less likely to receive on-the-job
training. In addition, temporary workers face significantly higher job strain, while this is
not the case for part-time workers.
NSW is associated with wage penalties. In most countries, temporary workers earn
significantly less than otherwise comparable workers in standard employment, controlling
for observed and unobserved individual characteristics. The wage penalty is slightly
higher for young temporary workers. Part-timers, mainly those on temporary contracts,
also tend to earn lower hourly wages than their standard full-time counterparts, while no
obvious wage penalty is found for those in permanent jobs. In addition, in many cases
temporary workers with no change in contract type tend to suffer greater earnings
instability.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 190 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Non-standard workers are more likely to be found in the lower part of the earnings
distribution, particularly in the lowest three deciles, and an increase in the share of nonstandard workers is associated with substantial wage penalties of 20% at the bottom of
the distribution (the bottom 40%) but not at the top, thereby increasing earnings
inequality. Adding households where non-standard workers are living to standard-worker
households increases household earnings inequality by 3 Gini points. On average across
countries, this contributes about 20% to cross-sectional household income inequality.
This contribution is higher in Australia and Austria (around 30%) and lowest in Estonia
(below 5%).
Non-standard workers are also more likely to live in lower income households. The
household constellation matters, however, a lot: low-earning non-standard workers are
much less likely to leave the bottom of the income distribution if they live with another
non-standard worker rather than with a standard worker.
More than half of all poor working-age households are households with at least one
worker. Among these “working poor” households, a large majority (60%) are made up of
NSW households. The average poverty rate for households where all earnings are drawn
from NSW is 21% while it is 2.5% for households where earnings from NSW are pooled
with earnings from standard work. Public transfers and taxes reduce poverty risks among
households comprising non-standard workers by some 35%.
There are statutory differences regarding the “access” and content of taxes and
benefits (e.g. the coverage or level) for workers in non-standard jobs with respect to
standard workers. The differences are larger and more widespread for self-employed
workers. The most common differences include the exclusion of self-employed workers
from unemployment benefits, and no or non-compulsory eligibility for work injury
benefits as well as differences in the rules on sickness and maternity benefits. With the
exception of a handful of countries (Ireland, Italy, New Zealand and the United
Kingdom), the effective differences in the content and extent of taxes and benefits for
part-time workers are related more to their particular circumstances (e.g. lower earnings
due to fewer hours of work) than to structural differences in policy rules.
In most countries, taxes and benefits reduce in-work poverty gaps for families relying
on earnings from non-standard jobs. Under similar circumstances, taxes and benefits are
more effective in reducing the poverty gap for part-time workers than for self-employed
workers. Taxes and benefits also have a considerable effect on the work incentives of
workers in non-standard jobs. On average, the results suggest that taxes and benefits
generate higher hurdles to moving from inactivity to part-time work than to increasing
work intensity or to moving from part-time to full-time work.
All in all, the analysis in this chapter suggests that the rise in NSW arrangements and
job polarisation have contributed to aggregate employment growth in the past, but have
also increased both individual wage and household income inequality. Tax and benefit
reforms therefore need to be focused on preventing in-work poverty among low-earning
households with non-standard workers while providing sufficient incentives to take up
and increase work efforts, and active labour market policies need to be designed to raise
the earnings potential of non-standard workers, especially the young and people with less
education.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 191

Notes
1.

Employers are excluded from the analysis since transitions between employers and
standard workers are likely to be small and employers differ from other workers in
their remuneration (receiving earnings as well as business income). In the OECD they
represent an average of 4% of total employment for the working-age population.

2.

Student workers and apprentices are excluded from the analysis, as they may increase
the share of part-time workers and temporary workers. They represent on average 2%
of total employment.

3.

While high levels of PMR could be detrimental to business activities, regulations can
be used to protect small-sized firms from large-sized competitors (Torrini, 2005).

4.

A first method to classify the skill level of a job follows a simple “task” approach to
classify skills into three broad categories based on the nature of the job task (i.e.
abstract, routine and non-routine manual) following Autor and Dorn (2013) among
others.

5.

A second method to classify the skill level of a job is to look at employment shifts
using the “job-based” approach – based on Eurofound (2008, 2012) – where a job is
defined as a particular occupation in a particular industry, and the skill of a job is
measured by the median hourly wages of workers within the job cell.

6.

The recent recession (2007-10) also plays an important role in reshaping the overall
changes in employment structure for some countries. In general, the observed
employment adjustments were amplified in the downturn.

7.

Some noticeable examples include Belgium, France, Germany and the Netherlands.

8.

The influx of migrant workers can also change the structure of employment as
foreign-born workers are more prevalent in the highest and the lowest quintile of
earnings and are often characterised by having NSW contracts.

9.

EU-SILC and income data for European countries are used in this and the subsequent
section since EU-LFS has no information on wages and income.

10.

In the presence of asymmetrical information, firms cannot discriminate between highand low-productivity workers. Therefore, they may use non-standard wage contracts
(by paying low initial wages) to create a probationary stage during which they can
evaluate workers’ performance. But another route is to offer efficiency wages and
allow workers to self-select into such jobs.

11.

The baseline regressions in Tables 4.1 and 4.2 include age groups, education, marital
status, the presence of children, limited health condition, region of residence,
occupation and the year dummies. Results for countries using national panels added
additional controls – industry, firm size and job tenure (for Australia, Germany and
the United Kingdom).

12.

The sum of the coefficients b(TE)+b(TE·age1529), for instance, captures the
difference in (log) wages for young workers between temporary workers and their
counterparts in standard jobs. The wage gap between young and prime-age

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 192 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
(reference) individuals among temporary workers can be obtained by
b(age1529)+b(TE·age1529. The wage differentials between other age/skill/contract
groups can be obtained in a similar way.
13.

Table 4.4 provides information on annual changes of contract and the unadjusted rates
of moving up/down or staying in the same earnings quintiles over any two
consecutive years. As only full-time employees are considered in the analysis (since
the previous section does not show wage penalties for part-time workers), there are
four possible scenarios in terms of a change of employment status: workers staying in
a standard job (SW-SW); workers moving from a standard to a temporary job (SWTE); workers moving from a temporary to a standard job (TE-SW); and workers
remaining in a temporary job (TE-TE). Earnings mobility is measured by examining
whether an individual experienced a change in the relative quintile position in the
distribution of annual earnings. Upward mobility occurs if an individual moves from
a lower to a higher earnings quintile from year t-1 to t. On the contrary, downward
mobility refers to a transition from a higher to a lower quintile, and stay indicates that
an individual remains in the same quintile over time.

14.

Note that the results here refer to marginal effects net of all observable
characteristics; the Korean exception may indicate some selectivity problems among
temporary workers there, as individuals who are less likely to transition into a regular
job – due to an unobserved effect – tend to select themselves into this group.
Additionally, because of a segmented labour market, accepting a non-standard job
may lead to stigma and give a negative signal to employers who may be less willing
to offer permanent jobs to those on non-standard jobs.

15.

This discussion draws on the OECD (2014) definition of job quality with three
sub-dimensions (level of earnings, labour market security and quality of the working
environment) and adds additional estimators.

16.

Only part-time workers and temporary workers are included in non-standard workers
when looking at the distribution of hourly wages because of inconsistencies in
obtaining the hourly wages of self-employed workers.

17.

Main income earner refers to the person who contributes the highest earnings in the
household.

18.

This may be the results of high-earning self-employed workers at the top of the
distribution.

19.

Equivalised household earnings are calculated as the sum of household labour
earnings (wages and self-employment incomes) from all household members,
dividing by the commonly used OECD equivalent scale (i.e. the square root of the
household size).

20.

Data for Korea could not be included in this section as it does not include information
on taxes.

21.

Note that the median income is calculated based on the entire population, not just the
working-age population.

22.

It is noteworthy that the poverty rate of SW households in Korea is high (12.4%),
compared with other countries. This is because many SW households in Korea
actually include a jobless adult member. The number of “pure” SW households –
where all adult members are standard workers – is relatively limited in Korea. In
2009, for instance, SW households that have both standard worker and non-workers

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 193

accounted for 35% of all working households, while this share is lower in other
OECD countries.
23.

Simulations for temporary workers cannot currently be incorporated into the taxbenefit models used here.

24.

Part-time workers are excluded from eligibility for job seeker’s benefit or allowance
in Ireland if they have worked less than three days a week; in Japan, if they have
worked less than 20 hours per week; and in Korea, if they have worked less than
60 hours per month or 15 hours per week.

25.

Simulations use two scenarios regarding the incidence of employer social insurance
contributions. In the first scenario, the effective burden of employer contributions is
assumed to fall exclusively on the employer. In the second scenario, the burden falls
exclusively on the employee, the underlying assumption being that in the long run
employers adjust wages in response to the level of employer contributions
(see Brittain, 1971; Vroman, 1974; and Beach and Balfour, 1983).

26.

See Annex 4.A4 for detailed descriptions of the OECD/EC tax-benefit simulation
models and EUROMOD.

27.

In Italy, PTRs are very low (3%) since no social assistance benefit is available. See
Figure 2.3 in OECD (2007).

28.

The indicators of work incentives used here are formally defined in Annex 4.A4.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 194 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

References
Acs, Z., D. Audretsch and D. Evans (1994), “Why Does the Self-employment Rate Vary
Across Countries Over Time?”, CEPR Discussion Paper, No. 871, London.
Amuedo-Dorantes, C. (2000), “Work Transitions Into and Out of Involuntary Temporary
Employment in a Segmented Market: Evidence from Spain”, Industrial and Labor
Relations Review, Vol. 53, No. 2, pp. 309–325.
Amuedo-Dorantes, C. and R. Serrano-Padial (2007), “Wage Growth Implications of
Fixed-Term Employment: An Analysis by Contract Duration and Job Mobility“,
Labour Economics, Vol. 14, pp. 829-847.
Australia Productivity Commission (2006), “The Role of Non-Traditional Work in the
Australian Labour Market”, Commission Research Paper, Melbourne, May.
Autor, D.H. (2001), “Why Do Temporary Help Firms Provide Free General Skills
Training?”, Quarterly Journal of Economics, Vol. 116, No. 4, MIT Press, pp. 14091448, November.
Autor, D.H. and D. Dorn (2013), “The Growth of Low-Skill Service Jobs and the
Polarization of the U.S. Labor Market”, American Economic Review, Vol. 103, No. 5,
pp. 1533-1597.
Autor, D.H. and S. Houseman (2005), “Do Temporary Help Jobs Improve Labor Market
Outcomes for Low-skilled Workers? Evidence from ‘Work First’”, NBER Working
Paper, No. 11743, National Bureau of Economic Research.
Bardasi, E. and J.C. Gornick (2008), “Working for Less? Women’s Part-time Wage
Penalties Across Countries”, Feminist Economics, Vol. 14, No. 1, pp. 37-72.
Beach, C.M. and F.S. Balfour (1983), “Estimated Payroll Tax Incidence and Aggregate
Demand for Labour in the United Kingdom”, Economica, Vol. 50, pp. 35-48.
Bentolila, S. and G. Saint-Paul (1994), “A Model of Labor Demand with Linear
Adjustment Costs”, Labour Economics, Vol. 1, No. 3-4, Elsevier, pp. 303-326,
September.
Blundell, R. and T. Macurdy (1999), “Labor Supply: A Review of Alternative
Approches”, in O.A. Ashenfelter (ed.), Handbook of Labor Economics, Elsevier
Science/North-Holland, Amsterdam.
Blundell, R., A. Bozio and G. Laroque (2011), “Labor Supply and the Extensive Margin”,
American Economic Review, Vol. 101, No. 3, pp. 482-486.
Booth, A.L. and M. Wood (2006), “Back-to-front Down-under? Part-time/Full-time
Wage Differentials in Australia”, IZA Discussion Paper, No. 2268, Bonn.
Booth, A.L., M. Francesconi and J. Frank (2002), “Temporary Jobs: Stepping Stones or
Dead Ends?”, Economic Journal, Vol. 112, No. 480, pp. 189–214.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 195

Bosio, G. (2014), “The Implications of Temporary Jobs on the Distribution of Wages in
Italy: An Unconditional IVQTE Approach”, Labour, Vol. 28, No. 1, pp. 64–86,
March.
Brewer, M., E. Saez and A. Shephard (2010), “Means-testing and Tax Rates on Earnings.
Dimensions of tax Design”, Mirrlees Review, pp. 90-174.
Brittain, J.A. (1971), “The Incidence of Social Security Payroll Taxes”, American
Economic Review, pp. 110-125.
Buddelmeyer, H. and M. Wooden (2011), “Transitions Out of Casual Employment: The
Australian Experience”, Industrial Relations: A Journal of Economy and Society,
Vol. 50, Issue 1, 109-130.
Buddelmeyer, H., W.S. Lee and M. Wooden (2010), “Low-Paid Employment and
Unemployment Dynamics in Australia”, Economic Record, Vol. 86, No. 272, pp. 2848, March.
Buddelmeyer, H., G. Mourré and M. Ward (2005), “Part-time Work in EU Countries:
Labour Market Mobility, Entry and Exit”, IZA Discussion Paper, No. 1550, Bonn.
Cahuc, P., O. Charlot and F. Malherbet (2012), “Explaining the Spread of Temporary
Jobs and its Impact on Labor Turnover”, IZA Discussion Paper, No. 6365, Bonn.
Cappellari, L. and S. Jenkins (2008), “The Dynamics of Social Assistance Receipt:
Measurement and Modelling Issues, with an Application to Britain”, OECD Social,
Employment and Migration Working Papers, No. 67, OECD Publishing, Paris,
http://dx.doi.org/10.1787/236346714741.
Chen, W.H., M. Förster, H. Levy and A. Llena-Nozal (2015, forthcoming), “Nonstandard Work and Inequality”, OECD Social, Employment and Migration Working
Papers, OECD Publishing, Paris.
De la Rica Goiricelaya, S. (2004), “Wage Differentials between Permanent and
Temporary Workers: Firm and Occupational Segregation”, DFAEII Working Papers
2003-06, University of the Basque Country, Department of Foundations of Economic
Analysis II.
Esteban-Pretel, J., R. Nakajima and R. Tanaka (2009), “Are Contingent Jobs Dead Ends
or Stepping Stones to Regular Jobs? Evidence from a Structural Estimation”, Labour
Economics, Vol. 18, No. 4, pp. 513-526.
Eurofound (2012), Trends in Job Quality in Europe, Publications of the Office of the
European Union, Luxembourg.
Eurofound (2008), “More and Better Jobs: Patterns of Employment Expansion in
Europe”, ERM Report 2008.
Feldman, N. and J. Slemrod (2007), “Estimating Tax Noncompliance with Evidence from
Unaudited Tax Returns”, Economic Journal, Vol. 117, pp. 327–352.
Firpo, S., N. Fortin and T. Lemieux (2007), “Decomposition Wage Distributions Using
Recentered Influence Functions Regressions”, mimeo, University of British Columbia.
Fortin, N., T. Lemieux and S. Firpo (2010), “Decomposition Methods in Economics”,
NBER Working Papers, No. 16045, Cambridge, United States.
Gagliarducci, S. (2005), “The Dynamics of Repeated Temporary Jobs”, Labour
Economics, Vol. 12, No. 4, pp. 429–448.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 196 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Gallie, D. (2013), Economic Crisis, Quality of Work and Social Integration, Oxford
University Press.
Genre, V., R. Gómez-Salvador and A. Lamo (2005), “European Women – Why Do(n’t)
They Work?”, Working Paper Series, No. 454, European Central Bank.
Goos, M. and A. Manning (2007), “Lousy and Lovely Jobs: The Rising Polarization of
Work in Britain”, Review of Economics and Statistics, Vol. 89, No. 1, MIT Press,
pp 118-133, February.
Görg, H., L. Killen and F. Ruane (1998), “Non-Standard Employment in Irish
Manufacturing: Do Firm Characteristics Matter?”, Labour, Vol. 12, No. 4, pp. 675–
699.
Gregg, P. and J. Wadsworth (1996), “It Takes Two: Employment Polarisation in the
OECD”, Discussion Paper, No. 304, Centre for Economic Performance, London
School of Economics.
Heckman, J. (1981), “Heterogeneity and State Dependence”, in S. Rosen (ed.), Studies in
Labor Markets, University of Chicago Press, pp. 91-140.
Heckman, J. (1974), “Shadow Prices, Market Wages, and Labor Supply”, Econometrica:
Journal of the Econometric Society, pp. 679-694.
Hirsch, B.T. (2005), “Why Do Part-time Workers Earn Less? The Role of Worker and
Job Skills”, Industrial and Labor Relations Review, Vol. 58, No. 4, pp. 525-551.
Houseman, S.N. and M. Osawa (2003), “Introduction to Nonstandard Work in Developed
Economies”, W.E. Upjohn Institute for Employment Research, Kalamazoo, United
States, pp. 1–14.
Houseman, S.N. and A.E. Polivka (1999), “The Implications of Flexible Staffing
Arrangements for Job Stability”, Upjohn Institute Staff Working Paper, No. 99-056,
May.
Ichino, A., F. Mealli and T. Nannicini (2008), “From Temporary Help Jobs to Permanent
Employment: What Can We Learn from Matching Estimators and their Sensitivity?”,
Journal of Applied Econometrics, Vol. 23, No. 3, pp. 305–327.
Immervoll H. and M. Pearson (2009), “A Good Time for Making Work Pay? Taking
Stock of In-work Benefits and Related Measures across the OECD”, OECD Social,
Employment and Migration Working Paper, No. 81, OECD Publishing, Paris,
http://dx.doi.org/10.1787/225442803245.
Jaumotte, F. (2003), “Female Labour Force Participation: Past Trends and Main
Determinants in OECD Countries”, OECD Economics Department Working Paper,
No. 376, OECD Publishing, Paris, http://dx.doi.org/10.1787/082872464507.
Kalleberg, A.L (2000), “Nonstandard Employment Relations: Part-time, Temporary and
Contract Work”, Annual Review of Sociology, Vol. 26, pp. 341-365.
Kalleberg, A.L., E. Rasell, N. Cassirer, B.F. Reskin, K. Hudson, D. Webster, E.
Appelbaum and R.M. Spalter-Roth (1997), Nonstandard Work, Substandard Jobs:
Flexible Work Arrangements in the U.S., Economic Policy Institute and Women’s
Research and Education Institute, Washington, D.C.
Leschke, J. (2011), “Labour Market Developments in the Light of the Crisis and the
Europe 2020 Strategy”, EU Anticipedia.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 197

Mertens, A., V. Gash and F. McGinnity (2007), “The Cost of Flexibility at the Margin.
Comparing the Wage Penalty for Fixed-term Contracts in Germany and Spain using
Quantile Regression”, Labour, Vol. 21, No. 4-5, Fondazione Giacomo Brodolini and
Blackwell Publishing Ltd, pp. 637-666.
Mosthaf, A. (2011), “Low-wage Jobs – Stepping Stones or Just Bad Signals?”, IAB
Discussion Paper, No. 11.
OECD (2014), OECD Employment Outlook,
http://dx.doi.org/10.1787/empl_outlook-2014-en.

OECD

Publishing,

Paris,

OECD (2011), Divided We Stand: Why Inequality Keeps Rising, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264119536-en.
OECD (2010), OECD Employment Outlook 2010: Moving Beyond the Jobs Crisis,
OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2010-en.
OECD (2009), OECD Employment Outlook 2009: Tackling the Jobs Crisis, OECD
Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2009-en.
OECD (2008a), OECD Employment Outlook, Chapter 2: “Declaring Work or Staying
Underground: Informal Employment in Seven OECD Countries”, OECD Publishing,
Paris, http://dx.doi.org/10.1787/empl_outlook-2008-en.
OECD (2008b), Growing Unequal? Income Distribution and Poverty in OECD
Countries, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
OECD (2007), Benefits and Wages 2007: OECD Indicators, OECD Publishing, Paris,
http://dx.doi.org/10.1787/ben_wages-2007-en.
OECD (2000), OECD Employment Outlook, Chapter 5: “The Partial Renaissance of SelfEmployment”, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook2000-en.
OECD (1999), OECD Employment Outlook,
http://dx.doi.org/10.1787/empl_outlook-1999-en.

OECD

Publishing,

Paris,

Pearson, M. and S. Scarpetta (2000), “An Overview: What Do We Know About Policies
to Make Work Pay?”, OECD Economic Studies, No. 31, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_studies-v2000-2-en.
Santangelo, G. (2011), “Do Temporary Contracts Cause Wage Discrimination?
A Quantile Treatment Effect Analysis for Europe”, mimeo.
Schuetze, H.J. (2000), “Taxes, Economic Conditions and Recent Trends in Male Selfemployment: A Canada-US Comparison”, Labour Economics, Vol. 7, No. 5, Elsevier,
pp. 507-544, September.
Stewart, M.B. (2007), “The Interrelated Dynamics of Unemployment and Low-wage
Employment”, Journal of Applied Econometrics, Vol. 22, No. 3, pp. 511-531.
Torrini, R. (2005), “Cross-country Differences in Self-employment Rates: The Role of
Institutions”, Labour Economics, Vol. 12, No. 5, Elsevier, pp. 661-683, October.
Tuzemen, D. and J.L. Willis (2013), “The Vanishing Middle: Job Polarization and
Workers’ Response to the Decline in Middle-Skill Jobs”, Economic Review, 2013 Q1,
pp. 5-32

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 198 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Uhlendorff, A. (2006), “From No Pay to Low Pay and Back Again? A Multi-State Model
of Low Pay Dynamics”, IZA Discussion Paper, No. 2482, Bonn.
Venn, D. (2011), “Earnings Volatility and its Consequences for Households”, OECD
Social, Employment and Migration Working Papers, No. 125, OECD Publishing,
Paris, http://dx.doi.org/10.1787/5kg3v00zgslw-en.
Vroman, W. (1974), “Employer Payroll Tax Incidence: Empirical Tests with Crosscountry Data”, Public Finance = Finances publiques, Vol. 29, No. 2, pp. 184-200.
Wenger, J. (2003), “Share of Workers in Nonstandard Jobs Declines”, Economic Policy
Institute Briefing Paper, No. 137.
Wooldridge, M. (2002), Econometric Analysis of Cross Section and Panel Data, MIT
Press, Cambridge, United States.
Yu, Wei-Hsia (2011), “Better Off Jobless? Scarring Effects of Contingent Employment in
Japan”, Social Forces, pp. 1-34.

Database references
OECD Income Distribution Database (IDD), www.oecd.org/social/income-distributiondatabase.htm.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 199

Annex 4.A1
Measuring job polarisation

The contributions to employment shifts due to the between- and within-sector
reallocation of workers can be identified using a simple decomposition approach.
Following Tuzemen and Willis (2013), the change in the employment share between any
𝐸
𝐸
two periods for a particular task group can be decomposed as: 𝑖,𝑠,2010 − 𝑖,𝑠,1995 =
𝐸𝑖,2010 𝐸𝑖,𝑠,2010
𝐸2010

(

𝐸𝑖,2010

−

𝐸𝑖,𝑠,1995 ) +
𝐸𝑖.1995

𝐸𝑖,𝑠,1995 𝐸𝑖,2010
𝐸𝑖.1995

(

𝐸 2010

−

𝐸𝑖,1995
𝐸 1995

𝐸2010

𝐸1995

), where E represents employment, i

represents the industry sector and s represents the task. Table 4.A1.1 reports the
decomposition of changes in the employment share for each of the three task groups. In
sum, this exercise reveals that the results are in line with the predictions of the skillbiased/task-biased technological change hypothesis.
Table 4.A1.1. Decomposition of changes in employment shares by task and sector, EU-23 countries average,
1995/98-2010

Task

Total

Manufacturing

Abstract
Between
Within
Routine
Between
Within
Non-routine manual
Between
Within

5.25
1.24
4.01
-8.59
-3.18
-5.40
1.82
1.03
0.79

-0.95
-1.87
0.92
-6.02
-5.11
-0.91
-0.45
-0.37
-0.08

Construction Wholesale Transport & Finance & Public & Education
& electricity
& hotels communication real estate community & health
0.37
-0.01
0.38
-0.46
-0.05
-0.41
0.12
0.00
0.13

0.32
0.06
0.27
-0.66
0.11
-0.77
0.43
0.04
0.39

1.98
0.40
1.57
-0.35
0.73
-1.07
0.18
0.64
-0.45

2.07
1.40
0.67
-0.28
1.11
-1.39
0.91
0.32
0.59

0.23
-0.39
0.62
-0.84
-0.28
-0.56
-0.41
-0.24
-0.17

1.23
1.66
-0.43
0.02
0.30
-0.28
1.04
0.65
0.39

Note: Abstract occupations (ISCO88: 12-34); Routine (ISCO88: 41-42, 52, 71-74, 81-82 and 93); Non-routine manual (ISCO88:
51 83 and 91). The sample excluded three industries (agriculture, mining and private households) as well as three occupational
groups (legislators, armed forces and farm labourers) due to the lack of consistent information over time. The overall sample is
restricted to workers aged 15-64, excluding employers as well as students working part-time.
Source: European Union Labour Force Survey (EU-LFS).
12 http://dx.doi.org/10.1787/888933209014

Job approach for polarisation
In line with Eurofound (2008, 2012), a job is defined as a particular occupation in a
particular industry (e.g. a manager in the hotel sector) using the international standard
classification of occupations (ISCO) and the European classification of economic
activities (NACE) from the labour force surveys. To measure job quality, each job is
assigned a value, based on the occupational median wages, by linking external
information on wages estimated from different data sources. Two external data sources
were used to estimate median wages for jobs. The main source is the Structure of
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 200 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Earnings Survey (SES), which provides accurate and harmonised data on earnings in
EU member states. It collects information from enterprises with at least ten employees
operating in all areas of the economy except public administration (and, in some
countries, also the education and health care sectors). At the time of writing, the SES data
are available for only two reference years: 2002 and 2006. For sectors not covered by the
SES, the median wage is supplemented by data from the European Survey of Income and
Living Conditions (EU-SILC).
We then allocate jobs to deciles in each country according to a job-wage ranking for
that country, weighted by the total number of employment within the job cell. That is, the
bottom job decile would capture 10% of all workers employed in the lowest-paid
occupations. Changes in employment share between the periods in each job decile can
then be computed to examine whether the employment structure has been polarised over
time.
Ideally, we would like a job to be defined in as much detail as possible by using
occupation and sector classifications at a two-digit or even finer level. In practice, this
may not be feasible as some of the combinations simply do not exist or contain very few
observations, especially in the external sources for wage data. Doing this would result in
imprecise estimates of wages for many job cells. In this study we use the combinations of
occupation at the two-digit level and industry at the one-digit level. This creates a matrix
of a very reasonable 338 non-agriculture “jobs” (26 occupational groups * 13 sectors). In
a separate specification we also performed a more detailed level of disaggregation (i.e.
754 jobs = 26 occupations * 29 sectors). However, due to a smaller sample size of the
SES, many occupation/sector combinations either do not exist or result in an imprecise
estimate of median hourly wages. We therefore discard such disaggregations, as the
ranking of jobs by wage cannot be established.
Labour force surveys (LFS) for three different years – 1995, 2007 and 2010 – are
used to construct the level and change in the employment shares in each job decile. Since
most LFS do not contain information on earnings (one exception being the
Canadian LFS, which collected information on hourly wages), the data is then augmented
with the European Union Structure of Earnings Surveys (SES) to obtain median hourly
wages for each job defined. Note that the SES is available only since the mid-2000s. This
means that the wage-to-job assignments will be the same for each of the three periods
under study (1995, 2007 and 2010). By doing this we assume that the rankings of jobs by
skill (as approximated by median wage) remain similar over time. In fact, previous
studies (e.g. Goos and Manning, 2007) have found considerable stability in the
occupational earnings structure over time.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 201

Annex 4.A2
Data sources for Sections 4.4 and 4.5

The following longitudinal household surveys are used for the analysis in the first
section of the chapter. All longitudinal datasets cover a wide range of subjects, including
personality traits, occupational and family biographies, employment, participation and
professional mobility, earnings and health.

British Household Panel Survey (BHPS)
The British Household Panel Survey1 (BHPS) is a nationally representative
household-based yearly survey which began in 1991, interviewing every adult member of
the sampled households. The wave 1 of the panel consisted of some 5 500 households and
10 300 individuals. Additional samples of 1 500 households in both Scotland and Wales
were added to the main sample in 1999, and in 2001 a sample of 2 000 households was
added in Northern Ireland. These same individuals are re-interviewed each successive
year and, if they split-off from original households to form new households, they are
followed and all adult members of these households are also interviewed.

EU Statistics on Income and Living Conditions (EU-SILC)
The European Union Statistics on Income and Living Conditions (EU-SILC)
instrument has collected annual data for 27 European Union countries, Croatia, Iceland,
Norway, Switzerland and Turkey since 2004 on a cross-sectional and longitudinal basis,
rotating every four years, for 130 000 households. Variables include information on
income, poverty, social exclusion and other living conditions. The EU-SILC does not rely
on a common questionnaire or a survey but on common guidelines and procedures, and
common concepts (household and income) and classifications aimed at maximising the
comparability of the information produced.

German Socio-Economic Panel (GSOEP)
The German Socio-Economic Panel (GSOEP) is an on-going household-based yearly
survey which began in 1984. The first wave consisted of 5 921 households containing a
total of 12 290 individual respondents who participated in “SOEP West”, containing only
West Germany. In 1990, 2 179 households with 4 453 members were surveyed for the
“SOEP East” sample.

Household, Income, Labour Dynamics in Australia (HILDA)
Household, Income, Labour Dynamics in Australia (HILDA) is an ongoing
household-based panel survey funded by the Department of Families, Community
Services and Indigenous Affairs. The survey started in 2001 and contains at the moment
seven waves. The wave 1 of the panel consisted of 7 682 households and
19 914 individuals.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 202 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

Korean Labor and Income Panel Study (KLIPS)
The Korean Labor and Income Panel Study (KLIPS) is an ongoing household survey
which has been conducted annually since 1998 and consists of 5 000 households and
13 000 individuals.

Japan Household Panel Survey (JHPS)
The Japan Household Panel Survey (JHPS) is a panel survey of around
4 000 households conducted by the Panel Data Research Center at Keio University. The
first survey was conducted in 2009 and is carried out annually. The survey topics include
household composition, income, expenditure, assets, and housing in addition to school
attendance, employment, and health conditions of respondents.
We are very grateful to Professor Yoshio Higuchi and Associate Professor
Kayoko Ishii from Keio University for providing the OECD Secretariat with the analysis
form the JHPS and their expert advice.

Note
1.

The BHPS was obtained through the UK data archive (www.data-archive.uk).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 203

Annex 4.A3
Additional tables and figures
Figure 4.A3.1. The marginal effect of previous labour force/contract status on the probability of transition
to unemployment
Women

Men

Australia

Austria

*

***
***

Unemployment t-1
*
**

Temporary t-1

Belgium

***

**

**

**

**

*
***

Part-time t-1
Self-employed t-1

***
***
***
***

Inactive t-1

***

Standard t-1 (ref.)
-5%

0%

5% 10% 15% 20% 25%

-5%

0%

5% 10% 15% 20% 25%

Czech Republic

***

***

***
***

Temporary t-1
Part-time t-1

0%

5% 10% 15% 20% 25%

Estonia

***

Unemployment t-1

-5%

France
***

***

***

***
***

*
***
***

.

Self-employed t-1

**

***

Inactive t-1

***

**

Standard t-1 (ref.)

-5% 0%

5% 10% 15% 20% 25%

-5%

0%

5% 10% 15% 20% 25%

Germany

Temporary t-1

***

***
***
***

Self-employed t-1

**
***

Inactive t-1

25%

***

Part-time t-1

0%

5% 10% 15% 20% 25%

Greece
***

Unemployment t-1

-5%

Hungary
***

***

**

***

*
**

*

***
**

***
***

***
**

Standard t-1 (ref.)

-5% 0% 5% 10% 15% 20% 25%

-5%

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

0%

5% 10% 15% 20% 25%

-5%

0%

5% 10% 15% 20% 25%

 204 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.A3.1. The marginal effect of previous labour force/contract status on the probability of transition to
unemployment (cont.)
Women

Men

Italy

Korea

***

Unemployment t-1
Temporary t-1

***
***

Part-time t-1

***
***

Self-employed t-1

***

Luxembourg

***
***

***

***
***

**

***
***

Inactive t-1

**
**

Standard t-1 (ref.)

-5%

0%

5% 10% 15% 20% 25%

-5%

0%

5% 10% 15% 20% 25%

***

***

5% 10% 15% 20% 25%

Slovak Republic
***

***

35%

***

***

***

***

*

**

Part-time t-1
Self-employed t-1

***

***

Temporary t-1

0%

Portugal

Poland

Unemployment t-1

-5%

**
***

Inactive t-1

**

**

**

Standard t-1 (ref.)
-5% 0%

5% 10% 15% 20% 25%

-5%

0%

Spain

Part-time t-1

5% 10% 15% 20% 25%

***

***
***

0%

***
***

***

***

Temporary t-1

-5%

United Kingdom

***

Unemployment t-1

5% 10% 15% 20% 25%

***

**

Self-employed t-1
***
**

***
***

Inactive t-1
Standard t-1 (ref.)

-5%

0%

5% 10% 15% 20% 25%

-5%

0%

5% 10% 15% 20% 25%

Note:
Marginal effects from lagged employment status on the probability of unemployment based on a random effects dynamic probit,
controlling for initial conditions. See Figure 4.11 for sample and controls. ***, **, * denote significance at the 1%, 5%, 10%
levels, respectively.
Source: British Household Panel Survey (BHPS, 2004-2009) for the United Kingdom, German Socio-Economic Panel (GSOEP,
2004-2012) for Germany, European Union Statistics on Income and Living Conditions (EU-SILC, 2004-2012) for other
European countries, Household, Income and Labour Dynamics in Australia (HILDA, 2004-2012) for Australia, Korean Labor &
Income Panel Study (KLIPS, 2004-2009) for Korea.
12 http://dx.doi.org/10.1787/888933208289

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 205

Figure 4.A3.2. Effect of non-standard work on log hourly wages, by decile
Australia

Austria

0.6

Canada

0.6

0.6

0.3

0.3

0.3

0.0

0.0

0.0

- 0.3

- 0.3

- 0.3

Coefficient
Confidence interval

- 0.6

- 0.6
0.1

0.2

0.3

0.4 0.5
Germany

0.6

0.7

0.8

0.9

- 0.6
0.1

0.2

0.3

0.4

0.5 0.6
Greece

0.7

0.8

0.9

0.6

0.6

0.6

0.3

0.3

0.3

0.0

0.0

0.0

- 0.3

- 0.3

- 0.3

- 0.6

- 0.6
0.1

0.2

0.3

0.4 0.5
Hungary

0.6

0.7

0.8

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.6

0.3

0.3

0.3

0.0

0.0

0.0

- 0.3

- 0.3

- 0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.9

0.2

0.3

Luxembourg

0.4

0.5

0.6

0.7

0.8

0.3

0.3

0.3

0.0

0.0

0.0

- 0.3

- 0.3

- 0.3

- 0.6

- 0.6
0.4

0.5

0.9

0.1

0.2

0.3

0.5

0.6

0.7

0.8

0.9

0.6

0.7

0.8

0.4

0.2

0.9

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.5

0.6

0.7

0.8

0.9

Portugal

0.6

0.3

0.8

Poland

0.6

0.2

0.7

0.1

0.9

0.6

0.1

0.6

- 0.6

- 0.6

- 0.6
0.3

0.5

Italy

0.6

0.2

0.3 0.4
Korea

Ireland

0.6

0.1

0.2

- 0.6
0.1

0.9

0.1

- 0.6
0.1

0.2

0.3

Spain

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

United Kingdom

0.6

0.6

0.3

0.3

0.0

0.0

- 0.3

- 0.3

- 0.6

- 0.6
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Source: Calculations based on European Union Statistics on Income and Living Conditions (EU-SILC, 2012),
Household, Income and Labour Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study
(KLIPS, 2009), Labour Force Survey (LFS, 2013) for Canada.
12 http://dx.doi.org/10.1787/888933208298

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 206 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY
Figure 4.A3.3. Household work patterns by standard work/non-standard employment, working households,
2010 or most recent year
Standard worker households

SW/Non-work

Mixed SW/NSW

Non-standard worker households

NSW/Non-work

100%

80%
60%

40%

20%

0%

Note: Standard work (Non-standard work) refers to households with all workers in standard (non-standard) employment;
SW/Non-work (NSW/Non-work) refers to households with the presence of both a standard (non-standard) worker and non-worker;
SW/NSW refers to households with both standard and non-standard workers. Countries are ranked by increasing shares of
households with at least one NSW.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012), Household, Income and Labour
Dynamics in Australia (HILDA, 2012), Korean Labor & Income Panel Study (KLIPS, 2009), Survey of Labour and Income
Dynamics (SLID, 2010).
12 http://dx.doi.org/10.1787/888933208303

Table 4.A3.1. Household income components by household work pattern, pooled 15 EU countries, 2010
Income component, in EUR (share)
Household work pattern

Jobless
Standard work only 1
Non-standard work only 1
SW/NSW 1
Total

Number of non-workers

Earnings

Capital

Pub.
transfers

5 053
(.31)
30 350
(1.22)
17 009
(.94)
31 680
(1.22)

1 386
(.1)
801
(.03)
1 259
(.05)
937
(.03)

8 732
(.71)
1 833
(.1)
3 688
(.24)
1 836
(.08)

-2 138
-(.12)
-9 115
-(.35)
-5 062
-(.23)
-9 158
-(.33)

0.37

1.54

0.57

0.58

0.66

0.66

0.83

0.41

23 063
(.99)

1 026
(.05)

3 554
(.24)

-6 980
-(.28)

0.59

0.76

Taxes

Number of Number of adult
kids<=17
non-workers

Note: Working-age households.
1. May include non-workers.
Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2012).
12 http://dx.doi.org/10.1787/888933209023

Table 4.A3.2. Impact on Gini by income source

Earnings of standard work households
Earnings of non-standard work households
Earnings of non-working households
Investment
Government transfers
Taxes
Total income

Sk

Gk

Rk

g*r

0.789
0.211
0.093
0.043
0.165
-0.302

0.554
0.799
0.897
0.892
0.683
-0.500
0.308

0.751
0.336
0.326
0.370
-0.043
-0.811

0.416
0.267
0.293
0.327
-0.023
0.407

Contrib.
s*g*r

% contrib.
s*g*r/G

0.327
0.057
0.028
0.017
0.000
-0.123
0.308

1.050
0.179
0.089
0.049
0.000
-0.377
1.000

Source: European Union Statistics on Income and Living Conditions (EU-SILC, 2010), Household, Income and Labour
Dynamics in Australia (HILDA, 2010), Korean Labor & Income Panel Study (KLIPS, 2008).
12 http://dx.doi.org/10.1787/888933209038

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY – 207

Annex 4.A4
Tax-benefit simulations

The OECD tax-benefit models
The aim of the OECD tax-benefit model project is to examine the effects of tax and
benefit systems on families. This is a long-standing activity in the Social Policy Division
of the Employment, Labour and Social Affairs Directorate. The results from this work are
widely used within the OECD, as well as by many external users including other
international organisations, academic researchers and national institutions that monitor
developments in social and fiscal policy. The model currently contains more than a
decade of information (2001-12) on the tax and benefit systems in place across 33 OECD
countries, plus an additional six EU non-OECD countries. Benefits covered in the model
include unemployment benefits, social assistance schemes, housing benefits, family
benefits and employment conditional benefits. The tax system covers personal income
taxes and social security contributions paid by employers and employees.
Using the tax-benefit model, the OECD produces regular updates for a number of key
indicators. This includes measures of work incentives, benefit generosity and income
adequacy. The results are presented in a standardised format to facilitate comparisons
across countries and over time. They capture the effects of taxes and benefits on the
incomes of working-age individuals and their families both in and out of work.
Recent work relying on the results of the tax-benefit model include analyses of
activation policies and work incentives, benefit adequacy, trends in government
redistribution and inequality, barriers to female employment, and countries’ policy
responses to the recent economic downturn (see www.oecd.org/social/benefits-andwages.htm for further details).

EUROMOD
The aim of EUROMOD is to estimate the effects of taxes (national and local income
taxes), social contributions (paid by employees, employers and self-employed) and
benefits (social assistance, family, housing and other income-related benefits) on
household incomes and work incentives for each country of the European Union. While
this is used in calculating the effects of existing policies, it is also used to evaluate the
effects of tax-benefit policy reforms and other changes on poverty, inequality, incentives
and government budgets. EUROMOD cannot take into account the numerous changes
occurring in the structure of the population or in the labour market but it can capture
changes in the average levels of market incomes and in tax-benefit policies (see
www.iser.essex.ac.uk/euromod/resources-for-euromod-users/country-reports for further
country-specific details).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 208 – 4. NON-STANDARD WORK, JOB POLARISATION AND INEQUALITY

Family types
In the present chapter, results are computed for a set of four typical single-earner
families plus the simple average of these families. The standard family types are: single
adult living alone, single-earner couple without children, single parent with two children
and single-earner couple with two children.

Measures of work incentives
Participation tax rate (PTR) is the proportion of earnings that are “taken away” due
to the combined effect of higher taxes and/or lower benefits when an individual moves
from inactivity or unemployment into work. It is defined in a similar way as the effective
marginal tax rate (EMTR):

PTR  1 
where

y
 ynet OW
ynet
 1  net IW
y gross
y grossIW
IW

y gross  y grossIW

stands

for in-work and
 y grossOW and y grossOW  0 .

OW

stands

for

out-of-work,

High levels of PTR indicate that a large share of earnings are “taken away” and
therefore are associated with small work incentives.
Transition tax rate (TTR) is the proportion of earnings that are “taken away” due to
the combined effect of higher taxes and/or lower benefits when an individual moves from
part-time work into full-time work:

TTR  1 

y
 ynet PT
ynet
 1  net FT
y gross
y grossFT  y grossPT

where FT stands for full-time work and PT stands for part-time work.
High levels of TTR indicate that a large share of the earnings increase is “taken
away” and therefore are associated with small work incentives.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 209

Chapter 5
Women, work and income inequality

This chapter first presents the trends in inequality between men and women in terms of
employment and earnings before discussing the earnings dispersion among male workers
and among female workers. The analysis shows that inequality in individual earnings is
driven primarily by increased wage dispersion among full-time full-year workers.
Looking at the household level, the chapter then presents how changes in work intensity
and skill level for women have affected the level of household income inequality. The
overall effect of changes in women’s employment has been to make the distribution of
income more equal.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of
such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements
in the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 210 – 5. WOMEN, WORK AND INCOME INEQUALITY

5.1.

Introduction and key findings
The increasing participation of women in the labour force has a significant impact on
economic growth: a 50% decrease in the gender gap in labour force participation rates has
been estimated to lead to an increase in the annual growth rate (GDP per capita) of
0.3 percentage point on average (OECD, 2012).1 Whether and how higher female labour
force participation affects inequality is, however, less clear-cut. The analysis in this
chapter shows that having more women in paid (full-time) work results in lower
household income inequality, but policies to increase the earnings potential of lowerearning women are needed to further strengthen this effect.
A clear consensus on the importance of the channels through which women’s
increased labour force participation rate affects household inequality is still missing. The
literature on the contribution of women’s earnings to household inequality shows mixed
results. Some studies found an equalising effect for married couples (Cancian et al., 1993;
Cancian and Reed, 1998). Other studies (Esping-Andersen, 2009; Shorrocks, 1983;
Lerman and Yitzhaki, 1985; and Karoly and Burtless, 1995) found that the increasing
correlation between husbands’ and wives’ earnings over time led to wives’ earnings
magnifying family income inequality. People’s tendency to live with spouses in groups at
similar earnings levels (“assortative mating”) and other changes in household structure,
such as the rise in single-headed households, have been previously found to increase
earnings inequality (OECD, 2011; Greenwood et al., 2014).
Recent studies point to an inverse relationship between female employment and
income inequality (e.g. Harkness, 2010). For a sample of 23 OECD countries, OECD
(2011) showed that the growth in female employment tended to have an equalising effect
on household earnings in all the countries studied, despite substantial differences in
employment levels and type of employment and in pay gaps across countries.
This chapter first documents the trends in gender gaps in terms of employment and
pay up to the most recent year. It then examines how changes in the distribution of female
earnings and hours of work have affected the evolution of individual earnings and
household income inequality for the period from the late 1980s up to the economic crisis.
While most of the literature looks at aggregate employment trends, the analyses below
decompose different employment types (full-time/part-time, full-year/part-year) to
analyse the impact of changing patterns of women’s employment in more detail. The
three pathways through which women’s labour market participation affect the distribution
of household income discussed here are: first, trends in individual gross earnings; second,
trends in household gross earnings; and, finally, total household disposable income (after
taxes and transfers).
Within this framework, inequality in individual earnings is assessed in terms of the
development of both between-group and within-group inequalities (Sections 5.2 and 5.3).
The former looks at gender gaps in wages and labour force participation, and the latter at
changes in the dispersion of earnings among male and female workers separately. The
second pathway, which describes the impact on the distribution of household earnings,
examines the correlation of earnings between husbands and wives as well as how this has
developed over time (Section 5.4). This last section uses a novel approach to
decomposition to link the explanatory gender-relevant labour market factors to household
income inequality.2

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 211

The key findings from the analysis in this chapter are:

5.2.



In the past 20 years, gender gaps in employment and earnings have declined in
most OECD countries, but they persist. Women are still less likely to being in
paid work, to progress in their career and are more likely to earn less in their job.



From the mid-1980s to the late 2000s, there was a trend towards greater earnings
inequality both among men and women, but the increase was lower for women
because female low-wage workers increased their hours more and also because
their relative earnings progressed faster than those of their male counterparts.



Increases in individual earnings inequalities between 1985 and the late 2000s were
driven by rising inequality mainly among full-time full-year workers, especially
among men. This means that widening earnings inequality is driven by a widening
gap in pay for full-time workers who work the whole year, rather than by a change
in the composition of employment, i.e. more part-time workers.



In the past two decades, there was an increase in the proportion of households
with women working full-time and having a high-skilled job. The former
contributed to a decline in household income inequality while the latter tended to
increase it slightly. Taken together, the overall effect was towards equalisation,
countering the trend for income inequality to increase.

Trends in employment and earnings gaps between men and women
Women’s employment rates have increased in all countries
The gender employment gap, defined as the difference in employment rates
between men and women, has narrowed in all OECD countries during the last twenty
years, except in Sweden where it was negligible already in the early 1990s (Figure 5.1):
on average it declined by almost 8 percentage points. The reductions in the gender
employment gap were not systematically related to its size. The largest reductions were
in Spain and Ireland, where the gap declined from around 30 percentage points in 1992
to below 10 percentage points in 2013. For Ireland, recent research suggests that most
of the narrowing of the gap was a result of the onset of the recession in 2008-09
(McGinnity et al., 2014). Gender employment gaps were closing in most OECD
countries during the early years of the economic crisis as a result of higher men’s
unemployment (OECD, 2013). Not surprisingly, the gender gap has not changed much
in the Nordic countries, where female employment rates have been high traditionally, at
70% or more; the gender employment gap is thus small, generally less than
5 percentage points. By contrast, in some southern European, Latin American and
Asian countries, it is above the OECD average of 16 percentage points. In Turkey, it is
even close to 40 percentage points.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 212 – 5. WOMEN, WORK AND INCOME INEQUALITY
Figure 5.1. Employment gender gap in 1992 and 2013
Percentage-point difference between men and women employment rates (persons aged 15-64)
1992

2013 (↗)

50

40

30

20

10

0

Note: Employment gender gap: difference in percentage-points between men and women employment rates (persons aged 15-64).
1992 refers to 1993 for the Czech Republic; 1994 for Austria and the Slovak Republic; 1996 for Chile; and 2002 for Slovenia.
Source: OECD estimates based on the OECD Short-Term Labour Market Statistics Database and the OECD Labour Force
Statistics Database.
12 http://dx.doi.org/10.1787/888933208315

Substantial gender differences remain in working hours
Even though gains in female educational attainment have contributed to improving
women’s conditions in the labour market, considerable differences remain in the type and
quality of jobs held by women and in their working hours. Women are much more likely to
work part-time. When accounting for working hours, the gender employment gap widens to
22 percentage points on average (Figure 5.2). The gap is larger in countries where the
overall employment gap is very large (Chile, Mexico and Turkey) but also in Switzerland,
where many women work part-time and the gap in full-time equivalent rates (FTER)3 is
three times larger than the gender gap in employment. Motherhood makes part-time work
much more likely, and the share of part-time work in total female employment is greater in
countries with significantly higher childcare costs (OECD, 2010).
Figure 5.2. Employment/population ratios (EPR) and full-time equivalent employment rates (FTER),
15-64 years old, 2013
%

EPR - Women (↘)

FTER - Women

FTER - Men

100
90
80

70
60
50
40
30

20
10
0

Note: The employment/population ratio (EPR) is defined as the proportion of employed in the working age population. The
full-time equivalent rate (FTER) is calculated as the employment/population ratio, multiplied by the average usual hours worked
per week per person in employment, then divided by 40.
Source: OECD Employment Database.
12 http://dx.doi.org/10.1787/888933208327

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 213

Labour markets are also segregated by occupation (“horizontal segregation”). Women
and men tend to work in different sectors, and this has not changed much since 2000.
Women dominate the service sector (83% of women, compared to 34% of men), while men
are overrepresented in industry (OECD, 2012). Within the service sector, gender gaps are
largest in health and social work, followed by education. In addition to horizontal
segregation, women also face a “glass ceiling” or “vertical segregation” because they do not
advance in their careers as fast or as far as men. On average, just over one-third of
managers are women in 2013, with small variations across countries (OECD Gender Data
Portal, www.oecd.org/gender/data/). In the public sector, where women are overrepresented
and account for 58% of the total sector workforce, career opportunities for women are more
comparable to men’s but women still make it to the top less often (OECD, 2012).

Gender pay gaps have narrowed but persist
Women are less likely than men to progress in their careers and more likely to be
employed in lower-paid occupations. These are all factors that contribute to the
persistence of a gender gap in pay. In all OECD countries, median wages for men are
higher than those for women, although the gap has narrowed over time. Even among fulltime employees, in 2013 women earned, on average, 15% less than men (Figure 5.3).
This is an improvement of four percentage points from 2000. The recent decline in gender
pay gaps is due to the reduction of disparities in the extra-wage components of pay, rather
than the wage itself. Gender wage gaps are highest in Korea, Estonia and Japan, at
between 26.6% and 36.6%.
Figure 5.3. Gender pay gaps over time
Gender pay gap in earnings for full-time employees 2000, 2005 and 2013 or last year available
%

2013 or last year available (↗)

2005

2000

50

40

30

20

10
0

Note: The wage gap is defined as the difference between male and female median wages divided by male median wages.
Estimates of earnings used in the calculations refer to gross earnings of full-time wage and salary workers. Self-employed and
unpaid family workers are not included in the calculations. Data for 2000 refer to 2001 for Israel and Poland and to 2002 for
Estonia, the Slovak Republic, Luxembourg and the Netherlands. Data for 2005 refer to 2006 for Estonia and the Netherlands and
to 2004 instead for Switzerland, Italy and Poland. The data for 2013 refer to 2010 for Estonia, France, Luxembourg, the
Netherlands, Slovenia, Switzerland and Turkey, to 2011 for Chile, Iceland, Germany and Israel, and to 2012 for Belgium,
Greece, Denmark, Spain, Poland, Italy, Sweden, Portugal, Austria and Finland.
Source: OECD Database on Earnings Distribution and OECD Secretariat estimates based on CASEN (Caracterizacion
Socioeconomica Nacional).
12 http://dx.doi.org/10.1787/888933208335

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 214 – 5. WOMEN, WORK AND INCOME INEQUALITY

5.3.

Trends in individual earnings inequality among male and female workers
This section analyses the changes in inequality within the two groups, men and women,
based on analyses of micro data of the Luxembourg Income Study.4 Greater diversity over
the life course in women’s employment patterns is likely to affect the distribution of
women’s earnings. A larger share of part-timers among women may mean that female
earnings are distributed more unequally than men’s. At the same time, there is little
empirical evidence on changes in female earnings inequality over time. Since changes in
annual earnings depend on both earnings and hours worked, this section discusses
differences and similarities in trends for both of these factors for men and women.

Lower but more rapidly increasing earnings inequality among men
Across the OECD, women’s annual earnings5 were slightly more unequally
distributed than those for men in the late 2000s, but in most countries the differences
were small (Figure 5.4). On average, the Gini coefficient (where 0 represents perfect
equality and 1 perfect inequality) was two percentage points higher among women in the
late 2000s. Germany showed the largest gender difference, with women’s Gini being
almost 8 points higher than men’s, followed by Mexico and Spain. On the other hand, in
Belgium, the Czech Republic and the United States there was no sizeable difference in
earnings inequality between men and women. In Denmark and Finland, by contrast,
earnings inequality was slightly lower among women.
Figure 5.4. Changes in annual earnings dispersion, workers with positive earnings
1990

Late 2000s (↘)
Panel A. Men

0.6

0.5

0.4

0.3

0.2

Panel B. Women
0.6

0.5

0.4

0.3

0.2

Note: Working-age (25-64) workers with positive annual earnings. Sample restricted to paid workers with positive wages. This
excludes self-employed and those employed without income. Earnings refer to net earnings for countries in square brackets and
to gross earnings for the other countries. The figures presented here include part-time, part-year employment.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208349

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 215

In almost two-thirds of OECD countries, the inequality in men’s earnings rose more
than that for women between the mid-1980s/early 1990s and the late 2000s (Figure 5.4).
Male earnings inequality rose by 17%, on average, compared to 9% for females, or by
5 Gini points for men and 3 points for women. In some countries, such as Australia and
Norway, earnings inequality declined for women while it increased for men. Similar
though less pronounced patterns emerge for Switzerland, the United Kingdom and the
United States. By contrast, earnings inequality increased by a far greater amount for
women than for men in Mexico.

Why did earnings inequality increase less for women?
Changes in individual annual earnings inequality depend on changes in both hourly
wages and annual hours worked among different groups along the income distribution.
The overall effect depends on whether changes in both components reinforce each other
or work in opposite directions. Figure 5.5 presents changes in real annual earnings and in
hours worked, by men and women at the bottom, middle and top of their respective
earnings distribution. In most countries, the gap between higher-paid workers and lowerpaid workers widened for men but not necessarily for women.
Figure 5.5. Percentage changes in real annual earnings and annual hours worked by earnings quintile
Panel A. Men
Changes in real annual earnings by earnings quintile

1-20% (↘)

80-100%

Changes in annual hours worked by earnings quintile

40-60%

1-20% (↘)

%

%

120
100
80
60
40
20
0
-20
-40

60

80-100%

40-60%

40
20
0

-20
-40
-60

Panel B. Women
Changes in real annual earnings by earnings quintile
1-20% (↘)

80-100%

Changes in annual hours worked by earnings quintile

40-60%

1-20% (↘)

%

%

120
100
80
60
40
20
0
-20
-40

60

80-100%

40-60%

40
20
0

-20
-40
-60

Note: Working-age (25-64) workers with positive annual earnings. “1-20%” refers to bottom earnings quintile; “40-60%” refers
to middle earnings quintile; “80-100%” refers to top earnings quintile. Earnings are CPI adjusted in 2005 national currency.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208351

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 216 – 5. WOMEN, WORK AND INCOME INEQUALITY
First, while at the top of the distribution trends in real earnings were similar for both
men and women, at the bottom of the distribution, women’s real earnings increased more
(or decreased less) than men’s in most countries.
Second, while men at the bottom of the earnings distribution tended to reduce their
working hours between the mid-1990s and the late 2000s, women in the bottom quintile
increased their hours or reduced them to a lesser extent than men in most countries. In the
United States, for instance, these developments were linked to stagnating educational
attainment among men and deteriorating employment prospects for less-educated males
(Autor and Wasserman, 2013). In Australia, hours worked have increased mostly among
part-timers, who are mostly women, in the lower part of the distribution, while working
hours for full-timers have been stable and employment rates for men have declined
(Greenville et al., 2013)

What counts more for earnings inequality: changes between men and women or
within different groups of male and female workers?
Changes in earnings inequality may reflect changes in the composition of
employment (e.g. increased share of part-time or part-year work), or it may reflect
widening wage dispersion within such employment-specific groups. For instance, male
wages may contribute to growing overall inequality because they become more dispersed,
while the composition effect, e.g. increased participation in part-time/part-year jobs, may
be the main channel through which women affect overall inequality. To test this
hypothesis, this section applies an analysis that decomposes earnings inequality
(measured by the Theil coefficient)6 by subgroup (between men and women from four
different employment groups, i.e. full-time full year, full-time part-year, part-time full
year and part-time part year).
Figure 5.6 reveals that changes in “within-inequality” (that is, wage dispersion within
the groups) were the key driver in explaining widening earnings inequality.7 In eight of
the ten countries where inequality increased, the within-group effect was the biggest
contributor to changes in inequality. In Australia, Canada and the United States, the
increase in wage dispersion within the groups was even large enough to outweigh a
significant negative (i.e. equalising) effect of the between-group component. In most
countries increases were primarily driven by growing wage dispersion among full-time
full-year (FTFY) workers, both male and female (except in the Netherlands where this
concerned only men). Contributions to changes in within-inequality from other subgroups
are generally very modest, with one exception: in Finland, inequality among part-time
full-year (PTFY) workers (mostly men) contributed most (see Annex 5.A2,
Table 5.A2.1).
The role of gender-specific changes in inequality for overall inequality is much less
important. The reason is that changes between different types of workers tend to cancel
each other out: while more women in part-time work reduced individual earnings
inequality in most countries studied, women in FTFY jobs contributed to an increase in
between-inequality because their earnings (relative to the national average) have
improved over time (Table 5.A2.1. in Annex 5.A2).8 Austria and Ireland are the only
countries where the between-group component outweighs the effect of within-group
inequality, in both countries again driven by full-time full-year workers.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 217

Figure 5.6. Contribution to changes in individual earnings inequality by gender
Panel A. Changes in inequality (Theil index), by between/within-group component
∆ in inequality (Theil) -within

∆ in inequality (Theil)-between

Total (↘)

Changes in inequality (Theil)
0.15

0.12
0.09
0.06
0.03
0

-0.03
-0.06
-0.09

Panel B. Changes in inequality (Theil index), by within-group component
Men

Women

Total (↘)

Changes in inequality (Theil)
0.15

0.12
0.09
0.06
0.03
0

-0.03
-0.06
-0.09

Note: Working-age (25-64) workers with positive annual earnings. Samples include only countries where information on both
weeks worked per year and hours worked per week is available. Changes in Panel B reflect the impact of the within component
for full-time full–year workers.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208364

5.4.

From individual to household earnings and the income distribution
Inequality in household earnings depends on the joint distribution of the earnings of
all persons in the family. Previous OECD work (2011) has shown that moving from the
individual to the household perspective significantly reduces earnings inequality, by
almost 9 Gini points in countries reporting gross income and by 12 points in countries
reporting net earnings.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 218 – 5. WOMEN, WORK AND INCOME INEQUALITY
Inequality of household earnings is influenced by labour market factors – the earnings
and employment intensity of household members – as well as the characteristics of
household members, i.e. their education level, whether they are single or couple
households or have children. These factors have changed over time, with more women
working but also more single-person households, and they have been found to contribute
to growing household earnings inequality. The following two sub-sections discuss
changes to the intra-household division of labour and how these may impact on
household earnings, and they present an analysis of the phenomenon of “assortative
mating”, a growing trend of “liking to live with like” (couples with similar education or
earnings). The third and last sub-section estimates the effect on overall income inequality.

Changes in female participation, work intensity and job nature within the
household
With increased female labour force participation, women's contributions to household
income have grown. From the mid-1980s/early 1990s to the late 2000s, the share of
female earnings rose by 12% in countries reporting gross earnings and by almost 20% in
countries reporting net earnings (Figure 5.7). The relative increases were largest in
Spain (43%), Luxembourg (40%) and Ireland (27%). By contrast, in some Nordic
countries there was either little change (Norway, Sweden) or a decline (Finland a 9%
decrease).
Figure 5.7. Female earnings as a share of total disposable household income,
mid-1980s/early 1990s – mid/late 2000s
Early year

Late year (↗)

%
60
50

Countries reporting
gross earnings

40

30
20

Countries reporting
net earnings

10
0

ISR CZE GBR NLD NOR DEU USA SVK CAN FIN SWE AUS DNK

LUX MEX AUT IRL BEL GRC ITA HUN FRA ESP POL

Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208372

Figure 5.8 shows that the proportion of households with a working woman (either
head or spouse) increased in nearly all OECD countries under study, on average by
9 percentage points, from 52% in the mid-1980s/early 1990s to 61% in the late 2000s
(Panel B). Luxembourg, Spain and Belgium saw the largest increase in this share: more
than 20 percentage points over the period studied. By contrast, the share of working-age
households with a working man remained relatively stable (at 76%) over time in most
countries (Panel A). Germany, nevertheless, witnessed a significant decline (by about
15 percentage points) in men’s employment participation among households during
this period.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 219

Figure 5.8. Changes in male/female labour participation among households
Panel.A. Households with a male worker
Early year

Panel.B. Households with a female worker
Early year

Recent year (↘)

%

%

100

100

80

80

60

60

40

40

20

20

Recent year (↘)

Note:
Sample restricted to households with a working-age head (25-64).
A male/female worker refers to someone who is either the head or spouse in a household.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208380

Similarly, employment intensity at the household level has also increased noticeably
as more women work full-time around the year than in the past (Figure 5.9). In Spain, for
instance, in 1995 only 22% (25%) of households had a woman working full-time (fullyear), compared to 42% (52%) in the year prior to the crisis. Dutch women are now
working more weeks per year, though not more hours per week. The share of households
with a full-time working woman in the Netherlands has remained low at 20% over time,
while the prevalence of women working full-year among households has gone up
significantly, from 32% in 1990 to nearly 60% in 2004. The equivalent share for men was
stable in most countries, although Germany and Hungary saw significant declines for fulltime workers and Australia for full-year working men in the same period.
In addition to increasing participation and work intensity, women have also
increasingly become more involved in high-skilled/high-paying occupations over time,
primarily driven by their rising levels of educational attainment. While this development
tends to reduce the gender pay gap, it may increase earnings dispersion among female
workers, and in turn affect the distribution of household income. Figure 5.10 reveals that
the proportion of households with a woman working in managerial-professional-technical
jobs has grown, on average, by 10 percentage points (from 14% to 24%) since the mid1980s/early 1990s. The increase is especially large (more than 15 percentage points)
among households in the United States, the Netherlands, Luxembourg, Spain and
Belgium. The same is true for men, albeit to a lesser degree. The share of households with
male high-skilled workers has risen most in Spain and Finland (more than 20 percentage
points over time).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 220 – 5. WOMEN, WORK AND INCOME INEQUALITY
Figure 5.9. Changes in male/female employment intensity among households
Panel A. Households with a male working full-time
Early year

Panel B. Households with a female working full-time
Early year

Recent year (↘)

%
100

%
100

80

80

60

60

40

40

20

20

0

0

Panel C. Households with a male working full-year
Early year

Recent year (↘)

Panel D. Households with a female working full-year
Early year

Recent year (↘)

%
100

%
100

80

80

60

60

40

40

20

20

0

0

Recent year (↘)

Note:
Sample restricted to working-age (25-64) households.
A male/female worker refers to someone who is either the head of a household or the spouse. Full-time refers to 35 or more
hours worked per week, full-year refers to 45 or more weeks worked per year.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208391

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 221

Figure 5.10. Changes in male/female job skill nature among households
Panel A. Households with a male in high-skilled job
Early year

Panel B. Households with a female in high-skilled job
Early year

Recent year (↘)

60

60

40

40

20

20

0

0

Recent year (↘)

Note: Sample restricted to working-age (25-64) households.
A male/female worker refers to someone who is either the head or spouse of a household. High-skilled jobs correspond to the
first three categories of the 1-digit ISCO (i.e. managers, professionals and technicians).
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208408

More couples with partners from a similar background
A further factor explaining changes in household income inequality is the way in
which families and households are formed. How a rise in women’s labour earnings
affects inequality in household earnings will also be influenced by the correlation
between the earnings of wives and husbands. There is increasing evidence that men and
women with similar characteristics are more likely to be married to each other, a
phenomenon described as “assortative mating” (OECD, 2011). Evidence from the United
States and Canada suggests that wives of high-paid husbands increased their employment
and their hours worked more than did the wives of lower-paid husbands (Junh and
Murphy, 1997; Morissette and Hou, 2008).9
Figure 5.11 presents one way to look at the effect of assortative mating, defined as
changes in wives’ employment rates by husband’s earnings decile, for the OECD average
(country-specific figures are shown in Figure 5.A1.1 in Annex 5.A1). Between the 1990s
and the late 2000s, in two-thirds of the countries female employment rates increased more
among women whose husbands were in the top earnings deciles than among those at the
bottom of the distribution. In one-third of the countries (Belgium, France, Greece,
Ireland, Italy, the Netherlands, the United Kingdom), the growth in employment rates of
wives of the top earners was at least twice as large as for the lower earners. In certain
countries, such as Italy and Mexico, the employment rates of wives of top earners
increased more than for wives of low earners at the end of the 2000s while the opposite
was true in the early 1990s.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 222 – 5. WOMEN, WORK AND INCOME INEQUALITY
Figure 5.11. Wives’ employment rates, by the husband’s earnings decile, OECD average

Early year (1990s)

%

Late year (2000s)

80
75
70
65
60
55
50

1

2

3

4

5

6

7

8

9
10
Husband's earnings decile

Note: Figures refer to couple households, with or without other people in the household. Only includes couples with one female
and one male. Unmarried, cohabiting partners are included. Both are of working age (25-64).
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208414

Figure 5.12 shows that in all countries (except Finland) the correlation between the
earnings of husbands and wives increased over this time period. While still at a relatively
low level, the data show an average increase in the correlation coefficient between the
earnings of husbands and wives from 0.18 to 0.22 over the past 20 to 25 years. The
increase in this correlation is particularly strong in Italy, Norway and Poland but also in
Denmark and Luxembourg, among others.
Figure 5.12. Correlation between the earnings of husbands and wives,
couple households with at least one person working
Early year

Recent year (↘)

Correlation coefficient
0.5
0.4
0.3

0.2
0.1
0

Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208422

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 223

How do labour market factors drive changes in household income inequality
over time?
What would be the level of overall income inequality if households today had the
same proportion of women working as households 20 years ago if the pay structure
remained at today’s level for households with a female worker. This sub-section
analyses the effect of changes in household employment patterns on household
disposable income, i.e. considering income components beyond earnings (see Box 5.1).
In particular, it looks at two questions: to what extent have changes in women’s
employment characteristics at the household level discussed in the earlier sections
contributed to the increase in income inequality? And what role did female
participation, employment intensity and job characteristics play in the upward trend in
household income inequality? To answer these questions, it is useful to simulate a
series of counterfactual distributions of household income. The counterfactual exercise
builds on the results from a decomposition approach based on Firpo et al. (2007) and
Fortin et al. (2010) and is described in Annex 5.A3.
Figure 5.13 suggests that in almost all countries the combined changes in women’s
work are found to exert a substantial effect in countering a rise in household income
inequality. These effects can be broken down into four components: women’s
employment rates, women’s work intensity, the share of females in skilled jobs, and the
overall returns to female employment (wage effect).
First, an increase in female employment is associated with a reduction in household
income inequality. The figure shows that if the proportion of households with a
working female had remained the same as in the mid-1980s, household income
inequality would have increased, on average across the OECD countries, by an
additional 0.8 point, i.e. an increase in the Gini from 28.2 to 31.6 instead of 30.8. The
equalising impact of rising female employment was larger in countries such as
Belgium, the Netherlands, Spain and the United Kingdom, where the share of workingage households with female workers increased greatly: this contributed to lowering
inequality by 2.5 or more Gini points in these countries. In contrast, given men’s
stability in terms of employment rates, the impact of male employment at the household
level was neutral overall (Annex 5.A3, Figure 5.A3.1).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 224 – 5. WOMEN, WORK AND INCOME INEQUALITY

Box 5.1. Components of household income by gender
It is important to keep in mind that while earnings constitute the bulk of household income, other income
components such as taxes and benefits play a role in the evolution of income. For instance, in the Nordic
countries the total share of social transfers represents close to 20% of household income (25% in Sweden),
while in Australia and the United States it constitutes 8% and 7%, respectively. The respective contributions of
earnings and other income sources to overall inequality can be estimated (see OECD, 2011), and it is interesting
to consider the extent to which the contributions of gender-specific income components can be estimated. The
figure below presents the results of a decomposition analysis by income source, separating the different
components that contribute positively or negatively to inequality. It examines the relative contributions of
gender-specific income components to income inequality using a decomposition by income source (following
Lerman and Yitzhaki, 1985). The Gini coefficient for total income, G, can be represented as 𝐺 = ∑𝐾
𝑘=1 𝐶𝑘 =
∑ 𝑆𝑘 𝐺𝑘 𝑅𝑘 where 𝑆𝑘 is the share of source k in total income, 𝐺𝑘 is the Gini coefficient corresponding to the
distribution of income from source k, and 𝑅𝑘 is the correlation of income from source k with the distribution of
total income. The relative contribution of source k to total income inequality can then be calculated as 𝐶𝑘𝑅 =
𝑆𝑘 𝐺𝑘 𝑅𝑘
. This approach relies on the assumption that flat-rate distributed benefits make a neutral or “zero”
𝐺
contribution to inequality, and thus do not account for a decreasing marginal benefit for transfers along the
income distribution.
The decomposition looks at gender-specific components for earnings and taxes components. For the analysis,
household income is divided into men’s labour income, women’s labour income, capital income, social transfers
(disaggregated into the components allocated to men and women) and overall household-related benefits, and taxes
paid by men and women. Some sources, in particular capital income and some social benefits, can be recorded only
at the household level. On average, four components contribute positively (i.e. increase) to inequality: men’s
earnings, women’s earnings, capital income and men’s related social benefits, although the effect of the latter is
almost nil. Men’s earnings are the largest contributor to inequality, while the role of women’s earnings is about
half that size. Taxes and social benefits tend to decrease inequality, especially men’s related taxes. Taking into
account the different components by gender, the share of inequality explained by men’s related income
components is close to 57% while the female share accounts for 37% of overall inequality.
Decomposition of household income inequality by income source, average across 13 OECD countries,
late 2000s
Benefits : non-gender specific (-4.4%)

Men's taxes
( -37.6%)

Women's taxes (-9.0%)

-100%

-50%

Capital income (10.7%)

Men's earnings (94.3%)

Women's
earnings
(47.8%)

Women's
benefits
(-1.8%)

Men's benefits (0.2%)

0%

50%

100%

150%

200%

Note: Sample includes only working-age (25-64) couple households in countries that report all components in gross terms
(Australia, Canada, Czech Republic, Denmark, Finland, Germany, Ireland, Netherlands, Norway, Slovak Republic, Sweden,
United Kingdom and United States).
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208438

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 225

Figure 5.13. Contribution to percentage point changes in Gini of household disposable income,
RIF decomposition results
Composition and wage structure effects (women)

0.100

∆ in % of households with female worker
∆ in % of households with female in skilled job
Change in Gini (↘)

∆ in % of households with FT, FY female worker
Combined wage effects (prior 3)

0.075
0.050
0.025

0.000
- 0.025
- 0.050
- 0.075

Note: Sample includes working-age (25-64) households. RIF: Recentered Influence Function (RIF) regressions. The wage
structure effects combine three covariates participation, work intensity and nature of job skill. FT: Full-time, FY: Full year.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208441

Second, an increase in the work intensity of women in the past two decades is also
associated with lower household income inequality in some – but not all – countries. In
Canada, Ireland, the Netherlands and the United States, having more households with a
female member working full-time and/or full-year was associated with a decline in the
Gini coefficient of household income by between 0.6 and 1 point. In turn, in the Czech
Republic and Finland the falling share of females in full-time or full-year jobs among
households has resulted in a 0.4 point rise in inequality.
Third, the rising proportion of households with females in high-skilled jobs is
associated with a small yet significant rise in household income inequality (0.3 point).
The impact of skill changes is larger in Belgium, the Netherlands, Spain and the United
Kingdom, where it led to an increase of between 0.7 and 1 point in inequality. The
increase in inequality driven by higher skill levels was more pronounced for men
(Annex 5.A3, Figure 5.A3.1), particularly in Belgium, Finland and Spain, where the
increased share of households with a male member employed in a managerial,
professional or technical job contributed 1.5 point to the increase in inequality over the
period. One possible explanation for why more households with high-skilled workers
drives inequality higher is the growing tendency of people with similar
education/earnings levels to live together (“assortative mating”), as described in
Section 5.2.
Finally, in almost all countries changes in the returns to women’s work (the wage
effect) are countering the rise in household income inequality, adding a further equalising
impact to that detected from the composition effect associated with the three changes
described above. In part, this may reflect a narrowing in the gender wage gap as a result
of increased women's educational attainment. It may also be due to an increase in
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 226 – 5. WOMEN, WORK AND INCOME INEQUALITY
women’s average job tenure. Female workers today are more likely to have more labour
market experience (and thus higher earnings) than 20 years ago. Moreover, changes in
policy parameters such as legislation mandating equal pay may also contribute to
reducing the gender gap. Such unobservable factors (or not controlled for) seem to play
an important role in equalising the distribution of household income and are likely to be
captured by the unexplained wage structure effect.
In sum, overall income inequality would have been higher if women’s employment
participation and work intensity had remained the same as 20 years ago, while changes in
the occupations in which women work have tended to aggravate inequality somewhat.
Changes in women’s wages accentuated the equalising impact of women on household
income inequality. The combined effect resulted in lowering income inequality by about
2 Gini points.

5.5.

Conclusion
This chapter explored the gender aspects of earnings and household income
inequality. Overall, it shows that the increased participation of women in the labour
market over the past 20 to 25 years up to the economic crisis has tended to make the
distribution of household earnings and income more equal.
Across the OECD, gender employment gaps and gender pay gaps have narrowed over
the past two decades – but on average, gender employment gaps persist at around 16%,
and gender pay gaps at around 15%. At the same time, earnings inequality within gender
groups has increased. Earnings inequality is higher among women than among men but
has been increasing at a slower pace, on average by 9% as compared to 17% for men.
This is explained by two facts: first, as opposed to men, lower-wage women increased
their working hours or reduced them to a lesser extent; second, lower-wage women saw
their wages rise more than those of their male counterparts.
As for trends in overall individual earnings inequality along the gender dimension,
increased wage dispersion among different groups of male and female workers (“withingroup” inequality) was the key driver. This was due in particular to growing wage
inequality among full-time full-year workers (especially men). On the other hand,
inequalities between male and female workers contributed much less. Among those
components, the growing relative shares of working women, in particular in part-time
work, tended to reduce overall individual earnings inequality in most countries.
For household income inequality, changes in female employment participation, work
intensity and the nature of job skills matter. If the proportion of households with a
working female had remained the same as 20 to 25 years ago (52% rather than 61%),
income inequality would have increased by almost 1 point more on average than it
actually did. The increasing work intensity of women was also associated with lower
income inequality in some – but not all – countries. On the other hand, the rising share of
households with women in high-skilled occupations generally increased income
inequality, but the effect was more modest. Taken together, the overall effect of changes
in the labour market for women was therefore to make the income distribution more equal
by reducing the Gini by 2 points, especially when not only compositional but also related
wage structure effects are taken into account. Having more households with women in
paid work, especially full-time work, means less income inequality. Therefore, policies
that focus on increasing the earnings potential of lower-paid women can reinforce the
equalising effect of women's labour market integration.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 227

Notes
1.

Convergence in participation rates are projected with the assumption that the male
participation rate remains constant at its 2010 level while the female participation rate
increases to converge with the same rate of men’s by 2030.

2.

The chapter focuses on the contribution of women’s paid employment to changes in
income inequality. It does not discuss the impact of women’s non-market or unpaid
work, which has a non-negligible impact on household’s income inequality as
discussed in Folbre et al. (2013).

3.

Differences in the full-time equivalent rates (FTER) represent the difference between
men and women if they were all working for 30 hours or more per week in their main
job.

4.

The analyses cover 25 countries and use two points in time for each country: an
“early” year, corresponding to the mid-1980s/early 1990s, and the “late year”,
generally corresponding to the mid/late 2000s. For Australia, the latest year available
is 2003 but, for later parts, the year 2001 is used as only the latter year contains
sufficient information to impute annual hours worked and hourly wage rates.

5.

The figures presented here include part-time, part-year employment.

6.

As in the case of the Gini coefficient, a value of zero for the Theil index indicates
perfect equality. The greater the value of the index, the greater is the level of
inequality. The properties of subgroup decomposability of Theil and other general
entropy indices are well documented (e.g. Shorrocks, 1984; Cowell, 2005).

7.

See Annex 5.A2 for a description of the decomposition method.

8.

The increased disequalising effect for women in FTFY is mainly due to the increase
in their relative group earnings, not due to changes in their relative employment share.

9.

Some concerns about endogeneity exist in the theory of assortative mating with
respect to the impact of household formation behaviour on labour market choices. In
this case, some of the correlation captured here would not be explained by pre-marital
household formation choices but rather by an increased norm for couple households
to choose to work in similar occupations and industries, within similar wage groups.
This caveat indicates that marital sorting may explain only part of the changes in
intra-household labour supply behaviour.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 228 – 5. WOMEN, WORK AND INCOME INEQUALITY

References
Autor, D. and M. Wasserman (2013), Wayward Sons: The Emerging Gender Gap in
Labour Markets and Education, Third Way, MIT.
Blinder, A. (1973), “Wage Discrimination: Reduced Form and Structural Estimates”,
Journal of Human Resources, Vol. 8, pp. 436-455.
Cancian, M. and D. Reed (1998), “Assessing the Effects of Wives’ Earnings on Family
Income Inequality”, Review of Economics and Statistics, Vol. 80, No. 1, pp. 73-79.
Cancian, M., S. Danziger and P. Gottschalk (1993), “Working Wives and the Distribution
of Family Income”, in S. Danziger and P. Gottschalk (eds.), Rising Tides: Rising
Inequality in America. Russell Sage Foundation, New York.
Cowell, F (2005), “Theil, Inequality Indices and Decomposition”, Working Paper No. 01,
ECINEQ, Society for the Study of Economic Inequality.
Esping-Andersen, G. (2009), The Incomplete Revolution: Adapting to Women’s New
Roles, Polity Press, Cambridge, United Kingdom.
Firpo, S., Fortin, N. and T. Lemieux (2009), “Unconditional Quantile Regressions”,
Econometrica, Vol. 77, No. 2, pp. 953-973.
Firpo, S., Fortin, N. and T. Lemieux (2007), “Decomposition Wage Distributions using
Recentered Influence Functions Regressions”, mimeo, University of British Columbia.
Folbre, N., J. Gornick, H. Connolly and T. Munzi (2013), “Women’s Employment,
Unpaid Work, and Economic Inequality”, in J.C. Gornick and M. Jäntti (eds), Income
Inequality; Economic Disparities and the Middle Class in Affluent Countries, Stanford
University Press, Stanford, pp. 234-260.
Fortin, N., T. Lemieux and S. Firpo (2010), “Decomposition Methods in Economics”,
NBER Working Papers, No. 16045, Cambridge, United States.
Gornick, J.C. and M. Jäntti (eds) (2013), Income Inequality; Economic Disparities and
the Middle Class in Affluent Countries, Stanford University Press, Stanford.
Greenville, J., C. Pobke and N. Rogers (2013), “Trends in the Distribution of Income in
Australia”, Productivity Commission Staff Working Paper, Canberra.
Greenwood, J., N. Guner, G. Kocharkov and C. Santos (2014), “Marry Your Like:
Assortative Mating and Income Inequality”, IZA Discussion Papers, No. 7895, Bonn.
Harkness, S. (2010), “The Contribution of Women's Employment and Earnings to
Household Income Inequality: A Cross-Country Analysis”, Luxembourg Income Study
Working Paper, No. 531.
Juhn, C. and K.M. Murphy (1997), “Wage Inequality and Family Labor Supply”, Journal
of Labor Economics, Vol. 15, pp. 72-97.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 229

Karoly, L.A. and G. Burtless (1995), “Demographic Change, Rising Earnings Inequality,
and the Distribution of Personal Well-Being, 1959-1989”, Demography, Vol. 32,
No. 3, pp. 379-406.
Lerman, R. and S. Yitzhaki (1985), “Income Inequality Effects by Income Source: A
New Approach and Application to the United States”, Review of Economics and
Statistics, Vol. 67, No. 1, pp. 151-156.
McGinnity, F., H. Russell, D. Watsony, G. Kingston and E. Kelly (2014), “Winners And
Losers?: The Equality Impact of the Great Recession in Ireland”, Economic and Social
Research Institute, accessed 14 May, 2014. Available at: www.equality.ie/research.
Morissette, R. and F. Hou (2008), “Does the Labour Supply of Wives Respond to
Husbands’ Wages? Canadian Evidence from Micro Data and Grouped Data”,
Canadian Journal of Economics, Vol. 41, No. 4, pp. 1185-1200.
Oaxaca, R. (1973), “Male-Female Wage Differentials in Urban Labor Markets”,
International Economic Review, Vol. 14, pp. 693-709.
OECD (2013), How’s Life? 2013: Measuring Well-being, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264201392-en.
OECD (2012), Closing the Gender Gap: Act Now, OECD Publishing, Paris, pp. 150-173,
http://dx.doi.org/10.1787/9789264179370-en.
OECD (2011), Divided we Stand: Why Inequality Keeps Rising, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264119536-en.
OECD (2010), OECD Employment Outlook 2010: Moving Beyond the Jobs Crisis,
OECD Publishing, Paris, p. 241, http://dx.doi.org/10.1787/empl_outlook-2010-en.
OECD (2008), OECD Employment Outlook 2008, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2008-en.
Shorrocks, A.F. (1984), “Inequality Decomposition by Population Subgroups”,
Econometrica, Vol. 52, No. 6, pp. 1369-1385.
Shorrocks, A.F. (1983), “The Impact of Income Components on the Distribution of
Family Incomes”, Quarterly Journal of Economics, Vol. 98, pp. 311-331.

Database references
OECD Short-Term Labour Market Statistics Database, http://dx.doi.org/10.1787/data00046-en
OECD Labour Force Statistics Database, http://dx.doi.org/10.1787/data-00046-en
OECD Employment Database, www.oecd.org/employment/database
OECD Database on Earnings Distribution, www.oecd.org/employment/database
Luxembourg Income Study (LIS) Database, www.lisdatacenter.org (mutilple countries:
microdata runs completed between 22 April 2013 and 5 February 2014.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 230 – 5. WOMEN, WORK AND INCOME INEQUALITY

Annex 5.A1
Additional figures
Figure 5.A1.1. Wives’ employment rates, by husband’s earnings decile
Panel A. Countries reporting gross earnings
Australia

Canada

1989

2003

1991

100%

100%

80%

80%

60%

60%

40%

40%

20%

1

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

Czech Republic
1992

2004

80%

80%

60%

60%

40%

40%

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

Finland
1991

2004

80%

80%

60%

60%

40%

40%

2

3

4

5

6

2004

5

6

1989
100%

1

7
8
9
10
Husband's earnings decile

7
8
9
10
Husband's earnings decile

Germany

100%

20%

6

1987
100%

1

5

Denmark

100%

20%

2007

7
8
9
10
Husband's earnings decile

20%

1

2

3

2007

4

5

6

7
8
9
10
Husband's earnings decile

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 231

Figure 5.A1.1. Wives’ employment rates, by husband’s earnings decile (cont.)
Panel A. Countries reporting gross earnings (cont.)
Israel

Netherlands

1992

2007

1990

100%

100%

80%

80%

60%

60%

40%

40%

20%

1

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

2004

4

1991

100%

80%

80%

60%

60%

40%

40%

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

2005

4

100%

80%

80%

60%

60%

40%

40%

1

2

3

4

5

6

6

1991

2007

100%

20%

5

7
8
9
10
Husband's earnings decile

United States

United Kingdom
1991

7
8
9
10
Husband's earnings decile

1992

2004

100%

1

6

Sweden

Norway

20%

5

7
8
9
10
Husband's earnings decile

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

20%

1

2

3

4

2007

5

6

7
8
9
10
Husband's earnings decile

 232 – 5. WOMEN, WORK AND INCOME INEQUALITY
Figure 5.A1.1. Wives’ employment rates, by husband’s earnings decile (cont.)
Panel B. Countries reporting net earnings

Austria
1994

Belgium
2004

1985

100%

100%

80%

80%

60%

60%

40%

40%

20%

1

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

France
1994

2005

80%

80%

60%

60%

40%

40%

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

Hungary
1991

2005

80%

80%

60%

60%

40%

40%

2

3

4

5

6

2007

5

6

1994
100%

1

7
8
9
10
Husband's earnings decile

7
8
9
10
Husband's earnings decile

Ireland

100%

20%

6

1995

100%

1

5

Greece

100%

20%

2000

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

2004

5

6

7
8
9
10
Husband's earnings decile

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 233

Figure 5.A1.1. Wives’ employment rates, by husband’s earnings decile (cont.)
Panel B. Countries reporting net earnings (cont.)
Italy

Luxembourg

1991

2008

1991

100%

100%

80%

80%

60%

60%

40%

40%

20%

1

2

3

4

5

6

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

Mexico
1989

2004

80%

80%

60%

60%

40%

40%

2

3

4

5

6

6

1992
100%

1

5

7
8
9
10
Husband's earnings decile

Poland

100%

20%

2004

7
8
9
10
Husband's earnings decile

20%

1

2

3

4

2004

5

6

7
8
9
10
Husband's earnings decile

Spain
1995

2007

100%

80%

60%

40%

20%

1

2

3

4

5

6

7
8
9
10
Husband's earnings decile

Note: Figures refer to couple households, with or without other people in the household. Only includes couples with one female
and one male. Unmarried, cohabiting partners are included. Both are of working age (25-64).
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208451

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 234 – 5. WOMEN, WORK AND INCOME INEQUALITY

Annex 5.A2
Decomposition of individual earnings inequality by subgroup

Section 5.3.2 applies an analysis that decomposes inequality (measured by the Theil
coefficient) by subgroup as:
I1 = ∑𝑘 𝑠𝑘 𝐼1𝑘 + ∑𝑘 𝑣𝑘 𝜆𝑘 𝑙𝑛 𝜆𝑘

[1]

where:
𝑠𝑘 is the share of income held by group k,
𝐼1𝑘 is Theil index for group k,
𝑣𝑘 is the population share of group k,
𝜆𝑘 is group k’s mean earnings relative to the population mean.
The equation decomposes overall earnings inequality into a “within” and a “between”
component that allows us to identify the composition and within-group inequality effects
by gender. The first term on the right hand of equation (1) represents the within-group
component, which is a weighted (by income share) sum of the subgroup inequality
values; the second term refers to the between-group component, which depends on
two factors: a) the size of the subgroup and b) its earnings relative to the national average.
The between-component measures the extent of earnings inequality as explained by the
difference between the mean earnings of different subgroups. In the case of the analysis
below, the subgroups refer to eight mutually exclusive working populations: two gender
groups and four employment groups (i.e. full-time full-year, full-time part-year, part-time
full-year and part-time part-year). Changes in aggregate inequality between two years can
then be expressed as:
𝑘
∆Theil = ∑𝑘 ∆𝑇ℎ𝑒𝑖𝑙𝑤
+ ∑𝑘 ∆𝑇ℎ𝑒𝑖𝑙𝑏𝑘 .

[2]

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 235

Table 5.A1.1. Decomposition of individual earnings inequality (Theil index)
Detailed contribution to changes in individual earnings inequality by gender and work status sub-groups
AUS
1985-01

AUT
1994-04

CAN
1987-07

DEU
1989-07

FRA
1994-05

FIN
1987-04

GRC
1995-07

IRL
1994-07

ITA
1987-08

NLD
1990-04

ESP
1995-07

USA
1986-07

Within
Men (FTFY)
Men (FTPY)
Men (PTFY)
Men (PTPY)
Women (FTFY)
Women (FTPY)
Women (PTFY)
Women (PTPY)

0.039
0.037
-0.002
0.001
-0.002
0.017
-0.002
-0.005
-0.005

0.014
0.009
0.000
0.000
0.002
-0.004
0.001
0.000
0.006

0.082
0.040
-0.003
0.005
0.000
0.041
-0.001
0.002
-0.002

0.037
0.024
-0.004
-0.002
-0.002
0.020
-0.003
0.011
-0.008

-0.007
-0.023
-0.007
0.022
0.000
-0.005
0.000
0.006
0.001

0.032
-0.005
0.000
0.023
0.002
0.000
0.002
0.009
0.002

0.060
0.029
0.001
-0.003
0.004
0.033
0.001
-0.006
0.001

0.012
-0.013
0.000
0.001
0.005
0.011
0.002
0.001
0.006

0.069
0.038
0.001
-0.002
0.000
0.030
0.001
0.002
-0.001

0.054
0.054
-0.003
-0.003
0.001
0.004
0.000
0.004
-0.002

-0.017
-0.028
-0.002
-0.001
0.004
0.010
0.000
-0.002
0.004

0.088
0.057
0.000
0.001
0.000
0.034
0.000
0.000
-0.004

Between
Men (FTFY)
Men (FTPY)
Men (PTFY)
Men (PTPY)
Women (FTFY)
Women (FTPY)
Women (PTFY)
Women (PTPY)

-0.025
-0.038
0.008
-0.007
0.007
-0.001
0.004
-0.010
0.013

0.020
0.039
0.001
-0.008
-0.005
0.020
0.001
-0.019
-0.010

-0.036
-0.052
0.005
-0.009
0.006
-0.004
0.006
-0.001
0.013

0.042
0.045
-0.002
-0.012
-0.001
0.032
0.004
-0.024
0.000

-0.009
-0.021
0.002
0.006
0.002
0.008
0.001
-0.007
0.000

0.021
0.012
-0.001
-0.006
-0.006
0.042
-0.002
-0.010
-0.007

0.030
0.045
-0.005
-0.005
-0.004
0.010
-0.006
0.002
-0.007

0.042
0.052
0.003
-0.014
-0.009
0.049
0.002
-0.030
-0.011

0.000
0.013
0.003
-0.012
0.001
0.003
0.001
-0.010
0.001

0.012
0.029
-0.002
-0.001
-0.001
0.015
0.001
-0.034
0.006

0.013
0.013
0.013
-0.004
-0.008
0.022
0.001
-0.012
-0.011

-0.038
-0.056
0.001
-0.013
0.004
0.010
0.006
0.000
0.010

Total change

0.015

0.034

0.046

0.078

-0.015

0.054

0.089

0.054

0.069

0.066

-0.004

0.051

Subgroup

Note: FTFY refers to full-time full-year, FTPY to full-time part-year, PTFY to part-time full-year, PTPY to part-time part-year.
12 http://dx.doi.org/10.1787/888933209045

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 236 – 5. WOMEN, WORK AND INCOME INEQUALITY

Annex 5.A3
A counterfactual decomposition analysis of changes in income inequality

For investigating the impact on inequality of the changing composition of household
employment due to the improved work status of women, we make use of a decomposition
methodology introduced by Firpo et al. (2007) and further developed in Fortin et al.
(2010). By using Recentered Influence Function (RIF) regressions (Firpo et al., 2009),
this methodology represents a generalisation of the Oaxaca-Blinder procedure (Blinder,
1973; Oaxaca, 1973), but can be applied to any distributional parameter other than the
mean, such as quantiles or the Gini.
The Oaxaca/Blinder (OB) decomposition approach allows splitting the total change in
household income inequality over time into two aggregate effects. The first, which is
referred to as the composition effect, is related to the overall changes in the distribution of
observed characteristics of households (e.g. the proportion of households with a female
worker), and the second is the wage structure effect, which captures the overall changes
in the return to households with a female worker. By this setting, the wage structure effect
also includes changes in the return to unobservable characteristics. It can be expressed as
follows.
Oaxaca/Blinder decomposition: ∆𝑌̅ = 𝑋̅𝑡−1 (𝛽̂𝑡 − 𝛽̂𝑡−1 ) + 𝛽̂𝑡 (𝑋̅𝑡 − 𝑋̅𝑡−1 )
(wage structure)

(composition)

Once the aggregate decomposition has been carried out, it is also possible to compute a
more detailed decomposition, subdividing both the composition and wage structure effects
into the contribution of each covariate. One important limitation of the OB decomposition
is that it applies only to differences in the mean. While there have been different methods
developed for decomposing general distributional statistics over the last 15 years, Firpo et
al. (2009) recently proposed a new comprehensive approach. Their idea is to use the
Recentered Influence Function (RIF) for the distributional statistic of interest instead of the
usual outcome variable as the dependent variable in a regression. Consider RIF (y ; ν) is the
influence function corresponding to an observed income y for the distributional statistic of
interest, ν(Fy). The RIF is defined as RIF (y ; ν) = ν(Fy) + IF (y ; ν). The conditional
expectation of the RIF (y;ν) can be modelled as a linear function of the explanatory
variables,
E[RIF(Y ; ν)   X] = X γ + ε
where the parameters γ represent the marginal effect of X on v, which can be
estimated by OLS. Once the RIF regression has been estimated, the estimated coefficients
can be used to perform the detailed decomposition in the same way as in the standard OB
decomposition. Given that RIF regressions for distributional statistics other than the mean
are unlikely to be linear, Fortin et al. (2010) suggest a reweighting procedure on the RIF
regression by reweighting the distribution of characteristics (X) in period 1 to have the
same distribution as in period 0. That is, it is possible to first estimate the counterfactuals
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 5. WOMEN, WORK AND INCOME INEQUALITY – 237

(X10, γ10) from the regression of RIF (Y1; γ) on the reweighted sample. Then the detailed
reweighted decomposition is thus obtained by the following two OB decompositions:
∆̂𝑋 = (𝑋̅1 − 𝑋̅10 ) 𝛾̂1 + 𝑋̅10 (𝛾̂1 − 𝛾̂10 ) = ∆̂𝑋,𝑝 + ∆̂𝑋,𝑒
∆̂𝑆 = 𝑋̅0 (𝛾̂10 − 𝛾̂0 ) + (𝑋̅10 − 𝑋̅0 ) 𝛾̂10 = ∆̂𝑆,𝑝 + ∆̂𝑆,𝑒
∆̂𝑂𝑣𝑒𝑟𝑎𝑙𝑙 = ∆̂𝑋,𝑝 + ∆̂𝑋,𝑒 + ∆̂𝑆,𝑝 + ∆̂𝑆,𝑒
where the composition effect ∆̂𝑋 is divided into a pure composition effect ∆̂𝑋,𝑝 and
the specification error ∆̂𝑋,𝑒 ; and similarly the wage structure effect ∆̂𝑆 is the sum of the
pure wage structure effect ∆̂𝑆,𝑝 and the reweighting error ∆̂𝑆,𝑒 .
In our exercises, we use this procedure to address women’s role in the context of
changes in household income inequality over time.1 Differences in the Gini coefficients
between any two periods are decomposed into the composition and the wage effects, as
mentioned. Nine groups of covariates of interest are considered in the decomposition:
demographic, education of the head, household structure, employment (men and women),
work intensity (men and women) and skilled occupation (men and women).2 We focus, in
particular, on three composition effects due to changes in women’s employment
characteristics. These include 1) the proportion of households with a female worker,
2) the proportion of households with a female working full-time and/or full-year, and
3) the proportion of households with a female employed in a skilled job. Composition
effects due to changes in men’s employment status are also discussed.
Figure 5.A3.1. Composition effect attributed to men’s work characteristics
∆ in % of households w/ male worker

∆ in % of households w/ FT, FY male worker

∆ in % of households w/ male in skilled job

Change in gini (↘)

0.100
0.075
0.050

0.025
0.000
- 0.025

- 0.050
- 0.075

Note: FT: Full-time, FY: Full year.
Source: OECD Secretariat calculations from the Luxembourg Income Study (LIS).
12 http://dx.doi.org/10.1787/888933208468

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 238 – 5. WOMEN, WORK AND INCOME INEQUALITY

Notes
1.

A Stata programme rifreg (www.econ.ubc.ca/nfortin/hmpgfort.htm) is used to fit a
regression model of the re-centered influence function (RIF) of a distributional
statistic of interest (in this case the Gini coefficient) of the marginal distribution of
household disposable income on a series of explanatory variables. Then a standard
Stata programme oaxaca is used to perform a OB decomposition using the RIF as
dependent variable.

2.

Demographic factors include age of the head, race of the head and regional dummies;
employment refers to whether a household has an adult member (man/woman)
working; work intensity indicates whether a household has an adult member
(man/woman) working full-time; occupation refers to whether a household has an
adult member in a skilled job, which corresponds to the first three categories of the 1digit ISCO (i.e. managers, professionals and technicians).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 239

Chapter 6
How does the concentration of household wealth compare across countries?1

This chapter describes the distribution of household wealth for 18 OECD countries, using
a database collected along a set of commonly agreed conventions and classifications.
Both the stocks of household wealth and their degree of concentration are compared
across countries. The analysis sheds light on the demographic characteristics of
households holding wealth, the composition of their assets, their debt as well as the
degree of over-indebtedness among low- and middle-income households. Changes in the
wealth distribution since the onset of the crisis are examined for a subset of countries.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of
such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements
in the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 240 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

6.1. Introduction and key findings
As stressed in the previous chapters, economic growth in the recovery from the 2008
financial crisis has been sluggish in many OECD countries, while market income inequality
has surged. Chapter 2 has highlighted that high and often growing income inequality and
lower social mobility could undermine growth. In this context, whether a high
concentration of household wealth cumulates to increase income inequality and further
undermine long-term growth is an open question. Answering this question first requires
examining the available evidence on the distribution of household wealth.2 This chapter
addresses this through a new set of data (mainly survey-based) collected by the OECD and
based on a set of commonly agreed conventions and classifications (OECD, 2013a).
While inequality issues have increasingly attracted the attention of policy makers and
the media – as witnessed by the debate that has followed the publication of Capital in the
21st Century by Thomas Piketty (2014) and by the very large number of studies
documenting inequality in the distribution of household income – there is far less
evidence on inequality in the distribution of household wealth, both within and between
countries. The main objective of this chapter is to fill this gap and to provide some
comparable information to support this discussion. To that end, the OECD has collected
data on the distribution of household wealth for 18 OECD countries, which extends
previous efforts in this field (OECD, 2008).3
This chapter is organised as follows. Section 6.2 summarises the evidence on the stocks
of household wealth across countries and their degree of concentration; this section also
provides evidence on the joint distribution of household income and wealth, so as to shed
light on how income and wealth are associated with each other at the household level.
Section 6.3 describes the demographic characteristics of households holding wealth.
Section 6.4 considers the composition of their assets, while Section 6.5 focuses on debt and
on the extent of over-indebtedness among low- and middle-income households, a factor that
has played a critical role as a “trigger” to the crisis in many countries, and which
contributed to the slowness of the recovery since then. Section 6.6 describes changes in the
wealth distribution since the onset of the crisis, while Section 6.7 concludes.
The key findings of the chapter are as follows:


Household wealth is much more unequally distributed than household income,
due to the very high levels of concentration at the top of the wealth distribution.
On average, across countries, the 10% of wealthiest households hold half of total
wealth, the next wealthiest 50% hold almost the other half, while the 40% least
wealthy households own little over 3%.



Levels of household wealth are generally higher among households with a head
who is better educated and aged 55 to 64. Non-financial assets, and in particular the
principal residence, constitute the most important assets for most households and
are less equally distributed than financial assets.



In some countries (e.g. Norway and the United States), a large share of
households have high levels of debt, which expose them to sizeable risks in the
event of sudden changes in asset prices, with implications for the vulnerability of
the economic system as a whole. While the incidence of debt increases with a
household's income, over-indebtedness is highest among middle-income groups
and among households headed by young people.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 241



The financial crisis has exacerbated wealth inequality at the top of the
distribution in four of the six countries where trend data are available: the
United States, the United Kingdom, the Netherlands and Italy. Inequality at the
bottom of the distribution increased in five of the six countries available. Over
the longer run, the increase in asset prices, especially financial assets, relative to
consumer prices has been the main driver of changes in the stock of household
wealth and its concentration.

6.2. The distribution of household wealth across OECD countries
This section describes the basic patterns in the distribution of household wealth across
OECD countries, based on the data collected for the OECD Wealth Distribution Database
(see Box 6.1 and Annex 6.A1). The section compares across countries the levels of the
mean and median household wealth, several measures of wealth concentration, and the
association between the distributions of income and wealth.
Box. 6.1. The OECD Wealth Distribution Database
The data included in this chapter have been collected by the OECD and are available to users on the OECD
dissemination platform (OECD.Stat). Estimates referring to the most recent year are currently available for 18 OECD
countries, while estimates referring to more than one year are available for six of them. For seven countries, estimates
have been obtained through a questionnaire sent and completed by national contact points in statistical offices and
agencies (e.g. central banks) that regularly collect micro-level information on the distribution of household wealth;
among these seven, estimates for Australia, Canada, Korea, the United Kingdom (limited to Great Britain) and the
United States are based on dedicated household wealth surveys, while those for the Netherlands and Norway are
based on tax and administrative records. For 11 countries (participants in the Eurosystem Household Finance and
Consumption Survey, henceforth HFCS), estimates for the most recent year available have been computed by the
OECD based on the public use file provided by the European Central Bank (complemented, in the case of Italy and
the Netherlands, by estimates for earlier years provided by national contact points). This OECD data collection builds
on previous efforts to provide more comparable measures of the distribution of household wealth as pursued by the
Luxembourg Wealth Study (Sierminska et al., 2006) and by the European Central Bank (2013).
The data used in this chapter have the following characteristics:



They refer to the distribution of non-financial and financial assets and liabilities across households
(rather than across persons or adults), with no adjustment made to reflect differences in household size
(which is the convention typically used when analysing the distribution of household income). The data
also refer to the assets and liabilities held by private households resident in the country. Countries may
differ in how “households” are defined (e.g. most HFCS countries rely on a relatively restricted
definition, which requires household members to maintain relations of financial interdependence).



Assets and liabilities are classified in a consistent manner across countries, based on the nomenclature
proposed by the OECD Guidelines for Micro-Statistics on Household Wealth (OECD, 2013a). This
nomenclature distinguishes between five categories of non-financial assets, eight categories of
financial assets, and three categories of financial liabilities. Among financial assets, the stock held in
the form of “pension schemes related to employment” is reported as a separate category: data on this
category are described in Box 6.3, while the data shown elsewhere in this chapter refers to a narrower
definition of household wealth that excludes these assets from the total.



Information is collected on the net wealth of households broken down by housing status (three types of
household), age of the household reference person (six groups), number of household members (five
groups), household type (six groups), education of the reference person (four groups), main source of
income (five groups), as well as by wealth and income quintiles (with additional breakdowns for the top
10%, 5% and 1% of the distribution). Information is also collected on the share of households holding
various types of assets and liabilities; on the mean value of assets and liabilities among households
holding them; on the joint distribution of household wealth and income across household quintiles; and
on the extent of over-indebtedness across households (based on two measures of over-indebtedness).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 242 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Box. 6.1. The OECD Wealth Distribution Database (cont.)
Despite efforts to ensure common treatments and classifications across countries, the micro data presented
in this chapter are affected by differences that limit their comparability. Three of the most important are:



Differences between countries in the year the data was collected (ranging between 2010 and 2013, for
the most recent observation), and in the detailed period of fieldwork within the year. These
differences are important, as the recent period featured significant swings in asset prices, which affect
the values of net wealth reported here. While data for the most recent year shown are all expressed in
constant prices referring to 2005, this does not remove this source of heterogeneity.



Differences in the degree of oversampling of rich households across countries.* These differences
(which range between no oversampling in the case of Australia and Canada to large oversampling in
the United States and Spain) do matter for international comparisons of both levels and
concentrations of household wealth as, in all countries, most wealth is typically held by the richest
households.



Pension wealth in the form of social security entitlements and defined benefit occupational pensions
are not included in the main wealth measures provided in this chapter. (Model-based estimates of
future pension wealth from public and mandatory private sources are shown in Annex 6.A3).

More generally, it should be stressed that all sources of information on household wealth have their own
limits (as described in the annexes to this chapter). The hope is that, as statistical offices and agencies collecting
wealth data become more aware of differences in their survey practices and of how these differences affect
international comparisons, a process of convergence in statistical practices would intensify in the future, in
particular as the OECD guidelines (OECD, 2013a) are more widely implemented.
* A measure of the degree of oversampling of rich households is shown in Table 6.A1.1; this is computed as the difference between
the share of wealthy households in the sample and the share in the population.

Levels of household wealth across countries
The levels of household wealth are described in Figure 6.1 through measures of the
mean and median household net wealth per household. The highest mean levels of wealth
are observed in Luxembourg, the United States, Canada, Australia, the United Kingdom
and Spain, while the Slovak Republic, Finland, Greece, Norway and the Netherlands
record the lowest levels. The net wealth of the median household may be considered as
more representative of the conditions of the typical household, and is less affected by
inaccurate measurements of wealth at the top of the distribution.4 Shifting from mean to
median net wealth does not change the ranking of countries dramatically (the correlation
between the two variables is equal to 0.79) but it does lead to strikingly large differences
in country ranking for the United States, Austria and Germany: the United States falls
from second in terms of mean net wealth to 16th in terms of median net wealth, while
Germany falls from 12th to 17th and Austria from 9th to 14th, suggesting that net wealth is
unequally distributed in these countries (see below).5

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 243

Figure 6.1. Mean and median net wealth per household in micro data
2010 or latest available year, values in 2005 USD
Mean (↘)

Median

700 000
600 000
500 000
400 000

300 000
200 000
100 000
0
LUX USA CAN AUS GBR ESP BEL

ITA AUT KOR FRA DEU PRT NLD NOR GRC FIN SVK

OECD18

Note: Wealth values are expressed in 2005 USD by, first, expressing values in prices of the same year (2005) through consumer
price indexes and, second, by converting national values into a common currency through the use of purchasing power parities
for household consumption.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208474

Household wealth can also be expressed in relative terms by looking at values
expressed as a multiple of household income. This measure intuitively corresponds to the
number of years a household could maintain its living standard (as measured by its
income) in the future by drawing down its accumulated wealth.6 Based on this measure,
the stock of net wealth varies between three and nine times the value of household
income on average (similar estimates are reported by Brzozowki et al., 2010; Domeij and
Floden, 2010; Fuchs-Schündeln et al., 2010; Jappelli and Pistaferri, 2010; Landais et al.,
2011; Maestri et al., 2014). A comparison of Figures 6.1 and 6.2 shows that the country
rankings do not change radically for most countries whether one considers absolute or
relative mean net wealth.7 Similarly, considering relative rather than absolute median net
wealth does not radically change the ranking of countries.8
Figure 6.3 presents the mean net wealth per household as measured from the micro
data based OECD questionnaire used for this chapter and in the National Accounts
statistics on household balance sheets.9 To compare the two sources, it is critical that the
same types of assets and liabilities are included in both. However, in 13 out of the
18 OECD countries covered in this chapter, OECD National Account statistics on the
non-financial assets of households exclude the value of land, which represents about 50%
of non-financial assets in the other five OECD countries where data are available.10 The
SNA values shown in Figure 6.3 are based on estimates of the value of land drawn from
the European Central Bank (2013) and on the amounts reported in household balance
sheets for the remaining countries. Conversely, the value of vehicles and other consumer
durables are excluded from micro-based measures of non-financial assets as they are not
considered as household wealth in the national accounts.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 244 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.2. Mean and median net wealth relative to household income
2010 or latest available year, values in 2005 USD
Ratio of median net wealth to income

Ratio of mean net wealth to income (↘)

10
9
8
7
6
5
4
3
2
1
0
ESP LUX

ITA

PRT GBR BEL AUS AUT USA SVK FRA GRC CAN DEU NLD

FIN

NOR

OECD17

Note: The ratio of mean net wealth to income is calculated by dividing the mean net wealth by the mean gross income. The ratio
of median wealth to income refers to the median net wealth divided by the mean gross income of the third quintile of net wealth,
which is a convenient proxy for the income of the households with median wealth. Data on household income for Korea are not
available.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208485

Figure 6.3. Net wealth per household in micro data and national accounts
2010 or latest available year, values in 2005 USD
Mean household net wealth - National Accounts

Mean household surveys (↘)
450 000
400 000
350 000
300 000

250 000
200 000
150 000
100 000
50 000
0

USA

CAN

AUS

GBR

ESP

BEL

ITA

AUT

FRA

DEU

NLD

FIN

SVK

OECD13

Note: Wealth values are expressed in 2005 USD: first, wealth values in different years are expressed in prices of the same year
(2005) through consumer price indexes; and second, national values are converted into a common currency through the use of
purchasing power parities for household consumption. For the 13 OECD countries shown, estimates of the value of land are
included in the National Account wealth measure, while estimates of the value of consumer durables (vehicles and others) are
excluded for the survey measure.
Source: OECD Wealth Distribution Database; OECD National Accounts Database; and European Central Bank (2013), “The
Eurosystem Household Finance and Consumption Survey: Results from the First Wave”, Statistics Paper Series, Vol. 2, April.
12 http://dx.doi.org/10.1787/888933208495

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 245

Overall, Figure 6.3 shows that, for the 13 countries for which comparable data can be
obtained, the values of mean household wealth reported in the two sources are fairly
similar (on average, net wealth per household is USD 275 000 when measured in the
SNA and USD 263 000 in the micro data). These results are in line with those obtained
from the European Central Bank (2013) and Fesseau et al. (2013). This comparison is
however only approximate, as other factors beyond those that could be controlled affect
the comparisons between macro and micro totals shown in Figure 6.3.11
Over a larger set of 17 OECD countries described in Table 6.1, net wealth as
measured by micro sources is similarly found to be 10% lower than National Account
estimates. This mostly reflects the under-estimation of financial assets in micro sources.
Liabilities also appear to be under-estimated in the micro data, but this effect is not large
enough to compensate for the first one. In countries for which the data is available,
estimates of non-financial assets appear to be broadly consistent across the two
data sources.
Table 6.1. Ratios of household assets and liabilities between micro sources and national accounts
2010 or latest available year, values in 2005 USD
Non-financial assets

Financial assets

Liabilities

Net wealth

Australia

1.0

..

0.7

1.4

Austria

1.1*

0.4*

0.4*

0.9*

Belgium

1.2*

0.6*

0.8*

0.9*

Canada

1.3

0.9

0.6

1.3

Finland

1.0*

0.4*

0.9*

0.8*

France

0.7

0.8

0.5

0.7

Germany

0.8*

0.4*

0.7*

0.7*

Greece

..

0.3

0.3

..

1.0*

0.2*

0.4*

0.8*

Lux embourg

..

0.4

0.7

..

Netherlands

Italy

0.8

0.8

1.0

0.8

Norw ay

..

1.0

1.0

..

Portugal

..

0.5

0.4

..

Slov ak Republic

0.8*

0.4*

0.4*

0.8*

Spain

0.8*

0.4*

0.6*

0.8*

United Kingdom

1.2

0.7

0.7

1.2

United States

1.1

1.0

0.9

1.1

OECD17

1.0

0.6

0.6

0.9

Note: “..” refers to non-available data. “*” refers to data from the European Central Bank (2013). SNA data on financial assets
exclude voluntary life insurance and private pensions and are not available for Korea.
Source: OECD Wealth Distribution Database and European Central Bank (2013).
12 http://dx.doi.org/10.1787/888933209053

Several factors may explain cross-country differences in levels of household wealth,
including demographic factors (i.e. the age structure of the population), the magnitude of
the savings rate as well as changes in asset prices relative to consumer prices. Regarding
demographic factors, Cowell (2013) conclude that, based on data from the Luxembourg
Wealth Study (www.lisdatacenter.org/our-data/lws-database/), the cross-country
differences in wealth distribution in Finland, Italy, Sweden, the United Kingdom and the
United States cannot be explained by differences in age, working status, household
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 246 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
structure, education or income: the authors calculate the wealth distributions that would
prevail if these countries had the same population distributions across several
characteristics, and find a similar ranking of countries in terms of net wealth. A similar
finding is also reported by Maestri et al. (2014).
Table 6.2 sheds light on the role played by asset prices and the savings rate. It shows
the average growth rate in the prices of dwellings and listed shares (relative to the
increase in the private consumption deflator) over the longest period of time available for
each country, as well as the national savings rate and the savings rate of households
(which is more relevant to the analysis but is typically available over a much shorter time
span). Based on these data, Figure 6.4 highlights a strongly significant positive
correlation between the median net wealth of households and the annual real growth rate
of house prices over the period covered by Table 6.2. This suggests that the rapid
appreciation of property prices vis-à-vis consumption prices in Australia, Belgium,
Canada, Spain and the United Kingdom has been a key factor leading to higher median
wealth compared to other OECD countries. Norway is the only country where the rapid
increase in house prices has not translated into higher median net wealth, possibly due to
a relatively lower household savings rate.12 Conversely, net wealth is relatively high in
Italy despite the recent fall in housing and stock prices, reflecting the high rate of
household savings and home ownership (Jappelli and Pistaferri, 2000). No significant
correlation appears between the real growth of asset prices, the national or the household
savings rate and measures of median wealth.
Table 6.2. Growth rate of asset prices and the savings rate, for the periods 1970-2012 and 2000-13
Period

Annual growth rate of real prices
Houses

Stocks

National savings rate

Households savings rate
Av erage

Period

Australia

1970-2012

2.8

0.3

9.4

9.4

1970-2012

Austria

2000-2013

1.5

3.3

11.3

9.4

1995-2012

Belgium

1986-2013

4.0

3.0

10.8

12.1

1995-2012

Canada

1970-2013

2.5

2.0

7.3

7.6

1981-2013

Finland

1970-2013

1.2

4.5

10.8

1.4

1995-2012

France

1970-2013

2.0

1.9

7.1

11.5

1978-2012

Germany

1970-2013

0.0

2.2

9.7

10.5

1995-2012

Greece

1997-2013

0.9

-5.3

-3.4

-5.4

2005-2012

Italy

1970-2013

1.2

-1.9

7.8

11.7

1990-2013

Korea

1986-2012

-1.1

3.5

20.4

4.4

2010-2013

Lux embourg

2007-2013

1.5

-15.7

3.4

10.0

2006-2012

Netherlands

1970-2013

2.0

1.6

14.1

9.7

1990-2012

Norw ay

1986-2013

2.7

7.1

18.0

4.2

1978-2013

Portugal

1988-2013

0.2

0.6

6.7

3.0

1995-2013

Slov ak Republic

2005-2013

1.8

-13.0

2.8

4.0

1995-2008

Spain

1987-2013

3.5

4.5

9.8

6.3

2000-2012

United Kingdom

1970-2013

3.3

1.7

4.2

1.9

1990-2012

United States

1970-2012

1.1

3.0

5.8

8.1

1970-2012

Sources: OECD National Accounts Statistics Database (2014).
12 http://dx.doi.org/10.1787/888933209066

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 247

Figure 6.4. Median net wealth and change in house prices
2010 or latest available year, values in 2005 USD

200 000

150 000

Median net wealth

250 000

AUS

ESP
BEL
GBR

ITA

CAN

KOR
GRC

FRA

PRT

100 000

AUT

NOR

SVK

FIN
50 000

DEU

USA

NLD

Annual growth rate of house prices

0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

Note: The median net wealth is expressed in 2005 USD; changes in house prices are measured by the annual growth rate of
house prices over the period 1970-2013 for most countries (refer to Table 6.2 for corresponding periods).
Sources: OECD Wealth Distribution Database and OECD National Accounts Statistics Database, 2014.
12 http://dx.doi.org/10.1787/888933208505

How does wealth inequality compare across countries?
In all OECD countries, the wealth distribution is much more concentrated at the top
than is the income distribution, and it is therefore much more unequal. Figure 6.5 shows
the average shares of household disposable income and wealth for each decile of the
respective distribution, across the subset of 18 OECD countries for which both data sets
are available.13 On average, the top decile of the distribution of household disposable
income accounts for about 25% of total household income, while the top decile accounts
for about 50% of the distribution of household wealth. The wealth share of the top centile
in the wealth distribution is almost as large as the income share of the top decile of the
income distribution.
Figure 6.5. Distribution of household disposable income and net wealth across deciles
Average of 18 OECD countries, 2010 or latest available year
Average share of income deciles across OECD countries (↗)

Average share of wealth deciles across OECD countries

60

%

50
40

Top 5 %

30

Top 1 %

20
10
0
- 10

D1

D2

D3

D4

D5

D6

D7

D8

D9

D10

Source: OECD Wealth Distribution Database and OECD Income Distribution Database.
12 http://dx.doi.org/10.1787/888933208517
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 248 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Other studies also find that wealth inequality is much greater than income inequality.
For instance, Maestri et al. (2014) calculate Gini indices over the wealth distribution of
between 0.55 and 0.80. However, this particular inequality index is not well-suited for the
study of the wealth distribution due to the large fraction of households with zero or negative
wealth; in these circumstances the Gini index is no longer bounded to one (Amiel et al.,
1996; Cowell, 2013). In addition, Morrisson and Murtin (2013) show that the percentage of
people with a zero outcome (e.g. wealth in this context) has a large influence on the level of
the Gini, which blurs the interpretation of cross-country differences or changes in inequality
over time. In practice, Table 6.3 shows that in seven countries among those analysed the
first quintile of the wealth distribution has negative wealth.
Table 6.3. Selected indicators of the distribution of household net wealth
2010 or latest available year, values in 2005 USD
Net household wealth, 2010 or latest available year
Average of
Mean

Median

Bottom

the

quintile

three middle

Ratio
Top quintile

Top 10%

Top 5%

Top 1%

quintiles

Ratio

(wealth of top 5% - (median wealth median wealth)/

bottom quintile)/

median

median

Australia

335 299

205 404

8 398

210 963

1 035 640

1 507 217

2 158 015

4 461 272

9.5

1.0

Austria

268 275

74 466

-7 578

106 874

1 051 613

1 684 393

2 659 268

6 560 503

34.7

1.1

Belgium

313 517

188 149

2 767

213 541

1 014 312

1 461 306

2 083 538

4 181 874

10.1

1.0

Canada

337 238

152 818

4 906

177 999

1 147 721

1 697 591

2 461 392

5 219 761

15.1

1.0

Finland

133 468

71 762

-7 636

82 060

440 670

611 660

831 889

1 685 064

10.6

1.1

France

211 330

106 497

- 868

125 263

778 342

1 152 393

1 686 153

4 145 304

14.8

1.0

Germany

191 554

53 896

-4 870

82 680

782 676

1 214 445

1 874 432

5 033 122

33.8

1.1

Greece

164 734

114 377

2 579

118 902

470 332

643 987

850 460

1 405 733

6.4

1.0

Italy

273 583

171 794

5 495

175 075

851 654

1 239 524

1 772 755

3 954 572

9.3

1.0

Korea

249 698

135 334

4 178

153 624

783 330

..

..

..

1.0

Lux embourg

644 850

360 251

2 125

366 654

2 208 512

3 403 744

5 325 290

14 800 000

13.8

1.0

Netherlands

172 004

34 194

-28 495

70 460

677 137

1 024 772

1 534 568

4 105 305

43.9

1.8

Norw ay

168 740

89 449

-43 260

100 325

586 002

845 275

1 221 517

3 124 719

12.7

1.5

Portugal

189 213

92 699

1 090

102 344

652 251

1 012 934

1 569 351

4 095 892

15.9

1.0

Slov ak Republic

112 608

86 642

18 470

90 820

278 106

374 125

498 284

902 533

4.8

0.8

Spain

324 087

205 238

10 422

217 442

987 304

1 434 765

2 049 090

5 003 514

9.0

0.9

United Kingdom

331 722

187 380

8 935

199 419

1 051 419

1 544 346

2 267 314

5 789 661

11.1

1.0

United States

411 044

56 724

-19 059

87 430

1 811 626

3 138 331

5 199 815

15 043 278

90.7

1.3

OECD18

268 498

132 615

-2 356

148 993

922 703

1 411 224

2 120 184

4 654 302

20.4

1.1

..

Note: “..” refers to non-available data. Values are expressed in 2005 USD based on purchasing power parities and consumer
price indexes.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209076

A more convenient, albeit partial, measure of wealth inequality within countries is the
ratio between mean and median net wealth. As is shown on Figure 6.6, mean net wealth is
2.5 times larger than median net wealth across the 18 OECD countries covered in this
chapter; the ratio between the two measures is above seven in the United States, around
five in the Netherlands, four in Germany and Austria, and close to two in most other
OECD countries. This compares to values of around 1.2 in the case of household income
for most OECD countries (1.1 in Nordic countries, 1.3 in the United States).
To complement the inequality measure for the entire wealth distribution shown
above, it is of interest to consider measures of inequality in the bottom and top part of the
wealth distribution. In Table 6.3, inequality in the top half of the wealth distribution is
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 249

measured by the difference in wealth between the top 5% and the median household, as a
share of median wealth; this relative difference is, on average, equal to 20 across the
18 OECD countries but much higher in the United States, the Netherlands, Austria and
Germany (with ratios of 91, 44, 35 and 34 respectively).14 Hence, the higher degree of
wealth inequality observed in these four countries is mainly due to high wealth
concentration at the top of the distribution. Conversely, the relative difference between
the wealth of the median household and the average wealth of households in the bottom
quintile of the distribution captures inequality at the bottom part of the distribution. On
average, this ratio is equal to 1.1, about one-twentieth of the relative difference between
the top 5% and the median. Hence, inequality at the bottom of the distribution is much
smaller than inequality at the top, as measured by these wealth relative differences.
Inequality in the lower half of the distribution is the largest, by decreasing order, in the
Netherlands, Norway and the United States.
Figure 6.6. Ratio of mean and median net wealth per household in survey data
2010 or latest available year
8
7
6
5
4
3
2
1
0
USA

NLD

AUT

DEU

CAN

PRT

FRA

NOR

FIN

KOR

LUX

GBR

BEL

AUS

ITA

ESP

GRC

SVK

OECD18

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208523

Table 6.3 also reports the absolute values of wealth at the top of the distribution in the
various countries. For both the top 10% and top 1% of wealthiest households, the largest
wealth levels are recorded in the United States, Luxembourg and Austria. Figure 6.7
presents the wealth shares of the top percentiles in the wealth distribution. On average,
across the 17 countries shown, the top 10%, 5% and 1% wealthiest households own 50%,
37% and 18% of total household wealth, respectively. By comparison, the bottom 60% of
the wealth distribution owns only 13% of total household wealth, on average. Based on
the latter measures, wealth is the most concentrated at the top of the distribution in the
United States, Austria, the Netherlands and Germany. In some but not all countries, these
measures probably underestimate the wealth shares of very rich households, as other
studies report larger figures for the top percentiles, although they are not necessarily
using the same concept for household wealth (Box 6.2).
The ownership of financial assets features prominently in the set of factors that have a
large influence on wealth inequality and most particularly on its concentration at the top.
As depicted on Figure 6.8, the real appreciation of stock prices is associated with a
relatively larger wealth share of the top 10%, and a similar relationship is found with the
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 250 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
top 1% and 5%.15 This suggests that higher prices of financial asset have contributed to
higher wealth inequality and concentration at the top end of the distribution. The data
presented here, however, capture only partially the rebound of asset prices since the end
of the Great Recession.
Figure 6.7. Wealth shares of top percentiles of the net wealth distribution
2010 or last available year
Top 5%

Top 10% (↘)

Top 1%

Bottom 60%

80
%
70
60
50
40
30
20
10
0

USA

AUT

NLD

DEU

PRT

LUX

CAN

NOR

FRA

GBR

FIN

AUS

ITA

BEL

ESP

GRC

SVK

OECD17

Note: The bottom 60% refers to the share of quintiles I, II and III in the total wealth.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208539

Figure 6.8. Wealth share of the top 10% and annual growth rate of real stock prices
2010 or latest available year

Net wealth of the top 10%

3500 000
3000 000

2500 000
2000 000
1500 000

USA

AUS
ITA

CAN AUT
GBR
BEL

1000 000

GRC

SVK

500 000

PRT

ESP

DEU

NLD

NOR
FRA
FIN

Annual growth rate of real stock prices (%)
0

-14

-12

-10

-8

-6

-4

-2

0

2

4

6

8

Note: The mean net wealth for the top 5% distribution is expressed in 2005 USD and the growth rate of stock prices is
provided for the period 1970-2012.
Source: OECD Wealth Distribution Database and OECD National Accounts Statistics Database, 2014.
12 http://dx.doi.org/10.1787/888933208541

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 251

Box 6.2. Top wealth shares across economic studies
A number of recent studies on household wealth – based on different indicators, units of analysis, data
sources and wealth concepts – suggest a higher degree of wealth concentration at the top of the distribution
than that found using the data collected by the OECD and presented in this chapter. The table below
summarises the available information on the wealth shares of the top decile and top centile (as well as the
average net wealth) provided by the Crédit Suisse Reports (2010, 2011, 2012, 2013, 2014), Pi ketty (2014)
and Vermeulen (2014).
Piketty’s study presents historical time series stretching back to the 19 th century on the level and
distribution of private wealth. The data on shares for the top 10% and top 1% begin in 1810 and cover
France, Great Britain, Sweden and the United States. Piketty’s estimates of the top wealth shares are
generally based on administrative records, e.g. tax returns (for France) and estate returns (for Great Britain).
For these two countries, the estimates of the wealth shares of the top 10% and 1% reported by Piketty are
slightly lower than those reported in this chapter, reflecting, inter alia, the inclusion of pension wealth held in
occupational pension funds in Piketty’s analysis. Data on the United States are from the SCF survey, which is
the source also used in this chapter, and as a consequence both estimates are relatively clos e.
Measuring wealth at the top of the wealth distribution through household surveys is intrinsically difficult,
as wealthy households typically under-report their wealth. Moreover, household surveys suffer from various
degrees of non-response, and the non-response bias is particularly large when looking at the top 1% of the
distribution (Vermeulen, 2014). The combination of under-reporting and under-sampling (or unit nonresponse) tends to reduce the measured degree of concentration at the top of the distr ibution. To address this
issue, Vermeulen relied on the same type of extrapolation used by Piketty and Crédit Suisse, assuming that
the distribution for the top 1% follows a Pareto distribution, which displays a heavy right tail. For the
countries covered by the three studies, estimates of the wealth share of the top 1% are close, but above those
reported in this chapter.
The under-sampling bias is addressed in 14 out of 18 OECD countries covered in this chapter, as
households from the top 10% of the distribution are typically oversampled in wealth surveys (and
11 countries oversample the top 1% percentile, see Table 6.A1.1). This raises the possibility that the
measured degree of wealth concentration at the top may simply reflect the design features of the surveys. To
gauge the influence of oversampling on wealth concentration, the figure below plots the relationships
between the wealth shares of the top 10% (left-hand panel) and of the top 1% (right-hand panel) and the
degree of oversampling. No significant correlation shows up when considering the top 10%, while the
positive correlation observed for the top 1% falls to zero when the United States is excluded from the
sample. These results suggest that the different sampling practices of the various national surveys do not
systematically bias cross-country comparisons of wealth concentration at the top of the distribution
presented in this chapter.
As a consequence, the result that the wealth shares of the top categories in the OECD Wealth Distribution
Database are lower than in other studies should reflect either under-reporting problems in household surveys
or differences in methodological assumptions (e.g. the use of households or individuals – as in Crédit Suisse
reports – as units of analysis, or the use of an extended net wealth concept in Piketty). This conclusion is
further highlighted by the fact that, although the same SCF source is used across the four studies for top
shares in the United States, OECD’s measures are ten percentage points below the others: differences in
scope and methodology across the various studies, rather than different in sources, appear to represent the
bulk of the difference in the estimated top shares.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 252 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Box 6.2. Top wealth shares across economic studies (cont.)
The concentration of household wealth at the top of the distribution in different studies
Top 10 %
OECD

TOP 1 %

Crédit Suisse

Piketty

(2014)

OECD

Average net wealth (USD)
OECD

Crédit Suisse

Piketty

(2014)

Vermeulen

(2012 or

Crédit Suisse

latest available year)

(2014)

1, 2

Australia

45%

51%

..

13%

21%

..

..

335 299

430 777

Austria

62%

64%

..

24%

29%

..

..

268 275

142 675

Belgium

44%

47%

..

13%

17%

..

..

313 517

181 083

Canada

50%

57%

..

15%

22%

..

..

337 238

176 848

Finland

45%

55%

..

12%

21%

..

..

133 468

112 937

France

50%

53%

62%

18%

21%

24%

19%

211 330

201 245

Germany

59%

62%

..

25%

28%

..

33%

191 554

211 049

Greece

39%

56%

..

8%

27%

..

..

164 734

111 405

Italy

45%

51%

..

14%

20%

..

20%

273 583

255 890

Lux embourg

51%

..

..

22%

n.a

..

..

644 850

340 836

Netherlands

60%

57%

..

24%

27%

..

17%

172 004

210 233

Norw ay

50%

66%

..

19%

29%

..

..

168 740

358 655

Portugal

53%

58%

..

21%

..

..

..

189 213

98 428

Slov ak Republic

33%

..

.;

8%

..

..

..

112 608

28 630

Spain

43%

56%

..

15%

27%

..

..

324 087

134 824

United Kingdom

47%

54%

71%

17%

23%

28%

..

331 722

292 621

United States

78%

75%

72%

38%

38%

34%

37%

419 815

347 845

Note: “..” refers to non available data.
1. Crédit Suisse reports average net wealth at individual levels.
2. For most countries, Credit Suisse estimated wealth levels for missing years by inflating data on the latest year available
with house prices, financial assets or GDP per capita growth.
Source: OECD Wealth Distribution Database; Crédit Suisse, Global Wealth Databook, various issues; Piketty, T. (2014),
Capital in the 21st Century; and Vermeulen (2014), “How Fat Is the Top Tail of the Wealth Distribution?”, ECB Working Paper
No. 1692.
12 http://dx.doi.org/10.1787/888933209130

Top wealth shares and the degree of oversampling
Panel A. Top 10% of the distribution

Panel B. Top 1% of the distribution
45%

USA

80
70

AUT

60

DEU

PRT

LUX
GBR

50

BEL

GRC

40

Share of wealth owned by top 1%

Share of wealth owned by top 10%

90%

SVK

ESP

FRA

FIN

30
20
10

USA

40
35
30
25

AUT
DEU
PRT LUX

20 GBR

FRA

BEL

15

FIN

10

GRC

5
0

0
-4

0

4

8

12

Oversampling rate on top 10%

16

20
%

0

200

400

600

800

1 000

Oversampling rate on top 1%

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208685

To provide a synthetic overview of wealth patterns among the sample of OECD
countries analysed here, countries are clustered along the two dimensions of average wealth
and wealth inequality. Figure 6.9 plots mean net wealth as a share of income (on the vertical

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 253

axis) against the ratio of the mean wealth of the top 5% to the median wealth as a measure of
wealth inequality (on the horizontal axis).16 This suggests four clusters of countries:


Austria, Germany, the United States and the Netherlands, with average or low
wealth-to-income ratios but high wealth inequality.



Spain, Italy, Luxembourg and Portugal, with high wealth-to-income ratios and
low wealth inequality.



Australia, Belgium, Canada, France, Greece, the Slovak Republic and the United
Kingdom, with average wealth-to-income ratios and low inequality.



Finland and Norway, with low mean wealth-to-income ratios and low inequality.
Figure 6.9. Mean net wealth and top wealth inequality
10

LUX
ITA

8

PRT

7

GBR
AUS
SVK

6

BEL
CAN

AUT

GRC

5
4

FIN
NOR

3

USA

FRA

OECD 18 (wealth levels)

Wealth levels (ratio of mean net wealth to income)

ESP
9

2

DEU
NLD

OECD17 (Top wealth inequality)

1
0
0

10

20

30

40

50

60

70

80

90

100

Top wealth inequality

Note: Wealth inequality is measured by the difference between the wealth of the top 5% and median wealth, which is then
divided by median wealth.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208556

What is the relationship between income and wealth inequality?
To shed light on how income and wealth inequalities are jointly determined and
interact with each other, the joint distribution of the two variables across households is
examined. As an example, Table 6.4 shows the joint distribution of income and wealth in
the United States in 2013: for each quintile of the wealth distribution, the table reports the
shares of households belonging to the various quintiles of the income distribution. A
perfect correlation between income and wealth would translate into a concentration of
people on the diagonal cells, while a zero correlation would spread wealth quintiles
equally across income quintiles. A priori, it may be expected that income and wealth are
positively correlated as higher wealth holdings, especially financial assets, generate
higher income from capital and, conversely, higher income can be used to accumulate
more wealth. But it is also possible that high wealth reflects elements that are only poorly
correlated to household income (such as inheritances or changes in asset prices). For the
United States, Table 6.4 shows that households in the bottom quintile of the wealth
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 254 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
distribution are five times more numerous in the bottom of the income distribution than in
the top quintile. Conversely, households in the top quintile of the wealth distribution are
ten times more numerous in the top of the income distribution than in the bottom quintile.
Table 6.4. The joint distribution of income and wealth in the United States, 2013
Wealth quintile I

Wealth quintile II

Wealth quintile III

Wealth quintile IV

Wealth quintile V

Total

Income quintile I

0.08

0.05

0.04

0.03

0.01

0.20

Income quintile II

0.05

0.06

0.04

0.03

0.02

0.20

Income quintile III

0.04

0.05

0.05

0.04

0.02

0.20

Income quintile IV

0.02

0.03

0.05

0.05

0.04

0.20

Income quintile V

0.01

0.01

0.03

0.04

0.11

0.20

Total

0.20

0.20

0.20

0.20

0.20

1.00

Note: For each income (wealth) quintile, the sum of values across wealth (income) quintile equals 0.20.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209082

The observation made for the United States can be generalised to all OECD countries.
As shown in Figure 6.10, low-wealth households are also more likely to be low-income
households (Panel A) while, conversely, high-wealth households are also typically highincome (Panel B). The concentration of households that are both income-poor and
wealth-poor is largest in Canada, Germany and Luxembourg (Panel A); while the
concentration of high-wealth and high-income households is largest in Canada, France,
Italy and the United States. The United Kingdom, the Netherlands and Australia display a
relatively low degree of correlation between income and wealth for both poor and rich
households.
A synthetic measure of association between income and wealth that considers all
groups of the two distributions is presented in Figure 6.11. It shows the rank correlation
across quintiles of households in the income and wealth distributions. The correlation
between the two distributions would be one if all households ranked in the bottom
quintile of the wealth distribution fell in the bottom quintile of the income distribution,
and similarly for all subsequent quintiles; a zero rank correlation between quintiles in the
income and wealth distributions would, on the contrary, imply that these shares were both
equal to 20%.17
In practice, the rank correlation is positive and significant in all countries. The rank
correlation across the 17 countries for which data are available is equal to 0.37, which
suggests a moderate degree of mobility across wealth and income. On average, about
36% of the households belonging to the bottom 20% of the income distribution are also
located in the bottom 20% of the wealth distribution, while 45% of households belonging
to the top 20% of the income distribution belong to the top 20% of the wealth
distribution. However, 18% of households in the bottom wealth quintile belong to the two
highest income quintiles, while a similar share of households in the top wealth quintile
belong to the two lowest quintiles of the income distribution. Consistently with the results
shown in Figure 6.10, the correlation between wealth and income is the highest in
Canada, Germany, Luxembourg, France and Italy, while it is the lowest in the United
Kingdom, the Netherlands and Australia.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 255

Figure 6.10. Income distributions of the bottom and top wealth quintiles across OECD countries
Panel A. Income distribution of the bottom wealth quintile1
Income quintile I (↗)

Income quintile II

Income quintile II

Income quintile IV

Income quintile V

FIN

ITA

BEL

0.2

0.15

0.1

0.05

0

GBR

NLD

AUS

ESP

PRT

NOR

GRC

SVK

USA

FRA

AUT

LUX

DEU

CAN

Panel B. Income distribution of the top wealth quintile
Income quintile I (↗)

Income quintile II

Income quintile III

Income quintile IV

FRA

DEU

Income quintile V

0.2

0.15

0.1

0.05

0

CAN

FIN

LUX

ITA

AUT

NOR

GRC

ESP

USA

PRT

SVK

BEL

AUS

NLD

GBR

1. Data on Canada for the quintile IV and V are not available separately.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208563

While the relationship between income and wealth inequality is positive
(i.e. countries with a larger degree of income inequality also tend to display a higher
concentration of wealth), this link is not very robust.18 Across the 17 OECD countries,
there is a positive and significant relationship between the income ratio of the 90 th and
10th percentiles of the income distribution and the wealth share of the top 1% in the
wealth distribution.19 Perhaps due to measurement problems or the small sample size, this
empirical relationship is, however, weak.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 256 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.11. Rank correlation between the distribution of net wealth and disposable income
Average

Lower bound

Upper bound

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

GBR

NLD

AUS

SVK

NOR

PRT

ESP

GRC

FIN

BEL

USA

AUT

ITA

FRA

LUX

DEU

CAN

OECD17

Note: The confidence interval (upper/lower bound values) is obtained from Monte Carlo simulations run on a smoothed
joint distribution of income and wealth within each country.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208576

6.3. Demographic characteristics of households with positive wealth
The demographic characteristics of the wealth distribution can be examined by
looking at the patterns of wealth holding by age, education, housing status, income source
and household type. Following the prevailing practice used in other studies of this issue,
the indicator used in this section is the mean net wealth of various household categories.20
Figure 6.12 (Panel A) shows a hump-shaped pattern of mean net wealth by age of the
household head. These curves (which are normalised relative to the wealth holding of
people in the 55-64 age group) reach a maximum at age 55-64 in 11 out of the
18 countries, and at age 65-74 in Belgium, Luxembourg, Norway and the United States.
On average, young households (those with a head under age 34) own about one-fifth as
much wealth as older households (55-64 year-old), with the largest disparities observed in
Germany, the Netherlands, Norway and the United States, and the lowest ones in the
Slovak Republic, Portugal and Spain.
Similar to overall wealth, mean housing wealth (net of liabilities) is relatively modest
among the youngest households, but housing wealth is more evenly distributed across age
groups than overall wealth (Figure 6.12, Panel B). Mean housing wealth increases with
the age of the household head until one reaches households with a head age 55-64, with a
moderate decline in net housing wealth for older households. The young age group holds
on average 40% as much housing wealth as the 55-64 age group, but only 20% as much
overall wealth. The gap in housing wealth is much smaller for households with the head
aged 35-44, who hold 80% as much housing wealth as older workers and retirees. Indeed,
home ownership in most countries rises quickly with the age of household heads before
their early 40s, reaching a maximum for household heads in their 50s and 60s before
declining after age 75, when their health deteriorates. Age disparities in housing wealth
are substantially higher in Germany and the United States.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 257

Mean net financial wealth is more unequally distributed across age groups than
housing wealth (Figure 6.12, Panel C), peaking between 55 and 64 years old (OECD,
2013b). Germany and the United States record the largest gap for the younger age group,
and also experience large gaps in mean wealth for the groups aged 35-44 and 45-54.
Older households (those with a reference person aged 75 and over) tend to use their
financial wealth to support their living standards during retirement; hence their wealth is
substantially lower than for older workers (those aged between 55 and 64). Converting
financial wealth into supplemental income during retirement is common among the
elderly,21 while housing wealth is typically left for bequests (OECD, 2013b): exceptions
to this pattern include Luxembourg and the Netherlands, where financial wealth is
continuously higher for each higher age cohort group.
Large differences in mean wealth holdings also appear when looking at other
household characteristics. Households headed by a college graduate have a net wealth
that, on average, is around 70% higher than households whose head has an upper
secondary education and more than three times (+230%) higher than those of households
with only a primary education (Figure 6.13, Panel A). In the United States, households
headed by college graduates own as much as 3.3 times the wealth of households with
upper secondary, a finding that is consistent with previous studies highlighting the rise in
the college premium and of a category of “working rich” in the United States (Piketty and
Saez, 2006). On average, households who own their principal residence (i.e. outright
owners) have 1.6 times the wealth of owners with a mortgage and 8.5 times the wealth of
renters (Panel B).
When looking at net wealth according to the number of persons in the household and
the presence of children, the highest mean wealth among those with a head of working
age (Figure 6.13, Panel C) is among those composed of two adults and no child (the sole
exceptions are Austria and Luxembourg), while single persons both with and without
children typically have very low wealth; among households with a retirement-aged head
(Panel D), those with two adults have higher net wealth than those with two working-age
adults and no children, while there is more diversity across countries among older
households composed of a single person. Finally, when considering the main income
source of the household (Panel E), in the large majority of countries the wealthiest
households are those whose main income is represented by property income: households
whose main income comes from self-employment have, on average, a mean wealth that is
twice as large as those who depend mainly on wages and salaries, but about 42% of the
mean wealth of households relying mainly on property income.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 0
1.4

1.2

1.2

1.0

1.0

0.8

0.8

0.6

0.6

0.4
0.4

0.2

0.0
0.2

AUS

AUS

AUS
AUT

AUT

AUT
BEL

BEL

BEL
CAN

CAN

CAN
FIN

FIN

FIN
FRA

FRA

FRA
DEU

DEU

DEU
GRC

GRC

GRC

ITA

0.0

1.2

1.2

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

ITA

ITA

0.0

1.4
1.4

1.2

1
1.2

1

0.8
0.8

0.6
0.6

0.4
0.4

0.2
0.2

0

-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+

1.6

1.4

-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+

1.6

-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+

-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+

0.0

-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+
-34
35-44
45-54
55-64
65-74
75+

258 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.12. Mean net wealth per household by age of the reference person and by broad component
Wealth holdings of different age groups, normalised relative to those of the 55-64 age group.
Panel A. Mean wealth

KOR

KOR

LUX

LUX

LUX

NLD

NLD

NLD

NOR

NOR

NOR

PRT

PRT

PRT

SVK

SVK

SVK

ESP

ESP

ESP

GBR

GBR

GBR

USA

Panel B. Mean housing wealth1

USA

Panel C. Mean financial wealth

USA

1. Data on net housing wealth refers to data on housing wealth net of liabilities.

Source: OECD Wealth Distribution Database.

12 http://dx.doi.org/10.1787/888933208581

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 259

Figure 6.13. Mean net wealth by education, housing status, household type and main income source
Panel A. Education of the household reference person1
Lower secondary or below (ISCED 0-2)

Upper & post-secondary (ISCED 3 & 4)

Tertiary (ISCED 5 & 6)

2.2

USA 3.3

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

AUS

AUT

BEL

CAN

FIN

FRA

DEU

GRC

ITA

LUX

NLD

NOR

PRT

SVK

ESP

USA

Panel B. Housing tenure
Outright owner

Owner with mortgage

Renter or other non-owner

4.5
4.0
3.5
3.0

2.5
2.0
1.5
1.0
0.5
0.0
-0.5

AUS

AUT

BEL

CAN

DEU

ESP

FIN

FRA

GRC

ITA

KOR

LUX

NLD

PRT

SVK

GBR

USA

Panel C. Household with a head of working age by type
Working-age head, two or more adults, with children
Working-age head, single person (adult or child)

Working-age head, two or more adults, no children
Working-age head, one adult, with children

2.0

1.5

1.0

0.5

0.0

AUS

AUT

BEL

CAN

DEU

ESP

FIN

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

FRA

GRC

ITA

KOR

LUX

NLD

NOR

PRT

SVK

GBR

USA

 260 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.13. Mean net wealth by education, housing status, household type and main income source (cont.)
Panel D. Household with a head of retirement age by type
Retirement-age head, two persons or more
Working-age head, two or more adults, no children

Retirement-age head, single person

2.0

1.5

1.0

0.5

0.0

AUS

AUT

BEL

CAN

DEU

ESP

FIN

FRA

GRC

ITA

KOR

LUX

NLD

NOR

PRT

SVK

GBR

USA

Panel E. Household by main income source
Property income

Wages and salaries

Self-employment income

12

USA 14.3

10
8
6
4
2
0

AUS

AUT

BEL

CAN

DEU

ESP

FIN

FRA

GRC

ITA

KOR

LUX

NLD

NOR

PRT

SVK

GBR

USA

1. Data on education is not available for Korea and United Kingdom.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208596

6.4. The composition of assets
The composition of household portfolios across different asset types differs
significantly across countries and across the wealth distribution. An overview of the
composition of net wealth by quintile is provided by Figure 6.14. On average, the mean
household wealth for households in the first wealth quintile includes USD 22 000 of nonfinancial assets, USD 5 000 of financial assets, and USD 3 000 of liabilities. The mean
values of both types of assets increase when moving along the wealth distribution, with
the fifth quintile holding a value of non-financial assets that is 30 times larger than the
value of the first quintile. The gap in financial assets is even wider: households in the top
wealth quintile have a mean value of financial wealth that is about 70 times the value of
those in the first quintile. Total liabilities also increase along the wealth distribution, but
they are only twice as large for the top quintile compared with the bottom one. The total
liabilities for the fifth quintile represent 7% of its total assets, while the corresponding
share for the first quintile is 11%. Financial wealth is very unevenly distributed in the
United States and Canada but much less so in Korea and the Slovak Republic.
Differences in mean real assets across quintiles are small in Norway and the Netherlands
(which is likely to be related to the high level of liabilities for less wealthy households)
but particularly large in Canada, France and Belgium.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 261

Figure 6.14. Wealth composition and average net wealth by quintile in OECD countries
2010 or latest available year, values in 2005 USD1
Non-financial assets

Financial assets

Australia

Liabilities

Austria

1400 000

1200 000
1000 000

Mean net wealth

Belgium

Canada
1400 000

1200 000

1200 000

1000 000

1000 000

800 000

800 000

600 000

600 000

400 000

400 000

200 000

200 000

0

0

0

-200 000

-200 000

1200 000
1000 000
800 000

800 000

600 000

600 000
400 000
200 000
0
-200 000

Q-I

Q-II

Q-III

Q-IV

Q-V

-200 000

Q-I

Finland

Q-II

Q-III

Q-IV

Q-V

400 000
200 000

Q-I

Q-II

France

600 000

800 000

400 000

Q-IV

Q-V

Q-I

Germany

1000 000

500 000

Q-III

600 000

300 000

Q-II

Q-III

Q-IV

Q-V

Q-III

Q-IV

Q-V

Q-IV

Q-V

Greece

1200 000

600 000

1000 000

500 000

800 000

400 000

600 000

300 000

400 000

200 000

200 000

100 000

400 000
200 000
200 000

100 000

0
-100 000

Q-I

Q-II

Q-III

Q-IV

Q-V

0

0

0

-200 000

-200 000

-100 000

Q-I

Italy

Q-II

Q-III

Q-IV

Q-V

Q-I

Korea

1200 000

1000 000

1000 000

800 000

800 000

Q-II

Q-III

Q-IV

Q-V

Q-I

Luxembourg
800 000

2200 000
1800 000

600 000

1400 000

400 000

1000 000

200 000

600 000

0

200 000

Q-II

Netherlands

600 000

400 000

600 000

400 000

200 000

200 000
0
-200 000

Q-I

Q-II

Q-III

Q-IV

Q-V

-200 000

Q-I

Norway

600 000

400 000

200 000

0

Q-I

Q-II

Q-III

Q-IV

Q-V

-200 000

Q-I

Portugal

800 000

-200 000

Q-II

0

Q-III

Q-IV

Q-V

400 000

700 000

350 000

600 000

300 000

500 000

250 000

400 000

200 000

300 000

150 000

200 000

100 000

100 000

50 000

0

0

-100 000

-50 000

Q-I

Q-II

Q-III

Q-III

Q-IV

-200 000

Q-V

Q-I

Slovak Republic

800 000

United Kingdom

Q-II

Q-IV

Q-V

Spain
1000 000

800 000
600 000
400 000
200 000
0
Q-I

Q-II

Q-III

Q-IV

Q-V

-200 000

Q-I

Q-II

Q-III

Q-IV

Q-V

OECD18
1200 000

2500 000

1000 000

Q-III

1200 000

United States

1200 000

Q-II

1000 000

2000 000

800 000

800 000
1500 000

600 000

600 000
1000 000

400 000

400 000
500 000

200 000

200 000

0

0

0

-200 000

-500 000

-200 000

Q-I

Q-II

Q-III

Q-IV

Q-V

Q-I

Q-II

Q-III

Q-IV

Q-V

Q-I

Q-II

Q-III

Q-IV

Q-V

1. Financial assets data are not available for Korea.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208607

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 262 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.15. Wealth composition and average net wealth for the top 10%, 5% and 1% in OECD countries
2010 or latest available year, values in 2005 USD
Liabilities

Financial assets

Australia

Non-financial assets

Austria

Mean net wealth

Belgium

Canada

7000 000
4500 000

6000 000
5000 000

3500 000

5500 000

4500 000

4500 000

3500 000

4000 000
2500 000

3500 000
2500 000

3000 000

2500 000

2000 000

1500 000

1500 000

1500 000

1000 000
500 000

500 000

-500 000

500 000

0
TOP 10 %

TOP 5 %

TOP 1 %

-1000 000

TOP 10 %

Finland

TOP 5 %

TOP 1 %

-500 000

TOP 10 %

France

TOP 5 %

TOP 1 %

-500 000

TOP 10 %

Germany

TOP 5 %

TOP 1 %

Greece
2000 000

2000 000

5500 000

4500 000
1500 000

3500 000

1000 000

4500 000

1500 000

3500 000

1000 000

2500 000
2500 000

500 000

500 000

1500 000
1500 000

0

-500 000

500 000

TOP 10 %

TOP 5 %

TOP 1 %

-500 000

0

500 000
TOP 10 %

Italy

TOP 5 %

TOP 1 %

-500 000

TOP 10 %

Luxembourg

TOP 1 %

-500 000

TOP 10 %

Netherlands

19000 000

4500 000

TOP 5 %

TOP 5 %

TOP 1 %

Norway
4000 000

5500 000

3500 000

4500 000

3500 000

3000 000

14000 000

2500 000

3500 000
2500 000

2000 000

9000 000

2500 000

1500 000

1500 000

1500 000

1000 000

4000 000
500 000

500 000

500 000

0
-500 000

-1000 000
TOP 10 %

TOP 5 %

TOP 1 %

Portugal

TOP 10 %

TOP 5 %

TOP 1 %

-500 000

TOP 10 %

Slovak Republic

TOP 5 %

TOP 1 %

-500 000

TOP 10 %

Spain

TOP 5 %

TOP 1 %

United Kingdom
7000 000

4500 000

900 000

3500 000

700 000

2500 000

5500 000

6000 000

4500 000

5000 000

3500 000

4000 000

500 000

3000 000

2500 000
1500 000

2000 000

300 000
1500 000

500 000

1000 000

100 000

-500 000

TOP 10 %

TOP 5 %

TOP 1 %

-100 000

500 000

TOP 10 %

TOP 5 %

TOP 1 %

-500 000

0

TOP 10 %

TOP 5 %

TOP 1 %

-1000 000

TOP 10 %

TOP 5 %

TOP 1 %

United States
18000 000
16000 000
14000 000
12000 000
10000 000
8000 000
6000 000
4000 000
2000 000

0
-2000 000

TOP 10 %

TOP 5 %

TOP 1 %

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208610

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 263

The importance of financial assets increases with the size of net wealth also at the
very top of the distribution (Figure 6.15). Mean financial wealth for the top 1% is
3.7 times higher than for the top 10%, while the gap is smaller for non-financial assets
(the level held by the top 1% is three times the level held by the top 10%). For the top
10% and top 5%, financial assets represent about 40% of total assets, while financial and
non-financial assets are equally important for the top 1%.

Non-financial assets
Table 6.5 describes the asset composition of household portfolios. With the exception
of the United States, non-financial assets constitute the most important category, with a
share of total assets of 75% on average; close to 90% in Greece and the Slovak Republic;
and 50% to 80% in most Nordic and southern European countries, as well as in Australia
and the United Kingdom. The principal residence is the most important asset owned by
households, but the cross-country variation in the value of the principal residence as a
share to total assets is wide, ranging from just over 30% in the United States to just over
70% in the Slovak Republic.22 At the same time, self-reported values for the principal
residence may be overestimated by households, in particular during a period of falling
housing prices. While a number of approaches may be used to correct this valuation bias
(e.g. information on the purchase price, mortgage value, property tax valuation and price
indexes), it is unclear to what extent they are used in various countries.
Part of the cross-country difference in the share of the principal residence in total
assets is related to household ownership rates and to the policies promoting them. For
instance, in Austria and Germany, fewer than 50% of households own their principal
residence (Andrews and Caldera Sanchez, 2011), while 90% of households do so in the
Slovak Republic. On the other hand, ownership rates in the United States are equal to
65%, while housing wealth is much lower than in most other countries with similar
ownership rates (e.g. Norway). Similarly, the home ownership rate is over 80% in Spain,
which translates into a wealth share of around 54%.
While the principal residence represents the bulk of non-financial assets, other real
estate property is also important, accounting on average for 17% of overall assets. In
Korea and Luxembourg, households hold over 30% of their assets in the form of “other
real estate property”, a share that is also high in Greece, Spain and Portugal. In some
southern European countries (Italy, Spain), holiday homes are common, and the period
between the mid-1990s and 2007 witnessed a surge in residential investment, mainly in
Greece and Spain (Andrews et al., 2011), with an ownership rate for other real estate of at
least 30%. On the other hand, other real estate property represents a small share of wealth
in the Netherlands and the Slovak Republic, amounting to less than 7% of total assets.
Other types of real assets beyond housing are of less importance and constitute
between 6% and 7% of total assets. While most households own a vehicle (close to 75%),
these represent only around 3.5% of overall wealth (slightly more in Finland and the
Slovak Republic). Other valuables are however significant in Australia and France (at
around 9% and 5% of assets, respectively).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 264 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Table 6.5. Household portfolio composition
Percentage share of total assets
United

United

Kingdom

States

80.7

79.7

48.5

74.1

53.9

54.8

31.5

51.2

6.7

23.7

9.3

12.8

17.4

4.0

5.5

2.6

2.3

3.4

3.2

0.1

0.9

0.6

0.4

13.3

0.8

4.5

28.8

26.8

24.4

12.9

19.3

20.3

51.5

25.9

4.9

9.0

12.5

8.8

6.3

5.3

9.7

7.0

7.6

0.7

1.2

0.1

0.2

0.0

0.2

2.0

2.0

0.9

..

2.3

1.7

1.3

0.5

0.2

0.8

1.1

7.6

2.0

8.3

..

3.0

2.2

0.8

11.9

4.5

9.0

..

4.5

6.7

0.2

0.5

..

0.8

0.9

9.6

0.8

0.0

0.9

2.9

5.7

2.3

0.0

0.4

..

0.1

0.0

..

0.0

0.4

0.8

0.6

14.2

2.1

1.2

0.2

0.6

..

0.3

0.7

1.8

0.8

0.4

0.7

0.8

1.1

0.9

7.5

5.7

0.5

0.9

..

2.1

13.1

0.6

1.3

0.9

1.6

3.3

9.4

4.2

Austria

Belgium

Canada

Finland

Non financial assets

80.7

59.1

66.6

62.9

81.7

72.2

65.8

88.3

81.7

73.3

85.9

71.2

73.2

75.6

87.1

Principal residence

51.0

43.7

51.6

43.1

54.6

47.6

40.9

55.9

60.8

36.3

51.9

61.3

60.9

47.8

74.4

Other real estate property

17.8

10.9

11.9

12.3

22.4

20.0

20.5

27.8

16.4

31.4

30.2

6.4

10.2

23.0

Vehicles

2.9

3.5

2.3

3.3

4.7

..

3.3

4.3

3.0

2.5

2.6

2.9

2.1

Other non financial assets

9.1

1.1

0.8

4.1

0.0

4.7

1.0

0.3

1.5

3.0

1.2

0.6

19.3

40.9

33.4

37.1

18.3

27.8

34.2

11.7

18.3

26.7

14.1

Deposits

5.7

10.4

11.4

4.6

7.8

6.5

9.4

5.6

4.7

..

Bonds and other debt securities

0.0

1.3

4.3

0.3

0.1

0.3

1.2

0.2

2.0

..

Mutual funds

3.7

2.0

3.8

3.2

1.7

1.1

2.2

0.2

1.0

2.8

24.3

4.3

2.7

3.3

8.4

13.0

4.8

Stocks

2.5

0.5

3.0

2.7

3.9

2.2

1.4

Unlisted shares and other equity

4.4

0.2

0.5

8.7

..

0.9

0.2

Other non pension financial assets

0.1

0.7

1.3

2.2

..

0.7

..

1.5

4.9

12.7

1.4

Financial assets

Net equity in ow n
unincorporated enterprises

Voluntary indiv idual life insurance
and priv ate pension funds
Liabilities

France Germany Greece

Italy

Korea Luxembourg Netherlands Norway Portugal

Slovak

Australia

Republic

Spain

OECD18

17.9

5.9

8.2

14.0

18.4

9.6

12.2

7.5

4.1

24.4

10.3

32.5

36.2

10.2

4.0

10.1

15.7

18.7

14.4

Principal residence loans

10.3

4.3

6.6

10.9

13.2

5.0

7.1

4.5

2.7

5.6

7.0

25.0

..

8.2

3.1

6.1

12.1

13.0

8.5

Other real estate loans

5.8

0.7

0.8

2.8

..

2.3

3.5

1.3

0.3

2.4

2.3

2.0

..

1.2

0.2

2.6

2.2

2.9

2.1

Other loans

1.8

1.0

0.9

0.3

5.1

2.4

1.5

1.6

1.1

16.4

1.0

5.5

..

0.8

0.8

1.4

1.5

2.8

2.7

Note: “..” refers to non-available data.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209098

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 265

Financial assets
Financial assets include deposit accounts, bonds, listed shares, mutual investment
funds, life insurance, investment and other financial assets.23 Household wealth in the
form of financial assets is high in the United States (52% of total assets) and in Canada
(37%), where a high share of households invest in the financial market, as well as in
Austria (40%).
Bank deposits are households’ most important financial asset, with almost all
households having a deposit (except in Greece). Deposits account for around 8% of total
assets (and 30% of financial assets) on average, with larger shares in Finland, the Slovak
Republic and the United Kingdom.
In addition to deposits, net equity in own unincorporated enterprises and voluntary
life insurance and personal pension funds are also sizeable, constituting on average
between 4% and 6% of overall wealth.24 Net equity in unincorporated enterprises is
particularly important in Austria, Germany, France and southern Europe (except Greece).
On the other hand, unlisted shares, other equity and individual life insurance and pension
funds represent the most important type of financial asset in English-speaking countries
(except the United Kingdom). Unlisted shares represent 4%, 9% and 14% of assets in
Australia, Canada and the United States, respectively. Life insurance and private pensions
account for a large share of financial assets in the Netherlands and Canada, and (to a
lesser extent) in the United States, France and Germany.

Pension wealth
Pension wealth is an important component of individual or household wealth. However,
in contrast to other wealth components, pension wealth has characteristics that make its
measurement very complex due to differences in the institutional arrangements of various
countries. Pension wealth measures the capitalised value of the flows of pension benefits
that people may expect to receive in the future; hence its value will reflect cross-country
differences in pension replacement rates (i.e. the ratio of pension benefits to previous
earnings), in the duration over which benefits are paid (which depend on country-specific
mortality rates), in retirement ages, and in the indexation of pensions in payment (Box 6.3).
Wealth from pension systems may take three different forms: 1) a first-tier of public
pensions ensuring a basic or minimum pension; 2) a second-tier, in the form of mandatory
or quasi-mandatory contributions to either private or public schemes; and 3) a third tier,
which comprises private voluntary pension savings. Across OECD countries, only Ireland
and New Zealand do not have mandatory second-tier schemes, while in 13 countries,
private pensions are either mandatory or quasi-mandatory. Private pension coverage tends
to be lower in countries where private pensions are voluntary (with coverage rates ranging
from 13% to 50% of the working-age population, as compared to coverage rates of around
or above 70% in countries with mandatory private pension systems (OECD, 2013b).
Household wealth surveys collect information on the amount of voluntary private
pensions outstanding (included among financial assets in Table 6.5 above); they
sometimes include the amounts outstanding in mandatory or quasi-mandatory
occupational pension plans (although reporting is less systematic); but they always
exclude first-tier public pensions, as well as pensions on account of mandatory
contributions to public pensions schemes. As the mix of different types of pension wealth
varies across countries, household wealth data of the type used in this chapter provide
only a partial picture of the amount (and distribution) of pension wealth across countries.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 266 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
An alternative approach to assessing the amount of household pension wealth is through
models that provide an estimate of the stream of pension benefits that people with
different characteristics may expect to receive in the future, based on the eligibility
conditions of the different pension systems. This approach is implemented by the OECD
pension models (OECD, 2013b), and the results are discussed in Annex 6.A3.
Box. 6.3. Measurement of household wealth from pensions
Pension wealth refers to “the discounted value of the lifetime flows of pension benefits measured at the
point of retirement” (OECD, 2013b). Contrary to the stream of pension benefits that are paid over time until
some contingency occurs, pension wealth is a “stock” concept generally measured at the point of retirement.
This stock measure is computed by discounting the lifetime flows of pension benefits, thus taking into
account the opportunity cost to delaying consumption. The present value of these flows depends on the
probability of being alive to claim the pension in a given period, which is inferred from mortality tables. Any
adjustment of pensions in payment that reflect changes in costs or standards of living should also be taken into
account (“indexation” or “uprating” policy).
Survey questionnaires on wealth generally do not include information on public pension wealth nor on
private pension wealth from defined-benefit occupational programmes. By contrast, they generally report some
information on private wealth from funded (generally private) pensions. This is, for example, the case for the
concept of “extended net wealth” in the OECD wealth questionnaire, which includes pension wealth in funded
pension schemes related to employment. The comparability of net extended wealth across countries is limited.
Even though funded private pensions related to employment might play an increasingly important role in the
income of future retirees, their coverage varies greatly across OECD countries. Private pensions are mandatory
or quasi-mandatory in 13 OECD countries. In most of them, payments occur as monthly benefits, which are
captured by income measures. For voluntary pensions, lump-sum withdrawals are more common.
Information on pension wealth derived from both social security schemes and defined-benefit occupational
pension programmes will be enhanced in the future as new rules for reporting pension liabilities in National
Accounts are implemented. In the recently endorsed standards for national accounts, SNA 2008 and ESA 2010,
the debate focused on the measurement of pension entitlements/liabilities, especially for programmes sponsored
by the government. The “compromise” solution finally agreed is that national statistical offices are required to
report on the value of all pension entitlements in an additional SNA table (Table 17.10). This table would cover
both pension entitlements that are recognised in the core system of national accounts and those that are not
recognised (implicit). Similarly, a new European regulation for the European System of Accounts (ESA 2010)
requires European countries to report their accrued-to-date pension liabilities starting from 2017. The collection
of these data might allow the computation of some sort of public pension wealth indicator in the future.

6.5. The incidence of debt and over-indebtedness
Debt and over-indebtedness across OECD countries
The importance of household liabilities (as a share of total assets) varies largely
across OECD countries, ranging from 4% in Italy and the Slovak Republic, to over 30%
in the Netherlands and Norway (Table 6.5). The higher level of debt in these latter
countries is often related to rising housing prices in the period preceding the crisis, which
led more households to take on higher mortgages to become homeowners. In Norway,
house prices rose by around 85% in real terms between 2000 and mid-2013, and
aggregate household debt is now over 200% of disposable income (OECD, 2014). In the
Netherlands, large tax deductions on interest payments encourage households to hold
mortgages as either interest-only or as a form of savings/investment where only interest is
paid until the end of the mortgage loan, when the capital is then reimbursed.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 267

Most household liabilities (two-thirds) are mortgages on the purchase of a principal
residence, but in some countries loans on other residential property and other types of loans
represent a substantial fraction of total liabilities. Close to 80% of household liabilities are
principal residence loans in Belgium, Canada, the Netherlands, the Slovak Republic and the
United Kingdom. In Australia, Germany and Spain, loans on other property constitute 30%
of total liabilities. In Korea, households also hold other types of loans with a value of 70%
of total liabilities (and more than 10% of total assets). In Korea, a large part of household
liabilities takes the form of a deposit paid to the landlord (the “chonsae” system, a rental
market system in which a tenant pays a deposit upfront with no additional periodic rent
payments, and receives the nominal value of the deposit from the landlord upon maturation,
see Cho, 2010).
While debt facilitates consumption smoothing over the life cycle, the concentration of
household debt on some households, even when matched by higher assets, may expose
them to high risks in the event of sudden changes in asset prices, with implications for the
vulnerability of the economic system as a whole. Figure 6.16 shows the percentage of
indebted and over-indebted households among the 18 countries for which data are
available: while no general definition of over-indebtedness currently exists, over-indebted
households are defined in Figure 6.16 as those with a debt-to-asset ratio exceeding 75%. On
average, about 55% of households are indebted, ranging from 25% in Italy to 80% in the
United States and Norway. Based on the criteria used here, about 12% of all households are
over-indebted across the 18 OECD countries for which data are available, ranging from 2%
in the United Kingdom and Italy to 24% in the United States and 30% in Norway.25
An alternative measure of over-indebtedness is the share of people with a ratio of debt
to income (rather than assets) above a given threshold. This measure offers a different
perspective of household over-indebtedness as it focuses on the ability to reimburse the
debt. The threshold used here is equal to three, implying that a household that saves onethird of its income each year would reimburse its debt in around 12 years based on
prevailing interest rates. This alternative measure (also shown in Figure 6.16) yields very
similar results to the previous one: on average, 11% of households are over-indebted,
ranging from 1% or less in the United Kingdom to 26% in Norway.26
Figure 6.16. Percentage of indebted and over-indebted households
% of indebted households (↗)

% of households with
debt-to-asset ratio above 75%

% of households
with debt to income ratio above 3

90%
80
70

60
50
40
30

20
10
0

ITA

SVK AUT GRC PRT BEL CAN FRA DEU ESP GBR LUX

FIN

NLD KOR AUS USA NOR

OECD18

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208624

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 268 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Demographic characteristics of indebted and over-indebted households
Figure 6.17 depicts the age characteristics of indebted and over-indebted households,
with over-indebtedness measured in terms of households’ debt-to-income ratio. Panel A
shows that indebtedness is almost equally spread across young (age 16-34) and prime-age
adults (35-44 and 45-54), while being marginal among households with a reference
person over age 65. In contrast, over-indebtedness falls disproportionately upon young
people: on average 40% of over-indebted households are headed by a person under
age 34, and 30% by one in the 35-44 age band. Over-indebtedness is therefore primarily a
concern for young people.
Figure 6.17. Age composition of indebted and over-indebted households
Panel A. Indebted households
16-34

35-44

45-54

55-64

65-74

75+

%
90
80
70

60
50
40
30
20
10
0
ITA

SVK

AUT GRC PRT

BEL

CAN

FRA DEU

ESP GBR LUX

FIN

NLD KOR AUS USA NOR

Panel B. Over-indebted households
16-34

35-44

45-54

55-64

65-74

75+

%
35

30
25
20

15
10
5
0
SVK GBR ITA GRC BEL KOR AUS ESP PRT FRA AUT LUX DEU CAN FIN NLD USA NOR

Note: Countries are ranked from left to right, in increasing order of total number of indebted households.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208635

Turning to the distribution of debt and over-indebtedness by education, Figure 6.18
shows that households headed by a person with an upper secondary or post-secondary
non-tertiary degree represent about half of indebted and over-indebted households, while
those with lower and tertiary education represent about one-quarter. However, in the
Netherlands and the United States, two countries with high levels of debt and
over-indebtedness, households with tertiary education are the most (over) indebted.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 269

Figure 6.18. Percentage of indebted and over-indebted households, by education
Panel A. Indebted households by education of household head
Lower

Upper & post-secondary

Tertiary

90
%
80
70
60
50
40
30
20
10
0
ITA

SVK

AUT

GRC

PRT

BEL

CAN

FRA

DEU

ESP

LUX

FIN

NLD

AUS

USA

NOR

Panel B. Over-indebted households by education of household head
Lower

Upper & post-secondary

Tertiary

%
35

30

25
20
15
10
5
0
SVK

ITA

GRC

BEL

AUS

ESP

PRT

FRA

AUT

LUX

DEU

CAN

FIN

NLD

USA

NOR

Note: Countries are ranked from left to right in increasing order of total of indebted households in Panel A and B.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208641

Figure 6.19 displays the average shares of indebted (Panel A) and over-indebted
households (Panel B) for each quintile (and the top percentiles) of the income distribution
across the 16 OECD countries for which data are available. In general, the share of
indebted households increases with income levels in all countries. On average, about onethird of households from the bottom income quintile have debts, as compared with more
than two-thirds for the top quintile and the top percentiles: this suggests the existence of
credit constraints for lower-income households. Perhaps due to this limited access to the
credit market, the largest share of over-indebted households is observed among the
“middle class”, namely quintiles II, III and IV, rather than among lower or higher-income
groups (Panel B): Norway, the Netherlands, Austria and Germany are exceptions to
this pattern.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 270 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Figure 6.19. Percentage of indebted and over-indebted households by income quintile and top distribution
Panel A. Indebted households
70

%

60
50
40

30
20
10
0

Quintile I

Quintile II

Quintile III

Quintile IV

Quintile V

Top 10 %

Top 5 %

Top 1 %

Top 10 %

Top 5 %

Top 1 %

Panel B. Over-indebted households
12
%
10

8
6
4
2
0
Quintile I

Quintile II

Quintile III

Quintile IV

Quintile V

Note: Data refer to OECD18 average.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208654

High levels of indebtedness are particularly problematic in countries where a large
fraction of households is indebted. To measure the depth of indebtedness, the median
debt-to-income ratio is used in Figure 6.20 to highlight how the depth of indebtedness
varies significantly across countries. The stock of household debt represents as much as
94% of household disposable income, on average, ranging from 11% in the United
Kingdom to more than 160% in Canada, Norway and the Netherlands. The Netherlands,
Norway and the United States combine a large fraction of households having debts and a
large level of indebtedness. Households in these countries (which are also those with the
highest levels of over-indebtedness) may thus be the most exposed to the risk of large
swings in asset prices. Conversely, Italy, Austria and Greece appear to be less vulnerable
to large asset price changes.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 271

Figure 6.20. Depth of indebtedness for indebted households
2010 or latest available year

Median debt-to-income ratio of indebted
households (%)

250

NLD

200

NOR
CAN
150

PRT
USA
ESP

100

AUS

LUX
BEL
ITA

GRC

50

DEU

AUT

SVK

FIN

FRA

GBR

0
0

10

20

30

40

50

60

70

80

90
%

Percentage of indebted households

Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208667

6.6. Trends in household wealth since the Great Recession
This section describes trends in household wealth in a few OECD countries where
time-series information is available. The impact of the 2008 financial crisis on stocks of
household wealth is examined with the data collected by the OECD, while long-term rises
in wealth-to-income ratios are described in Box 6.4 drawing on the data and analysis in
Piketty (2014).
Two major patterns emerge from the examination of trends in household wealth in the
six OECD countries for which comparable data are available before and after 2008
(Australia, Canada, Italy, the Netherlands, the United Kingdom and the United States,
shown in Table 6.6).


First, the mean net wealth per household has changed in very different ways
across countries in the aftermath of the Great Recession. Mean net wealth has
increased in Australia and Canada, but it has declined in Italy, the Netherlands,
the United Kingdom and the United States.



Second, trends in wealth inequality have been different across countries
throughout the Great Recession. In Australia and Canada, the median net wealth
has increased at a faster rate than the wealth of the upper percentiles, so that
inequality at the top of the distribution has receded. Conversely, the median net
wealth has fallen at a faster rate than the net wealth of the top percentiles in Italy,
the Netherlands and the United States, implying an increase in wealth inequality
at the top of the distribution. Finally, median net wealth has decreased in the
United Kingdom, while the net wealth of the top percentiles has increased, so
that inequality at the top of the wealth distribution has unambiguously risen.
Inequality at the bottom of the wealth distribution has increased in all countries
except the United Kingdom.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 272 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Table 6.6. Changes in mean net wealth between 2006 and 2012
Annual percentage changes
Mean

Median

Bottom
quintile

Middle three
quintiles

Top
quintile

Top
10%

Top
5%

Top
1%

Top w ealth Bottom w ealth
inequality
inequality

Observed
period

Australia

1.0

1.0

-4.0

1.1

1.0

0.8

0.3

-2.7

-0.7

0.3

2006-2012

Canada

5.1

5.9

3.4

6.0

4.7

4.1

3.5

1.3

-1.8

0.1

2005-2012

Italy

-1.3

-2.2

-25.3

-2.1

-0.8

-0.6

..

..

..

0.4

2006-2012

Netherlands

-1.6

-6.0

1.9

-3.6

-0.8

-0.5

-0.3

0.3

9.3

2.8

2006-2012

United Kingdom

-0.1

-2.5

5.0

-1.9

1.1

2.0

3.1

8.4

7.2

-0.3

2006-2012

United States

-2.3

-7.1

-26.4

-5.6

-1.7

-1.5

-1.5

-1.2

9.9

4.0

2007-2013

Note: Data for Canada and on United States refer to the period 2005-12 and 2007-13, respectively. Top wealth inequality refers to
the difference between the mean wealth of the top 5% and the median wealth, divided by the median wealth. Bottom wealth
inequality refers to the difference between median wealth and the mean wealth of the bottom quintile, divided by the median wealth.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209105

Figure 6.21 describes wealth changes by group of households, showing that the least
wealthy groups have taken the biggest hit from the Great Recession. In most countries,
the change in net wealth has been lower for households whose head was younger than 34
(Panel A), and for households with low educational attainment (Panel B). Box 6.4
provides an historical perspective on changes in private wealth, based on the estimates
provided in Piketty (2014).27
Figure 6.21. Change of mean net wealth between 2006 and 2012
Annual percentage changes
Panel A. Households headed by a young and old reference person 1
% change for young

% change for old

10
%
8
6
4
2
0
-2
-4
-6
-8
- 10

AUS

CAN

ITA

GBR

USA

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 273

Figure 6.21. Changes in mean net wealth by household characteristics, 2006 and 2012 (cont.)
Panel B. Households whose reference person has low or high education 2
% change for lower skilled

% change for highly skilled

10
%
8

6
4

2
0

-2
-4
-6
-8
- 10

AUS

CAN

ITA

USA

1. Young household heads are those under 34 years of age, while the old heads are those above 65.
2. The low education group refers to household heads with lower secondary education or below (ISCED 0-2), while the high
education group refers to household heads with tertiary education (ISCED 5 & 6).
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933208671

Box 6.4. Piketty’s historical perspective on trends in household wealth
In a longer-term perspective, the evidence presented by Piketty (2014) suggests that both private wealth and
its concentration have increased markedly over the last four decades. As shown in the figure below, private
wealth as a share of national income almost doubled in eight OECD countries since 1970. Similarly, according
to Piketty’s estimates, the concentration of wealth (as captured by the shares of the top decile and the top centile
of the wealth distribution) has increased in France, Sweden, the United Kingdom and the United States
since 1970. The rise recorded over this period stands in sharp contrast to the long-term decline recorded in the
interwar period.
Private wealth in eight OECD countries since 1970, as a percentage of national income
United States

Japan

Germany

France

800
%
700

600
500
400
300
200

100
0

1970

1975

1980

1985

1990

1995

2000

2005

2010

Source: Piketty, T. (2014), Capital in the 21st Century, Harvard University Press.
12 http://dx.doi.org/10.1787/888933208692

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 274 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Box 6.4. Piketty’s historical perspective on trends in household wealth (cont.)
Piketty (2014) identifies four factors that could explain this long-term increase in private wealth since the
1970s, while a fifth one accounts for persistent deviations from this long-term trend:



First, as private wealth is measured as a share of national income, the slowdown of economic growth
in the aftermath of the 1973 oil shock mechanically increased this share.



Second, over the long-term, savings rates have remained relatively stable and translated into a higher
stock of private wealth.



Third, the price of assets in the form of property or stocks dramatically increased (relative to
consumer prices) in a majority of countries from 1980 onwards.



Fourth, privatisation programmes in the 1980s and the 1990s induced a transfer of assets (which
includes firm’s net wealth) from the public to the private sector.



Fifth, real estate bubbles in Japan in the 1980s and stock market bubbles in the United States in
2000-01 entailed a generally persistent deviation in private wealth’s share of national income from its
long-term growth path.

The empirical analysis described in Annex 6.A2 confirms the relevance of the factors listed above for the
eight OECD countries for which Piketty provides data since 1970. Taking stock of the long-term series of
private wealth available in Piketty (2014), analysis shows significant correlations between, on the one hand, the
stock of private wealth and, on the other hand, per capita GDP growth and asset prices (i.e. price indices of
quoted shares and dwellings relative to the price of private consumption). Conversely, there is no robust
correlation between the stock of private wealth and the national savings rate; this may reflect the erratic changes
in national savings rates. As expected, economic growth is negatively related to private wealth, while increases
in stock or housing prices relative to private consumption prices are associated with an increase of private
wealth.

6.7. Conclusion
The main objective of this chapter has been to highlight patterns in the distribution of
household net wealth within a sample of OECD countries that regularly collect statistics
on this issue. This provides a basis for comparative analyses, extending previous OECD
work (OECD, 2008) and complementing academic studies that have provided a long-term
perspective for a smaller number of countries (Piketty, 2014). While several factors limit
the comparability of the wealth data used in this chapter, a number of stylised patterns
emerge from this review.


First, household wealth is much more concentrated than household income: the
10% of wealthiest households hold about 50% of all household wealth on
average in the sample of OECD countries, while the 10% of people with the
highest income earn about 25% of total income. At the bottom of the wealth
distribution, 40% of households hold little over 3% of all household wealth.
Looking at the joint distribution of income and wealth, most wealthy households
are high-income households, and vice versa. But there is also a significant
minority of wealthy households that report low income, and a similar share of
low-wealth households that report high income, with significant differences
across countries.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 275



Second, households holding wealth share some common characteristics. They are
generally headed by a person who is in the 55-64 age band and has a highereducation degree. In terms of the composition of net wealth, non-financial assets,
and in particular the principal residence, constitute the most important assets for
most households (75% on average and up to 90% in some countries); only a
minority of households hold financial assets, whose degree of concentration is
generally higher than in the case of non-financial assets.



Third, there are significant differences across countries in terms of debt and overindebtedness, which has implications for the economic well-being of households
and for the vulnerability of the economic system as a whole. Even after the crisis,
there appears to be a large proportion of households with debts reaching very
high levels relative to income in countries such as the United States, Norway and
the Netherlands.



Finally, over the longer run, the increase in the price of shares and housing
relative to consumer goods was one of the main drivers of higher household
wealth, and this has likely increased wealth concentration over the last 40 years.
This trend was not reversed by the financial crisis in the United States, the
Netherlands and the United Kingdom, where wealth inequality (at the top) has
grown markedly, while smaller changes have been observed in Australia, Canada
and Italy.

Beyond these broad empirical findings, which should inform discussion on how best
to respond to concerns about greater inequality in economic well-being, significant
challenges confront the statistical community in terms of improving the information base
on the level and distribution of household wealth, through initiatives to:


Collect information on the distribution of wealth in those countries that currently
lack statistics in this field



Improve the comparability of the available information through better alignment
to the existing international guidance in this field.28

Through the collection of the data presented in this chapter (which the OECD plans to
pursue regularly in the future), the OECD has engaged in a process similar to the one
followed for statistics on income distribution, which – over a period of around 20 years –
led to a much improved comparative reporting by countries. While the comparative
information on household wealth used in this chapter has its limits, the hope is that – as
statistical offices and agencies collecting wealth data become more aware of differences
in their survey practices – a significant convergence in statistical reporting will occur in
the near future.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 276 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Notes
1

Helpful comments on a previous version of this chapter were provided by Martine
Durand, Stefano Scarpetta, Marco Mira d’Ercole, Michael Förster, Hervé Boulhol and
Kate Lancaster (all from the OECD) as well as Tim Smeeding (University of
Wisconsin-Madison); by Martin Schuerz (Austrian National Bank), Philip Vermeulen
(European Central Bank), Martin Tambour (Statistics Germany), Anna Gunnarsdottir
(Statistics Iceland), Chris Kim (Statistics Korea), Conceição Veiga (Statistics
Portugal); and by country delegates to the OECD Committee on Statistics and
Statistical Policy. The OECD wishes to acknowledge the essential contribution of a
number of country specialists who contributed to gathering the empirical evidence
presented in this chapter. Veli-Matti Törmälehto, Statistics Finland, contributed to the
design of the standard tabulations used to collect country data, and to the choice of
comparable definitions and methodological approaches. Estimates drawn from the
micro-data files of various countries were provided by: Caroline Daley and Heather
Burgess, Australian Bureau of Statistics; Paul Roberts and Brian Murphy, Statistics
Canada; Rosalinda Coppoletta and Maud Romani, INSEE; Giovanni D’Alessio and
Andrea Neri, Bank of Italy; Woonjoo Suh and You Kwon Park, Korean National
Statistical Office; Wim Bos, Statistics Netherlands; Jon Epland, Statistics Norway;
Alan Newman and Elaine Chamberlain, Office for National Statistics, United
Kingdom; and Arthur B. Kennickell, Federal Reserve Board, United States. The
chapter also uses data from the Eurosystem Household Finance and Consumption
Survey. The OECD wish to acknowledge their essential contribution.

2.

Household wealth (the concept used in this chapter) is different from private wealth
(which includes the wealth of private corporations) and, a fortiori, from total wealth
(which includes the wealth of other institutional sectors, such as the public sector).
The concept of wealth used in this chapter refers to assets that are owned and
transferrable across units. Hence, it differs from other concepts of wealth (i.e. assets
that are not owned and not transferrable, such as human capital).

3.

The countries included in the chapter are Australia, Austria, Belgium, Canada,
Finland, France, Germany, Greece, Italy, Korea, Luxembourg, the Netherlands,
Norway, Portugal, the Slovak Republic, Spain, the United Kingdom and the United
States. In the case of the United Kingdom, the geographical coverage of the data is
limited to Great Britain (excluding Northern Ireland). In the case of the Netherlands,
the data on wealth levels and distribution are based on administrative records, while
the data on the demographic characteristics of households with wealth and on the
composition of assets and liabilities are based on survey data.

4.

However, the composition of the median household may differ across countries.
D’Alessio et al. (2013) argue that cross-country differences in mean net wealth are
less sensitive to differences in household size.

5.

These results are broadly consistent with findings from other sources and economic
studies, even if the methodologies differ. See Annex 6.A2 for a discussion.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 277

6.

The income concept used in this chapter to compute wealth-to-income ratios is that of
“gross” household income, i.e. household income before income and wealth taxes
paid by the household (but after monetary transfers received), as collected in the
wealth sources used here. While a measure of household disposable income (i.e. net
of taxes and contributions) would have been more appropriate, this measure is only
available for a minority of countries (Australia, Canada, Finland, Italy, Norway and
the United States).

7.

The correlation of country rankings between absolute and relative mean net wealth
is 0.66. However, the ranking of the United States, Canada, Portugal and Italy
changes significantly (the United States falls from the second position for absolute
mean wealth to the 10th position for relative mean wealth and Canada from third to
13th, while Portugal rises from 12th to fourth and Italy from eighth to third.

8.

The correlation of country rankings between absolute and relative median net wealth
is 0.75. Relative median wealth is calculated as the ratio of median net wealth to the
average disposable income of households in the third quintile of the wealth
distribution, which is used as a proxy for the income of the household with median
net wealth.

9.

The definition of a household differs across the two sources, as national accounts data
combine households and non-profit institutions serving households, while micro-data
refer to private non-institutional households.

10.

The National Accounts of most OECD countries measure the value of dwellings
based on the Perpetual Inventory Method (PIM) applied to historical investment
costs; changes in house prices are hence reflected in changes in the value of the land
underlying the dwelling.

11.

For example, the System of National Accounts (SNA) data shown in Figure 6.3
include the wealth of non-profit institution serving households, while the micro-data
sources refer to private households.

12.

Data provided by Statistics Netherlands confirm that changes in median net wealth
largely reflect changes in the value of the principal residence: for instance, the median
net wealth increased by 24% between 2006 and 2008, while the value of the principal
residence increased by 10% over the same period; conversely, median net wealth
declined by 43% between 2008 and 2012, a period when the value of the principal
residence fell by 11%. The correlation between the two series over the period 2006-12
is 0.88.

13.

Income data are drawn from the OECD Income Distribution Database, and refer to
distribution across individuals (rather than across households, as in the case of the
wealth distribution) based on a concept of “equivalised” (i.e. with an adjustment for
household size) disposable income (rather than non-adjusted amounts as in the case of
the wealth distribution).

14.

Data on the wealth share of the top 10, 5 and top 1% of the distribution are not
available for Korea.

15.

These correlations are significant at a 5% confidence level.

16.

Other wealth-concentration measures (such as the share of the top 1% or top 20%)
highlight the same pattern but in a less clear-cut way. Focusing on inequality at the
top of the wealth distribution also avoids using a wealth ratio with a denominator (the
average wealth of the less-wealthy households) that is close to zero or even negative.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 278 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
17.

This synthetic measure and a confidence interval are obtained from Monte-Carlo
simulations run on a smoothed joint distribution of income and wealth within each
country.

18.

In theory, the correlation of ranks in the income and wealth distributions and their
respective degree of inequality are completely independent of the statistical measures.
A large correlation of ranks could co-exist with low inequality in the marginal
distributions of income and wealth, and vice-versa.

19.

Due to the small number of countries, this correlation is however significant only at a
10% confidence level. When one looks at the wealth shares of the top 5% and top
10%, the p-values of the latter correlation drop to 0.12 and 0.15.

20.

Age and education are those of the household reference person.

21.

This study uses cohort data and not longitudinal data that would follow households as
they age.

22.

This ratio is significantly higher among the middle class (e.g. it is equal to 72%
among households in the third quintile of the wealth distribution in the United States).

23.

In the classification used in this chapter, financial assets include life insurance and
private pension wealth. See Box 6.3 for a more detailed discussion.

24.

While “net equity in own unincorporated enterprises” is considered a non-financial
asset in National Accounts, the OECD Guidelines for Micro Statistics on Household
Wealth (OECD, 2013a) recommends including this among financial assets, alongside
the value of shares in listed companies.

25.

In principle, it would be important to disentangle non-secured debt from secured debt,
namely debt contracted for the acquisition of an asset in the form, for instance, of real
estate or human capital. However, such a breakdown of liabilities into “pure debt”
and “leveraged debt” is not available in the micro-data sources used for this chapter.

26.

The correlation between the two measures is 0.74.

27.

The concept of private wealth used by Piketty (2014) aggregates household and
corporate wealth, as compared to the narrower concept of household wealth used in
this chapter. This loss of comparability is balanced by the long-term perspective
provided by Piketty’s estimates.

28.

No statistical standard currently exists to inform the efforts of the statistical
community in this field. A first step in this direction is represented by the OECD
Guidelines for Micro Statistics on Household Wealth (OECD, 2013a).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 279

References
Amiel, Y., F. Cowell and A. Polovin (1996), “Inequality among the Kibbutzim”,
Economica, Vol. 63, No. 250, London School of Economics and Political Science,
pp. 63-85.
Andrews, D. and A. Caldera Sánchez (2011), “Drivers of Homeownership Rates in
Selected OECD Countries”, OECD Economics Department Working Papers, No. 849,
OECD Publishing, Paris, http://dx.doi.org/10.1787/5kgg9mcwc7jf-en.
Andrews, D., A. Caldera Sánchez and Å. Johansson (2011), “Housing Markets and
Structural Policies in OECD Countries”, OECD Economics Department Working
Papers, No. 836, OECD Publishing, Paris, http://dx.doi.org/10.1787/5kgk8t2k9vf3-en.
Beer, C., P. Mooslechner, M. Schürz and K. Wagner (2006), “Das Geldvermögen privater
Haushalte in Österreich: eine Analyse auf Basis von Mikrodaten‘, Geldpolitik und
Wirtschaft, Vol. 2, pp. 101-119.
Brzozowki, M., M. Gervais, P. Klein and M. Suzuki (2010), “Consumption, Income and
Wealth Inequality in Canada”, Review of Economic Dynamics, Vol. 13, No. 1,
pp. 52-75.
Cho, Sang-Wook, S. (2010), “Household Wealth Accumulation and Portfolio Choices in
Korea”, Journal of Housing Economics, Vol. 19, No. 1, Elsevier, pp. 13-25, March.
Cowell, F. (2013), “Accounting for Cross-country Differences in Wealth Inequality”,
Gini Discussion Paper, AIAS, Amsterdam, August.
Crédit Suisse (2014), Global Wealth Database 2014, Research Institute Crédit Suisse,
Zurich.
Crédit Suisse (2013), Global Wealth Database 2013, Research Institute Crédit Suisse,
Zurich.
Crédit Suisse (2012), Global Wealth Database 2012, Research Institute Cérdit Suisse,
Zurich.
Crédit Suisse (2011), Global Wealth Database 2011, Research Institute Crédit Suisse,
Zurich.
Crédit Suisse (2010), Global Wealth Database 2010, Research Institute Crédit Suisse,
Zurich.
D’Alessio, G., R. Gambacorta and G. Ilardi (2013), “Are Germans Poorer than Other
Europeans?”, Vox, 24 May 2013.
Domeij, D. and M. Floden (2010), “Inequality Trends in Sweden 1978-2004”, Review of
Economic Dynamics, Vol. 13, No. 1, Elsevier for the Society for Economic Dynamics,
pp. 179-208, January.
Eizinger, C., M. Kalmár, G. Kernbeis, M. Prammer-Waldhör and M. Wagner-Pinter
(2008), Vermögungsbildung und Reichtum in Österreich, Vienna.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 280 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
European Central Bank (2013), “The Eurosystem Household Finance and Consumption
Survey: Results from the First Wave”, Statistics Paper Series, Vol. 2, April.
Fesseau, M., F. Wolff and M.L. Mattanetti (2013), “A Cross-country Comparison of
Household Income, Consumption and Wealth between Micro Sources and National
Accounts Aggregates”, OECD Statistics Directorate Working Paper, No. 52, OECD
Publishing, Paris, http://dx.doi.org/10.1787/5k3wdjrnh7mv-en.
Fuchs-Schündeln, N., D. Krueger and M. Sommer (2010), “Inequality Trends for
Germany in the Last Two Decades: A Tale of Two Countries”, Review of Economic
Dynamics, Vol. 13, No. 1, pp. 103-132.
Jappelli, T. and L. Pistaferri (2000), “Does Consumption Inequality Track Income
Inequality in Italy?”, Review of Economic Dynamics, Vol. 13, No. 1, Elsevier for the
Society for Economic Dynamics, pp. 133-153, January.
Landais, C., T. Piketty and E. Saez (2011), Pour une révolution fiscale, un impôt sur le
revenu pour le XXe siècle, Le Seuil/République des idées.
Maestri, V., F. Bogliacino and W. Salverda (2014), “Wealth Inequality and the
Accumulation of Debt”, Chapter 4 in W. Salverda et al. (eds.), Changing Inequalities
in Rich Countries: Analytical and Comparative Perspectives, Oxford University Press.
Morrisson, C. and F. Murtin (2013), “The Kuznets Curve of Human Capital Inequality:
1870-2010”, Journal of Economic Inequality, Vol. 11, No. 3, Springer, pp. 283-301.
OECD (2014), OECD Economic Survey: Norway 2014, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_surveys-nor-2014-en.
OECD (2013a), OECD Guidelines for Micro Statistics on Household Wealth, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264194878-en.
OECD (2013b), Pension at a Glance 2013: OECD and G20 Indicators, OECD
Publishing, Paris, http://dx.doi.org/10.1787/pension_glance-2013-en.
OECD (2008), “How is Household Wealth Distributed? Evidence from the Luxembourg
Wealth Study”, Chapter 10 in Growing Unequal? Income Distribution in OECD
Countries, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
Piketty, T. (2014), Capital in the 21st century, Harvard University Press.
Piketty, T. and E. Saez (2006), “How Progressive is the U.S. Federal Tax System? A
Historical and International Perspective”, Journal of Economic Perspectives, Vol. 21,
No. 1, pp. 3-24.
Piketty, T. and E. Saez (2003), “Income Inequality In The United States, 1913-1998”,
Quarterly Journal of Economics, Vol. 118, No. 1, MIT Press, pp.1-39.
Sierminska, E., A. Brandolini and T. Smeeding (2006), “The Luxembourg Wealth Study
– A Cross-country Comparable Database for Household Wealth Research”, Journal of
Economic Inequality, Vol. 4, pp. 375–383.
Vermeulen, P. (2014), “How Fat Is the Top Tail of the Wealth Distribution?”, ECB
Working Paper No. 1692, Frankfurt, July.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 281

Database references
OECD Wealth Distribution Database (forthcoming)
OECD National Accounts Database, www.oecd.org/std/na.
OECD Income
database.htm.

Distribution

Database,

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

www.oecd.org/social/income-distribution-

 282 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Annex 6.A1
Methodological features of household wealth data

Data sources and characteristics
Table 6.A1.1 describes the data sources used for this study, the national institution
responsible for data collection as well as a few characteristics of the data. For most of the
countries covered, the data are based on household surveys, which implies that, as in the
case of all household surveys, surveys on wealth suffer from data quality problems due to
sampling and misreporting. The high skewedness of the wealth distribution makes
sampling errors especially important. Misreporting is typically caused by the fact that the
wealthiest households are less likely to respond, in particular about financial wealth. In
most countries, the wealthiest households are more likely to invest in financial assets but
are also those with the least propensity to respond to surveys (OECD, 2013a).
Adjustments are typically made to compensate for errors of sampling and
misreporting. To correct for item non-response, most countries use imputation methods
(Table 6.A1.1). In the HFCS survey, such adjustments include both multiple and simple
imputation. The HFCS dataset provides five imputed values for every missing value
corresponding to a variable entering the composition of household wealth. Data included
in the OECD dataset are based on the average of the five imputed values.
Some fraction of the population such as the wealthiest households might not be well
captured by the survey, which will affect measures of the wealth at the top of the
distribution, as well as of mean wealth levels and wealth distribution.1 Various
possibilities exist to improve the measurement of the wealth data at the top of the
distribution, for example by using information from other sources such as the Forbes List
of very wealthy people, or fitting the top end of the wealth distribution with a Pareto
distribution, following Vermeulen (2014). Several countries conducting wealth surveys
(but not all the countries reviewed here) rely on oversampling the very rich (as indicated
in Table 6.A1.1), with oversampling based on different methods: for example, Austria
oversampled households in Vienna because of a higher non-response rate; Spain and
France used individual data on net wealth based on other sources; and Belgium and
Germany oversampled based on regional-level income.

Comparison of OECD Wealth Distribution Database with other sources
Several non-official compilations of household wealth data have become available in
recent years. Among these, the Crédit Suisse report is the most comprehensive in terms of
country coverage and wealth indicators. The data from Crédit Suisse differ in various
aspects from the data used here, which are based on standardised tabulations provided
from national statistical offices and central banks. First, the Crédit Suisse data aim to
describe the global distribution of household wealth, across individuals, countries and
regions in the world, as opposed to the much narrower remit of the data presented in this

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 283

chapter. Second, they are based on a combination of SNA household balance sheet and
survey data, aimed at providing a consistent description of both the level and the
distribution of household wealth.
The Crédit Suisse estimates of levels of household wealth (which are reported per
adult person, rather than per household) are mainly based on national accounts. This
implies that, for countries where both balance sheet and survey data are available, balance
sheet data are preferred, while surveys are used for those countries where these are the
only source of available information, and econometric and other techniques are used for
other countries. Also, when the available data from national sources are not up-to-data,
Crédit Suisse constructs estimates of current wealth by adjusting previous data by indexes
of house prices, financial assets or GDP per capita growth. The Crédit Suisse data on the
distribution of household wealth are based on survey data for around 20 countries,
estimated from income distribution data for most countries, and for the remaining
countries imputed based on data from other countries with similar income in the same
region; wealth values across the distribution are scaled to match the levels of total
household wealth and adjusted in the top tail to match the number of Forbes billionaires
by country.
The wealth levels reported by Crédit Suisse (2010, 2011, 2012, 2013, 2014) are
broadly similar to those reported in Figure 6.1, but there are some noticeable exceptions.
In both sources, mean wealth is higher in Luxembourg, the United States, the United
Kingdom and Canada, and lower in the Slovak Republic, the Netherlands and Greece.
However, there are significant differences in country rankings across the two sources for
Spain, Norway and Australia.
Comparative information on wealth distribution has been reported recently by both
Crédit Suisse and a range of academic studies (e.g. Piketty, 20142; Maestri et al. 2013).
When looking at wealth concentration at the top of the distribution, most of these studies
identify the United States as one of the most unequal countries, but there is less consensus
for the other countries. For instance, Crédit Suisse (2014) reports a very high wealth
concentration at the top of the distribution for Norway relative to other countries, which is
even higher than those reported in this chapter. Conversely, Maestri et al. (2013) report
above-average Gini indices of wealth inequality for Australia, Canada, Norway and the
United States, countries that are also identified as high-inequality in this chapter.
Estimates of wealth inequalities for individual countries are strongly affected by specific
methodological choices. For example, studies by Beer et al. (2006) and Eizinger et al.
(2008) for Austria report very large wealth concentrations at the top of the distribution
(68% for the top 10%, and 34% for the top 1%) after including privately-held company
assets in total household wealth.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 284 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Table 6.A1.1. Data sources and characteristics
Source
Australia

Organisation

Frequency

Years used

undertaking the survey

of collection

in the analysis

Sample size
(number of
households)

Surv ey of Income and Housing (SIH)

Australian Bureau of Statistics

Ev ery 2 y ears

2006, 2012

Austria

Household Finance and Consumption Surv ey (HFCS-AT)

Oesterreichische Nationalbank

Ev ery 2 or 3 y ears

2010

Belgium

Surv ey of the Financial Behav iour of Households (HFCS-BE)

National Bank of Belgium

Ev ery 2 or 3 y ears

2010

~ 11 400

Canada

Surv ey of Financial Security (SFS)

Statistics Canada

Ev ery 6 or 7 y ears

1999, 2005, 2012

Finland

Household Finance and Consumption Surv ey (HFCS-FN)

Bank of Finland / Statistics Finland

Ev ery 2 or 3 y ears

France

Enquête Patrimoine (HFCS-FR)

INSEE

German Panel on Household Finances (HFCS-GE)

Germany
Greece
Italy
Korea
Lux embourg
Netherlands
Norw ay
Portugal
Spain
United Kingdom
United States

2

Oversampling rate

rich households

on top 10%1

item
non-response

..

Yes

0.01

Yes

0.22

Yes

0.05

Yes

~ 20 000

0.69

No

..

Yes

2010

~ 13 500

0.82

Yes

0.07

Yes

Ev ery 2 or 3 y ears

2010

~ 24 000

0.69

Yes

0.13

Yes

Deutsche Bundesbank

Ev ery 2 or 3 y ears

2010

~ 20 000

0.19

Yes

0.12

Yes

Household Finance and Consumption Surv ey (HFCS-GR)

Bank of Greece

Ev ery 2 or 3 y ears

2010

~ 6 500

0.47

No

..

Yes

Surv ey of Household Income and Wealth (HFCS-IT)

Bank of Italy

Ev ery 2 or 3 y ears

2010

~ 15 500

0.52

No

..

Yes

Surv ey of Household Finances (SHF)

Statistics Korea

Ev ery 5 y ears

2013

~ 20 000

0.91

No

..

No

Household Finance and Consumption Surv ey (HFCS-LX)

Banque Centrale du Lux embourg

Ev ery 2 or 3 y ears

2010

~ 5 000

0.20

Yes

0.05

Yes

2010

–

–

–

–

Yes

~ 2 000

0.58

Wealth Statistics

Central Bureau of Statistics

Annual

Netherlandsche Bank

Ev ery 2 or 3 y ears

Income Statistics for Households

Statistics Norw ay

0.84

Imputation for

No

Surv ey on the Financial Situation of Households (HFCS-PG) Banco de Portugal / Statistics Portugal

Slov ak Republic

Oversampling of

rate

Yes

DNB Household Surv ey (HFCS-NL)

~ 14 500

Response

Annual

2012

–

–

–

–

No

Ev ery 2 or 3 y ears

2010

~ 8 000

0.64

Yes

0.02

Yes

Household Finance and Consumption Surv ey (HFCS-SV)

Národná banka Slov enska

Ev ery 2 or 3 y ears

2010

~ 2 000

..

Yes

0.01

Yes

Financial Surv ey of Households (HFCS-SP)

Banco de España

Ev ery 2 or 3 y ears

2010

~ 12 000

0.57

Yes

0.19

Yes

Wealth in Great Britain (WGB)

Office for National Statistics
Board of Gov ernors of

Ev ery 2 y ears

2012

~ 20 000

0.64

Yes

0.05

Yes

the Federal Reserv e Sy stem

Ev ery 3 y ears

2007, 2010, 2013

~ 9 000

0.68

Yes

0.15

Yes

Surv ey of Consumer Finances (SCF)

Note: “..” means “not available” ; “-“ means “does not apply”.
1. The oversampling rate is calculated as the difference between the number of wealthy households in the sample and in the population.
2. Data on the United Kingdom are limited to Great Britain.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209113

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 285

Annex 6.A2
Long-term determinants of private wealth in eight OECD countries

The long-term determinants of the stock of private wealth, as constructed by Piketty
(2014), have been assessed in eight OECD countries (Australia, Canada, France,
Germany, Italy, Japan, United Kingdom, United States) over the period 1970-2010.
Piketty’s wealth series relate to private wealth (including firms’ wealth) as a share of
GDP, rather than to the narrower concept of household wealth used in this chapter. When
describing his “second fundamental law of capitalism”, Piketty argues that private wealth
converges towards a fundamental level that is determined by the ratio of the savings rate
and the rate of economic growth. In addition, Piketty identifies a number of factors that
may lead to an appreciation of financial and property assets relative to consumer prices.
To capture the importance of the factors identified by Piketty, the following econometric
model has been used:

Yi ,t  ai  dt  Si ,t  gi ,t  pia,t  pih,t  ui ,t
where Yi ,t stands for private wealth, S i ,t for gross national savings as a share of GDP,
a
g i ,t for the rate of economic growth, p i ,t for an index of stock market prices divided by
the deflator of private consumption, p ih,t for an index of house prices divided by the
deflator of private consumption, u i ,t for a statistical residual, a i for country fixed-effects
and d t for time dummies. The results are described in the Table 6.A2.1.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 286 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?
Table 6.A2.1. Long-term determinants of private wealth in eight OECD countries, 1970-2010
(1)

(2)

(3)

Dependent variable is private capital
as a share of GDP per capita
Gross national sav ings rate

-0.054***

-0.014

-0.007

(0.008)

(0.010)

(0.009)

-5.389***

-5.673***

-5.006***

(0.946)

(1.261)

(1.158)

Relativ e house prices

1.890***

1.315***

1.015***

(0.157)

(0.174)

(0.150)

Relativ e stock prices

0.696***

0.487***

0.491***

(0.082)

(0.105)

Economic grow th

Lagged dependent v ariable

(0.094)
0.334***
(0.049)

Country dummies

Yes

Yes

Yes

Time dummies

No

Yes

Yes

N

298

298

274

0.91

0.94

0.99

R

2

Note: To avoid unit roots problem, the fifth lag of the dependent variable is used.
Estimates are significant at the 10% level *, 5% level ** and 1% level ***.
Source: OECD Wealth Distribution Database.
12 http://dx.doi.org/10.1787/888933209125

The estimates suggest that private wealth as a share of GDP per capita is negatively
related to the pace of economic growth, while the national savings rate displays no robust
association with it. The relative prices of dwellings and stocks always have strongly
significant positive signs. When household-level variables are used for savings and
income, the effect of relative house price remains significant; this regression, however, is
based on many fewer observations and the results are therefore not reported here.
Overall, these results indicate that there is empirical support for most of the factors
identified by Piketty (2014), with the exception of the national savings rate. This latter
finding is probably explained by the fact that the regression uses country dummies to
identify the effects of the change of variables over time, and that the savings rate typically
follows erratic movements from one year to another around some constant threshold.
As a second step, it is possible to calculate the contributions of economic growth,
relative house prices and relative stock prices to changes in the ratio of private wealth to
GDP over time. To do so, decennial averages of the three factors (over the periods
1970-79, 1980-89, 1990-99, and 2000-10) are calculated first; then, their change between
the 1970s and the 2000s is multiplied by the corresponding coefficients reported in the
table above, and related to the observed change in the private wealth-to-GDP ratio. This
analysis suggests that economic growth, relative house prices and relative stock prices
each explain about 20% of the variation in the share of private wealth-to-GDP ratio on
average among the eight OECD countries between the 1970s and the 2000s. The rest of
the change in private wealth reflects unobserved factors that are not taken into account in
the above regression. Overall, the cumulative effects of the three factors explain 86% and
98% of the increase in private wealth in the United Kingdom and in the United States
respectively. The increase in house prices explains a larger share of private wealth’s
increase in the United Kingdom (38%) and in the United States (45%), while stock prices
account for 32% of the increase in private capital in the United States.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 287

Notes
1.

Nevertheless, the impact on the most important indicators produced by the survey for
the other fractions of the population should be limited.

2.

Piketty’s (2014) historical analysis is mainly based on a combination of national
accounts and tax data. For some countries, the existence of wealth and estate tax
returns since early times allows examining historical trends on wealth concentration.
Estate returns used for taxing inheritances generally contain a complete listing of the
assets and debts of the deceased, as well as information on his/her demographic
characteristics. However, data of this type are typically available only irregularly and
for a limited number of countries. Also, estate data reflects the wealth of the deceased
population, who are predominantly older persons, implying that the living and the
deceased population are not directly comparable. To use estate returns to derive
information on the distribution of wealth among the living, researchers use “mortality
multipliers” (i.e. the inverse of mortality rates by age groups, with some adjustments
to weights to reflect differential mortality by gender and social status.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 288 – 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES?

Annex 6.A3
Estimations of future pension wealth

An alternative approach to assessing the amount of household pension wealth is
provided by models that provide an estimate of the stream of pension benefits that people
with different characteristics may expect to receive in the future. The indicator presented
in the OECD reports Pensions at a Glance and shown below refers to the projected
pension wealth of future retirees and is based on the lifetime flows of all pension benefits
(public and private) for a “typical individual”. It does not illustrate the current pension
wealth of real individuals. Calculations use country-specific mortality rates and a uniform
discount rate of 2%. According to OECD pension models, the net pension wealth for
individuals entering the labour market in 2012 (net of the taxes and social security
contributions paid on pension income) is highest in Luxembourg (at 12 and 14 times
average annual earnings for men and women, respectively) and lowest in Mexico, the
United Kingdom and the United States (Figure 6.16). At the same time, voluntary private
pensions (which are not considered in OECD pension models) are significant sources of
income in countries such as Canada, Ireland, the United Kingdom and the United States.
In most OECD countries, mandatory or quasi-mandatory public pension plans
account for almost the totality of pension wealth (Figure 6.17). Mandatory private
pensions account for about 50-60% of the total mandatory pension package in Australia,
Denmark, Israel and the Netherlands, and are significantly larger in Chile, Iceland and
Mexico.
Figure 6.A3.1. Net pension wealth for future retirees as a multiple of individual annual gross earnings,
OECD countries
Men (↗)

Women

16
14
12

10
8

OECD (women)
OECD (men)

6
4
2
0

Source: OECD (2013), Pensions at a Glance 2013: OECD and G20 Indicators, OECD Publishing, Paris,
http://dx.doi.org/10.1787/pension_glance-2013-en.
12 http://dx.doi.org/10.1787/888933208701

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 6. HOW DOES THE CONCENTRATION OF HOUSEHOLD WEALTH COMPARE ACROSS COUNTRIES? – 289

Figure 6.A3.2. Balance between public and private provision of mandatory pensions
Percentage share of pension wealth
Public (↘)

Private

%

100
90
80
70
60
50
40
30
20
10
0

Source: OECD (2013), Pensions at a Glance 2013: OECD and G20 Indicators, OECD Publishing, Paris,
http://dx.doi.org/10.1787/pension_glance-2013-en.
12 http://dx.doi.org/10.1787/888933208713

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 291

Chapter 7
Inequality and fiscal redistribution in emerging economies1

This chapter looks at income inequality and fiscal redistribution in emerging economies.
The first part describes overall levels and trends in inequality as well as in social
spending and taxation in selected emerging economies, comparing them to those typically
recorded in the OECD area. It also highlights some prominent redistributive policies
recently implemented in major emerging economies and OECD key partner countries.
The second part of the chapter provides an in-depth analysis of redistribution in seven
middle-income countries that are part of the Commitment to Equity (CEQ) project:
Brazil, Chile, Colombia, Indonesia, Mexico, Peru and South Africa. Using a common
method of fiscal incidence analysis, it examines the redistributive impact and effect on
poverty of fiscal policy, comprising direct taxes, cash transfers, net indirect taxes and inkind benefits in the form of education and health services.

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities.
The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and
Israeli settlements in the West Bank under the terms of international law.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 292 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

7.1. Introduction and key findings
This chapter looks at income inequality and redistribution in emerging economies.
Levels of inequality in these countries tend to be higher than in most OECD countries,
although there is a great diversity in trends over the recent decade. Notably, middleincome countries in the Latin American region have recorded promising signs of
decreasing inequality (Lustig et al., 2012), and the trend towards greater inequality
appears to have come to a halt in a number of other countries, including China (OECD,
2015a).
In contrast to recent trends in much of the OECD area where austerity measures have
been implemented in the frame of fiscal consolidation, most emerging economies have
strengthened social protection and intensified redistribution measures in order to address
concerns about high levels of poverty and inequality. In some countries, income
redistribution measures have become a key pillar of the growth model (OECD, 2014a).
Section 7.2 of this chapter summarises differences in the overall levels and trends in
inequality as well as in social spending and taxation in emerging economies, comparing
key indicators with those typically recorded in the OECD area. It also highlights some
prominent examples of redistribution policies recently implemented in major emerging
economies and OECD key partners.
The bulk of the chapter (Section 7.3) examines the redistributive impact of fiscal
policy for seven middle-income countries that are part of the Commitment to Equity
(CEQ) project,2 for a year around 2010: Brazil, Chile, Colombia, Indonesia, Mexico, Peru
and South Africa.3 “Fiscal policy” in this chapter is defined to include the following main
fiscal interventions: direct taxes, direct cash transfers, net indirect taxes and transfers inkind (benefits in the form of education and healthcare services). The analysis uses a
common methodology across countries, which is designed to be as comprehensive as
possible without sacrificing detail in any particular component of the analysis. The
method of fiscal incidence analysis is described in detail in Lustig and Higgins (2013)
and summarised in Box 7.2 and Annex 7.A1. In essence, the method uses an “accounting
approach” to estimate the average incidence of fiscal interventions. As is common in
most such benefit-tax incidence analyses, the results do not take into account behavioural,
lifecycle, or general equilibrium effects.
This chapter makes several important contributions. First, it updates and extends the
discussion on inequality trends and redistribution policies in major emerging economies
and OECD key partner countries. Second, results from the fiscal incidence analysis are
comparable across countries because the seven country studies in Section 7.3 apply a
common methodology. Third, this analysis presents results not just for inequality but also
for poverty. Fourth, the analyses estimate the contributions of the main fiscal
interventions to the overall reduction in inequality. Finally, sensitivity analyses present
results for an alternative scenario in which contributory pensions are considered as
deferred income rather than as a government transfer.
The main findings can be summarised as follows.


Inequality levels are generally higher in emerging economies. However, in
contrast to the OECD area, some follow a downward trend, especially in the
Latin American region – although this decline slowed as from 2010, especially in
terms of poverty alleviation. Other countries, such as China, Indonesia and South
Africa, have become more unequal over the long run.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 293



The level of social spending and taxation is still relatively low in emerging
economy countries, though there is large country diversity. Social spending
levels are very low in India and Indonesia, while in Brazil, Latvia and Russia
they are just below the OECD average. Tax levels follow a similar pattern.
General consumption taxes (mainly VAT and sales taxes) account for the bulk of
tax revenue, and their share is considerably larger than the OECD average.



Redistribution policies have been strengthened in many EEs, with several
countries making their cash transfers more generous and others aiming at
widening their coverage of unemployment benefits and health insurance, and
expanding educational opportunities. At the same time, there is considerable
scope to make the tax system more progressive and to increase tax revenues by
promoting formal employment and enlarging the tax base.



Among the seven country case studies in Section 7.3, overall fiscal policy lowers
income inequality to various degrees, with the largest redistributive effect in
South Africa and the smallest in Indonesia. South Africa’s result can be
attributed to the combination of a large redistributive effort with transfers
targeted to the poor and direct taxes targeted to the rich. In spite of this, South
Africa remains the most unequal of the seven countries.



Income redistribution tends to be higher in more unequal countries to start with:
redistribution is considerable higher in countries with higher market income
inequality such as South Africa and Brazil than in countries with relatively lower
inequality such as Indonesia and Peru.



As expected, the level of income redistribution is related in part to the size of the
budget allocated to social spending (as a share of GDP). Redistribution is
considerably larger in countries with high social spending, such as Brazil and
South Africa, than in Colombia, Indonesia and Peru, where social spending is
more limited. However, differences across countries suggest that institutional,
political and demographic factors also affect the level of redistribution.



Direct taxes and direct transfers generally exercise an equalising force. Indirect
taxes are equalising in Chile, Mexico and Peru, neutral in the case of South
Africa but increase inequality in Brazil, Colombia and Indonesia.



Contributory pensions are equalising in Brazil, Colombia and Indonesia and
unequalising in Mexico and Peru, and very slightly so in Chile.



Total per capita spending on public education tends to be higher for poorer
households (i.e., pro-poor) in all countries except for Indonesia, where the per
capita benefit is roughly the same for all households. Government spending on
tertiary education increases with income in all countries, but only in Indonesia it
increases inequality.



Health spending is pro-poor (that is, per capita spending declines with income) in
Brazil, Chile, Colombia and South Africa. In Mexico, the per capita benefit is
roughly the same across the income scale. In Indonesia and Peru, health spending
per person tends to increase with income but still reduces inequality.



Although education and health spending have the highest redistributive effect of
the different components of fiscal policy, the existing information cannot
disentangle to what extent the progressivity or pro-poorness of education and
health spending is a result of differences in household or personal characteristics

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 294 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
that could explain a more intense use by poorer households (e.g., having more
children and worse health) or the “opting-out” of those better-off.


While fiscal policies overall unambiguously reduce income inequality, in terms
of poverty reduction, the outcome is less auspicious. In Chile, Indonesia, Peru
and South Africa poverty after cash transfers, net direct taxes and net indirect
taxes is lower than market income poverty. In Colombia, however, income
poverty increases after taxes and cash transfers are taken into account, a result
driven by the impact of indirect taxes. Also, in Brazil income poverty would be
higher if public pensions are considered as deferred income rather than a public
transfer, which means that a portion of the poor who are not pensioners are net
payers into the fiscal system.

This chapter is organised as follows. Section 7.2 provides an overview of inequality
and redistribution policies in emerging economies. Section 7.3 offers a detailed analysis
of the redistributive effect of fiscal policies in seven middle-income countries.
Section 7.2 is divided into four sub-sections. The first gives an overview of inequality
levels and trends in the seven middle-income countries studied in detail in Section 7.3 as
well as other major emerging economies in comparison to the OECD area. The second
sub-section presents the composition of public social spending and public revenue
patterns for the same set of countries. The third sub-section highlights recent examples of
redistribution policies that have been developed and implemented in selected countries
and the final sub-section concludes.
Section 7.3 is divided into seven sub-sections sections which provide a detailed
analysis of the impact of fiscal policy on inequality and poverty in seven middle-income
emerging economies. The section also examines the degree of “pro-poorness” of
government spending on education and health, and a final sub-section concludes.

7.2. Inequality trends and policy responses in emerging economies (EEs)
Income inequality in emerging economies
As documented in Chapter 1 of this report, income inequality has risen in most OECD
countries over the past three decades, in some cases reaching historical levels. Emerging
economies – where inequality levels are often higher than in OECD economies – have
also followed this pattern during the earlier decades of the 1980s and 1990s; but there
have been more promising trends over the past decade in some of them.
There is great variation in inequality levels between the world’s main regions (as
defined in Ferreira et al., 2015; Lopez-Calva and Lustig, 2010). Latin America and the
Caribbean is the region with by far the highest level of inequality. Inequality is also above
average in Sub-Saharan Africa. It stands at an average level in East Asia, while eastern
Europe, Central and South Asia are the most equal geographical areas. Among emerging
and developing economies, generally speaking, inequality is lower in low-income
countries, and the highest in upper-middle income countries.
Based on the latest data available, the level of inequality in emerging economies is
higher than in the OECD region on average, and close to or above the level in highinequality OECD countries. Figure 7.1 compares the Gini coefficients for the OECD
average, as well as for the OECD's most equal and most unequal countries, with the
inequality levels in the seven countries studied in Section 7.3 and a selection of key
emerging economies. While the average Gini coefficient is 0.32 in the OECD area – 0.25
among the most equal countries and 0.4 among the most unequal ones – it is between 0.34
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 295

and 0.4 in India, Indonesia and the Russian Federation; around 0.45 to 0.56 in China, Peru,
Mexico, Chile, Brazil and Colombia; and reaches the highest level in South Africa (0.67).
It is important to bear in mind in this comparison that not all data are perfectly
comparable. The data for the OECD countries, the Russian Federation and Colombia
(blue bars) are based on the OECD standardised income concept, while other data are
based on different concepts, in particular for Indonesia and India, which are consumptionbased and likely to underestimate the level of inequality (see Box 7.1).
Figure 7.1. Gini coefficients in selected emerging economies and OECD countries, total population
Level of income inequality (Gini coefficient), 2013 or latest year available

Latest data (OECD Income Distribution Database)

Latest data (Secondary data sources)

Mid-2000s

Mid-1990s

Mid-1980s

0.70
0.65
0.60
0.55
0.50

0.45
0.40
0.35
0.30
0.25

0.20

Note:
Data for Indonesia, Argentina, Peru, Brazil, South Africa, India and China (grey bars) are based on secondary data sources and
are not strictly comparable with OECD Income Distribution Database (IDD) data (blue bars). The Gini coefficients are based on
equivalised incomes for OECD countries, Colombia, Latvia and the Russian Federation and per capita incomes for other partner
countries, except India and Indonesia for which per capita consumption was used. Mid-1980s data for Argentina and Mexico
refer to 1986 and 1984 respectively. Mid-1990s data for Mexico, Peru and Indonesia refer to 1994, 1997 and 1996 respectively.
Mid-2000s data for Mexico, Chile and Russian Federation refer to 2004, 2006 and 2008.
Source: OECD Income Distribution Database for OECD countries, Latvia, Russian Federation and Colombia; World Bank,
Poverty and Inequality Database for India; Statistics Indonesia (Susenas) for Indonesia; OECD (2015), All on Board: Making
Inclusive Growth Happen in China, OECD Publishing for China; SEDLAC (Socio-Economic Database for Latin America and
the Caribbean) Database for Argentina, Brazil and Peru. For South Africa, data refer to Leibbrandt, M., I. Woolard, A. Finn and
J. Argent (2010), “Trends in South African Income Distribution and Poverty since the Fall of Apartheid”, OECD Social,
Employment and Migration Working Papers, No. 101, OECD Publishing, Paris; and to Finn, A. and M. Leibbrandt (2013),
“Mobility and Inequality in the First Three Waves of NIDS”, SALDRU Working Paper, No. 120, University of Cape Town,
NIDS Discussion Paper, No. 2013/2 for 2012.
12 http://dx.doi.org/ 10.1787/888933208729

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 296 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
While inequality has been continuously increasing in most OECD countries in the last
three decades, it has tended to decrease since the mid-1990s in several emerging
economies, especially in Latin America and the Caribbean (Lopez-Calva and Lustig,
2010). Inequality decreased in particular in Peru, Mexico, Brazil and Argentina (since the
2000s), and to a lesser extent in Chile. Nevertheless, the decline of inequality in this
region tended to be much more modest as from 2010, especially in terms of poverty
alleviation. Inequality also decreased in the Russian Federation. By contrast, Indonesia
and South Africa have become more unequal over time. Inequality has also increased
over the long run in China, especially in the years following the opening-up of the
economy in the 1990s, but has tended to stabilise since 2010 (OECD, 2015a). Inequality
in India was broadly stable.
These developments took place in a context of increasing real household incomes in
many countries, such as the Russian Federation, Brazil, India and South Africa. In China,
rapid economic expansion did not translate into equivalent gains in aggregate household
disposable income until around 2010 (OECD, 2015a). This illustrates that income growth
as such is not a guarantee for smoothing the income distribution.
The underlying forces driving inequality trends in the emerging economies are
strongly related to the role of the labour market, as is also the case in the developed
economies. One of the main drivers of declining income inequality in a number of
countries has been the reduction in labour earnings inequality associated with decreasing
wage dispersion and/or increasing numbers of adults participating in the labour force. In
Argentina, Brazil, Mexico and Peru, wage effects are found to account for a great deal of
the reduction in inequality (Lopez-Calva and Lustig, 2010). In contrast, in the Russian
Federation, labour market changes might have hampered the decrease of inequality (ILO,
2015).
The reduction in inequality in Latin America results from a joint reduction of nonlabour income inequality, due to the increase and better targeting of government transfers,
and – above all – labour income inequality, due to a decline in the gap between high- and
low-skilled earnings (the so-called “education premium”, see Lopez-Calva and Lustig,
2010). In Brazil, for example, but also in Argentina, Mexico and Peru, the traditionally
high education premium has narrowed. The reduction of the education premium has been
driven – among other reinforcing factors – by increased access to education (Lopez-Calva
and Lustig, 2010; OECD, 2014a). In contrast, in South Africa, the increase in income
inequality is related to the combination of rising unemployment and an increase in wage
disparities (Leibbrandt et al., 2010; ILO, 2015).
While the underlying causes of pressures on the income distribution – technological
change, globalisation, structural labour market changes – are characteristic of OECD and
non-OECD countries alike, some key drivers are specific to emerging economies. First,
the structure of the labour market differs from developed economies. In particular, the
informal sector is sizeable. The share of informal workers in total employment is above
60% in Colombia and Peru, 50% in Brazil and Mexico, 40% in Argentina and Chile;4 it
reaches 80% in India, 70% in Indonesia, and 30% in China and South Africa.5 The
informal sector has a major impact on income distribution through many channels, such
as wage penalties, low career progression, exclusion from the labour market and social
protection regulations (OECD, 2011a). High shares of workers outside the formal sector
implies lower progressivity of personal income tax and social transfers (see for example
OECD, 2015d). Since the crisis, the risk of becoming informal increased, particularly for
disadvantaged groups (OECD, 2010a).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 297

Box. 7.1. Inequality measurement in emerging economies
The baseline measure to estimate income inequality in OECD countries is the net equivalised household
disposable income, with income standard concepts and definitions following the Guidelines of the Canberra
Group Handbook on Household Income Statistics (United Nations, 2011). The resulting indicators are available
at the OECD Income Distribution Database (www.oecd.org/social/income-distribution-database.htm). This
database also covers a number of OECD member and non-member emerging economies. Assessing the extent of
income inequality and its evolution over time in other emerging economies is challenging due to differences in
the standard concepts and definitions and in the underlying databases.
First, in some cases (e.g. India, Indonesia), inequality measures are based on consumption, which generally
results in lower levels of inequality than do income-based measures (Morelli et al., 2015). Deininger and Squire
(1996) suggest that, on average, Ginis computed on a consumption expenditure basis are some 6.6 points lower
than when computed on an income basis. Moreover, differences in the definitions of consumption itself make
international comparisons even more difficult. Among the main obstacles, it could be mentioned the variation in
the number of consumption items, whether survey participants record their consumption themselves or are
interviewed, the length of the recall period, and differences in the methods used to impute housing, durables and
home production, which alters the impact of in-kind consumption.
Second, the definition of income can also vary – and usually does – between different datasets, depending,
for instance, on whether and how in-kind income, imputed rents and home production are treated, and whether
specific income sources such as remittances, private transfers, or property income are properly captured.
Further, incomes may be reported on a net or gross income basis. In the latter case (such as for the data on
Argentina), inequality measures are based on pre-tax income and do not reflect the impact of redistribution
policies. This typically leads to higher levels of reported inequality.
Finally, as a welfare indicator, income can be adjusted for economies of scale with so-called equivalence
scales (see www.oecd.org/els/soc/OECD-Note-EquivalenceScales.pdf). While all OECD estimates rely on
calculations of equivalised household incomes, the data available for most emerging economies use income per
capita as a standard, which assumes away any economies of scale within households.
Several projects were recently developed to improve cross-country comparisons in the field of inequality, with
the aim of building “secondary inequality data sets”. Not only do they collect existing data sources, but they also
reshape the data to improve their comparability (for example, by trying to reach common definitions when there is
no ex-ante comparability). They also propose best second-order choices among alternative data sources when no
first best option can be isolated, as is usually the case. Prominent examples of such databases are the Luxembourg
Income Study (LIS), the United Nations University-World Institute for Development Research (WIDER), the
World Income Inequality Database (WIID), the World Income Distribution (WYD), PovCal and the World
Development Indicators (WDI) World Bank datasets, and the Chartbook of Economic Inequality Data (Atkinson
and Morelli, 2014). The Journal of Income Inequality (2015 special issue) and Förster and Tóth (2015) provide a
detailed review and discussion of the strengths and weaknesses of these and other datasets.

Access to education and educational quality also substantially influence income
inequality trends in emerging economies. PISA results show that several emerging
economies still lag considerably behind OECD countries with regard to educational
achievement. Significant gaps remain also in educational outcomes for different socioeconomic backgrounds. They hamper the employability and competence of the workforce
and add to greater overall inequality.
Gender, race and discrimination against migrants are important factors behind the
high levels of earnings inequality in many emerging economies. In India, for instance,
occupational segregation weighs on gender inequality (Kapsos et al., 2014). In South
Africa, income poverty persists at acute levels for the African and Coloured racial groups
and in urban areas (Leibbrandt et al., 2010). In South Africa, inequality within the labour
market – due to high unemployment, informality and earnings inequality – have
contributed for the labour market playing a negative role in the distribution of income.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 298 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
Last, widespread regional divides play an important role in some emerging
economies. They often overlap with regional imbalances due to unequal resource
endowments, distance from markets, long-standing power imbalances, ethnic and racial
disadvantage and institutional weaknesses. In China, for example, spatial inequality
remains high, reflecting the persistently vast differences between urban and rural incomes
within provinces. However, recent developments indicate that poorer provinces have
grown faster than richer ones in recent years (OECD, 2015a), contributing to a decline in
inequality. In Indonesia, persistent regional disparities, exacerbated by a process of
decentralising policy during the 2000s, caused substantial – and increasing – differences
in inequality levels between geographical areas and persistent disparities between levels
of rural and urban poverty (Miranti et al., 2013).

Budget size, social spending and taxation
The redistributive potential of a country is determined first and foremost by the size
and composition of its budget and how government spending is financed. Public social
spending as a share of GDP tends, on average, to be considerably lower in emerging
economies than in OECD countries. Also, there is a wide variation across countries,
ranging from comparatively low levels in Indonesia and India6 to levels close to the
OECD average in Brazil, Latvia and Russia. Public social spending in China is
considerably lower than in the OECD and other EEs, but comparable to average social
spending in the Asia/Pacific region (OECD, 2014b).
Public spending on cash benefits (such as old-age pensions and income support to the
working-age population7) is on average much lower in emerging economies than in OECD
countries and explains the bulk of the differences in overall public social spending.
Furthermore, spending on cash benefits tends to vary much more than does spending on
health and education both across emerging economies and in comparison to the OECD
average. Spending on cash benefits is particularly high in Brazil, Latvia and Russia. In
Brazil, public pension benefits account for most of such spending. In fact, public pension
spending as a share of GDP in Brazil is higher than the OECD average. This is quite
remarkable since the Brazilian population is relatively young, with eight persons aged 20-64
to each person aged 65 or more, double the OECD average (OECD, 2014a, OECD, 2014d).
Similarly, in Colombia, where the population is also young and pension coverage is low,
pensions account for the largest share of public spending on cash benefits (OECD, 2015b).
In South Africa, public spending includes important support programmes for the
working-age population, such as the Community Work Programme, the Extended Public
Works Programme and the Child Support Grant. Public pension spending is low in
comparison with some other emerging economies, as it is focused on social pensions with
basic amounts paid to the poor elderly (OECD, 2014c). Contributory pensions are
available only to public servants who belong to the Government Employees Pension Fund
(Inchauste et al., 2015). Peru allocates relatively little to income redistribution through its
signature cash transfer Juntos. In Indonesia, a considerable part of public social
protection tends to be provided through energy subsidies rather than cash transfers (Afkar
et al., 2015). However, the energy subsidies were scrapped in early 2015 and Indonesia
plans to use such social savings for development.
In most countries spending on health and education ranges between 3% and 5% of
GDP, which is slightly less than the OECD average of around 6%. The main exceptions
are South Africa, which spends considerably more on education, and India and Indonesia,
which spend significantly less on health (see Figure 7.2).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 299

Figure 7.2. Size and composition of government budgets
Panel A. Public primary and social spending as a share of GDP in 2012 or latest year available
Old age

Other

Old age and other

Health

Education

30
25

20
15
10
5
0

OECD

Average

BRA

ZAF

COL

CHL

MEX

PER

IDN

LVA

RUS

CHN

IND

CHN

IND

Panel B. Composition of total government revenue as a share of GDP in 2010
Other

Indirect taxes

Social contributions

Direct taxes

40
35
30
25

20
15
10

5
0

OECD

Average

BRA

ZAF

CHL

MEX

COL

PER

IDN

RUS

LVA

Note:
Panel A: For OECD countries, data for Australia, Canada, Chile, Israel, Korea, New Zealand, the United States refer to 2012 otherwise they
refer to 2011. Data for Brazil and Colombia refer to 2010, and to 2009 for China, India and Indonesia. Data on public spending on
Education are not available for Greece, Luxembourg and Turkey, data are therefore not comparable with other countries. Data for Brazil
refer to Federal social expenditure from Ipea (Instituto de Pesquisa Econômica Aplicada) report (Abrahão de Castro, J., J. Aparecido Carlos
Ribeiro, J. Valente Chaves and B. Carvalho Duarte (2012), Gasto Social Federal: prioridade macroeconômica no período 1995-2010,
No. 9, Brasília, September) except for health where data are for general government expenditure on health from WHO. Data exclude social
spending by local authorities, including on non-federal civil servants. Benefits for civil servants (mostly pensions) amounts to 2.3% of GDP
(from the description of the Brazilian system from Ministério da Previdência Social, 2009, Overview of Social Welfare in Brazil, 2nd
Edition, January as in www.previdenciasocial.gov.br/arquivos/office/3_091113-150152-707.pdf%20–%202009 – 2009), it shows that there
are as many local government pensioners as civil servant pensions. Hence an estimation of 2 to 2.5% of GDP from the data in not taken into
account. The policy areas covered include old age, survivors, incapacity-related benefits, family, health, active labour market policies,
unemployment, housing and other social policy areas. Old age refers to cash old age and survivors pensions.
Panel B: Data for South Africa are not comparable as they do not include social security. Data for India and Indonesia refer to cash taxes
collected by general government as a percentage of GDP.
Source: Panel A: OECD (2014), Social Expenditure Database (www.oecd.org/social/expenditure.htm). Data for South Africa are from
National Budget 2014, Estimates of National Expenditure, National Treasury and World Health Organisation (WHO). Data for China,
Indonesia and India refer to Asian Development Bank’s Social Protection Index (SPI) Database except for Health where they refer to World
Health Organisation Global Health Expenditure Database (WHO) and Education from China Census Bureau and Unesco respectively. Data
for the Russian Federation refer to Federal State Statistics service Rosstat. Data for Peru refer to ECLAC (United Nations Economic
Commission for Latin America and the Caribbean). Data for Brazil refer to Federal social expenditure from Ipea (Instituto de Pesquisa
Econômica Aplicada). Panel B: OECD Revenue Statistics 2014, China, India and Indonesia: Government Finance Statistics, IMF, 2010,
Latvia: Eurostat and Euromod 2010; Brazil, Chile, Mexico, Colombia, Peru: Revenue statistics in Latin America, OECD 2015; South
Africa, National Treasury, 2013/2014; Russian Federation, Government at a Glance OECD (2013) and Federal state statistics service, 2012.
12 http://dx.doi.org/10.1787/888933208730

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 300 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
Although still below the OECD level, on average tax revenue as a proportion of
national income has increased steadily in Latin American countries8 and China since the
1990s, as well as in Russia, India and Indonesia up until the 2007/08 economic crisis. There
is wide variation in the tax burden across emerging economies, from 12% of GDP in
Indonesia to 33% in Brazil (Figure 7.2, Panel B). Except for South Africa, general
consumption taxes (mainly VAT and sales taxes) account for the largest share of tax
revenue, considerably higher than the OECD average. Taxes on income and profits are the
second-largest source of revenue, with corporate income taxes usually playing an important
role (larger than in most OECD countries). The third largest source of public revenues is
social security contributions. In addition to taxes, Brazil, Colombia and Mexico rely on
non-tax revenues from natural resources (OECD/ECLAC/CIAT/IDB, 2015).
Among the seven middle-income countries studied in detail in Section 7.3, given the
size and structure of their public budgets, Brazil and South Africa have the largest amount
of resources at their disposal to engage in fiscal redistribution. At the other end of the
spectrum are Indonesia and Peru. Whether Brazil and South Africa achieve higher
redistributive outcomes, however, depends on how the burden of taxation and the benefits
of social spending are distributed. This will be discussed in Section 7.3.

Redistributive policies in emerging economies: An overview
While the subsequent section provides an in-depth comparative analysis of the
redistributive impact of fiscal policy in seven selected emerging economies, previous
OECD work has discussed the role of policies in tackling high inequality in emerging
countries more generally, and in a broader and more qualitative way (OECD, 2010b,
2011a). This work suggested that direct fiscal redistribution (direct taxes and cash
benefits) had a limited but increasingly strong role in easing inequality. High levels of
self-employment and sizeable informal sectors have up to now been limiting the capacity
of the tax authorities to verify taxpayers’ declared income, and there have also been
administrative bottlenecks in tax collection on personal income. On the other hand, and
related to the former, lower levels of social protection and social safety nets have put
brakes on the capacity of policy to reduce high disparities.
In this context, the policy focus on simultaneously reducing inequality while
promoting inclusive employment deals with four key areas, in which many emerging
economies have recently redoubled their efforts (OECD, 2014d):
1.

Improving social protection and moving towards a more universal welfare system.

2.

Increasing equality of opportunity through public health and education services.

3.

Better incentives for more formal employment and enhancing job quality.

4.

Expanding the redistributive capacity of the tax system.

Improving social protection
All emerging economies have in place cash transfers that play an important role in
equalising distribution and alleviating poverty. These programmes can be unconditional,
such as the Child Support Grant (CSG) in South Africa or the Dibao minimum
subsistence benefit in China, or conditional, such as Brazil’s Bolsa Familia and Mexico’s
Prospera (former Oportunidades) schemes, which provide cash transfers to low-income
households that are conditional on school attendance and health check-ups. In Indonesia,
a similar conditional cash transfer (the Bantuan Siswa Miskin, BSM) exists together with
a cash transfer targeting vulnerable households and promoting the completion of higher
education (the Program Keluarga Harapan, PKH; OECD, 2015d).
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 301

Cash transfers are an important instrument to redistribute resources to the lower end
of the distribution. Conditional cash transfers may be particularly well suited to reducing
inequality and promoting social mobility in emerging economies (EEs). The recent
refinements to the programmes in Mexico and Brazil have further increased the
programmes’ generosity, effectively lifting all participants whose income was below the
national poverty line above that threshold. There is also evidence that such programmes
have contributed positively to growth, by freeing women from their caring duties and
bringing them into the labour force (Del Valle Suarez, 2014).
Social welfare programmes could be further strengthened by appropriate targeting
and payment mechanisms, better management structures and better design of the
conditionalities. In practice, however, the task of appropriately identifying the population
in need is difficult. In addition, there are often trade-offs between reducing exclusion
errors and improving efficiency. In South Africa, the take-up rate of CSG is only 60%. In
the case of the Indonesian Program Keluarga Harapan (PKH), which delivers a cash
benefit to the poorest families with children and/or with pregnant women, only 13% of
households who live below the poverty line received the actual benefit (Miranti et al.,
2013). To tackle such issues, some countries such as Indonesia rely on “proxy” meanstests that use available information on household characteristics while efforts to develop
single registries of vulnerable households could result in better cost-effectiveness (OECD,
2015d). South Africa and Brazil use income declarations, which may be less effective as
they are more prone to errors or under-declaration.
In many emerging economies, the coverage of the unemployment insurance system
needs to be widened to help workers cushion against a transitory loss of income. The
share of the unemployed effectively covered by unemployment benefits is much lower
than in advanced economies, generally less than 15%, except for Chile and the Russian
Federation where the share ranges from 20 to 25%, respectively (OECD, 2011b).
Furthermore, these schemes tend to be less generous than the OECD average, with lower
replacement rates and shorter duration of entitlements in most emerging economies
(OECD, 2015a). In many countries income support for the unemployed is restricted to
formal workers: in Brazil, it applies only to formal workers who are dismissed without
just cause and workers who lost their jobs when their firms closed down, thus excluding
the majority of the unemployed.
There have been many initiatives to widen coverage of unemployment in recent years.
For instance, in 2014 Mexico announced the introduction of a nation-wide unemployment
benefit scheme. In other countries with unemployment insurance, fragmentation remains a
challenge. In China, many workers covered by insurance do not receive benefits, and many
local governments do not require employers to affiliate migrant workers to unemployment
insurance schemes. As a result, the unemployment benefit system covers less than half of the
urban working population. As a policy response, provincial governments have been asked to
take more responsibility for social insurance from local city schemes (OECD, 2015b).

Increasing equality of opportunities through public health and education services
Improving equitable access to health care services is an important element for greater
equality of opportunities. Recently, China, India, Indonesia and South Africa committed
themselves to providing universal health care for their populations. Often, the main
challenges in achieving this include the organisation of adequate financing and improving
the delivery of healthcare. Indonesia created an insurance programme for the poor in 2005
and in 2014 launched a universal health care programme, Jaminan Kesehatan
Nasional (JKN), to be rolled out by 2019 (EIU, 2015). In India, the expansion of the national
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 302 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
health insurance RSBY system for the poor is an important step, as large out-of-pocket
expenses associated with private hospital stays have long been a barrier to health care access
(OECD, 2014b). In 2012, China achieved near universal health care coverage, but migrant
workers typically pay higher out‐of‐pocket expenses, and the country still needs to broaden
the scope of health care to provide a more comprehensive set of services (OECD, 2015b).
Another important policy avenue to decrease disparities is to continue investing in
policies that promote the up-skilling of the workforce. Argentina and Brazil are
interesting examples of countries that have been successful over the past two decades in
promoting more equal access to education, while broadening the distribution of school
attainment (Lopez-Calva and Lustig, 2010). In both countries, the expansion of basic
education – supported by non-school family policies to improve early childhood health
and nutrition programmes, and progress in the service infrastructure – has contributed to
narrowing the earnings gap between skilled and low-skilled workers. Investments in
education in India and Indonesia have likewise increased access to education, even
though progress so far in reducing income gaps has been less tangible, particularly among
the most disadvantaged. India still has low literacy levels but it is now approaching nearuniversal enrolment in elementary education, spending on secondary education has risen
significantly in recent years, and the 2009 “Secondary Education for All Action Plan”
aims to provide universal access to secondary education by 2017 (OECD, 2014b).
Overall, in many EEs investment in education has paid off in terms of educational
attainment but the quality of education is still lagging. While this may be partly related to
low public spending in many countries, efficiency of public spending on education is also
key as, for instance, in Indonesia public spending, increased substantially in the past decade
but outcomes have remained largely static (OECD, 2015e). Countries that have improved
their performance in PISA, like Brazil, or Colombia, for example, have established policies
to improve the quality of their teaching staff (OECD, 2014e). In this respect, soft skills such
as teachers’ expectations of their students’ futures but also the type and level of teacher
certification were found to matter for Latin American educational outcomes (OECD,
2015f). In secondary education, performance seems also to be affected by students’ socioeconomic background and the way the school operates and there is thus a need to improve
both equity and quality.
Recent OECD work suggests that wider access to vocational pathways in secondary
education can keep youth who are disaffected with academic education engaged in the
educational process (Quintini and Manfredi, 2009). More vocational education could be a
particularly interesting option for emerging economies, insofar as this could not only
improve nationwide graduation rates, but also play a considerable role in smoothing the
transition from school to work. Yet, only low percentages of students are generally involved
in vocational education: for instance, no more than 10% of upper secondary students attend
vocational courses in India and Mexico (OECD-ILO, 2011a). The National Policy on Skill
Development in India is an interesting example in this respect: it encompasses the creation of
a private-public partnership to strengthen the engagement of industry in skills development
and promotes greater employer involvement in the country’s Industrial Training Institutes
(OECD-ILO, 2011b). In Brazil, the Federal Pronatec programme launched in 2011 aims to
expand the federal network of technical schools, to provide free training places for youth
from disadvantaged backgrounds and to include bursaries and loans (OECD, 2013).
Youth who leave school without an upper secondary qualification as well as youth
facing multiple disadvantages find it difficult to access stable, formal employment. In
recent years, access to safety nets in advanced economies has been made conditional on
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 303

an active job-search and training. In Mexico, the Oportunidades programme has been
replaced by a new programme, Prospera. Children can now receive scholarships for
college or technical college; and those looking for a job have priority in the National
Employment Service (OECD, 2015d). In Argentina and a number of other Latin
American countries, youth programmes that combine training with other services, such as
job readiness and job-search assistance, have shown that these improve the employment
and earnings prospects of participants (Elias et al., 2004; Pagés et al., 2009).

Better incentives for formal employment and improved job quality
Faced with high levels of informality, one important priority is to widen the coverage of
the formal sector in order to enhance the distributive capacity of the tax system. This would
require special emphasis on improving revenue-collection procedures and strengthening
labour inspections (OECD, 2011a). But this requires sufficient resources to enable the
labour inspection system to carry out its work effectively. The number of workers per
labour inspector remains very high in countries like Colombia, Turkey and Mexico (30 000,
39 000 and about 192 000 respectively while the ILO benchmark for transition countries is
20 000 workers per inspector (OECD 2015a). Labour inspectors should be adequately
qualified and able to use modern statistical techniques in order to be more effective. These
techniques include the statistical profiling of workers and firms who are most at risk of
informality and selective targeting of enforcement actions. Almeida and Carneiro (2011),
for instance, show that frequent labour inspections in Brazil have made it “more” attractive
to be a formal worker. In addition to better enforcement, reforms that increase transparency
and the efficiency of how revenues are used (to which the quality of public services is vital)
might help to increase tax morale and therefore tax compliance (Daude et al., 2012).
Both employers and employees need to have better incentives for formal
employment. The costs of formality could be lowered for employers and the selfemployed by simplifying tax and administrative systems. Over the past two decades,
Brazil adopted a number of policy measures to reduce the costs of formality, such as the
“Simples Law” that introduced a more progressive tax structure and simplified the
collection of taxes and social security contributions. It is estimated that these measures
contributed to the formalisation of 500 000 micro-enterprises accounting for 2 million
jobs from 2000 to 2005 (Delgado et al., 2007). In India, recent efforts at the central
government and state levels to modernise labour regulations and reduce compliance
requirements are steps in the right direction to help the expansion of formal employment
(OECD, 2014b). The Mexican government has started to address the problem of
informality through the creation of a new tax regime for micro and small enterprises,
which seeks to promote the formalisation of the workforce, and through a comprehensive
strategy launched in 2014 to “Go Formal”, which pools the benefits and resources of
different programmes. Mexico’s new fiscal regime for small firms (RIF), which started to
operate in January 2014 and replaced an earlier small-taxpayer regime (REPECOS),
substantially reduces the obligations on personal, social security and value added and
excise tax in the initial decade of operation, to induce informal firms to regularise their
status and start paying taxes (OECD, 2015d).

Expanding the redistributive capacity of the tax system
Enhancing the distributive capacity of the tax system would require an emphasis on
improving revenue collection procedures and strengthening the extent to which taxpayers
comply voluntarily with their obligations. A focus on the fight against corruption would
also help improve tax collection. In Colombia, for instance, the tax administration has
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 304 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
little effective control over customs administration due to a lack of personnel and other
constraints, which has led to very high levels of VAT evasion on imports. More
inspection of taxpayers requires more technical capacity and personnel to take advantage
of information technologies to detect potential tax fraud. Tax fraud penalties could also
be increased, by following general practice in OECD countries and by making domestic
and offshore tax evasion a criminal offence (OECD, 2015c).
In the mid-term, greater redistribution in emerging economies requires a change in the
structure of the tax system. Special attention should be given to striking a better balance
between tax revenue from personal income and property taxes, on the one hand, and
consumption taxes, on the other. Broadening tax bases could also contribute to meeting
objectives on efficiency, growth and distribution. South Africa is an example of a country
that, since the end of apartheid, focused on broadening the tax base and building an efficient
tax administration to generate the resources it needed to progressively expand the social
safety net for the poor. As a result, South Africa relies more on direct taxes – personal
income tax (PIT), corporate income tax, and payroll tax – than do other similar EEs, while
the top earners pay a much higher fraction of the overall PIT (Inchauste et al., 2015).
Finally, in many EEs the personal income tax could be made more progressive and
exemptions could be removed for higher income groups. Colombia took a step in this
direction with the 2012 reform, which implemented an alternative minimum personal
income tax (IMAN) that acts as a cap on some exemptions, increasing the effective tax
rate paid by high-income households (OECD, 2015c). India had ambitious plans to
reform income tax and make its system more progressive, via a proposed Direct Tax
Code Bill that aimed at reducing the corporate income tax rate to 30%, removing some
tax allowances and widening income brackets. For individuals, currently the lowest
marginal tax rate is applied at over 2.5 times the average wage, or three times GDP per
capita, a very high level compared to Brazil, China, Indonesia and South Africa
(Gandullia et al., 2012). The Direct Tax Code Bill lapsed, however (OECD, 2014b).
China has recently strengthened redistribution by enacting measures for better tax
collection and implementing property taxes. The guidelines issued in 2013 also called for
increased social expenditure (from 36% of government outlays in 2011 to 38% by 2015),
with an emphasis on low-income regions through intergovernmental transfers. As in
South Africa, effective implementation at the local level is likely to be critical to
effectively functioning redistributive policies, especially in big, expanding urban areas.

Conclusions
Emerging economies face generally higher levels of inequality than OECD countries.
Income inequality rose in most emerging economies during the decades of the 1980s and
early 1990s, as it did in most OECD countries. However, the trend has been more
contrasted since the 2000s. Latin America and the Caribbean, which remains the most
unequal region by far, became less unequal, as was the case in Russia. Indonesia, China
and South Africa faced increases in inequality over the long run but this trend seems to
have come to a halt in the latter two countries in the recent decade.
While the underlying causes of pressure on the income distribution – technological
change, globalisation, structural labour market change – affect OECD and non-OECD
countries alike though through different channels, some key drivers are specific to
emerging economies. In particular, the size and persistence of the informal sector, access
to and the quality of education, discrimination based on gender or ethnic origin or
widespread regional divides all heavily weigh on inequality.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 305

Many emerging economies have progressed in enhancing their social protection
systems (unemployment compensation, social assistance programmes such as cash
transfers and healthcare benefits). Improving educational attainment and the school-towork transition are other avenues for making a long-term impact on inequality and
poverty. Sizeable informal sectors together with administrative bottlenecks on the
collection of personal income tax limit its redistributive capacity. Some important areas
for future reform include promoting formalisation through tax simplification, better
monitoring and enforcement, and improving incentives for both employers and workers.
In addition, emerging economies face the challenge of broadening the tax base while
making taxation more progressive.

7.3. Fiscal policy and income redistribution in Brazil, Chile, Colombia, Indonesia,
Mexico, Peru and South Africa9
Fiscal policy components, income inequality and redistribution
Different components of fiscal policy contribute to changes in income inequality
across countries in different orders of magnitude. Figure 7.3 shows the level of inequality,
measured by the Gini coefficient, for different income definitions: market income,
disposable income, post-fiscal income and final income.10 Following Box 7.2, each
income definition includes a set of components of fiscal policy. The order in which the
components appear may or may not correspond to an institutional sequence. For example,
to include the component of net indirect taxes after disposable income makes sense
institutionally since individuals are likely to pay consumption taxes based on their
disposable – and not their market income. However, whether education and health
spending should go at the end of the income accounting framework or not depends more
on the fact that it is desirable to separate the cash from the in-kind component of social
policy rather than at which point these transfers in kind exercise their effect on the
individual’s decision to consume them.
Box 7.2. Measuring fiscal incidence and income accounting framework
Fiscal incidence analysis assesses the distributional impact of a country’s taxes and transfers. Essentially,
fiscal incidence analysis allocates taxes (personal income tax and consumption taxes, in particular) and public
spending (social spending in particular) to households or individuals so that one can compare their impact on the
distribution of income (Box 7.4 discusses indicators to measure the impact of taxes and transfers on the
distribution of income). Transfers include both cash transfers and benefits in kind such as publicly-provided
government services in education and healthcare. Transfers may also include consumption subsidies such as
food, electricity and fuel subsidies.

Income concepts
The diagram below summarises the four income concepts used throughout this chapter. Starting from
market income (mainly, income from labour and capital), each new income concept adds another element of the
fiscal system to the previous one. Hence, disposable income subtracts direct personal income taxes and adds
cash transfers, post-fiscal income subtracts indirect taxes and adds subsidies and final income adds public health
and education. The particular issue of including public contributory pensions with market income or cash
transfers (i.e. disposable income) is discussed in Box 7.3.
In broad terms, disposable income measures how much income individuals may spend on goods and
services as well as save. Post-fiscal income measures how much individuals are able to actually consume. For
example, a given level of disposable income could mean different levels of actual consumption depending on
the size of indirect taxes and subsidies. Final income includes the value of public services in education and
health if individuals would have had to pay for those services at the average cost to the government. For details
regarding assumptions for the seven countries analysed here, see Lustig (2015b).

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 306 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Box 7.2. Measuring fiscal incidence and income accounting framework (cont.)
The income accounting framework

TRANSFERS
ES

Market income
Wages and salaries, income from capital, private
transfers (remittances, private pensions, etc.)
before taxes, social security contributions and
government transfers AND contributory social
insurance old-age pensions ONLY in the case in
which pensions are treated as deferred income

Personal income taxes AND
employee contributions to social
security ONLY in the case that
contributory pensions are treated
as transfers

-

Direct cash and near cash transfers:
conditional and unconditional cash
transfers, school feeding programmes,
free food transfers, etc.

TAXES

+

Disposable
income
Indirect subsidies: energy, food
and other general or targeted price
subsidies

+

-

Indirect taxes: VAT, excise taxes and other
indirect taxes

Post-fiscal (or
consumable) income

In-kind transfers: free or subsidised
government services in education
and health

Co-payments, user fees

+

Final income

Income measurement
In this chapter, the welfare indicator used in the fiscal incidence analysis is income per capita. This indicator
differs from the one used in other chapters – based on equivalised income (see Box 7.1). The analysis used here
is point-in-time and does not incorporate behavioural or general equilibrium effects. It is a first-order
approximation that measures the average incidence of fiscal interventions. The analysis is based on economic
rather than statutory tax incidence. For example, it is assumed that personal income taxes and contributions by
employees and employers are borne by labour in the formal sector. Individuals who are not contributing to
social security are assumed to pay neither direct taxes nor contributions. Consumption taxes are fully shifted
forward to consumers. The analyses take into account the lower consumption tax incidence associated with
own-consumption, rural markets and informality. Finally, it is worth noticing that CEQ data is aimed at
incidence analysis and does not necessarily coincide with that found in other sources, in particular National
Accounts.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 307

Box 7.2. Measuring fiscal incidence and income accounting framework (cont.)
Due to data limitations and country specificities, particular data adjustments had to be applied in some
countries. In Indonesia, the household survey reports consumption rather than income. Disposable income was
imputed as equal to consumption and market income was generated “backwards” applying a “net to gross”
conversion (see Immervoll and O’Donoghue, 2001). Furthermore, the Indonesian survey does not include
individuals with income levels beyond the threshold at which direct taxes begin to apply (see Afkar et al., 2015).
In the data for South Africa, Free Basic Services are assumed as direct transfers. The only contributory pensions
in South Africa are for public servants who must belong to the Government Employees Pension Fund. Since the
government made no transfers to the GEPF in 2010/11, there is no scenario in which contributory pensions are
treated as a transfer. Also, survey data on own-consumption (which is part of market income) were not
considered reliable in the case of South Africa (see Inchauste et al., 2015). In Chile, contributions to the old
(pay-as-you-go) pension system are not available as a separate item in National Accounts (Ruiz-Tagle and
Contreras, 2014). See Annex 7.A1 for details about the household surveys used.

Imputing income from public health and education
The imputation of income from public health and education relies on a series of assumptions in three areas:
i) how to value public services (usually done via a production cost approach); ii) how to allocate these services
across households (either with an “actual consumption approach” or an “insurance value approach” in the case
of health); and iii) how to account for differences in needs (use of particular equivalence scales).
In this chapter, the value of education was imputed as the benefit accrued to an individual going to public
school, estimated as the per beneficiary input costs (for example, the average government expenditure per
primary school student obtained from administrative data is allocated to the households based on how many
children are reported attending public school at the primary level). In the case of health, the approach was
analogous. The value of public health is imputed as the average cost to the government of delivering healthcare
services to the beneficiaries. In the case of Colombia, however, the method used was to impute the insurance
value to beneficiary households rather than base the valuation on utilisation of healthcare services. Thus, this
approach measures how much a household would have had to pay for the public service at the full cost to the
government. No adjustments for differences in needs were made here.
Similarly, OECD (2008; 2011a) examined the distributive effects of in-kind public transfers (services in
health, education, housing and care) in OECD countries. Such analysis typically makes use of a methodology of
“extended income”, i.e. the value of public services is imputed into household income. Broadening disposable
income to account for services have been found to increase households' economic resources by as much as 30%
to 40% in OECD countries, with notable effects on income inequality and poverty.
Differences in the quality of services between but also within countries can typically not be accounted for.
Also, consumers may value services quite differently from what they cost. Furthermore, the indirect effects of
services (such as the increase in income through higher labour supply) cannot be taken into account. These
issues are discussed in detail in Verbist et al. (2012).
Source: Lustig, N. and S. Higgins (2013), “Commitment to Equity Assessment (CEQ): Estimating the Incidence of Social
Spending, Subsidies and Taxes. Handbook”, CEQ Working Paper, No. 1, July 2011, revised January 2013, New Orleans,
United States.

As can be observed in Figure 7.3, in Colombia, Indonesia and Peru, fiscal income
redistribution is quite limited while in South Africa, Chile and Brazil, it is of a relevant
magnitude. Mexico is in the middle of these two groups. One can observe that South
Africa is the country that redistributes the most but it still remains the most unequal of all
seven. It is interesting to note that although Brazil, Chile and Colombia start out with
similar market income inequality, Brazil and Chile reduce inequality considerably while
Colombia does not. Similarly, Mexico and Peru start out with similar levels of market
income inequality but Mexico reduces inequality by more. Indonesia is the less unequal
of all seven and fiscal redistribution is also the smallest in order of magnitude.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 308 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
Figure 7.3. Fiscal policy components and income inequality
Panel A. Gini coefficient for market, disposable, post-fiscal and final income (pensions as transfers) 2010

Market income

Disposable income

Post-fiscal income

Final income

0.8
0.75
0.7

0.65
0.6
0.55
0.5

0.45
0.4
0.35
0.3

BRA

CHL

COL

IDN

MEX

PER

ZAF

Panel B. Gini coefficient for market, disposable, post-fiscal and final income (pensions as market income) 2010

Market income

Disposable income

Post-fiscal income

Final income

0.8
0.75
0.7

0.65
0.6
0.55
0.5

0.45
0.4
0.35
0.3

BRA

CHL

COL

IDN

MEX

PER

ZAF

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. For
Indonesia, the fiscal incidence analysis was carried out adjusting for spatial price differences. This adjustment, however, does
not affect figures on this figure. In South Africa, the only statutory contributory pension scheme is the Government Employees
Pension Fund, which is mandatory for officials in national and provisional government. These pensions were not included in the
analysis for South Africa and are not shown here. The scenario for South Africa assumed free basic services are direct transfers.
Source: Lustig, N. (2015), “Inequality and Fiscal Redistribution in Middle Income Countries: Brazil, Chile, Colombia,
Indonesia, Mexico, Peru and South Africa. Evidence from the Commitment to Equity Project (CEQ), CEQ Working Paper, No.
31, Center for Inter-American Policy and Research and Department of Economics, Tulane University and Inter-American
Dialogue.
12 http://dx.doi.org/10.1787/888933208740

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 309

The largest change in inequality occurs between post-fiscal and final income, thus
suggesting that health and education are the main drivers of income redistribution in these
countries. This is not surprising given the fact that governments spend more on education
and health than on direct transfers and pensions. The difference between post-fiscal and
final income inequality is particularly large in Brazil, South Africa and Chile, where they
account for almost 10 percentage points of Gini coefficient. This is somewhat expected
since spending on education and health tends to be larger than on direct transfers and
pensions in most of these countries (see Figure 7.3, Panel A). However, given the strong
assumptions underlying the imputation of the value of in-kind transfers on household
final income (see Box 7.2) such finding must be considered with caution.
Comparing results in Panel A (where pensions are counted as public transfers) and
Panel B (where pensions are counted with market income) reveal that the patterns of
inequality decline are similar whether one looks at the scenario in which contributory
pensions are considered deferred income (and, thus, part of market income) or with
pensions as transfers. In Brazil and Colombia, and to a lesser extent in Indonesia, the
redistributive effect is larger when pensions are treated as a transfer, while in Mexico and
Peru it is somewhat lower. Pensions as transfers have most of an effect in Brazil where
redistribution between market and disposable income would be 2.4 percentage points less
if pensions were not counted as transfers (see also Box 7.3).
Box 7.3. Public contributory pensions as deferred income or social transfers
Public pensions from pay-as-you-go contributory systems may be considered either as deferred income
(i.e., private savings) or as a social transfer (i.e., a public cash transfer). If considered as deferred income, public
pensions are included as part of market income and therefore excluded from the fiscal incidence analysis. If
considered as social transfers, public pensions are added, together with other social cash transfers, to disposable
income and therefore are included as one of the components of fiscal income redistribution. For consistency,
when pensions are treated as deferred income, pension contributions are accounted as private savings and
deducted from market income, while when pensions are treated as social transfers, pension contributions are
included as part of market income and deducted from disposable income as a direct tax.
It is important to note that the treatment of contributory pensions not only affects the amount of spending
and how it gets redistributed, but also the ranking of households by original income or pre-fiscal income. For
example, in the scenario in which contributory pensions are considered a government transfer, households
whose main (or sole) source of income is pensions will have close to zero income before taxes and transfers and
hence will be ranked at the bottom of the income scale. When contributory pensions are treated as deferred
income, in contrast, households who receive contributory pensions will be placed at a (sometimes considerably)
higher position in the income scale. Thus, the treatment of contributory pensions in the incidence exercise may
have significant implications for the order of magnitude of the “pre-fisc” and “post-fisc” inequality and poverty
indicators.
Source: Lustig, N. and S. Higgins (2013), “Commitment to Equity Assessment (CEQ): Estimating the Incidence of Social
Spending, Subsidies and Taxes. Handbook”, CEQ Working Paper, No. 1, July 2011, revised January 2013, New Orleans,
United States.

The redistributive effect of fiscal policy: Do more unequal countries redistribute
more?
Income redistribution tends to be higher in more unequal countries to start with:
redistribution is considerable higher in countries with higher market income inequality
such as South Africa, Brazil and Chile than in countries with relatively lower inequality,
such as Indonesia, Peru and Mexico (see Figure 7.4, Panel A). Among these countries,
Colombia stands as an outlier with a rather low degree of redistribution given its high
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 310 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
level of market income inequality. Previous studies also generally suggest a positive
correlation between market income inequality and measures of redistribution. Lustig
(2015a) finds this in an analysis for thirteen developing countries. An OECD study
(2011a, Chapter 7) illustrates that more market income inequality tends to be associated
with higher redistribution, for a sub-set of OECD countries, both within countries (over
time) and across countries. The findings in Ostry et al. (2014) confirm that this
relationship is particularly strong among OECD countries and weaker, although still
significant, in non-OECD countries.
Differences in redistribution change the relative ranking of countries by inequality level.
Figure 7.4, Panel B displays the levels of income inequality before (horizontal axis) and after
(vertical axis) accounting for fiscal policies. Fiscal policies reduce inequality in all countries,
as none of the countries fall above the 45°-line, which indicates no effect from redistribution.
While South Africa is the most unequal country and Indonesia the least unequal country
based on either definition of income, due to lower redistribution Colombia and Peru are
more unequal than Brazil, Chile and Mexico, once fiscal policies are considered.
Figure 7.4. Inequality and redistribution, 2010
Panel A. Redistribution and market income inequality
Redistribution
0.20

Panel B. Final income inequality and market income
inequality
Final income
0.8

0.18

ZAF

BRA

0.16

0.7

0.14

CHL

ZAF

0.12

0.6
COL

0.10

MEX

0.08

0.5

0.06

PER

0.04

PER

COL

CHL BRA

MEX
0.4

IDN

IDN

0.02
0.00
0.3

0.4

0.5

0.6

0.7

0.8

Market income inequality

0.3
0.3

0.4

0.5

0.6

0.7
0.8
Market income inequality

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. For
Indonesia, the fiscal incidence analysis was carried out adjusting for spatial price differences. This adjustment, however, does
not affect figures on this figure. The only contributory pensions in South Africa are for public servants who must belong to the
Government Employees Pension Fund; they were not included in the analysis for South Africa and are not shown here. The
scenario for South Africa assumed free basic services are direct transfers. Redistribution measures the difference between Gini
of market and final incomes.
Source: Lustig, N. (2015), “The Redistributive Impact of Government Spending on Education and Health: Evidence from
13 Developing Countries in the Commitment to Equity Project”, Chapter 17 in B. Clements, R. de Mooij, S. Gupta and M. Keen
(eds.), Inequality and the Role of Fiscal Policy: Trends and Policy Options, International Monetary Fund, Washington,
forthcoming.
12 http://dx.doi.org/10.1787/888933208758

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 311

As expected, the level of income redistribution and the size of the budget allocated to
social spending (as a share of GDP) are associated. However, differences across countries
suggest that institutional factors such as the composition and design of such policies and
their interaction with socio-economic circumstances also affect the level of redistribution.
Figure 7.5 presents the level of redistribution and social spending measured in the
CEQ database. Redistribution is considerably larger in countries with high social
spending, such as Brazil and South Africa, than in Colombia, Indonesia and Peru, where
social spending is more limited. Given the level of social spending, income redistribution
is particularly high in South Africa and Chile.
Figure 7.5. Redistribution and social spending, 2010
Redistribution
0.20
0.18

ZAF

BRA

0.16
0.14
CHL

0.12
0.10
0.08

MEX

0.06
PER

0.04

COL

IDN

0.02
0.00
0%

5%

10%

15%

20%

25%

Social spending as % of GDP

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. For
Indonesia, the fiscal incidence analysis was carried out while adjusting for spatial price differences. This adjustment, however,
does not affect figures in this figure. The only contributory pensions in South Africa are for public servants who must belong to
the Government Employees Pension Fund; they were not included in the analysis for South Africa and are not shown here. The
scenario for South Africa assumed free basic services are direct transfers. Redistribution measures the difference between Gini
of market and final incomes.
Source: Lustig, N. (2015), “The Redistributive Impact of Government Spending on Education and Health: Evidence from
13 Developing Countries in the Commitment to Equity Project”, Chapter 17 in B. Clements, R. de Mooij, S. Gupta and M. Keen
(eds.), Inequality and the Role of Fiscal Policy: Trends and Policy Options, International Monetary Fund, Washington,
forthcoming.
12 http://dx.doi.org/10.1787/888933208766

Redistribution in comparison to the United States and the European Union
How does fiscal redistribution in these seven middle income countries compare to
that in the United States and the European Union? Although there is no data for an exact
comparison between these groups of countries, results based on somewhat similar
methodologies are available. In the case of the European Union, the EU tax-benefit model
(EUROMOD) estimates the distributive incidence of fiscal policies of 27 member
states.11 In the case of the United States, results are drawn from Higgins et al. (2015,
forthcoming). The scope of policies included by EUROMOD (direct taxes, social security
contributions and direct transfers) is equivalent to that used here to measure the
redistributive effect from market to disposable income. Post-fiscal (i.e., after indirect
taxes) and final income (i.e., after health and education) analyses are not possible.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 312 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
There are three important differences between the United States and the European
Union on the one hand and the seven middle-income countries analysed here, on the
other. First, market income inequality is considerably higher for the middle income
countries than in the European Union and the United States. One important aspect to note,
however, is that taking contributory pensions as either deferred income or as social
transfers (see Box 7.3) changes considerably the extent of market income inequality and
the redistributive effect in the European Union and, somewhat less, in the United States
while it makes a relatively small difference in the seven middle income countries
(Figure 7.6). Second, the redistributive effect is much larger in the EU countries and, to a
lesser extent, in the United States. Third, while in the European Union and the United
States, contributory pensions have a significant equalising effect, in some of the seven
middle income countries pensions have a small or even unequalising effect.
Figure 7.6. Redistributive effect in Brazil, Chile, Colombia, Indonesia, Mexico, Peru, South Africa,
the European Union and the United States
Change in Gini points: Market to disposable income, 2010
Redistribution, pensions as deferred income

Redistribution, pensions as transfers

Gini market income, pensions as deferred income (right scale)

Gini market income, pensions as transfers (right scale)

Redistribution
0.25

Gini of market income
0.9
0.8

0.20

0.7
0.6

0.15

0.5
0.4

0.10

0.3
0.2

0.05

0.1

0.00

0
IDN

COL

PER

MEX

BRA

CHL

ZAF

Avg

USA

EU-27

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012b. For
Indonesia, the fiscal incidence analysis was carried out while adjusting for spatial price differences. This adjustment, however,
does not affect the figures on this figure. The only contributory pensions in South Africa are for public servants, who must
belong to the Government Employees Pension Fund; they were not included in the analysis for South Africa and are not shown
here. The scenario for South Africa assumed free basic services are direct transfers.
Source: Lustig, N. (2015), “Inequality and Fiscal Redistribution in Middle Income Countries: Brazil, Chile, Colombia,
Indonesia, Mexico, Peru and South Africa. Evidence from the Commitment to Equity Project (CEQ), CEQ Working Paper, No.
31, Center for Inter-American Policy and Research and Department of Economics, Tulane University and Inter-American
Dialogue. EUROMOD statistics on Distribution and Decomposition of Disposable Income, accessed at
www.iser.essex.ac.uk/euromod/statistics/ using EUROMOD version No. G2.0. Higgins, S., N. Lustig, W. Ruble and T.
Smeeding (2015), “Comparing the Incidence of Taxes and Social Spending in Brazil and the United States”, Review of Income
and Wealth, forthcoming.
12 http://dx.doi.org/10.1787/888933208778

The contribution of taxes and cash transfers to changes in income inequality
Figure 7.7 shows the marginal contributions of direct and indirect taxes and cash
transfers to changes in inequality between market and post-fiscal income.12 Contributions
that reduce inequality (equalising) are presented with a positive sign and contributions
that increase inequality (unequalising) with a negative sign (see Box 7.4). Direct taxes
and transfers have an equalising effect in all countries. The equalising effect of direct
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 313

transfers is particularly high in South Africa and Brazil. Income redistribution through
direct taxes is also considerable in South Africa, Chile and Brazil.
The redistributive effect of direct transfers is higher than that of direct taxes in most
of the countries analysed. Only in Mexico and Peru is the marginal contribution of direct
taxes higher than that of direct transfers. Similar results are found in the OECD context.
In most OECD countries, direct transfers play a higher redistributive role than direct
taxes. A noticeable exception is the United States, where direct taxes contribute more to
income redistribution than direct transfers (OECD, 2008, OECD, 2015b).
In all analysed countries, the contribution of indirect taxes and transfers to income
redistribution is much lower than that of direct taxes and transfers, except for Indonesia
where direct taxes are not available and direct transfers are small. Indirect taxes have an
unequalising effect in Brazil, Colombia and Indonesia, while the marginal contribution of
indirect taxes is equalising in Chile, Mexico and Peru and neutral in South Africa. This
result is quite remarkable since indirect taxes are usually found to be regressive and to have
an unequalising effect when measured against the distribution of income (OECD/KIPF,
2014). Furthermore, in the case of Chile and South Africa this takes place despite the fact
that indirect taxes are regressive in these countries (see Lustig et al., 2015). This apparently
paradoxical result is consistent with “Lambert’s conundrum” (see Lambert, 2001 and
Annex 7.A1). If taxes are regressive vis-à-vis the original income but progressive with
respect to the less unequally distributed post-transfers income, regressive taxes may exert an
equalising effect over and above the effect of progressive transfers (Lambert, 2001).
Figure 7.7. Fiscal policy components and income redistribution
Marginal contributions of fiscal policy types to income redistribution (in Gini points), 2010
Direct taxes

Direct transfers

Indirect taxes

Indirect subsidies

Redistributive effect

0.09

0.079

0.08
0.07
0.06
0.05

0.045
0.037

0.04

0.031

0.03

0.015

0.02
0.01

0.007

0.006

COL

IDN

0
-0.01

BRA

CHL

MEX

PER

ZAF

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. For
Indonesia, the fiscal incidence analysis was carried out while adjusting for spatial price differences. This adjustment, however,
does not affect figures on this figure. The only contributory pensions in South Africa are for public servants, who must belong to
the Government Employees Pension Fund; they were not included in the analysis for South Africa and are not shown here. The
scenario for South Africa assumed free basic services are direct transfers.
Source: Lustig, N. (2015), “Inequality and Fiscal Redistribution in Middle Income Countries: Brazil, Chile, Colombia,
Indonesia, Mexico, Peru and South Africa. Evidence from the Commitment to Equity Project (CEQ), CEQ Working Paper, No.
31, Center for Inter-American Policy and Research and Department of Economics, Tulane University and Inter-American
Dialogue.
12 http://dx.doi.org/10.1787/888933208789
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 314 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

The contribution of health and education to changes in income inequality
Health and education reduce income inequality in all countries and disproportionately
benefit the less well-off in most countries. Panel A in Figure 7.8 presents the
concentration coefficients of health and education spending for each country (see
Box 7.4). The distribution of public health spending is progressive in all countries (i.e. the
concentration coefficient is lower than the Gini coefficient of market income). However,
in Indonesia and Peru (and to a lesser extent Mexico), health spending per capita tends to
increase with household income (i.e. pro-rich). On the other hand, health spending is propoor (i.e., spending per capita tends to decrease with income) in Brazil, Chile, Colombia
and South Africa. Such pro-poor distribution is particularly large in Chile.
Total spending on education is pro-poor in all countries except for Indonesia, where
the amount per capita is approximately the same throughout the income distribution.
Public spending on pre-school education tends to be pro-poor in all countries for which
there is data, particularly in South Africa. Primary school spending is pro-poor in all
countries. Public spending on secondary education is pro-poor in Brazil, Chile, Colombia,
Peru and South Africa, neutral in Mexico and pro-rich (although progressive) in
Indonesia. Public spending on tertiary education is pro-rich in all countries. However, it is
only in Indonesia that it is also regressive. Compared with their respective levels of
market income inequality, spending on tertiary education is most progressive in South
Africa, followed by Colombia and Chile.
Results regarding the pro-poorness of spending on education and health are
encouraging, as they suggest that use of public education and health services is quite
intensive among low-income households. However, these results do not say anything
about the quality of such services.13 If the quality of the schooling and healthcare
provided by the government is low, distortive patterns will produce major obstacles to the
equalisation of opportunities. For example, one of the reasons why tertiary education
mostly benefits the middle-classes and the rich is that their children are better equipped to
pass the entrance examination, as they receive better (often private) primary and
secondary education.
Furthermore, the existing information cannot disentangle the extent to which the
progressivity or pro-poorness of education and health spending is a result of differences
in household or personal characteristics that could explain a more intense use by poorer
households (e.g., having more children and worse health) or simply the “opting-out” of
the better-off.
Besides its distribution, in the case of these sizeable transfers, the final redistributive
effect of a transfer also depends on its size. Panel B in Figure 7.8 presents health and
education spending as a share of household market income. While in Brazil and South
Africa this result is driven primarily by the level of spending on education and health and
its progressivity, in Chile the level of spending is lower but its progressivity, especial in
health spending, is much higher.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 315

Figure 7.8. Concentration coefficients and budget shares for education and health
Panel A. Concentration coefficients for education and health, 2010
Health

Tertiary education

Secondary school

Middle school

Primary school

Pre-school

Education

0.8
0.6

0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1

-1.2

Health Education Health Education Health Education Health Education Health Education Health Education Health Education
BRA
CHL
COL
IDN
MEX
PER
ZAF

Panel B. Education and health spending as a share of household market income, 2010
Health

Education

9%
8%
7%
6%
5%

4%
3%
2%
1%
0%

Health Education Health Education Health Education Health Education Health Education Health Education Health Education
BRA
CHL
COL
IDN
MEX
PER
ZAF

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. For
Indonesia, the fiscal incidence analysis was carried out while adjusting for spatial price differences. This adjustment, however,
does not affect figures on this figure. The only contributory pensions in South Africa are for public servants, who must belong to
the Government Employees Pension Fund; they were not included in the analysis for South Africa and are not shown here. The
scenario for South Africa assumed free basic services are direct transfers.
Source: Lustig, N. (2015), “The Redistributive Impact of Government Spending on Education and Health: Evidence from
13 Developing Countries in the Commitment to Equity Project”, Chapter 17 in B. Clements, R. de Mooij, S. Gupta and M. Keen
(eds.), Inequality and the Role of Fiscal Policy: Trends and Policy Options, International Monetary Fund, Washington,
forthcoming.
12 http://dx.doi.org/10.1787/888933208792

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 316 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Fiscal policy and the poor
The above discussion has concentrated on the impact of fiscal policy on inequality. This
section focuses on the impact of fiscal policy on poverty. Although related, inequality and
poverty outcomes do not necessarily go in the same direction. The method most widely
used to assess the effect of fiscal policy on poverty is the change in the headcount ratio.
This method compares the percentage of the population below the poverty line before and
after taxes and transfers. While the OECD generally measures poverty using a relative
poverty line (see Chapter 3), in the context of emerging economies an absolute poverty line
is usually used. Here, a poverty line equivalent to USD PPP 2.50 per day is used. This is a
rather low income level and is used by Ferreira et al. (2013) to measure “extreme poverty”.
Unlike the discussion on income inequality and redistribution, the analysis is based on
comparisons between market income and post-fiscal income, thus measuring the impact of
direct and indirect taxes and cash transfers, but excluding health and education.
Fiscal policies in terms of cash transfers and direct and indirect taxes14 reduce income
poverty in most, but not all, the countries analysed. Income poverty rates decline
significantly after accounting for fiscal policies in Chile, Mexico, Brazil and South Africa.
However, in Colombia income poverty increases after taxes and cash transfers are taken
into account. Similarly, in Brazil income poverty also rises if public pensions are excluded
from public social spending (i.e. considered as deferred income, see Box 7.3). These results
are mainly driven by the regressive impact of indirect taxes (see Figure 7.7). Higgins and
Pereira (2014) suggest that in the case of Brazil, this result is due to the heavy taxation on
basic foodstuffs.
Figure 7.9. Fiscal policy and reduction of extreme poverty
Percentage change in headcount ratio from market to post-fiscal Income 2010
Poverty change due to fiscal policy

Poverty change due to fiscal policy (excluding pensions) (↘)

10%
5%
0%
-5%
-10%

-15%
-20%
-25%

-30%
-35%

BRA

COL

IDN

PER

MEX

ZAF

CHL

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. Poverty is
measured with the international poverty line of USD PPP 2.50/day (with 2005 PPP). Countries are ranked by the change in
poverty for the scenario where contributory pensions are included in market income. Data for Indonesia is consumption-based,
but income-based for the rest. The South Africa scenario assumes Free Basic Services as direct transfers. The fiscal incidence
analysis for South Africa does not include a scenario with contributory pensions as transfers, and hence is not shown above.
Source: Lustig, N. (2015), “Inequality and Fiscal Redistribution in Middle Income Countries: Brazil, Chile, Colombia, Indonesia,
Mexico, Peru and South Africa. Evidence from the Commitment to Equity Project (CEQ), CEQ Working Paper, No. 31, Center for
Inter-American Policy and Research and Department of Economics, Tulane University and Inter-American Dialogue.
12 http://dx.doi.org/10.1787/888933208806

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 317

Who gains and who loses with fiscal policies? Figure 7.10 shows the distribution of net
payers (i.e. those who pay more in taxes than they receive in transfers) and net recipients
(i.e. those who receive more in transfers than they pay in taxes) across market income
categories. These categories are based on Ferreira et al. (2013) and Lopez-Calva and OrtizJuarez (2014), who use them to identify the “extreme poor” (USD 0 to 2.50 per day), the
“moderate poor” (USD 2.50 to 4 per day), “the vulnerable” (USD 4 to 10 per day) and the
“middle class” (USD 10 to 50 per day). The results presented below take public pensions as
deferred income and thus exclude them from the public social spending on cash transfers
(see Box 7.3).
The distribution of net payers and net recipients indicates that lower income
households tend, on average, to benefit from fiscal policies. In particular, in all countries
the most vulnerable income groups (those earning less than USD 1.25 or USD 2.50 per
day) are net recipients. Above that income threshold, the net effect of fiscal policies
differs considerably across countries. In Brazil, only those in “extreme poverty” are on
average net recipients of fiscal policies. Households in “moderate poverty” are net
contributors to the fiscal system, mainly due to indirect taxes. As already mentioned,
consumption taxes on basic foodstuffs have a strong negative effect on the incomes of
low-income households in Brazil.
In Chile, Colombia, Mexico, Peru and South Africa, households in “moderate
poverty” are also net recipients. In Indonesia, the “middle class” are also, on average, net
recipients. Due to the combination of low direct and indirect taxes and untargeted energy
subsidies, only the “rich” are net contributors to the fiscal system in Indonesia.
Figure 7.10. Winners and losers of the fiscal system
Net receivers and net payers to the fiscal system, 2010
Net receivers

Net payers

Indonesia
Mexico
Chile
Colombia
Peru

South Africa
Brazil
y<1.25

1.25<=y<2.5

2.5<=y<4

4<=y<10

10<=y<50

y>=50
y= USD per day

Note:
Brazil, Chile and Peru refer to 2009, Colombia, Mexico and South Africa refer to 2010 and Indonesia refers to 2012. Poverty is
measured with the international poverty line of USD PPP 2.50/day (with 2005 PPP). Countries are ranked by the change in
poverty for the scenario where contributory pensions are included in market income. Data for Indonesia is consumption-based,
but income-based for the rest. The South Africa scenario assumes Free Basic Services as direct transfers. The fiscal incidence
analysis for South Africa does not include a scenario with contributory pensions as transfers, and hence is not shown above.
Source: Lustig, N. (2015), “Inequality and Fiscal Redistribution in Middle Income Countries: Brazil, Chile, Colombia, Indonesia,
Mexico, Peru and South Africa. Evidence from the Commitment to Equity Project (CEQ), CEQ Working Paper, No. 31, Center for
Inter-American Policy and Research and Department of Economics, Tulane University and Inter-American Dialogue.
12 http://dx.doi.org/10.1787/888933208810
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 318 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Box 7.4. Measuring redistribution, progressivity and “pro-poorness”
A typical indicator of the redistributive effect of fiscal policy is the difference between the Gini coefficient
of market income before and after taxes and transfers. If the redistributive effect is positive (negative), fiscal
policy is equalising (unequalising).

Measuring the contribution of individual fiscal policies to redistribution
There are several ways of calculating the contribution of a particular fiscal intervention to the change in
inequality (or poverty), taking account of path dependency. The most commonly used in the literature are the
methods of marginal contribution and sequential contribution. A less commonly used measure is the “total
average contribution”, which is calculated by considering all the possible paths and taking, for example, the socalled Shapley value (Shorrocks, 2013).
The sequential contribution is calculated as the difference between inequality indicators with fiscal
interventions ordered in a path according to their presumed institutional design. For example, if direct transfers
are subject to taxation, the sequential contribution would first measure the effect of transfers on market income
and then of taxes on market income plus transfers (usually known as gross income). This approach requires
detailed knowledge about the interactions between the different fiscal interventions. OECD (2011a, Chapter 7)
and Immervoll and Richardson (2011) adopted a sequential approach to examine the redistributive effects of
taxes and transfers, identifying the most suitable sequence for each country. However, getting the institutional
sequencing right is close to impossible – especially as the analysis becomes more granular in terms of individual
components of fiscal policy. Moreover, the issue of path dependency described in the annex may not go away
even if the institutional sequencing is perfect.
The marginal contribution is calculated by taking the difference between the inequality indicator without
and with a given policy with regard to a fixed income concept. * This indicator is equivalent to asking the
question: what would inequality be if the system did not have a particular tax (or transfer) or if a tax (or
transfer) was modified? Would inequality be higher, the same or lower with the tax (or transfer) than
without it?
One drawback of the marginal contribution method is that the sum of all the marginal contributions is not
equal to the total redistributive effect, due to the interaction between policies, thus limiting the accuracy of the
magnitude of each contribution (Shorrocks, 2013).
In this chapter, marginal contributions are calculated based on decile rather than sample micro-data. Since
this measure ignores intra-decile inequality, the overall redistributive effect level is larger than the one
computed using micro-data (and shown in the sections above). Furthermore, the decile-based analysis assumes
no re-ranking. Since public pensions produce a substantial re-ranking, the results based on pensions as transfers
are not included. For the scenario with pensions as market income, the extent of re-ranking caused by
redistributive policy is small, so the decile-based analysis is a good approximation.

Progressivity and pro-poorness
The distribution of public spending on healthcare and education is measured here using the concentration
coefficient, which is an indicator calculated in a way analogous to the Gini coefficient (see Lambert, 2001). A
transfer is defined as progressive whenever its share of market income tends to fall as market income rises,
which is the case when the concentration coefficient is lower than the Gini coefficient for market income.
Conversely, a transfer is regressive whenever its share of market income tends to increase as market income
rises, which is the case when the concentration coefficient is higher than the Gini coefficient for market income.
If a transfer as a share of market income is constant (i.e. proportional), it is defined as neutral.
A transfer is defined as pro-poor whenever its absolute amount tends to fall with market income (i.e.
progressive in absolute terms) – this takes place when the concentration coefficient is negative. Conversely, a
transfer is defined as pro-rich whenever its absolute amount tends to rise with market income (i.e. regressive in
absolute terms) – this takes place when the concentration coefficient is positive.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 319

Box 7.4. Measuring redistribution, progressivity and “pro-poorness” (cont.)
Note that in the analysis presented here, households are ranked by per capita market income and no
adjustments are made to their size because of differences in the composition. In some analyses, the pro-poorness
of education spending, for example, is determined by using children as the unit of analysis instead of all
household members. Given that, on average, poorer families have a larger number of children, the observation
that concentration curves are pro-poor is a reflection of this fact and not that poorer families receive more
resources per child.
*. The marginal contribution should not be confused with the marginal incidence, the latter being the incidence of a small
change in spending. The marginal contribution is not a derivative. Note that, because of path dependency, adding up the
marginal contributions of each intervention will not be equal to the total change in inequality. Clearly, adding up the
sequential contributions will not equal the total change in inequality either. One approach that has been suggested to
calculate the contribution of each intervention so that they add up to the total change in inequality is to use the Shapley
value. The studies analysed here do not have estimates for the latter.
Source: Lustig, N., A. Enami and R. Aranda (2015), “The Analytics of Fiscal Redistribution”, Chapter in N. Lustig and S.
Higgins (eds.), Commitment to Equity Handbook: Estimating the Redistributive Impact and Pro-poorness of Fiscal Policy,
forthcoming.

Conclusions
Section 7.3 of this chapter examines the redistributive impact of fiscal policy in
Brazil, Chile, Colombia, Indonesia, Mexico, Peru and South Africa. In particular, it
addresses the following questions: What is the impact of fiscal policy on inequality and
poverty? What is the contribution of direct taxes and transfers, net indirect taxes and
spending on education and health to the overall reduction of inequality? How pro-poor is
spending on education and health?
In order to analyse the impact of fiscal policy on income inequality it is useful to
separate the “cash” portion of the system (direct and indirect taxes, direct transfers and
indirect subsidies) from the “in kind” portion (the monetised imputed value of the use of
government education and health services). The results show that the reduction in
inequality induced by the cash portion of the fiscal system is quite heterogeneous, with
South Africa redistributing the most and Indonesia the least. Redistribution levels are
determined primarily by the amount of resources devoted to (collected from) direct
transfers (direct taxes) and their progressivity, and the presence of unequalising net
indirect taxes.
While the cash portion of the net fiscal system is always equalising, the same cannot
be said with respect to poverty. In Brazil and Colombia, the headcount ratio measured
with the international extreme poverty line of USD PPP 2.50 per day is higher for postfiscal income than for market income. In these two countries, fiscal policy increases
poverty, meaning that a significant number of households with income around the poverty
line are made poorer (or poor) by taxes and transfers. This startling result is primarily the
consequence of high consumption taxes on basic goods.
The equalising effect of direct taxes is higher than of direct transfers in Chile, Mexico
and Peru, while the converse is true in Brazil, Colombia, Indonesia and South Africa. Net
indirect taxes increase inequality in Brazil, Colombia, Indonesia and South Africa but
reduce it in Chile, Mexico and Peru. If taken as social transfers, public contributory oldage pensions are equalising in Brazil, Colombia and Indonesia but unequalising in Chile,
Mexico and Peru. Education and health have an equalising and rather large effect in all

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 320 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
countries. Total spending on education decreases with income (i.e., is pro-poor) in all
countries except for Indonesia, where it is neutral in absolute terms. Pre-school and
primary school tend to be pro-poor in all countries, and particularly so in South Africa.
Secondary school is progressive in all countries and pro-poor in most. Spending on
secondary school per capita is neutral in Mexico, and it increases with income in
Indonesia. Public spending on tertiary education per capita increases with income (i.e.., is
pro-rich) in all countries. However, only in Indonesia it is also regressive (and
unequalising). Compared with their respective levels of market income inequality,
spending on tertiary education is most progressive in South Africa, followed by Colombia
and Chile.
Public health spending is progressive in all countries. However, in Indonesia and Peru
(and to a lower extent Mexico) health spending per capita tends to increase with
household income (i.e. pro-rich). On the other hand, health spending is pro-poor in Brazil,
Colombia, South Africa and, particularly, in Chile.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 321

Notes
1.

Section 7.3 of this chapter and its annexes are authored by Nora Lustig, Tulane
University. Nora Lustig would like to thank Luis Felipe Munguia and Rodrigo
Aranda, for their research assistance.

2.

www.commitmentoequity.org/, see Annex 7.A1.

3.

The background information for the analyses comes from the following fiscal
incidence analyses: Brazil (Higgins and Pereira, 2014), Chile (Ruiz-Tagle and
Contreras, 2014), Colombia (Melendez, 2014), Indonesia (Jellema et al., 2014),
Mexico (Scott, 2014), Peru (Jaramillo, 2014) and South Africa (Inchauste et al.,
2015); and Lustig, Pessino and Scott (2014) and Lustig (2015a and b).

4.

Source: SEDLAC, http://sedlac.econo.unlp.edu.ar/eng/index.php.

5.

Source: ILO, http://laborsta.ilo.org/informal_economy_E.html.

6.

The available data on public social spending in India and Indonesia is likely to
underestimate the public social effort, as outlays by state and other local governments
are under-reported (OECD, 2014a).

7.

For a discussion on the treatment of pensions in the income accounting framework,
see Box 7.2 and Box 7.3.

8.

Revenue and spending differences between OECD and LAC economies are smaller if
natural resources and private pension schemes are taken into account (see
OECD/ECLAC/CIAT/IDB, 2015, Equivalent Fiscal Pressure).

9.

This section and its corresponding annex are authored by Nora Lustig, Tulane
University. Nora Lustig would like to thank Luis Felipe Munguia and Rodrigo
Aranda, for their research assistance.

10.

Other measures of inequality such as the Theil index or the 90/10 ratio are available
in the individual studies. Requests should be addressed directly to the authors.

11.

The model for Croatia is currently being developed under the project
EUROMODupdate2
(www.iser.essex.ac.uk/euromod/developingeuromod/euromodupdate2).

12.

Pensions are counted here with market income rather than with transfers, as the
results are not available under both definitions.

13.

OECD/CAF/ECLAC (2014) offers a detailed analysis of quality indicators for
education systems in Latin American Countries.

14.

The effect of in-kind benefits of education and health services cannot be taken into
account in this analysis.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 322 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

References
Abrahão de Castro, J., J. Aparecido Carlos Ribeiro, J. Valente Chaves and B. Carvalho
Duarte (2012), Gasto Social Federal: prioridade macroeconômica no período 19952010, No. 9, Brasília, September.
Afkar, R., J. Jellema and M. Wai-Poi (2015), “The Distributional Impact of Fiscal Policy
in Indonesia”, Background document, Tulane University and World Bank.
Almeida, R. and P. Carneiro (2011), “Enforcement of Labor Regulation and Informality”,
IZA Discussion Papers, No. 5902, Institute for the Study of Labor (IZA), Bonn.
Atkinson, A.B. and S. Morelli (2014), “Chartbook of Economic Inequality”, ECINEQ
Working Paper, Vol. 324.
Barrera-Osorio, F., M. Bertrand, L. Linden and F. Perez-Calle (2011), “Improving the
Design of Conditional Transfer Programs: Evidence from a Randomized Education
Experiment in Colombia”, American Economic Journal: Applied Economics, Vol. 3,
No. 2, pp. 167-195.
Daude, C., Gutiérrez, H. and Á. Melguizo (2012), “What Drives Tax Morale?”, OECD
Development Centre Working Papers, No. 315, OECD Publishing, Paris,
http://dx.doi.org/10.1787/5k8zk8m61kzq-en.
Deininger, K. and L. Squire (1996), “A New Data Set Measuring Income Inequality”,
World Bank Economic Review, Vol. 10, No. 3, pp. 565–591.
Del Valle Suarez, A. (2014), “From Caring to Work: The Labor Market Effects of NonContributory Health Insurance”, Unpublished manuscript.
Delgado, G., A.C. Querino, A. Campos, F. Vaz, L. Rangel and M. Stivali (2007),
“Avaliação do Simples: Implicações à formalização previdenciária”, Texto para
Discussão, No. 1277, IPEA.
Ebrill, L, M. Keen, J.P. Boudin and V. Summers (2001), The Modern VAT, International
Monetary Fund, Washington, DC.
Elias, V., F. Ruiz-Nuñez, R. Cossa and D. Bravo (2004), “An Econometric Cost-Benefit
Analysis of Argentina’s Youth Training Program”, Research Network Working Paper,
No. R-482, Inter-American Development Bank, Washington, DC.
Engel, E., A. Galetovic and C. Raddatz (1999), “Taxes and Income Distribution in Chile:
Some Unpleasant Redistributive Arithmetic”, Journal of Development Economics,
Vol. 59, No. 1, pp. 155-192.
EIU – Economist Intelligence Unit (2015), “Universal Healthcare Coverage in Indonesia
– One Year On”, London.
EUROMOD Statistics on Distribution and Decomposition of Disposable Income,
accessed at www.iser.essex.ac.uk/euromod/statistics/ using EUROMOD version
No. G2.0.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 323

Ferreira, F.H.G., N. Lustig and D. Teles (eds.) (2015), “Appraising Cross-National
Income Inequality Databases”, Journal of Economic Inequality, Special Issue,
forthcoming.
Ferreira, F.H.G., J. Messina, J. Rigolini, L.F. Lopez-Calva, M.A. Lugo and R. Vakis,
(2013), Economic Mobility and the Rise of Latin American Middle Class, World
Bank: Washington, DC.
Finn, A. and Leibbrandt, M. (2013), “Mobility and Inequality in the First Three Waves of
NIDS”, SALDRU Working Paper, N. 120, NIDS Discussion Paper, No. 2013/2 for
2012, SALDRU, Cape Town, University of Cape Town.
Förster, M. and I. György Tóth (2015), “Cross-Country Evidence of the Multiple Causes
of Inequality Changes in the OECD Area”, Chapter 19 in Handbook of Income
Distribution, Vol. 2B.
Gandullia, L., N. Iacobone and A. Thomas (2012), “Modelling the Tax Burden on Labour
Income in Brazil, China, India, Indonesia and South Africa”, OECD Taxation
Working
Papers,
No.
14,
OECD Publishing,
Paris,
www.oecdilibrary.org/taxation/modelling-the-tax-burden-on-labour-income-in-brazil-chinaindiaindonesia-and-south-africa_5k8x9b1sw437-en.
Higgins, S. and C. Pereira (2014), “The Effects of Brazil’s Taxation and Social Spending
on the Distribution of Household Income”, in N. Lustig, C. Pessino and J. Scott (eds.),
The Redistributive Impact of Taxes and Social Spending in Latin America, Special
Issue, Public Finance Review, May, Vol. 42, No. 3 and CEQ Master Workbook:
Brazil, November 4.
Higgins, S., N. Lustig, W. Ruble and T. Smeeding (2015), “Comparing the Incidence of
Taxes and Social Spending in Brazil and the United States”, Review of Income and
Wealth, forthcoming.
ILO – International Labour Office (2015), “Global Wage Report 2014/15 – Wage and
Income Inequality”, Geneva.
Immervoll, H. and C. O’Donoghue (2001), “Imputation of Gross Amounts from Net
Incomes in Household Surveys: An Application using EUROMOD”, EUROMOD
Working
Paper
EM1/01,
University
of
Essex,
https://www.iser.essex.ac.uk/research/publications/working-papers/euromod/em101.pdf.
Immervoll, H. and L. Richardson (2011), “Redistribution Policy and Inequality Reduction
in OECD Countries: What Has Changed in Two Decades?”, IZA Discussion Paper
No. 6030, Bonn, October.
Inchauste, G., N. Lustig, M. Maboshe, C. Purfield and I. Wollard (2015), “The
Distributional Impact of Fiscal Policy in South Africa”, Policy Research Working
Paper, No. 7194, World Bank, February.
Jaramillo, M. (2014), “The Incidence of Social Spending and Taxes in Peru”, in N.
Lustig, C. Pessino and J. Scott (eds.), The Redistributive Impact of Taxes and Social
Spending in Latin America, Special Issue, Public Finance Review, May, Vol. 42, No.
3 and CEQ Master Workbook: Peru, June 21, 2013.
Journal of Economic Inequality (2015, forthcoming), “Special Issue: A Review of
International Inequality Indicator Datasets”.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 324 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
Kakwani, N.C. (1977), “Measurement of Tax Progressivity: An International
Comparison”, Economic Journal, pp. 71-80.
Kapsos, S., A. Silberman and E. Bourmpoula (2014), “Why Is Female Labour Force
Participation Declining So Sharply in India?”, ILO Research Paper, No. 10,
International Labour Office.
Lambert, P. (2001), The Distribution and Redistribution of Income, Third Edition,
Manchester University Press.
Leibbrandt, M., I. Woolard, A. Finn and J. Argent (2010), “Trends in South African
Income Distribution and Poverty since the Fall of Apartheid”, OECD Social,
Employment and Migration Working Papers, No. 101, OECD Publishing, Paris,
http://dx.doi.org/10.1787/5kmms0t7p1ms-en.
Lopez-Calva, L. and N. Lustig (eds.) (2010), Declining Inequality in Latin America: A
Decade of Progress, Brookings Institution Press, Washington, DC.
López-Calva, L.F. and E. Ortiz-Juárez (2014), “A Vulnerability Approach to the
Definition of the Middle Class”, Journal of Economic Inequality, Vol. 12, No. 1,
pp. 23-47.
Lustig, N. (2015), “The Redistributive Impact of Government Spending on Education and
Health: Evidence from 13 Developing Countries in the Commitment to Equity
Project”, Chapter 17 in B. Clements, R. de Mooij, S. Gupta and M. Keen (eds.),
Inequality and the Role of Fiscal Policy: Trends and Policy Options, International
Monetary Fund, Washington, forthcoming.
Lustig, N. (2015b), “Inequality and Fiscal Redistribution in Middle Income Countries:
Brazil, Chile, Colombia, Indonesia, Mexico, Peru and South Africa. Evidence from
Commitment to Equity Project (CEQ)”, CEQ Working Paper, No. 31, Center for
Inter-American Policy and Research and Department of Economics, Tulane University
and Inter-American Dialogue.
Lustig, N. and S. Higgins (2013), “Commitment to Equity Assessment (CEQ): Estimating
the Incidence of Social Spending, Subsidies and Taxes. Handbook”, CEQ Working
Paper, No. 1, July 2011, revised January 2013, New Orleans, United States.
Lustig, N., A. Enami and R. Aranda (2015), “The Analytics of Fiscal Redistribution”,
Chapter in N. Lustig and S. Higgins (eds.), Commitment to Equity Handbook:
Estimating the Redistributive Impact and Pro-poorness of Fiscal Policy, forthcoming.
Lustig, N., L. Lopez-Calva and E. Ortiz-Juarez (2012), “Declining Inequality in Latin
America in the 2000s: The Cases of Argentina, Brazil, and Mexico”, CGD Working
Paper, No. 307, Center for Global Development, Washington, DC.
Lustig, N., C. Pessino and J. Scott (eds.) (2014), “The Redistributive Impact of Taxes and
Social Spending in Latin America”, Special Issue, Public Finance Review, May,
Vol. 42, No. 3.
Melendez, M. (2014), CEQ Master Workbook: Colombia, November 21, Tulane
University.
Miranti, R., Y. Vidyattama, E. Hansnata, R. Cassells and A. Duncan (2013), “Trends in
Poverty and Inequality in Decentralising Indonesia”, OECD Social, Employment and
Migration
Working
Papers,
No.
148,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/5k43bvt2dwjk-en.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 325

Morelli, S., T. Smeeding and J. Thompson (2015), “Post-1970 Trends in Within-Country
Inequality and Poverty: Rich and Middle-Income Countries”, Chapter 8 in Handbook
of Income Distribution, Vol. 2B.
OECD (2015a), OECD Employment Outlook 2015, OECD Publishing, Paris,
forthcoming.
OECD (2015b), All on Board: Making Inclusive Growth Happen in China, OECD
Publishing, Paris, forthcoming.
OECD (2015c), OECD Economic Surveys: Colombia 2015, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_surveys-col-2015-en.
OECD (2015d), OECD Economic Surveys: Mexico 2015, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_surveys-mex-2015-en.
OECD (2015e), OECD Economic Surveys: Indonesia 2015, OECD Publishing, Paris.
OECD (2015f), Latin American Economic Outlook 2015, OECD Publishing, Paris.
OECD (2014a), “Social Expenditure Update – Social spending is falling in some
countries, but in many others it remains at historically high levels”, OECD Publishing,
Paris, www.oecd.org/social/expenditure.htm.
OECD (2014b), OECD Economic Surveys: India 2014, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_surveys-ind-2014-en.
OECD (2014c), Society at a Glance: Asia/Pacific 2014, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264220553-en.
OECD (2014d), Society at a Glance: OECD Social Indicators. The Crisis and its
Aftermath, OECD Publishing, Paris, http://dx.doi.org/10.1787/soc_glance-2014-en.
OECD (2014e), “PISA 2012 Results in Focus: What 15 years old know and what they
can do with what they know”, OECD, Paris, www.oecd.org/pisa/keyfindings/pisa2012-results-overview.pdf.
OECD (2013), OECD Economic Surveys: Brazil 2013, OECD Publishing, Paris,
http://dx.doi.org/10.1787/eco_surveys-bra-2013-en.
OECD (2011a), Divided We Stand: Why Inequality Keeps Rising, OECD Publishing,
Paris, http://dx.doi.org/10.1787/9789264119536-en.
OECD (2011b), OECD Employment Outlook 2011, OECD Publishing, Paris,
http://dx.doi.org/10.1787/empl_outlook-2011-en.
OECD (2010a), OECD Employment Outlook 2010: Moving Beyond the Jobs Crisis,
OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2010-en.
OECD (2010b), Tackling Inequalities in Brazil, China, India and South Africa – The Role
of Labour Market and Social Policies, OECD Publishing, Paris,
http://dx.doi.org/10.1787/9789264088368-en.
OECD (2008), Growing Unequal? Income Distribution and Poverty in OECD Countries,
OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264044197-en.
OECD/CAF/ECLAC (2014), Latin American Economic Outlook 2015: Education, Skills
and
Innovation
for
Development,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/leo-2015-en.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 326 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES
OECD/ECLAC/CIAT/IDB (2015), Revenue Statistics in Latin America and the
Caribbean, OECD Publishing, Paris, http://dx.doi.org/10.1787/rev_lat-2015-en-fr.
OECD-ILO (2011a), “Giving Youth a Better Start”, Policy note for the G20 Meeting of
Labour and Employment Ministers, Paris, 26-27 September.
OECD-ILO (2011b), “The National Policy on Skill Development”, G20 Country Policy
Briefs: India, prepared for the G20 Meeting of Labour and Employment Ministers,
Paris, 26-27 September.
OECD/Korea Institute of Public Finance (2014), “The Distributional Effects of
Consumption Taxes in OECD Countries”, OECD Tax Policy Studies, No. 22, OECD
Publishing, Paris, http://dx.doi.org/10.1787/9789264224520-en.
Ostry, J., A. Berg and C. Tsangarides (2014), “Redistribution, Inequality, and Growth”,
IMF Staff Discussion Note, SDN/14/02.
Pagés-Serra, C., P. Gaelle and S. Scarpetta (2009), “Job Creation in Latin American and
the Caribbean. Recent Trends and the Policy Challenges”, Inter-American
Development Bank, Washington, DC.
Quintini, G. and T. Manfredi (2009), “Going Separate Ways? School-to-Work
Transitions in the United States and Europe”, OECD Social, Employment and
Migration
Working
Papers,
No. 90,
OECD
Publishing,
Paris,
http://dx.doi.org/10.1787/221717700447.
Ruggles, P. and M. O’Higgins, (1981a), “The Distribution of Public Expenditure among
Households in the United States”, Review of lncome and Wealth, Vol. 27.
Ruggles, P. and M. O’Higgins (1981b), “The Distribution of Public Expenditures and
Taxes among Households in the United Kingdom”, Review of lncome and Wealth,
Vol. 27.
Ruiz-Tagle, J. and D. Contreras (2014), CEQ Master Workbook: Chile, Commitment to
Equity (CEQ), New Orleans, August.
Scott, J. (2014), “Redistributive Impact and Efficiency of Mexico’s Fiscal System”, in N.
Lustig, C. Pessino and J. Scott (eds.), The Redistributive Impact of Taxes and Social
Spending in Latin America, Special Issue, Public Finance Review, May, Vol. 42,
No. 3.
Shorrocks, A.F. (2013), “Decomposition Procedures for Distributional Analysis: A
Unified Framework Based on the Shapley Value”, Journal of Economic Inequality,
pp. 1-28.
United Nations (2011), Canberra Group Handbook on Household Income Statistics,
Second edition, United Nations, Geneva.
Verbist, G., M.F. Förster and M. Vaalavuo (2012), “The Impact of Publicly Provided
Services on the Distribution of Resources: Review of New Results and Methods”,
OECD Social, Employment and Migration Working Papers, No. 130, OECD
Publishing, Paris, tp://dx.doi.org/10.1787/5k9h363c5szq-en.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 327

Database references
OECD Income
database.htm.

Distribution

Database,

www.oecd.org/social/income-distribution-

OECD Social Expenditure Database, www.oecd.org/social/expenditure.htm.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 328 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Annex 7.A1
The CEQ project and underlying household survey data

Directed by Nora Lustig since 2008, the CEQ was designed to analyse the impact of
taxation and social spending on inequality and poverty in individual countries, and
provide a roadmap for governments, multilateral institutions, and nongovernmental
organisations in their efforts to build more equitable societies.
The main purpose of CEQ is to inform governments of how their fiscal policy affects
their equity goals, recommend practical measures, and enhance accountability and
transparency through better data collection and evaluation systems. In order to achieve
this, CEQ assessments use incidence analysis and a specially-designed diagnostic
questionnaire to address three questions:


How much inequality and poverty reduction is being accomplished through
social spending, subsides and taxes?



How equalising and poverty reducing are specific taxes and transfers?



Within the limits of fiscal prudence, what could be done to increase redistribution
and poverty reduction through changes in taxation and spending?

The CEQ assessment provides the first comprehensive analysis of how taxes and
social spending (including indirect subsidies and taxes, and education and health
expenditures) affect income inequality and poverty in a country. The assessments are
comparable across countries and over time. The country-level studies examine the
distributional effects of individual programmes and policy measures – and the net effect
of each country's mix of policies and programmes. The results will give policy makers,
multilateral institutions, and nongovernmental groups the data and analysis necessary to
determine what changes in tax and spending policies will lead to greater equality and
poverty reduction.
The CEQ diagnostic questionnaire examines whether governments collect and
allocate enough resources to support a minimum living standard for all; whether they
collect and redistribute fiscal resources progressively; whether their expenditures are
fiscally sustainable and their programmes of reasonable quality; and whether they are
responsibly transparent, i.e., they collect and make public sufficient information, and are
subject to independent evaluations.
The CEQ provides new opportunities for civil society organisations to monitor the
distributional effects of government taxes and spending. It will also provide a critical
source of information and analysis for international donors regarding external resources
needed to meet specific goals in low-income countries. And the CEQ results can also be
used to monitor the situation of minority and other excluded groups – including for
example, Afro-descendants and indigenous groups – and determine the extent to which
government taxes and spending affect their incomes and well-being.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 329

The CEQ project is an initiative of the Center for Inter-American Policy and
Research (CIPR) and the Department of Economics, Tulane University, the Center for
Global Development and the Inter-American Dialogue. The project’s main output is the
CEQ assessment, a methodological framework designed to analyse the impact of taxation
and social spending on inequality and poverty in individual countries.
Since its inception, the CEQ has received financial support from the Bill & Melinda
Gates Foundation, the Inter-American Development Bank (IADB), the World Bank, the
United Nations Development Programme's Regional Bureau for Latin America and the
Caribbean (UNDP/RBLAC), the African Development Bank, the International Fund for
Agricultural Development (IFAD), the Development Bank of Latin America (CAF), the
Canadian International Development Agency (CIDA), the Norwegian Ministry of Foreign
Affairs, and the General Electric Foundation.
The surveys used in the fiscal incidence analysis are the following:
Country
Brazil
Chile
Colombia
Indonesia
Mexico
Peru
South Africa

Survey
Pesquisa de Orçamentos Familiares
Encuesta de Caracterización Social (CASEN)
Encuesta de Calidad de Vida
Survei Sosial-Ekonomi Nasional
Encuesta Nacional de Ingreso y Gasto de los Hogares
Encuesta Nacional de Hogares
Income and Expenditure Survey and National Income Dynamics Study

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

Year
Data
2009
Income
2009
Income
2010
Income
2012
Consumption
2010
Income
2009
Income
2010-2011
Income

 330 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Annex 7.A2
Lambert’s conundrum
(Excerpts from Lustig et al., 2015, forthcoming)
Suppose one observes that fiscal policy has an equalising effect. Can one measure the
influence of specific taxes (direct vs. indirect, for example) or transfers (direct transfers
vs. indirect subsidies or in-kind transfers, for example) on the observed result?1 A
fundamental question in the policy discussion is whether a particular fiscal intervention
(or a particular combination of them) is equalising or unequalising. In a world with a
single fiscal intervention (and no re-ranking), it is sufficient to know whether a particular
intervention is progressive or regressive to give an unambiguous response, using typical
indicators of progressivity such as the Kakwani index.2 In a world with more than one
fiscal intervention (even in the absence of re-ranking), this one-to-one relationship
between the progressivity of a particular intervention and its effect on inequality breaks
down. As Lambert (2001) so eloquently demonstrates it, depending on certain
characteristics of the fiscal system, a regressive tax can exert an equalising force over and
above that which would prevail in the absence of that regressive tax.3
An example borrowed from Lambert (2001, Table 11.1, p. 278) helps illustrate this
point in the case of a regressive tax (Table 7.A2.1). The table below shows that “…taxes
may be regressive in their original income… and yet the net system may exhibit more
progressivity” than the progressive benefits alone. The redistributive effect for taxes only
in this example is equal to -0.0517, highlighting their regressivity.4 Yet, the redistributive
effect for the net fiscal system is 0.25, which is higher than the redistributive effect for
benefits only equal to 0.1972. If taxes are regressive vis-à-vis the original income but
progressive with respect to the less unequally distributed post-transfers income,
regressive taxes may exert an equalising effect over and above the effect of progressive
transfers.5
Table 7.A2.1. Lambert’s conundrum
1

2

3

4

Total

Original income x

10

20

30

40

100

Tax liability t(x)
Benefit level b(x)

6
21

9
14

12
7

15
0

42
42

Post-benefit income
Final income

31
25

34
25

37
25

40
25

142
100

Source: Lambert, P. (2001), The Distribution and Redistribution of Income, Third Edition, Manchester
University Press, Table 11.1, p. 278.
12 http://dx.doi.org/10.1787/888933209145

Note that Lambert’s conundrum is not equivalent to the well-known (and frequently
repeated) result that efficient regressive taxes can be fine as long as, when combined with
transfers, the net fiscal system is equalising.6 The surprising aspect of Lambert’s
conundrum is that a net fiscal system with a regressive tax (vis-à-vis market income) is
more equalising than without it.7
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES – 331

The implications of Lambert’s “conundrum”8 in real fiscal systems are quite
profound. It means that in order to determine whether a particular intervention (or, a
particular policy change) increases inequality or reduces inequality – and by how much –
one must resort to numerical calculations that include the whole system. As Lambert
mentions, his example is “not altogether farfetched”.9 Two renowned studies in the 1980s
found this type of result for the United States and the United Kingdom.10 It also made its
appearance in a 1990s study for Chile.11 In the present analysis, Lambert’s conundrum is
found for indirect taxes in the cases of Chile and to a smaller extent in South Africa. This
counter-intuitive result is the consequence of path dependency: a particular tax can be
regressive vis-à-vis market income but progressive vis-à-vis the income that would
prevail if all the other fiscal interventions were already in place.

IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015

 332 – 7. INEQUALITY AND FISCAL REDISRIBUTION IN EMERGING ECONOMIES

Notes
1.

Note that the influence of specific interventions may not be equalising, even if the
overall effect of the net fiscal system is.

2.

The Kakwani index for taxes is defined as the difference between the concentration
coefficient of the tax and the Gini for market income. For transfers, it is defined as the
difference between the Gini for market income and the concentration coefficient of
the transfer. See, for example, Kakwani (1977).

3.

See Lambert (2001), pp. 277 and 278. Also, for a derivation of all the mathematical
conditions that can be used to determine when adding a regressive tax is equalising or
when adding a progressive transfer is unequalising, see Lustig et al. (2015,
forthcoming).

4.

Since there is no re-ranking, the R-S equals the difference between the Ginis before
and after the fiscal intervention.

5.

Note that Lambert uses the term progressive and regressive differently than other
authors in the literature on theoretical and empirical incidence analysis. Thus, he calls
“regressive” transfers that are equalising. See the definitions in earlier chapters of his
book.

6.

As Higgins and Lustig (2015) mention, “efficient taxes that fall disproportionately on
the poor, such as a noexemption value added tax, are often justified with the argument
that spending instruments are available that are better targeted to the pursuit of equity
concerns” (Keen and Lockwood, 2010, p. 141). Similarly, Engel et al. (1999, p. 186)
assert that “it is quite obvious that the disadvantages of a proportional tax are
moderated by adequate targeting” of transfers, since “what the poor individual pays in
taxes is returned to her”. Ebrill et al. (2001, p. 105) argue that “a regressive tax might
conceivably be the best way to finance pro-poor expenditures, with the net effect
being to relieve poverty”.

7.

It can also be shown that if there is re-ranking, a pervasive feature of net tax systems
in the real world, making a tax (or a transfer) more progressive can increase post-tax
and transfers inequality. In Lambert’s example, regressive taxes not only enhance the
equalising effect of transfers, but making taxes more progressive (i.e.., more
disproportional in the Kakwani sense) would result in higher (!) inequality; any
additional change (towards more progressivity) in taxes or transfers would just cause
re-ranking and an increase in inequality.

8.

This is Lambert’s choice of words (p. 278).

9.

Quotes are from Lambert (2001), p. 278.

10.

Ruggles and O’Higgins (1981a) for the United States and Ruggles and O’Higgins
(1981b) for the United Kingdom.

11.

Engel et al. (1999). Although the authors did not acknowledge this characteristic of
the Chilean system in their article, in a recent interaction with the lead author, it was
concluded that the Chilean system featured regressive albeit equalising indirect taxes.
IN IT TOGETHER: WHY LESS INEQUALITY BENEFITS ALL © OECD 2015