DELSA/ELSA/WD/SEM(2005)1 

 

OECD SOCIAL,  EMPLOYMENT  AND MIGRATION WORKING PAPERS 

22

 

 
 
 
 
 
 
 
 
 

Income Distribution and Poverty in OECD Countries 

in the Second Half of the 1990s 

 

Michael Förster and Marco Mira d'Ercole 

 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

Unclassified DELSA/ELSA/WD/SEM(2005)1

 

 

Organisation de Coopération et de Développement Economiques 

   

Organisation for Economic Co-operation and Development 

 10-Mar-2005 

___________________________________________________________________________________________

English - Or. English 

DIRECTORATE FOR EMPLOYMENT, LABOUR AND SOCIAL AFFAIRS 

EMPLOYMENT, LABOUR AND SOCIAL AFFAIRS COMMITTEE 

 

 
 
  
 

 

OECD SOCIAL, EMPLOYMENT AND MIGRATION WORKING PAPERS No. 22 
 
INCOME DISTRIBUTION AND POVERTY IN OECD COUNTRIES IN THE SECOND HALF OF THE 
1990S 
 

Michael Förster and Marco Mira d'Ercole  
 

 

JEL Classification: D3, I3 

 

 
 

 

JT00180148 
 

Document complet disponible sur OLIS dans son format d'origine 
Complete document available on OLIS in its original format 

 

DELSA/EL

SA/WD/SE

M(2005)1 

Un
cl

assi

fi

ed
 

Eng

lis

h - O

r. Eng

lish 

Cancels & replaces the same document of 18 February 2005 

 

 

DELSA/ELSA/WD/SEM(2005)1 

 

2

DIRECTORATE

 

FOR

 

EMPLOYMENT,

 

LABOUR

 

AND

 

SOCIAL

 

AFFAIRS

 

OECD SOCIAL, EMPLOYMENT AND MIGRATION 

WORKING PAPERS 

 
 
This series is designed to make available to a wider readership selected labour market, social policy and 
migration studies prepared for use within the OECD. Authorship is usually collective, but principal writers 
are named. The papers are generally available only in their original language – English or French – with a 
summary in the other. 
 
Comment on the series is welcome, and should be sent to the Directorate for Employment, Labour and 
Social Affairs, 2, rue André-Pascal, 75775 PARIS CEDEX 16, France. 
 

The opinions expressed and arguments employed here are the responsibility 

of the author(s) and do not necessarily reflect those of the OECD 

 

 
 

Applications for permission to reproduce or translate  

all or part of this material should be made to: 

 

Head of Publications Service 

OECD 

2, rue André-Pascal 

75775 Paris, CEDEX 16 

France 

 

Copyright OECD 2005

 

 DELSA/ELSA/WD/SEM(2005)1 

 

3

EXECUTIVE SUMMARY 

This report provides evidence on income distribution and poverty in 27 OECD countries over the 

second half of the 1990s, using data that correct for many of the features that limit cross-country and inter-
temporal comparisons in this field. Patterns for income distribution and relative poverty in the second half 
of the 1990s — a period of significant improvement in labour market conditions in most OECD countries 
— conform to many of the longer-term trends identified in previous OECD analysis, but also highlight 
some significant departures. 

•

 

Inequality in the distribution of household disposable income among the total population 
increased slightly over the second half of the 1990s, continuing the trend of the previous decade.  

•

 

Relative poverty, measured with respect to a threshold set at half of median income, affected in 
2000 around 11% of the OECD population, with an increase since the mid-1990s that is similar 
to that of the previous decade. Absolute income poverty, which had declined by more than one-
third in the decade from the mid-1980s to the mid-1990s, fell by close to one-fourth in the five 
years to 2000.  

•

 

Following steady increases in inequality in the distribution of market income among the 
population of working age in previous decades, several OECD countries reversed or halted this 
trend in the second half of the 1990s. Impacts on inequality in the distribution of disposable 
income and relative poverty have so far been muted because of a reduction in the effectiveness of 
public transfers and taxes in reaching those with greater needs.  

•

 

Relative poverty is, in most countries, most common among children than among the entire 
population, and this increased further in the second half of the 1990s. While child poverty rates 
are lower in countries with higher level of maternal employment, there is much diversity in 
country experiences, suggesting that specific factors increase risks of destitution for children in 
some OECD countries.  

•

 

Income of the elderly, relative to that of the rest of the population, stopped improving in the 
second half of the 1990s. Their poverty rates, measured using a relative threshold, also increased 
in several OECD countries, mainly reflecting changes in public transfers and taxes.  

DELSA/ELSA/WD/SEM(2005)1 

 

4

RESUME 

Ce rapport examine la distribution des revenus et la pauvreté dans 27 pays de l’OCDE pour la 

deuxième moitié de la décennie 90, sur la base de données corrigées d’une grande partie des paramètres 
qui handicapent les comparaisons transnationales et intertemporelles dans ce domaine. L’évolution de la 
distribution des revenus et de la pauvreté au cours de la deuxième moitié de la décennie 90 – période 
d’amélioration notable de la situation du marché du travail dans la plupart des pays de l’OCDE – s’inscrit 
pour une grande part dans le prolongement des tendances à long terme qui se dégageaient des analyses 
précédentes, mais présente aussi quelques écarts notables par rapport à celles-ci. 

•

 

L’inégalité de la distribution du revenu disponible des ménages sur l’ensemble de la population 
s’est légèrement accentuée dans la seconde moitié de la décennie 90, prolongeant la tendance 
observée au cours de la décennie précédente. 

•

 

La pauvreté relative, mesurée par rapport à un seuil fixé à la moitié du revenu médian, touchait 
en 2000 environ 11 % de la population de l’OCDE, soit une élévation depuis le milieu des 
années 90 analogue à celle observée dans la décennie précédente. Quant à la pauvreté absolue, 
après un recul de plus d’un tiers dans la décennie précédente, elle a baissé de près d’un quart dans 
les cinq années considérées. 

•

 

Après une période d’accroissement constant au cours des décennies précédentes, l’inégalité des 
revenus marchands dans la population d’âge actif s’est réduite ou stabilisée dans plusieurs pays 
dans la deuxième moitié des années 90. Les incidences de ces changements de tendance sur 
l’inégalité dans la distribution du revenu disponible et sur la pauvreté relative ont été atténuées 
par la baisse d’efficacité des systèmes et d’imposition de transferts pour les catégories les plus 
défavorisées. 

•

 

Dans la plupart des pays, la pauvreté relative est plus répandue chez les enfants que dans 
l’ensemble de la population, et ce phénomène s’est encore accentué dans la deuxième moitié des 
années 90. Si les taux de pauvreté des enfants sont plus faibles dans les pays où le taux d’emploi 
des mères est plus élevé, la situation est très diverse selon les pays, ce qui donne à penser que les 
risques de pauvreté des enfants sont accrus dans certains pays de l’OCDE par des facteurs 
spécifiques. 

•

 

Les revenus des personnes âgées, par rapport au reste de la population, ont cessé de s’améliorer 
au cours de la période. Leurs taux de pauvreté, mesurés par rapport à un seuil relatif, ont par 
ailleurs augmenté dans plusieurs pays de l’OCDE, principalement en raison des modifications 
apportées aux systèmes et d’imposition de transferts. 

 DELSA/ELSA/WD/SEM(2005)1 

 

5

ACKNOWLEGMENTS 

This report has been made possible by the collaboration of experts from member countries who have 

provided data from national datasets based on common assumptions and definitions.  

Data were provided by Peter Saunders and Peter Siminski, Social Policy Research Center, for 

Australia; Gudrun Biffl, Austrian Institute for Economic Research, for Austria; Brian Murphy and Shawna 
Brown, Statistics Canada, for Canada; Aleš Ka

ň

ka and Jaromir Kalmus, Czech Statistical Office, for the 

Czech Republic; Peter Bach-Mortensen and Lars Pantmann, Ministry of Finance, for Denmark; Heikki 
Viitamäki, Government Institute for Economic Research, for Finland; Saïda Khalfaoui, Pascal Chevalier 
and Nadine Legendre, Institut National de la Statistique et des Études Économiques, for France; Markus 
Grabka, German Institute for Economic Research, for Germany; Theodore Mitrakos, Bank of Greece, for 
Greece; Márton Medgyesi and Péter Szivós, TÁRKI Social Research Institute, for Hungary; Brian Nolan 
and Bertrand Maître, Economic and Social Research Institute, for Ireland; Gaetano Proto and Marco Di 
Marco, Istituto Nazionale di Statistica, for Italy; Yoshihiro Kaneko and Katsuhisa Kojima, National 
Institute of Population and Social Security Research, and Atshuiro Yamada, Keio University, for Japan; 
Frédéric Berger, Centre d' Etudes de Populations, de Pauvreté et de Politiques Socio-Economiques / 
International Networks for Studies in Technology, Environment, Alternatives, Development, for 
Luxembourg; Ana Laura Pineda Manriquez and Rodrigo Tadeo Negrete Prieto, Instituto Nacional de 
Estadística Geografia e Informatica, for Mexico; Hans de Kleijn, Central Bureau of Statistics, for the 
Netherlands; Caroline Brooking, Statistics New Zealand, for New Zealand; Jon Epland, Statistics Norway, 
for Norway; Malgosia Kalbarczyk, Centre for Social and Economic Research, for Poland; Daniel Mota, 
Instituto Nacional de Estatística, for Portugal; Olga Canto Sanchez, Universidade de Vigo, for Spain; 
Thomas Pettersson, Ministry of Finance, for Sweden; Anne Cornali-Schweingruber and Ueli Oetliker, 
Office fédéral de la statistique, for Switzerland; Murat Karakas and Ozlem Sarica, State Institute of 
Statistics, for Turkey; Asghar Zaidi, London School of Economics and Social Disadvantage Research 
Centre, Oxford, for the United Kingdom; John Coder, Sentier Research, Annapolis, for the United States.  

Japanese data are based on tabulations prepared in the context of the "Research on Policy Planning 

and Evaluation Project" (H14-Seisaku-012), financed out of the Health and Labour Sciences Research 
Grants, courtesy of Yoshihiro Kaneko. 

Michael Förster and Marco Mira d'Ercole are, respectively, Administrator and Principal Administrator 

at the OECD Social Policy Division. Useful comments on previous drafts of this paper were provided by 
Martine Durand, John P. Martin, Mark Pearson and Peter Whiteford, from the OECD Directorates for 
Employment, Labour and Social Affairs, as well as by delegates at the OECD Working Party on Social 
Policy. All remaining errors are the sole responsibility of the authors. The opinions expressed in this paper 
do not necessarily reflect those of the OECD or of its Member countries.  

DELSA/ELSA/WD/SEM(2005)1 

 

6

TABLE OF CONTENTS 

EXECUTIVE SUMMARY .............................................................................................................................3

 

RESUME.........................................................................................................................................................4

 

INCOME DISTRIBUTION AND POVERTY IN OECD COUNTRIES IN THE SECOND HALF OF THE 
1990S...............................................................................................................................................................8

 

1.

 

Introduction.......................................................................................................................................8

 

2.

 

Trends in inequality and poverty for the entire population in the second half of the 1990s .............9

 

2.1.

 

Levels of income inequality ........................................................................................................9

 

2.2.

 

Trends in income distribution....................................................................................................12

 

2.3.

 

Levels and trends in income poverty.........................................................................................18

 

3.

 

Labour markets, taxes and benefits: distributive effects on the population of working-age...........23

 

3.1.

 

Recent trends in income distribution and relative poverty for the working-age population .....24

 

3.2.

 

The influence of labour markets................................................................................................25

 

3.3.

 

The role of taxes and public transfers........................................................................................28

 

3.4.

 

Accounting for changes in poverty rates since the mid-1990s ..................................................30

 

4.

 

Poverty and inequality among children and households with children...........................................32

 

4.1.

 

Levels and trends in relative income and poverty .....................................................................33

 

4.2.

 

The influence of household structure, mothers employment and benefit systems ....................33

 

5.

 

Income adequacy in old age: effects of pension reforms on the retirement-age population...........37

 

5.1.

 

Levels and trends in relative income and poverty among the elderly .......................................37

 

5.2.

 

Public pension systems and their impacts on the elderly population ........................................43

 

5.3.

 

Distributive patterns of public transfers and private capital income .........................................45

 

BIBLIOGRAPHY .........................................................................................................................................50

 

ANNEX 1. CHARACTERISTICS OF THE DATA USED IN THE ANALYSIS.......................................52

 

ANNEX 2. POVERTY THRESHOLDS USED IN THE ANALYSIS.........................................................60

 

ANNEX 3. SUPPORTING TABLES ...........................................................................................................61

 

 
 
Tables 

Table 1. 

Overall trends in income inequality: summary results for the entire population...................14 

Table 2. 

Gains and losses of income share by income quintile: entire population, mid-1990s to early 
2000 .......................................................................................................................................15 

Table 3. 

Trends in real household income at different points of the income ladder............................18 

Table 4. 

Levels and trends in the Gini coefficient of market income inequality among the working-
age population .......................................................................................................................26 

Table 5. 

Poverty rates in households with children under different household structure, mid-1990s 
and 2000 ................................................................................................................................36 

 

 

 DELSA/ELSA/WD/SEM(2005)1 

 

7

Figures 

Figure 1.

 

Gini coefficients of income concentration in 27 OECD countries, most recent year............10

 

Figure 2.

 

Actual and perceived inequalities in the distribution of income ...........................................11

 

Figure 3.

 

Relation between perceived inequalities and views about government's role in reducing 

 them .......................................................................................................................................12

 

Figure 4.

 

Trends in absolute poverty rates............................................................................................20

 

Figure 5.

 

Relative poverty rates among the entire population ..............................................................21

 

Figure 6.

 

Relative poverty rates at different income thresholds ...........................................................22

 

Figure 7.

 

Income gaps of people living in relative poverty ..................................................................23

 

Figure 8.

 

A composite measure of relative poverty in OECD countries, 2000.....................................23

 

Figure 9.

 

Income inequality among the population of working age .....................................................24

 

Figure 10.

 

Trends in inequality of market and disposable income among the working-age population 25

 

Figure 11.  Relative poverty rates at the level of market income among the working-age population, 
 

non-employment of individuals and joblessness ...................................................................27 

Figure 12.

 

Structure of poverty in households headed by a working-age head, by work attachment of 

 household 

members ...............................................................................................................28

 

Figure 13.

 

Relative poverty among households with a working-age head and social spending.............29

 

Figure 14.

 

Effects of taxes and transfers in reducing relative income poverty.......................................29

 

Figure 15.

 

Changes in relative poverty rates among households with a working-age head by 

 

components, mid-1990s to 2000............................................................................................32

 

Figure 16.

 

Relative poverty rates for children and the entire population................................................33

 

Figure 17.

 

Relative disposable income of households with children, 2000............................................34

 

Figure 18.

 

Relative poverty rates in households with children and single-parent households, 2000 .....35

 

Figure 19.

 

Poverty among children and employment rates among mothers, 2000.................................35

 

Figure 20.

 

Poverty rates before and after taxes and transfers, households with and without children, 

 2000 .......................................................................................................................................37

 

Figure 21.

 

Quasi replacement rates for persons aged 66 to 75 ...............................................................38

 

Figure 22.

 

Family structure among individuals living in households with an older head ......................39

 

Figure 23.

 

Gini coefficient of income inequality among the elderly ......................................................42

 

Figure 24.

 

Relative poverty rates among the elderly ..............................................................................42

 

Figure 25.

 

Structure of poverty among persons living in households with a retirement-age head .........43

 

Figure 26.

 

Relative poverty among the elderly and pension systems .....................................................44

 

Figure 27.

 

Changes in relative poverty rates among households with a retirement age head by 

 

components, mid-1990 to 2000 .............................................................................................45

 

Figure 28.

 

Distributive shape of public transfers and private capital income to the elderly and of 

 

disposable income to the working-age population, 2000 ......................................................47

 

Figure 29.

 

Income composition among the older population by income groups, OECD average 2000.49 

 

Boxes 

Box 1. Survey measures of high income ...................................................................................................16

 

Box 2. Assessing poverty: consumption, assets and income measures .....................................................19

 

Box 3. Housing costs and poverty outcomes .............................................................................................41

 

 

DELSA/ELSA/WD/SEM(2005)1 

 

8

 

INCOME DISTRIBUTION AND POVERTY IN OECD COUNTRIES 

IN THE SECOND HALF OF THE 1990S 

1. Introduction 

1. 

This paper extends to the end of the 1990s the analysis of income distribution and poverty 

provided in Burniaux 

et al

. (1998) and Förster and Pearson (2002).

1

 It illustrates some of the findings from 

a third wave of country submissions based on national household surveys and other micro datasets. These 
submissions are based on a standard questionnaire, which uses common assumptions and definitions to 
increase the degree of cross-country comparability (Annex 1). The data are based on the concept of 
equivalised disposable income of individuals (

i.e.

 the disposable income of households, adjusted for 

household size) broken down by gross income components (

i.e.

 before payment of direct taxes and social 

security contributions levied on individuals) and presented for a variety of socio-demographic 
characteristics of individuals and households. Efforts have been made to maximise the country coverage of 
the data — available for 27 OECD countries — so as to allow the identification of common trends.  

2. 

The use of a common methodology and definitions allows us to overcome many of the 

comparability issues that limit cross-country and inter-temporal comparisons of income distribution and 
poverty among OECD countries. The data used in this paper, however, are not without limits. First, priority 
has been given to maximise country coverage rather than to collect continuous time series for each 
individual country. Second, while common definitions are used, the underlying data differ in some respects 
that escape standardisation: they are based on household surveys in most cases, but rely on a combination 
of survey and administrative data for a few countries (Belgium, Denmark, Sweden); further, even when 
household surveys are used, data for different countries are affected by differences in survey design, 
response rates and imputation methods used to integrate the survey data.

2

 Third, the extent to which 

income and population data from household surveys match independent estimates (

e.g.

 from national 

accounts) may differ across countries and over time, possibly distorting comparisons. Finally, the period 
covered ends around the year 2000 — a cyclical peak in most OECD countries. 

3. 

Beyond these methodological aspects, other limits relate to the significance 

for policy

 of the 

comparisons undertaken in this paper. In this respect, two issues are important. First, the focus in this paper 
is on income distribution among individuals and households 

at a point in time

: in other words, the data 

used do not allow us to distinguish between persistent and temporary conditions,

 

or to track how conditions 

                                                      

1

  

While, for most countries, the data in this paper extend the series used in previous OECD publications, 
revised estimates are presented for Denmark, Germany, Hungary, Japan (based on a different survey), 
Mexico and the United States (based on the internal version of March 

Current Population Survey)

. In 

addition, this report presents for the first time data for the Czech Republic, reunified Germany, 
Luxembourg, New Zealand, Poland, Portugal, Spain and Switzerland. No updates were available, at the 
time of writing, for Belgium and Spain. Updates refer to the year 2000 for most countries but to 1998/99 
for Australia; 1999 for Austria and Greece; 2001 for Germany, Luxembourg, New Zealand and 
Switzerland; and 2002 for the Czech Republic, Mexico and Turkey.  

2

  

Although "weighting" of responses is used to provide a more representative picture of the population in 
each country, this does not eliminate biases that can accompany low response rates. 

 DELSA/ELSA/WD/SEM(2005)1 

 

9

change over the different stages of each individual’s life-course. Second, this paper presents evidence on 
the impact of taxes and public transfers on income inequality and poverty irrespective of the goals of 
policies. In practice, the policies that most affect income distribution and poverty at a point in time pursue 
a variety of objectives – providing social insurance against various contingencies, improving economic 
efficiency, mobilising additional labour resources — rather than redistributing income 

per se. 

In these 

cases, the crucial issue is to design policies that achieve multiple goals simultaneously, and to shift trade-
offs when such goals conflict with each other. 

4. 

The paper is organised as follows. The first section identifies some of the stylised facts that 

characterise the experience of OECD countries over the second half of the 1990s in terms of income 
inequality and poverty among the entire population, comparing these with longer-term trends described in 
previous OECD studies. The second section looks at the experience of the population of working-age, with 
a special focus on the role of labour markets, taxes and welfare systems. The third section looks at the 
experience of children and of households with children. The last section considers the elderly population, 
and how changes in pension systems have affected their well-being and poverty risks. 

2. 

Trends in inequality and poverty for the entire population in the second half of the 1990s 

5. 

This section, which looks at data for the 

entire

 population, highlights both cross-country 

differences in the 

levels

 of the various inequality and poverty indicators, and ways in which 

changes

 over 

the second half of the 1990s differ from the long-term trends identified in previous OECD analysis. One 
problem, for analysis of changes over time, is that inequality and poverty indicators for individual 
countries refer to specific years that may differ in terms of the cyclical position of each country. In theory, 
changes between these years may not be fully representative of underlying trends. In practice, however, a 
comparison with "commonly used" measures of income inequality for several OECD countries suggests 
that this consideration is of limited importance for most countries.

3

 

2.1. 

Levels of income inequality 

6. 

Our starting point for assessing distributive patterns prevailing at the end of the XXI

th

 century in 

OECD countries is represented by levels of inequality in the distribution of equivalised disposable income. 
Figure 1 displays one widely used summary indicator of income inequality – the Gini coefficient of income 
concentration

4

 – in 27 OECD countries in the latest year available. Four groups of countries can be 

distinguished in terms of increasing levels of inequality: 

•

 

The four Nordic countries (Denmark, Sweden, Finland and Norway), together with Austria, the 
Czech Republic, Luxembourg and the Netherlands, all display Gini coefficients at around 26 (at 
least 15% less than the OECD average value).  

                                                      

3

  

Annual time-series of "commonly used" measures of income inequality in nine OECD countries — shown 
in Atkinson (2002) — display relatively minor variations around the trend (with the exception of Italy), 
suggesting that the years used in this paper are quite representative of the values prevailing over the period 
considered. 

4

  

The Gini coefficient is defined as the area between the Lorenz curve (which plots cumulative shares of the 
population, from the poorest to the richer, against the cumulative share of income that they receive) and the 
45

o

 line, taken as a ratio of the whole triangle. The values of the Gini coefficient range between 0 in the 

case of “perfect equality” (each share of the population gets the same share of income) and 100 in the case 
of “perfect inequality” (all income goes to the share of the population with the highest income).  

DELSA/ELSA/WD/SEM(2005)1 

 

10

•

 

Other Continental European countries, together with Hungary, Canada, Spain, Ireland and 
Australia show higher values of the Gini coefficient than those in the first group — between 27 
and 30.5 — but still below the average value for the OECD as a whole. 

•

 

New Zealand, the United Kingdom and the United States, as well as Greece, Portugal, Italy

5

, 

Japan and Poland, record values of the Gini coefficient between 31 and 36 — above the OECD 
average. 

•

 

Mexico and Turkey, with values of their Gini coefficients of around 45, are clear outliers in this 
league table – the difference between their Gini coefficient and that of Poland (the third highest 
country) is close to the difference between Poland’s and that of lower inequality countries. 

A simple OECD average of the Gini coefficient is 30.6 (29.4 when Mexico and Turkey are excluded).  

Figure 1.  Gini coefficients of income concentration in 27 OECD countries, most recent year 

10

15

20

25

30

35

40

45

50

DE

N

S

W

E

N

LD

A

UT

C

ZE

LU

X

FIN

NO

R

SW

I

B

EL

FR

A

G

E

R

HU

N

C

AN

SP

A

IR

L

A

US

JP

N

UK

G

NZ

L

G

R

C

IT

A

P

O

R

US

A

P

O

L

T

UR

M

E

X

OECD average

 

Note:

 The income concept used is that of disposable household income, adjusted for household size (e=0.5). Gini 

coefficients multiplied by 100.  "Most recent year" refers to the year 2000 in all countries except 1999 for Australia, 
Austria and Greece; 2001 for Germany, Luxembourg, New Zealand and Switzerland; and 2002 for the Czech Republic, 
Mexico and Turkey; In the case of Belgium and Spain (countries shaded in the figure), the data refer to 1995. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

7. 

The Gini coefficient is just one measure of income concentration. Analysis of three additional 

income inequality indicators – the squared coefficient of variation (SCV), the mean-log deviation (MLD) 
and the inter-decile ratios

6

 (Annex Table 3) – suggests, however, that the four broad country groups 

                                                      

5

  

In the case of Italy, income from financial assets is excluded from disposable income to ensure 
comparability with results from previous years. Its inclusion raises the Gini coefficient by around 7% (from 
33.4, shown in Figure 1, to 35.7). 

6

  

The P90/P10 inter-decile ratio is the ratio of the lower bound value of the top income decile to that of the 
first. The squared coefficient of variation is the variance of average income of each decile, divided by the 
square of the average income of the entire population. The mean log deviation is the average value of the 
natural logarithm of the ratio of mean income to the income of each decile. All these summary indicators 
have different upper and lower bounds: the squared coefficient of variation has a lower bound of 0 and 
upper bound of infinity, while the mean log deviation and inter-decile ratio have a lower value of 1 and no 
upper bound. Also, each index differs in its sensitivity to changes at various points in the distribution. 
Relative to other indices, the Gini coefficient is less sensitive to changes at the two extremes of the 
distribution, while the Mean Log Deviation is more sensitive to changes at the bottom of the distribution, 
and the Squared Coefficient of Variation is more sensitive to changes at the top.  

 DELSA/ELSA/WD/SEM(2005)1 

 

11

identified above are robust to the choice of the summary indicator.

7

 In general, Nordic countries together 

with the Czech Republic and Luxembourg consistently display lower levels of income inequality. Also 
included in the lower-inequality group, according to the indicator used, are some other Western European 
countries (Austria and the Netherlands in the case of MLD; and Austria, France, Germany and Netherlands 
in the case of SCV). Hungary, other Western European countries and Japan tend to cluster around the 
middle of the ranking for all the indicators used, while Switzerland records below-average values for the 
Gini coefficient, decile ratio and MLD but scores somewhat higher on the SCV index (indicating greater 
concentration at the very top and greater equality across other parts of the income distribution). Higher 
inequality levels are consistently recorded by some Anglo-Saxon countries (especially New Zealand, the 
United Kingdom and the United States) and Southern European countries, while Turkey and Mexico 
record the highest inequality whatever the indicator used.  

8. 

Country differences in these various inequality indicators, however, do not match closely 

measures of individuals' 

perceptions

 of whether income inequality is "too high" in the country where they 

live. Figure 2 plots data on the share of individuals that agree with the statement that income inequality is 
too large, based on surveys undertaken in 1999 under the 

aegis 

of the International Social Science 

Programme, and values of the Gini coefficient of income inequality discussed above. While in all OECD 
countries a majority of respondents agreed with the view that income differences were "too large" at the 
end of the 1990s, there are large differences across countries, from around 60% in the United States to 90% 
or more in Hungary, Italy and Portugal. The share of respondents perceiving income inequality as too large 
is lowest in the United States — where the Gini coefficient is above the OECD average — and highest in 
Portugal, where the Gini coefficient is also high. A weak association between real and perceived 
inequalities also holds when using other inequality indicators. This suggests that, beyond actual inequality, 
other factors play a significant role in shaping these perceptions.

8

 Perceptions about income inequalities, 

however, critically shape attitudes towards government policies aimed to reduce them (Figure 3). 

Figure 2.  Actual and perceived inequalities in the distribution of income 

50

60

70

80

90

100

AU

S

AU

T

CA

N

CZ

E

FR

A

GE

R

HUN

IT

A

JP

N

NZ

L

NO

R

P

O

L

PO

R

SP

A

S

W

E

UK

G

US

A

P

e

rc

e

iv

e

d

 in

e

q

u

a

lit

ie

s

 (

%

)

20

25

30

35

40

G

in

i c

o

e

ff

ic

ie

n

t o

f 

in

c

o

m

e

 in

e

q

u

a

lit

y

 

Note:

 Perceived inequalities in 1998 (shown as bars on the left-hand axis) are measured by the share of respondents who agree or 

strongly agree with the statement "differences in income are too large"; data for Italy refer to 1992. Actual inequalities are measured 
by the Gini coefficient of inequality in 2000 (1995 in the case of Italy) (shown as diamonds on the right-hand axis). 

Source:

 Data from the International Social Science Programme and the OECD questionnaire on income distribution. 

                                                      

7

  

Data collected in previous waves also suggest that the grouping of countries in terms of levels of income 
inequality is relatively unaffected by different choices about scale economies in consumption.  

8

  

Surhcke (2001), who analyses a range of determinants of perceptions towards income inequality, reports 
greater "tolerance" towards inequality among individuals with higher income, those who experienced 
upwards income mobility over the past ten years, those who believe that people get rewarded for effort, 
intelligence and skills, as well as among men, youths, and those living in smaller families. In his results, 
tolerance towards inequality is lower in countries with higher income inequality, and in those that 
underwent a transition from socialist regimes. 

DELSA/ELSA/WD/SEM(2005)1 

 

12

Figure 3.  Relation between perceived inequalities and views about government's role in reducing them 

AUS

AUT

CAN

CZE

FRA

GER

UKG

HUN

JPN

NZL

NOR

POL

POR

SPA

SWE

USA

20

40

60

80

100

60

65

70

75

80

85

90

95

100

perceived inequalities

gov

e

rnm

ent

 role in r

e

duc

ing

 inequa

lities

 

Note:

 Data on government's role in reducing inequalities refer to the share of respondents who agree or strongly agree 

with the statement "it is the responsibility of governments to reduce inequalities". 

Source:

 International Social Science Programme, 1992 and 1999. 

2.2. 

Trends in income distribution 

9. 

Table 1 summarises trends in the distribution of equivalised disposable income over three 

different time periods, based on movements in the value of the Gini coefficient. While, in general, the 
statistical significance of a given change in the Gini coefficient depends on the sample size and the design 
of different surveys, conventional benchmark are used in Table 1: changes in Gini coefficients between 
zero and 2% are characterised as "no change"; changes between 2% and 7% as "small"; changes in excess 
of 7% as “moderate"; and changes in excess of 12% as “strong". Countries are attributed to different 
columns of Table 1 according to the size of changes in the Gini coefficient over different time periods.

9

 

Main patterns, based on this conventional classification, are as follow: 

•

 

No common trend in income inequality is evident over the period from 

the mid-1970s to mid-

1980s

. Across the seven countries for which information is available, the Gini coefficient of 

income inequality increased in three, and narrowed in four. A simple average across the seven 
OECD countries shows a decline of 3.2% in the Gini coefficient of inequality – despite some 
pronounced movements in the various countries.  

•

 

There is stronger evidence of a common trend across OECD countries in the period from 

the mid-

1980s to the mid-1990s

. Over this decade, the Gini coefficient decreased in three of the 

25 countries (only slightly in two of them), remained stable in five, and increased in the 
remaining 17 (by significant amounts in most of them).

10

 A simple average across the countries 

                                                      

9

  

Because of its shorter length, however, a smaller cumulative percentage change of the Gini coefficient in 
the most recent period does not necessarily imply a deceleration in underlying trends. 

10

  

Part of the significant increase in inequality in Norway may be explained by a major tax reform 
implemented in 1992. This reform expanded the tax base and, as a result, some formerly “invisible” capital 
income was identified in the data. 

 DELSA/ELSA/WD/SEM(2005)1 

 

13

for which data are available over this period shows an increase of around 6% in the Gini 
coefficient of income inequality. 

•

 

No common trend in income inequality is apparent over the most recent period, 

from the mid-

1990s to 2000

. The Gini coefficient declined in five of the 24 countries for which information is 

available, was stable in 10, and increased in the other 9 (in most cases by small amounts).

11,12

 A 

simple average across the 20 countries for which data are available since the mid-1980s shows a 
small rise in the Gini coefficient of income inequality in the second half of the 1990s (1%), as 
compared to stronger increases in the previous decade.

13

 As the largest increases in the Gini 

coefficient occurred in countries characterised by low inequality, dispersion in inequality across 
countries narrowed in this period, in contrast to trends over previous decades.

14

  

                                                      

11

  

In the case of Mexico, however, national data based on the same household survey used in this paper point 
to an increase in inequality in the period from 1996 to 2000, followed by declines from 2000 to 2002. 

12

  

For the United Kingdom, the data on income distribution and poverty used in this note (based on the 

Family Expenditure Survey) 

differ from those commonly used in most national discussions (based on the 

Family Resource Survey). 

While the latter survey has a sample size three times larger than the former, 

results are available for a shorter period. Recent trends based on the 

Family Resource Survey, 

as 

summarised by Brewer 

et al. 

(2004), point to a significant increase in the Gini coefficient of income 

inequality from 1996/97 to 2000/01, followed by small reductions in the two following years.

 

13

  

These "average" changes in inequality over the two periods are influenced by opposite movements in 
Mexico and Turkey, which recorded an increase in inequality from the mid-1980s to mid-1990s and strong 
declines between the mid-1990s and 2000. When excluding those two countries, the average Gini 
coefficient increased by 5% over the first period and by 2% over the past five years.  

14

  

The standard deviation of the Gini coefficient declined by around one fifth over the most recent period, 
after having increased by a similar amount in the period from the mid-1980s to the mid-1990s.  

DELSA/ELSA/WD/SEM(2005)1 

 

14

Table 1.  Overall trends in income inequality: summary results for the entire population 

 

Strong 

decline 

Moderate 

decline 

Small 

decline 

No change 

Small increase 

Moderate 

increase 

Strong increase 

Mid-1970s 

to mid-

1980s 

Greece Finland 

Sweden 

Canada 

 

Netherlands 

United States 

United Kingdom 

Mid-1980s 

to mid-

1990s 

 Spain 

Australia 

Denmark 

Austria 

Canada 

France 

Greece 

Ireland 

 

Belgium 

Germany 

Luxembourg 

Japan 

Sweden 

Czech Rep. 

Finland 

Hungary 

Netherlands 

Norway 

Portugal 

United Kingdom 

United States 

Italy 

Mexico 

New Zealand 

Turkey 

Mid-1990s 

to 2000 

 Mexico 

Turkey 

France 
Ireland 
Poland 

 

Australia 

Czech Rep. 

Germany 

Hungary 

Italy 

Luxembourg 

Netherlands 

New Zealand 

Portugal 

United States 

Austria 

Canada 

Denmark  

Greece  

Japan 

Norway 

United Kingdom 

 Finland 

Sweden 

Note:

 "Strong decline/increase" denotes a change in income inequality above +/- 12%; "moderate decline/increase" a 

change between 7 and 12%; "small decline/increase " a change between 2 and 7%; "No change" changes between +/- 
2%. Results are based on the values of the Gini coefficient in four reference years which may vary among countries. 
"2000" data refer to the year 2000 in all countries except 1999 for Australia, Austria and Greece; 2001 for Germany, 
Luxembourg, New Zealand and Switzerland; and 2002 for the Czech Republic, Mexico and Turkey; "Mid-1990s" data 
refer to the year 1995 in all countries except 1993 for Austria; 1994 for Australia, Denmark, France, Germany, Greece, 
Ireland, Japan, Mexico and Turkey; and 1996 for the Czech Republic and New Zealand; "Mid-1980s" data refer to the 
year 1983 for Austria, Belgium, Denmark and Sweden; 1984 for Australia, France, Italy and Mexico; 1985 for Canada, 
Japan, the Netherlands, Spain and the United Kingdom; 1986 data for Finland, Luxembourg, New Zealand and 
Norway; 1987 for Ireland and Turkey; 1988 for Greece; and 1989 for the United States. For the Czech Republic, 
Hungary and Portugal, the period mid-80s to mid-90s refers to early to mid-90s. 

Source:

 Computations from OECD questionnaire on distribution of household incomes. 

10. 

Different measures of inequality can give different indications about changes in inequality. 

Annex Table 3 shows Gini coefficients alongside the three other summary indicators of income inequality 
previously used: the income ratio between the top and bottom decile (P90/P10 ratio), the mean log 
deviation of equivalised disposable income, and the squared coefficient of variation. On average, two of 
the four indicators point to declines in inequality in the second half of the 1990s, following increases in the 
preceding decades; and to moderate increases when excluding Mexico and Turkey. With respect to the 
most recent period, all indices point to increases in income inequality in nine countries: Canada, Czech 
Republic, Denmark, Finland, Greece, Japan, New Zealand (where only two indicators are available), 
Sweden and the United Kingdom; and to declines in seven countries: France, Germany (old and new 
Länder together), Italy, Mexico, Poland, Portugal and Turkey (only two indicators available). For the 
remaining eight countries (and the old Länder of Germany), the various inequality indices move in 
different directions. Over the previous decade, indicators suggested an unambiguous increase in inequality 
in 13 countries (values in bold in Annex Table A.3), an unambiguous decrease in two (values in italics), 
and movements in different directions in 9.  

11. 

Differences in the direction of changes in inequality provided by the different indicators partly 

reflect the different weight that each indicator gives to different portions of the distribution. It is therefore 
important to look closer at changes at different point of the distribution. When, for instance, persons in the 
middle of the distribution lose ground relative to those at the bottom and top, a “hollowing out” of the 
distribution occurs. Conversely, a widening of the distribution could reflect those at the bottom becoming 
poorer, those at the top improving their situation, or a combination of the two.  

 DELSA/ELSA/WD/SEM(2005)1 

 

15

12. 

Previous OECD analyses have noted that, in the period from the mid-1980s to mid-1990s, 

changes in income distribution were dominated by movements at the higher end of the spectrum: in 16 out 
of 22 OECD countries persons in the top quintile increased their share of disposable income, while in eight 
countries those in the bottom quintile lost ground (moderately) relative to the average (Förster and Pearson, 
2002). Developments in the most recent period (Table 2) are less clear-cut: persons in the bottom quintile 
lost ground slightly in seven countries, and gained slightly in two (Mexico and Turkey); those in the top 
quintile increased their share of household disposable income in eleven countries (substantially so in 
Finland and Sweden), while losing considerable ground in four. On the other hand, middle-income groups 
gained significantly at the expense of both lower and higher incomes groups in Ireland, and of higher 
incomes groups in Mexico, Poland and Turkey. In a majority of countries, and on average, income shares 
in the bottom, middle and top quintiles were broadly unchanged from the mid-1990s to 2000. Assessments 
of changes in the distribution at the top of the scale, however, critically depend on how accurate are survey 
measures of high income and on confidentiality limits applied by statistical agencies (Box 1).  

Table 2.  Gains and losses of income share by income quintile: entire population, mid-1990s to early 2000 

Bottom quintile

Middle quintiles

Top quintile

Australia

=

=

=

Austria

-

=

+

Canada

=

-

+

Czech Republic

=

=

=

Denmark

=

-

+

Finland

-

-

+++

France

=

=

=

Germany

=

+

=

Greece

=

-

+

Hungary

=

=

=

Ireland

-

+++

---

Italy

=

=

=

Japan

-

-

+

Luxembourg

=

=

=

Mexico

+

+++

---

Netherlands

=

=

=

New Zealand

=

=

=

Norway

=

-

+

Norway

=

+++

---

Portugal

=

=

=

Sweden

-

-

+++

Turkey

=

+++

---

United Kingdom

=

-

+

United States

=

=

=

OECD (unweighted)

=

=

=

 

Note:

 The table shows percentage point changes in the shares of equivalised disposable income received by each quintile of the 

population. +++ denotes an increase of more than 1.5 percentage points in the share of disposable income received by the each 
quintile group. + denotes increase of between 0.5 and 1.5 percentage point. = denotes changes between -0.5 and +0.5 percentage 
points. - denotes decrease between 0.5 and 1.5 percentage point. --- denotes decrease of more than 1.5 percentage points. 

Source: 

Calculations from OECD questionnaire on distribution of household incomes. 

DELSA/ELSA/WD/SEM(2005)1 

 

16

Box 1. Survey measures of high income 

 

How good are the data used in this report at measuring high incomes? The short answer is "probably not 

very”, although this varies among countries. Quality of data on high incomes depends on how broad is the income 
concept used, and on confidentiality norms applicable to persons with very high income. With regard to the first issue, 
the main feature is whether the income concept encompasses income sources that disproportionately accrue to the 
very rich. Capital gains are generally excluded from the income surveys of most OECD countries. Similarly, changes in 
the remuneration package of managers (e.g. the growing importance of stock options) have increased the importance 
of flows that are likely to be poorly or not recorded in household surveys. With regard to the second issue, the main 
feature is whether survey data “top code” income or earnings beyond a given threshold. Use of “top-coded” values will 
underestimate the extent of inequality at a point in time, while an increase in the share of the population that is “top 
coded” will dampen the recorded rise in income inequality. 

 

While problems in the measurement of high income due to the exclusion of income streams that are most 

important for individuals at the top of the scale are common to most countries, “top coding” is a specific feature of the 
data for Japan and the United States. In the case of Japan, the data used here exclude persons with income three 
times above the standard deviation (1.6% of the total population in 1995 and 1.3% in 2001). In the case of the United 
States, most analysis of income distribution — including the one presented in Burniaux 

et al. 

(1998) and Förster and 

Pearson (2002) — rely on the "public use" version of the March supplement to the 

Current Population Survey. 

These 

data are however affected by confidentiality limits imposed upon different income sources and by changes over time in 
the methodology used by the Census Bureau with respect to top-coding. While analysts (including previous OECD 
work) have applied various adjustments to the public-use data to improve temporal consistency, these adjustments are 
only proximate. To remedy for his weakness, the US data used in this paper are based on the Census Bureau "internal 
use" version of the March CPS.*  

 

Information about the importance of the narrow income concept used in household surveys for measuring 

high income in the United States is available from studies by the US Congressional Budget Office (CBO), which 
combines the Census data used in this paper with tax records to track income trends near the top of the distribution.** 
The CBO data show large real income gains for the top quarter and top 1% of the US population during the 1980s, and 
real losses for individuals in the bottom 50% of the population; over the 1990s, although real income gains are spread 
more widely than in the previous decade, income gains at the very top continued to exceed those at the bottom. Tax-
based data for 4 OECD countries on the share of the equivalised disposable income accruing to the top 1% of the 
population (as a share of the income of the top decile), as presented in Atkinson

 

(2002) and reproduced below, 

suggest large increases in this share in all countries except France.  

Income share of the highest 1% in the share of the top decile of the population 

 

Source: Atkinson (2002).

 

----------- 
* 

The US data used in this paper remain affected, however, by changes in the maximum values recorded in the survey questionnaire 

(in 1979, 1985 and 1993) and by processing limits imposed by the Census Bureau to minimize the impacts of recording errors and 
prevent volatility of annual statistics. According to Welniak (2003), changes in recording limits had no effect on the Gini coefficient of 
(non-equivalised) household (pre-tax) income in 1979 and 1984, while increasing it by 2% in 1993.  
** CBO also adds to the Census income data in-kind and non-cash-income (such as income from food stamps, housing assistance 
and health insurance coverage), tax-financed wage supplements (such as Earned Income Tax Credits) and deduct payments for 
income and payroll taxes. The CBO income concept refers to individuals and is adjusted for economies of scale in consumption. 

 DELSA/ELSA/WD/SEM(2005)1 

 

17

13. 

Changes in income shares of different quintiles, as shown in Table 2, do not reflect the real 

income 

changes

 experienced by persons at different points of the income ladder. These absolute changes 

are a function of both trends in income inequality (how persons at different points of the income ladder 
fare relative to others) and of the overall pace of growth of household disposable income. Table 3 shows 
information on the annual rate of growth in equivalised disposable income in real terms (i.e. deflated by the 
increase in the consumer price index), based on the household surveys used in this paper, for people at 
different points in the income distribution over the decade since the mid-1980s and the 5-year period since 
the mid-1990s. Two main features emerge:  

•

 

First, most OECD countries experienced a higher real income growth in the second half of the 
1990s, relative to the previous decade, when data are averaged across all individuals: the only 
exceptions are Japan, Mexico and Turkey — where equivalised disposable income declined in 
real terms since 1995 — and the Netherlands and the United States — where gains are lower than 
in the previous period.

15

 In all other countries, the faster pace of real growth in equivalised 

disposable income on average

 

implied that persons at all income levels experienced stronger 

gains than in the previous period.  

•

 

Second, differences in the pace of income growth across the distribution are often significant. At 
the lower end of the income distribution, for example, the decline in 

average 

real household 

income recorded in Japan in the period 1995-2000 seems to have mainly affected persons in the 
bottom two deciles. Persons in the top two deciles recorded larger gains than those at the bottom 
in most OECD countries in the second half of the 1990s, but the differences are smaller than 
those recorded in the previous decade. 

14. 

It should be stressed, however, that the patterns highlighted by Table 3 are shaped by the specific 

features of the data and definitions used. First, the income concept used in household surveys differs in 
important respect from that embodied in the national-accounts measures conventionally used in analysis of 
living standards, and changes in the "coverage" of the survey data can distort trends over time.

16

 Second, 

changes in equivalised disposable income are affected by both the overall trends in household income and 
by changes in household size across different income deciles.

17

 

                                                      

15

  

In the case of the United States, national accounts data show stronger gains in real per capita household 
income in the second half of the 1990s relative to the previous decade. 

16

  

Analysis by Siminski 

et al 

(2003) suggests that the Australian data based on the 

Household Expenditure 

Survey

 used in this report are characterised by relatively low population estimates until the mid-1990s, and 

by estimates of gross income in 1975-76 that are well above those in later years; survey data for 1975-76 
also produce relatively high estimates of per capita wages and salary income (as well as own-business 
income and income tax) and low estimates of per capita income from government transfers. Both effects 
skew the

 

income distribution away from the bottom end and dampen growth in income relative to later 

periods. Because of these considerations, Australian data from the 

Household Expenditure Survey 

of 

1975/76, as presented in the previous OECD reports, have been omitted from the present analysis. 

17

  

Because of declines in average household size recorded by most OECD countries over this period, the 
gains in equivalised income shown in Table 3 are lower than those for per capita income. 

DELSA/ELSA/WD/SEM(2005)1 

 

18

Table 3.  Trends in real household income at different points of the income ladder 

Bottom 2 

deciles

Middle 6 

deciles

Top 2 

deciles

Average

Bottom 2 

deciles

Middle 6 

deciles

Top 2 

deciles

Average

Australia

0.1

-0.3

-0.4

-0.3

1.8

2.5

2.2

2.3

Belgium

1.1

0.5

1.0

0.7

..

..

..

..

Canada

0.3

-0.2

-0.1

-0.1

0.8

1.6

2.7

2.0

Czech Republic

..

..

..

..

0.4

0.6

0.7

0.6

Denmark

1.0

0.7

0.4

0.7

0.6

1.0

1.6

1.1

Finland

0.8

0.8

1.6

1.1

2.3

3.6

5.4

4.0

France

1.2

0.8

1.1

0.9

0.0

0.1

-0.2

0.0

Germany

0.6

1.3

1.4

1.3

0.4

0.7

0.6

0.6

Greece

0.3

0.1

0.1

0.1

3.0

2.9

3.8

3.3

Hungary

..

..

..

..

1.8

2.4

2.1

2.2

Ireland

3.1

2.5

2.4

2.5

5.2

7.7

5.4

6.6

Italy

-1.5

0.3

1.0

0.5

2.8

1.8

2.2

2.0

Japan

0.7

1.6

1.8

1.6

-1.9

-0.8

0.0

-0.7

Luxembourg

1.9

2.0

2.3

2.1

2.5

2.4

2.7

2.5

Mexico

0.6

1.0

2.8

2.1

1.1

0.3

-1.5

-0.7

Netherlands

0.5

1.5

1.7

1.5

2.6

2.3

2.1

2.3

New Zealand

-1.2

-0.6

1.3

0.2

1.3

2.3

2.3

2.3

Norway

-0.4

0.3

0.9

0.5

6.6

5.2

6.3

5.7

Poland

..

..

..

..

2.3

2.4

0.7

1.6

Portugal

..

..

..

..

5.0

4.1

4.4

4.3

Spain

3.1

2.4

1.9

2.3

..

..

..

..

Sweden

0.4

0.7

0.9

0.8

1.3

2.7

4.5

3.2

Switzerland

..

..

..

..

6.0

1.8

0.4

1.6

Turkey

-1.0

-1.0

1.7

0.5

0.2

0.4

-2.2

-1.0

United Kingdom

0.8

1.5

1.9

1.6

2.3

2.6

3.6

3.0

United States

1.1

0.9

1.6

1.2

0.7

0.9

0.5

0.7

OECD-20

0.6

0.8

1.3

1.0

1.6

2.0

2.1

1.9

Average annual change mid-1980s to mid-1990s

Average annual change mid-1990s to 2000

 

Note:

 Survey data on household income have been deflated by the change in the consumer price index in each country. Data for 

Germany refer to old Länder. Exact years are specified in the note to Table 1. 

Source: 

Calculations from OECD questionnaire on distribution of household incomes.  

2.3. 

Levels and trends in income poverty 

15. 

Changes in income distribution are of greater concern when they affect those at the bottom of the 

income scale. Over the second half of the 1990s, the greater attention paid to poverty in many OECD 
countries

18

 has been accompanied by changes in the way poverty is conceptualised, 

i.e.

 as a multi-

dimensional phenomenon that stretches beyond income to include inadequate access to learning, housing, 
poor health, and recourse to debt to meet ordinary living expenses. Consumption-based measures of 
poverty, based on peoples’ access to various essential goods and services, have been used in national and 

                                                      

18

  

At the World Summit on Social Development, held in 1995, governments agreed to the commitment of 
eradicating absolute poverty by a target date set by each country. At the regional level, the European 
Council agreed in December 2000 that the “fight against poverty and exclusion” should be pursued through 
the definition of commonly-agreed objectives for the European Union, the development of national action 
plans to meet these objectives, and the periodic reporting of progress through statistical indicators" (the 20 
so-called “Laeken indicators”). At the national level, some countries have adopted explicit targets to reduce 
or eradicate poverty, either for the whole population (e.g. Ireland, in the context of its [2000] “National 
Anti-Poverty Strategy”) or for selected sub-groups (e.g. the United Kingdom). Poverty goals have also 
often been set through “national sustainable development strategies” adopted by many countries in the 
wake of the 2002 World Summit on Sustainable Development. 

 DELSA/ELSA/WD/SEM(2005)1 

 

19

regional contexts (Box 2) while subjective measures of poverty are explored in others.

19

 While these 

alternative measures are important for in-depth poverty assessment

20

, low income remains the dimension 

that is more suited for cross-country and times-series comparisons aimed at identifying common trends. At 
the same time, however, poor reporting of means-tested benefits and other survey features may also imply 
that data quality is low, in some countries, for those at the lower end of the income distribution. 

Box 2. Assessing poverty: consumption, assets and income measures 

 

The case for assessing poverty and inequality on the basis of 

consumption

 rather than current income is 

strong. As consumption streams are more closely related to long-term income, they avoid counting as poor many 
families that, when suffering temporary income falls, are able to maintain a constant standard of living by lower savings 
or higher borrowing. Studies that have relied on consumption data typically report lower rates of poverty and inequality, 
relative to income-based estimates. Evidence, limited to the United States, suggests that greater income inequality 
among individuals from 1972 to 1998 has not been matched by higher inequality in consumption (Krueger and Perri, 
2002), as higher borrowing and lower asset holding allow to smooth income variations. Estimates for the United States 
also show that around one quarter of all households had asset holdings in 1999 insufficient to meet basic needs for 
more than three months (based on the 

Panel Study of Income Dynamics

), with this share increasing to 40% when 

home equity is excluded (Caner and Wolff, 2004).  

 

Data on consumption and ownership of different goods have been used to develop direct indicators of 

material well-being, measured in terms of ownership of appliances and electronic goods, housing and neighbourhood 
conditions, access to community services, and ability to meet basic needs (Census Bureau, 2003). Questions 
designed to asses the success of families in making ends meet are included in the 

European Community Household 

Panel Survey

 and in the US 

Survey of Income and Program Participation. 

In practice, however, the difficulties in 

deriving quality data on consumption and assets remain daunting, in particular in an international context. These 
difficulties relate to both the treatment of durable goods and work-related expenditures, as well as to how to aggregate 
measures of deprivation in specific areas (e.g. housing, heating) into a single index. More generally, consumption-
based measures of poverty may be criticised on the ground that they relate to actual behaviour,

 

as shaped by 

individuals' preferences, rather than to the resources

 

constraining their decisions.

 

 

16. 

Even when limiting assessments of poverty to low income, differences in country practices are 

substantial. These differences are reflected in the poverty thresholds used to identify the poor. While some 
countries rely on an 

absolute

 threshold – typically the cost of a minimum basket of goods and services 

deemed to be required to assure minimum living standards, indexed over time

21

 – others use measures of 

                                                      

19

  

For example, surveys for EU countries, based on questions about the extent to which the disposable 
(weekly) income of respondents falls below what they deem necessary to make ends meet, show 
“subjective” levels of poverty distinctly greater than those based on “objective” measure (based on a 60% 
disposable income threshold) but also little changes in country rankings (Gallie and Paugman, 2002).  
Förster 

et al. 

(2003) compare subjective poverty, income poverty and multiple deprivation measures across 

18 countries of the enlarged EU: their results suggest that differences between “old” and “new” EU 
member states are less pronounced for subjective poverty than for deprivation and income poverty.  

20

  

Alternative measures of poverty typically display little overlap with income-based measures. Data from the 
1999 “Survey on Poverty and Social Exclusion in Britain”, reviewed by Bradshaw and Finch (2003), show 
that despite the similar level of poverty on three different definitions — "income poverty", i.e. equivalent 
disposable income before housing costs less than 60% of the median (19%); "subjective poverty", i.e. 
respondents declaring that their household income is “a little” or “a lot” lower than the income they regard 
as required to keep households out of poverty (20%); and "deprivation"

, 

i.e. the proportion of households 

who cannot afford four or more (survey-based) “perceived necessities” (17%) — only 16% of all 
respondents are poor on at least two dimensions, and less than 6% are poor on all three of them. 

21

  

For example, the official poverty line in the United States is defined as the costs of an adequate diet, 
multiplied times three (i.e. based on the assumption that food represents around 

⅓

 of household 

expenditure), adjusted annually for inflation. For a family of four, this poverty line was about half of the 
median disposable income of families of that type in the early 1960s, while it is now a little over one 
quarter. Mexico introduced in 2001 three poverty lines based on absolute thresholds (food poverty, poverty 
of capabilities and asset poverty), each of which is representative of different levels of unsatisfied needs. 

DELSA/ELSA/WD/SEM(2005)1 

 

20

the 

relative 

income of different groups. The use of either absolute or relative thresholds for measuring 

income poverty has important implications for policy, 

inter alia 

for assessing the role of economic growth 

in reducing poverty.

22

 However, both absolute and relative income poverty provide important information 

to policy makers, and they are ultimately complementary. Moreover, while in theory these two approaches 
to the measurement of income poverty represent poles of a continuum of combinations, in practice the 
greater the commitment of governments to reducing poverty in all its dimensions, the lesser the conflict 
between them becomes. 

17. Use 

of 

absolute

 thresholds poses difficult methodological issues in the context of cross-country 

comparisons (Förster 1994). One way to illustrate how the extent of “absolute” poverty has changed over 
time is to use a relative income threshold in a base year for each country and to keep it unchanged in real 
terms

23

. On this measure, all OECD countries have achieved significant reductions in absolute poverty 

since the mid-1980s (Figure 4). By 1995, absolute poverty was only one-sixth of the level it had reached 
ten years earlier in Ireland and Spain — countries that have undergone radical economic transformations 
— and close to 70% lower in Sweden. The decline has continued in the second half of the 1990s, with the 
Czech Republic, Germany and Japan as the sole exceptions. On average, across 15 OECD countries for 
which this information is available, absolute poverty rates have declined by more than one-third in the 
period from the mid-1980s to the mid-1990s, and by close to one-fourth in the (shorter) period from the 
mid-1990s to 2000. 

Figure 4.  Trends in absolute poverty rates 

M id- 19 9 0 s   =  1.0 0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

AU

S

BE

L

CZ

E

DE

N

FI

N

FR

A

GE

R

G

RC

HUN

IR

L

IT

A

JP

N

ME

X

NL

D

NO

R

PO

L

P

OR

S

PA

S

WE UK

G

US

A

O

E

CD (

15

)

Mid-1980s

Mid-1990s
Around 2000

 

Note:

 Levels of absolute poverty rates in the mid-1990s are set equal to 1. Absolute poverty, as here defined, refer to 

thresholds set at 50% of median equivalised disposable income in the "base year", kept constant in real terms in the 
following years. The "base year" differs across countries (mid-1980s for most countries, except the Czech Republic, 
Hungary and Poland — where it refers to the mid-1990s — and Australia and the United States — where it refers to 
the mid-1970s). The OECD average value is the unweighted average of 15 OECD countries for which information is 
available both in the mid-1980s and in 2000. Exact years are those specified in the note to Table 1. 

Source

: Calculations from OECD questionnaire on distribution of household incomes. 

                                                      

22

  

When poverty is defined with reference to an absolute threshold, higher economic activity (i.e. a shift to 
the right in the distribution of household income) will reduce poverty, although by decreasing amounts. As 
an illustrative example, Freeman (2001) calculates that — when income is assumed to be normally 
distributed — a 0.1 point increase of mean income (relative to its standard deviation) would reduce 
absolute poverty by 3.2 points when 30% of the population is in the bottom tail of the distribution, by 2.6 
points when 20% of the population is in the bottom tail, and by 1.6 points when only 10% of the population 
is in that tail; he concludes that "the impact of an increase in income on [absolute] poverty falls roughly in 
half as poverty drops from 30% to 10%, due simply to the shape of the income distribution".  

23

  

This is also the idea behind one of the 20 EU social inclusion indicators (“Laeken indicators”): the at-risk-
of-poverty rate anchored at a moment in time (year t-3, uprated by inflation over the three years). 

 DELSA/ELSA/WD/SEM(2005)1 

 

21

18. 

In some respects, however, measures of absolute income poverty are overly restrictive.

 

Indeed, 

when family budgets are directly examined to determine the amount of resources needed to afford "decent 
living conditions", the range of expenditure items broadens beyond the necessities typically considered by 
absolute income measures: child-care costs, for example, hardly qualify as an essential item for meeting 
the basic needs of a person living on benefits, but become important when the same persons is expected to 
work. Also, the notion of relative income poverty seems to be better able to reflect risks that some 
individuals are excluded from the goods and services that are customary in any given society, which is an 
important dimension of social exclusion — a notion that has gained a central role in the social policy 
agendas of several OECD countries.

24

 For these reasons, a 

relative

 measure of poverty (most often based 

on a threshold of half of the median income for the entire population, Annex 2) is used below as the main 
poverty measure.  

19. 

Figure 5 shows levels of the poverty headcount, using a 50% median disposable income 

threshold. In 2000, the average poverty rate across 20 OECD countries — those for which information is 
available since the mid-1980s — was 10.6%, with an increase from the level recorded in the mid-1980s 
(9.4%) and in the mid-1990s (10.0%). The poverty rate, on this definition, increased over the second half 
of the 1990s by more than one percentage point in Australia, Austria, Finland, Ireland, Japan, New Zealand 
and Sweden, while it declined by one point or more in Norway, Italy and Mexico (from higher levels in the 
latter two countries). Beyond differences in trends, cross-country differences in the levels of poverty rates 
remain significant, ranging from less than 5% in the Czech Republic and Denmark to values above 15% in 
Ireland, Japan, the United States, Turkey and, in particular, Mexico. 

Figure 5.  Relative poverty rates among the entire population 

 

0%

4%

8%

12%

16%

20%

AU

S

AU

T

BE

L

C

AN

C

ZE

D

EN

FI

N

FR

A

GE

R

GR

C

H

U

N

IRL

ITA

JP

N

LU

X

ME

X

NL

D

N

ZL

N

OR

PO

L

PO

R

SP

A

SW

E

SW

I

TU

R

U

KG

US

A

OE

C

D

 (2

0)

mid-1980s
mid-1990s
2000

  

Note:

 Poverty rates are defined as the share of individuals with equivalised disposable income less than 50% of the 

median for the entire population. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

20. 

The relative poverty rates shown in Figure 5, while providing a convenient benchmark for cross-

country comparisons, are however limited in two main respect. First, relative poverty is measured with 
respect to an arbitrary threshold. When large proportions of the population are clustered just around this 
threshold, small changes in their income can lead to large changes in headcount rates. To examine the 
sensitivity of results to alternative choices of the poverty line, Figure 6 shows poverty rates measured with 
respect to both the 50% thresholds and the 60% line that is now commonly used by EUROSTAT as one of 

                                                      

24

  

The notion of social exclusion is broader than that of poverty. Atkinson (1998) identifies three common 
elements shared by all definitions of social exclusion: 

relativity 

(

i.e.

 exclusion can only be judged by 

looking at a person’s, or group’s, circumstances relative to those of others in a given place); 

agency 

(

i.e.

 

exclusion implies an act by some agents, either the excluded person herself or third parties); and 

dynamics 

(

i.e.

 exclusion reflects not just current circumstances, but dim prospects for the future). 

DELSA/ELSA/WD/SEM(2005)1 

 

22

their main indicators (“at-risk-of-poverty rate”). Figure 6 suggests that, in all OECD countries reviewed, a 
significant share of the population (8% or more in Ireland, Australia and New Zealand) is clustered 
between the 50% and 60% thresholds. In Germany, Hungary and the United States, the increase in poverty 
rates measured with respect to the 50% threshold over the second half of the 1990s largely reflected a 
decline in the number of persons with income between 50 and 60% of the median; conversely, in Australia, 
Denmark and many other countries, the increase in poverty measured with respect to the 50% threshold 
over the same period was accompanied by a similar increase in poverty rates based on the higher threshold. 
Persons with equivalised disposable income below 60% of the median may not be counted as poor when 
assessed with respect to more conservative thresholds, but still face difficulties in making ends meet.

25

 

Figure 6.  Relative poverty rates at different income thresholds 

 

0%

5%

10%

15%

20%

25%

30%

A

U

S

A

U

T

B

E

L

C

A

N

C

Z

E

D

E

N

F

IN

F

R

A

G

E

R

G

R

C

H

U

N

IR

L

IT

A

JP

N

L

U

X

M

E

X

N

L

D

N

Z

L

N

O

R

P

O

L

P

O

R

S

P

A

S

WE

S

W

I

T

U

R

U

K

G

U

S

A

O

E

C

D

 (

2

4

)

Increase in poverty when threshold
moves from 50% to 60%

Poverty at 50%

Mid-1990s

2000

 

Note:

 Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

21. 

Second, the headcount is just one dimension of poverty. Also important is the income level of 

poor individuals. Poverty gaps – the extent to which the average income of the poor falls below the 50% 
median income threshold – declined in the second half of the 1990s in about half the countries 
(considerably in Australia, New Zealand and Sweden) while increasing in the other half (considerably in 
Austria and Ireland, Figure 7). Overall, across the 24 countries shown for 2000, average disposable income 
of the poor was 29% lower than the poverty line. A synthetic measure of poverty, which takes into account 
both poverty rates and gaps (Teekens and Zaidi, 1990), indicates that the income transfer needed to raise 
all those living below the poverty line to that level ranged in 2000 between a high of 7% of (equivalised) 
disposable income in Mexico and a low of less than 1% in the Czech Republic and Luxembourg (Figure 8). 
Survey measures of income at the lower end of the spectrum are, however, less reliable than headcounts.  

                                                      

25

  

The proportion of people falling below the 50% threshold, as a share of those falling below the 60% line, is 
between 50% and 60% in most OECD countries, ranging between 50% or less in the Nordic countries and 
the Netherlands and 70% or more in Japan, Mexico, Turkey and the United States. 

 DELSA/ELSA/WD/SEM(2005)1 

 

23

Figure 7.  Income gaps of people living in relative poverty 

 

0%

10%

20%

30%

40%

AU

S

AU

T

BE

L

CA

N

C

ZE

DE

N

FIN FR

A

GE

R

GR

C

HU

N

IR

L

IT

A

JP

N

LU

X

ME

X

NL

D

N

ZL

NO

R

PO

L

PO

R

SPA SW

E

SW

I

TU

R

UK

G

U

SA

OE

C

D

 (20)

mid-1980s
mid-1990s
2000

 

Note:

 Income poverty gaps are defined as the difference of the average income of the poor and the national poverty 

threshold, in percent of that threshold. Thresholds are set at 50% of the median income for the entire population. Exact 
years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

Figure 8.  A composite measure of relative poverty in OECD countries, 2000 

0%

2%

4%

6%

8%

CZ

E

LU

X

DEN FI

N

S

WE NL

D

NO

R

FR

A

HU

N

BE

L

NZ

L

SW

I

PO

L

UKG AU

T

A

U

S

GE

R

C

A

N

P

O

R

IRL

G

R

C

SP

A

TU

R

IT

A

JP

N

US

A

ME

X

OECD average

 

Note:

 The composite poverty index is the poverty rate multiplied by the poverty gap. It measures the size of the income 

transfer that would be required to raise all those in poverty up to the poverty threshold of 50% of median equivalised 
disposable income. Data for Belgium and Spain refer to 1995. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

3. 

Labour markets, taxes and benefits: distributive effects on the population of working-age 

22. 

Determinants of income inequality and poverty differ across groups, reflecting their demographic 

characteristics (

e.g.

 the importance of lone parenthood) and the composition of their resources (

e.g.

 the 

extent to which individuals depend on earnings or government transfers for their daily living). This section 

DELSA/ELSA/WD/SEM(2005)1 

 

24

presents information about the population of working age (18 to 65).

26

 First, it briefly describes the extent 

to which changes in income distribution and relative income poverty for this group depart from those 
observed for the entire population. Second, it discusses the relative importance of

 

labour market conditions 

and of tax and welfare systems in shaping these trends. Third, it presents a simple decomposition of 
changes in relative poverty for the working-age population according to its main proximate determinants. 

3.1. 

Recent trends in income distribution and relative poverty for the working-age population 

23. 

Persons of working age represent the largest share of the total population in all countries 

reviewed. As a result – unsurprisingly – the main trends in inequality and poverty among this group largely 
follow those described for the entire population. The Gini coefficient of income inequality among the 
population of working-age was stable in 2000 relative to the mid-1990s on average. This, however, mainly 
reflected the strong declines recorded in Mexico, Turkey and Ireland; in most OECD countries, the Gini 
coefficient increased in this period, with the increase being around 10% or more in Finland and Sweden 
(Figure 9).  

Figure 9.  Income inequality among the population of working age 

0

10

20

30

40

50

60

A

U

S

A

U

T

B

E

L

C

A

N

C

Z

E

D

E

N

F

IN

F

R

A

G

E

R

G

R

C

H

U

N

IR

L

IT

A

J

P

N

L

U

X

M

E

X

N

L

D

N

Z

L

N

O

R

P

O

L

P

O

R

S

P

A

S

W

E

S

W

I

T

U

R

U

K

G

U

S

A

OE

C

D

 (

 2

0

)

mid-80s

mid-90s

2000

 

Note:

 Data for Canada and Sweden are adjusted for breaks in series. Data for Germany refer to old Länder. Exact 

years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

24. 

Findings are similar when looking at relative poverty, measured with respect to a 50% median 

threshold. On average, across 24 countries, 8.7% of the working-age population had income of less than 
half of the median around 2000 (1 ½ points lower than for the total population), with little change relative 
to the mid-1990s. Relative poverty rates increased in most countries during this period, including in some 
of the countries (e.g. Greece) that recorded declines in poverty rates for the entire population.  

                                                      

26

  

This age range differs from that used in other OECD publications to identify the working-age population in 
order to assure consistency with internationally-agreed definitions of children (persons aged less than 18). 

 DELSA/ELSA/WD/SEM(2005)1 

 

25

3.2. 

The influence of labour markets 

25. 

Market income (earnings, self-employment and capital income) represents the largest component 

of the disposable income of the working-age population

27

 and has been — over the two decades to the mid-

1990s — the main driver of changes in income inequality (Förster and Pearson, 2002).

28

 Indeed, inequality 

in market income had been widening 

everywhere 

in the decade from the mid-1980s to the mid-1990s, a 

widening that translated in several OECD countries into larger inequalities in the distribution of disposable 
income (Figure 10, 1

st

 panel). In this respect, the second half of the 1990s represents a significant departure 

from previous trends. The employment gains that most OECD countries experienced in the second half of 
the 1990s

29

 have led to an increase in the share of earnings and self-employment in the disposable income 

of the working-age population (on average by around 2.5 points) and to a corresponding decline in the 
share of capital incomes and public transfers. Partly as a result, the Gini coefficient of market income

 

inequality for the population of working-age declined in around one third of the countries over the second 
half of the 1990s, while it increased only marginally in most other countries: only in the Czech Republic, 
Germany, Japan and Norway did market income inequality increase significantly in the second half of the 
1990s (Table 4). 

Figure 10. 

Trends in Inequality of market and disposable income among the working-age population 

 

Note:

 Gini coefficients for market (vertical axis) and disposable income (horizontal axis) among the working-age 

population. Arrows pointing to the upper-right corner in each panel indicate an increase in inequality for both market 
and disposable income. Data for Germany refer to old länder. Data for Canada and Sweden take account of breaks in 
series in the mid-1990s. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

                                                      

27

  

For a number of countries (Austria, Greece, Hungary, Luxembourg, Mexico, Poland and Spain), tax data 
are not available separately. The analysis on market versus disposable incomes presented in this and later 
sections is therefore restricted to a sub-set of countries.  

28

  

The Gini coefficient for market income increased by around 9% in the period from the mid-1970s to the 
mid-1980s (across 7 OECD countries) and by 11% in the period from the mid-1980s to the mid-1990s 
(across 14 countries), while it was almost stable (an increase of less than 1%) in the second half of the 
1990s. 

29

  

The employment to population ratios among the working-age population, across the OECD area as a 
whole, increased from 64.3% in 1995 to 65.7% in 2000. Two-thirds of the countries registered an increase 
in employment rates, with increases of 5 points of more in Ireland, the Netherlands, Spain, New Zealand 
and Norway (OECD, 2003). 

DELSA/ELSA/WD/SEM(2005)1 

 

26

Table 4.  Levels and trends in the Gini coefficient of market income inequality among the working-age 

population 

Percentage point changes 

 Levels 

2000 

mid-80s to mid-

90s 

mid-90s to 2000 

Australia 42.1 

3.4 

-0.2 

Canada 

39.0 2.2  0.1 

Czech Republic  

40.4 2.8  3.3 

Denmark 

35.5 5.2  0.7 

Finland 

37.1 7.6  -1.1 

France 

40.3 2.2  -1.0 

Germany 

39.3 0.3  3.0 

Ireland 

39.1 ..  .. 

Italy 

45.6 7.2  -0.8 

Japan 

36.2 2.9  2.3 

Netherlands 

33.2 0.4  -4.5 

New Zealand 

43.0 6.6  0.2 

Norway 

36.3 4.7  2.2 

Portugal  

43.3 3.5  0.2 

Sweden 

37.5 6.9  0.1 

Switzerland 

32.4 ..  .. 

United Kingdom 

43.2 3.4  0.8 

United States 

42.0 4.1  0.2 

OECD (16) 

39.6 4.0  0.3 

 

 

Note:

 Data for Germany refer to old Länder. For Czech Republic and Portugal, “mid 80s” refer to 1992 and 1990, 

respectively. Exact years are those specified in the note to Table 1. 

Source: 

Calculations from OECD questionnaire on distribution of household incomes 

26. 

The influence of labour market conditions on inequality and poverty at the level of market 

income reflects changes in both the weight of earnings, self-employment and capital income, and in the 
degree of inequality of each component. In almost all countries covered in this paper, self-employment 
and, in particular, capital income are more unequally distributed than earnings. Annex Table A.4 provides 
one measure of the degree of inequality in the distribution of the individual components of market income, 

i.e.

 the share of earnings, self-employment income and capital income going to the bottom, top and middle 

quintiles of the working-age population. In most countries, the bottom quintile of the working-age 
population receives between 4 and 5% of all earnings (3% or lower in Australia, New Zealand and the 
United Kingdom) while the top quintile receives some 40%. Both self-employment income and capital 
income are more concentrated towards both tails of the distribution than earnings (between 6 and 7% 
accrue to the bottom quintile, and as much as 50% goes to the top quintile). Over the second half of the 
1990s, however, the share of earnings going to the lower quintile increased in many countries, often after 
declines in the previous decade.

30

 

27. 

The relationship between labour markets, on one side, and income inequality and poverty, on the 

other, is crucial for social policies. At the level of individuals, higher employment increases the well-being 
of those at greatest risk of social exclusion and poverty. Across countries, those with lower non-

                                                      

30

  

This is not the case of self-employment and capital incomes: their share going to the bottom (and often the 
middle) quintile decreased in most countries in the late 1990s, while that going to the top quintile increased 
(Anglo-Saxon countries being an exception), continuing the trend of the previous decade. 

 DELSA/ELSA/WD/SEM(2005)1 

 

27

employment rates (in particular for women) experience lower poverty rates 

at the level of market income

 

among the population of working age (Figure 11, panel a). The relationship is stronger when account is 
taken of the way in which employment is distributed among households, as countries with similar levels of 
employment (

e.g.

 Japan and Australia, where employment-to-population ratios are slightly above 70%) 

show large differences in the share of the working-age population living in households where no adult 
works (ranging from less than 3% in Japan to more than 13% in Australia).  

Figure 11. 

Relative poverty rates at the level of market income among the working-age population, non-

employment of individuals and joblessness 

Inactivity among individuals

AL

AT

CA

CZ

DK

FN

FR

GE

GR

HU

IR

ITA

JP

LX

MX

NL

NZ

NW

PL

PG

SW

SZ

UK

US

y = -0.07x + 0.13

R

2

 = 0.02

0%

4%

8%

12%

16%

20%

50%

60%

70%

80%

P

o

v

e

rt

y rat

e

Non employment rates

Jobless households

AL

AT

CA

CZ

DK

FN

FR

GE

GR

HU

IR

ITA

JP

LX

MX

NL

NZ

NW

PL

PG

SW

SZ

UK

US

y = -0.20x + 0.10

R

2

 = 0.07

0%

4%

8%

12%

16%

20%

0%

5%

10%

15%

20%

P

o

v

e

rt

y rat

e

Share of the population in jobless households 

7

Inactivity among individuals

US

UK

SZ

SW

PG

NO

NZ

NL

JP

IT

IR

GR

GE

FR

FN

DK

CZ

CA

AL

10%

15%

20%

25%

15%

25%

35%

45%

P

o

v

e

rt

y rat

e

Non-employment rates

Jobless households

US

UK

SZ

SW

PG

NO

NZ

NL

JP

IT

IR

GR

GE

FR

FN

DK

CZ

CA

AL

10%

15%

20%

25%

0%

4%

8%

12%

16%

P

o

v

e

rt

y rat

e

Share of the population in jobless households 

  

Note:

 Relative poverty rates of individuals aged 18 to 65. Non-employment rates of persons aged 16 to 64, from labour 

force surveys. Joblessness is the share of the total population living in households with a working-age head and where 
no one works. Exact years are those specified in the note to Table 1.

 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

28. 

While cross-country differences in the extent of non-employment and joblessness go a long way 

towards explaining differences in market income inequality, the 

quality of jobs

 also matter, especially for 

those in the lower tail of the distribution. Jobs are increasingly diverse, and some of them provide little 
protection from risks of poverty. While the risk of falling into poverty is much higher for households with 
no adult in employment than for those where someone works, households with one or more workers 
represent a very substantial proportion of the income-poor in all OECD countries.

31

 Even households with 

two or more workers are not immune from the risk of inadequate income, especially in Austria, Greece, 
New Zealand, Portugal, Switzerland and the United States. Further, since 1995, the share of income-poor 
households with at least one worker has increased in around half of the countries shown in Figure 12.  

                                                      

31

  

However, this partly reflects the inclusion — among household with workers — of the self-employed, 
whose income is often under-recorded in household surveys, as well as of people with part-time and part-
year jobs. With reference to the latter, persons who worked full-time full-year in the United States in 1999 
had an (absolute) poverty rate of 2.6%, as compared to 13% for those who worked either part-time or for 
only part of the year and 20% for those who did no paid work at all during the year (Freeman, 2001). 

DELSA/ELSA/WD/SEM(2005)1 

 

28

Figure 12. 

Structure of relative poverty in households headed by a working-age head, by work attachment 

of household members 

0%

4%

8%

12%

16%

20%

A

U

S

A

U

T

C

A

N

C

Z

E

D

E

N

F

IN

F

R

A

G

E

R

G

R

C

H

U

N

IR

L

IT

A

JP

N

L

U

X

M

E

X

N

L

D

N

Z

L

N

O

R

P

O

L

P

O

R

S

P

A

S

W

E

U

K

G

U

S

A

O

E

C

D

-2

2

Two or more workers
One worker
No workers

Mid-1990s

2000

 

Note:

 The height of each bar represents the poverty rate (using a 50% threshold) of persons living in households with a 

head of working age in each country. Data for Germany refer to old Länder. Exact years are those specified in the note 
to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

3.3. 

The role of taxes and public transfers  

29. 

The relationship between employment levels and relative poverty at the level of 

disposable 

income

, however, does not follow simple patterns: in other terms, countries where non-employment 

(among individuals) and joblessness (at the household level) are lower do not consistently show lower rates 
of relative income poverty. This is because, beyond the levels

 

and

 

quality of jobs, government policies play 

a significant role in accelerating or moderating trends in income distribution and poverty. The relationship 
between government policies and poverty outcomes is striking: across countries, relative poverty rates 
among the working-age population are lowest where (non-health) social spending on the working-age 
population is highest (Figure 13).

 32

 Within each country, the combined effect of the tax and benefit 

systems is to lift out of relative income poverty more than half of the population at risk, on average 
(Figure 14). This effect, which ranges between around one-fourth of those below the poverty threshold 
before taxes and transfers in the United States and more than two-thirds in Denmark, declined however 
over the second half of the 1990s in most OECD countries, as the growth of real benefits most often lagged 
that of median disposable income.  

                                                      

32

  

Other means through which governments influence poverty and income inequality include policies aimed 
at changing the distribution of skills among the population (in particular, at increasing the earnings 
potential of those most exposed to poverty risks), at supporting the earnings of workers at the bottom of the 
pay scale (for example through minimum wage provisions), and at addressing the specific barriers to 
labour farce participation faced by disadvantaged groups. 

 DELSA/ELSA/WD/SEM(2005)1 

 

29

Figure 13. 

Relative poverty among the working-age population and social spending, 2000 

AUSAUT

CAN

CZE

DEN

FIN

FRA

GER

GRC

HUN

IRL

ITA

JPN

LUX

MEX

NLD

NZL

NOR

POL

POR

SWE

SWI

TUR

UKG

USA

0

4

8

12

16

20

0

4

8

12

16

Non-health public social spending tow ards w orking-age population (%GDP)

P

o

v

e

rt

y

 r

a

te

 (%)

 

Note:

 Social spending is defined as public social spending excluding health, old-age and survivor benefits, as a share 

of GDP. Poverty rates are measured with respect to a threshold set at half of the median equivalised household 
disposable income. Exact years are those specified in the note to Table 1. 

Source:

 OECD Social Expenditure database and data from the OECD income distribution questionnaire. 

Figure 14. 

Effects of taxes and transfers in reducing relative income poverty 

0%

10%

20%

30%

AU

S

BE

L

CA

N

CZ

E

D

EN

FI

N

FR

A

GE

R

IRL

IT

A

JP

N

NL

D

N

ZL

NO

R

PO

R

SW

E

SW

I

UK

G

U

SA

OE

C

D

 (1

7)

Effect of taxes and transfers in reducing poverty
Poverty rate, disposable income

Mid-1990s

2000

 

Note:

 The "light" bar is the poverty rate at the level of disposable income, the "dark" bar is the arithmetic difference 

between the poverty rates at the levels of market and disposable income. Exact years are those specified in the note to 
Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

30. 

Both taxes paid by the population of working-age (income and payroll taxes) and public transfers 

received by the same group reduce inequalities in the distribution of disposable income: taxes are mainly 

DELSA/ELSA/WD/SEM(2005)1 

 

30

paid by the richer, while public transfers mainly accrue to the poorer. An indicator of the relative role of 
taxes and transfers in reducing income inequalities ("pseudo"-Gini coefficient of each of the income 
components) shows that, in all countries, taxes are more concentrated than transfers, although the extent to 
which this is the case declined over the second half of the 1990s.

33

 The extent of concentration of public 

non-pension transfers to persons at the bottom quintile of the income scale increased significantly in the 
second half of the 1990s in most countries, continuing a trend already observed in the previous decade, 
while declining only in Greece, Germany, Luxembourg, Ireland and Poland (Annex Table A.5). In turn, the 
proportion of non-pension transfers accruing to the six middle income quintiles decreased in most 
countries and on average, while changes at the top have been much less pronounced. Changes have been 
smaller in the case of taxes, with greater concentration of taxes at the top of the income distribution in a 
majority of countries, but reductions in Australia, France, Ireland, Japan, Portugal and the United States.  

3.4. 

Accounting for changes in poverty rates since the mid-1990s 

31. 

Although both taxes and public transfers reduce income inequality and poverty at a point in time 

– for a given distribution of market income – they also distort decisions of private agents in terms of 
employment and work efforts. Marginal effective tax rates, which are one cause of these distortions, are 
typically high at both ends of the income distribution, and they may contribute to poverty traps among 
many individuals relying on benefits as well as to reductions in work effort, or attempts to escape taxation, 
by individuals with high earnings. Reforms implemented by several OECD countries during the second 
half of the 1990s (generally in the form of earnings top-up for low-paid workers, and of greater pressures 
put on persons relying on benefits to take up suitable employment offers) have aimed at reducing these 
distortions so as to improve work incentives for individuals with low income. 

32. 

How have these reforms affected changes in poverty? Efforts to address this question have 

typically followed two tracks. The first relies on individual records, assessing what poverty rates would be 
today if the structure of wages, hours of work, and government benefits had remained at some base-year 
level; while this approach does not account for behavioural changes following reforms, it allows tracking 
the same individual over time.

34

 A second approach, which is easier to implement when comparing a large 

number of countries, relies on aggregate data.

35

 This approach is used here to account for changes in 

relative poverty rates among individuals living in households with a working-age head. A simple shift-
share analysis is used to decompose changes in relative poverty rates into three components:  

•

 

the part due to changes in market-income poverty for each group, while keeping constant both 
population structure and the effectiveness of taxes and transfers in reducing poverty;  

                                                      

33

  

The only exceptions are Germany, New Zealand, Norway and the United States. Italy is the only country 
where public transfers to persons of working age are unequally distributed (i.e. higher income groups 
receive a larger share than lower-income ones), reflecting the importance of earnings-related pensions 
(which tend to increase with income) received by persons of working-age.  

34

  

Based on this approach, Dickens and Ellwood (2001) argue that demographic conditions (e.g. a greater 
incidence of single-parent households), earnings structure (e.g. wider earnings distributions) and work 
efforts (i.e. the combined effect of changes in activity rates and hours worked) account for a similar share 
of the increase in relative poverty in the United Kingdom from 1979 to 1999, while greater generosity in 
government benefits contributed to reduce poverty rates over the same period; in the United States, the 
increase in relative poverty over the same period mainly reflected demographic changes and, to a less 
extent, changes in earnings structure; higher work efforts contributed to lower poverty, while changes in 
government benefits did not exert significant influence in either direction. 

35

  

Most often, studies using aggregate data regress poverty rates against a range of possible determinants, and 
use results to compare situations at two points in time. However, results from this type of analysis have 
been found to be typically unstable and sensitive to the specification used. 

 DELSA/ELSA/WD/SEM(2005)1 

 

31

•

 

the part due to changes in the effectiveness of taxes and transfers in reducing market-income 
poverty for each group, for given population structure and market-rate poverty; and  

•

 

the part due to changes in population structure, for given market-income poverty and 
effectiveness of tax and transfers in reducing poverty of that group.

36

  

While mechanical decompositions of this type cannot reflect the complex interrelations between each 

pair of variables

37

, and the effects of various factors on the income of each individual, they do provide a 

convenient summary of the relative role played by various factors.

38

 At the same time, however, the 

detailed breakdown used for this decomposition may imply that results are affected by the small sample 
size of surveys. 

33. 

Figure 15 shows results applied to changes in relative poverty rates for persons living in 

households with a head of working-age over the second half of the 1990s, with population broken down by 
work attachment of adult members (i.e. no adult in the household working, only one adult working, and 
two or more adults working). In the case of Australia, for example, relative poverty rates of persons living 
in households with a head of working age increased by 1 point in the second half of the 1990s (shown with 
a "diamond" in Figure 15): changes in market-income poverty and in the relative size of the three types of 
households (a slight decline in the share of persons in households where no one works) contributed to 
reduce poverty rates (by 0.4 points in the first case, marginally in the second) but their effect was more 
than offset by a (1.4 point) reduction in the poverty-reducing effect of taxes and transfers. (The sum of the 
3 bars for each country equals the change in the relative poverty rate).  

34. 

Figure 15 suggests that, while reforms to taxes and transfers systems introduced in the second 

half of the 1990s may have contributed to higher employment and lower market-income poverty in several 
countries, their effects were often offset by a smaller impact of taxes and transfers in reducing poverty. 
Overall, despite much diversity in country experiences in terms of overall changes in poverty rates, higher 
market-income and changes in population structure (declines in the share of workless households and 
increases in that of two-worker households) contributed to lower total poverty in a majority of countries 
(with the exception of Germany, Japan, New Zealand and Portugal, in the first case; and of Finland and 
Germany in the second) over this period. However, in most cases these positive developments were offset 
by a reduction in the effect of taxes and transfers in reducing poverty (with the exceptions of France, 

                                                      

36

  

In this exercise, the aggregate poverty rate, at the level of disposable income, is expressed as a weighted 
sum of group-specific poverty rates, with these rates expressed as the product of market-income poverty 
and of a coefficient expressing the effect of taxes and transfers in reducing market-income poverty. 

PR

t 

= 

Σ

 

PR

i

t 

= 

Σ

 

[PR(MI)

 i

t 

* (1 – 

β

 )

 i

t 

] * 

α

i

t 

where PR is the poverty rate at times t, at the level of disposable income, for each group 

i

; PR(MI) is the 

poverty rate at times t, at the level of market income, for each group; (1- 

β

) is the poverty-reducing effect 

of taxes and transfers for each group; and 

α

 is the population share. When analysing changes over time in 

the poverty count, changes in one variable are multiplied by the average value (between two points in time) 
of the other two variables (to avoid explicit consideration of interaction terms between each pair of 
variables). 

37

  

Changes in benefit level, for example, may encourage previously inactive individuals to take-up jobs, 
leading to positive effects (i.e. a reduction is poverty) for both household structure (decline in workless 
households) and market-income poverty (higher earnings as former benefit recipients enter employment). 

38

  

Danziger and Gottshalk (1995) apply a shift-share decomposition to changes in (absolute) poverty in the 
United States from 1949 to the early 1990s; they concluded that changes in income for each demographic 
group dominate, while changes in the demographic composition played a minor role. 

DELSA/ELSA/WD/SEM(2005)1 

 

32

Germany

39

, Greece, Italy and Norway).

40

 In interpreting these results, it should be noted that a "smaller" 

poverty-reducing effect of net public transfers may reflect an increase in real benefits lagging that of 
median disposable income, and lower benefit take-up, rather than cuts in the value of benefits in real terms. 

Figure 15. 

Changes in relative poverty rates among households with a working-age head by components, 

mid-1990s to 2000 

-4%

-2%

0%

2%

4%

6%

AU

S

C

AN

DE

N

FIN

FR

A

G

ER

GR

E

HU

N

IT

A

JP

N

N

LD

NZ

L

N

O

R

POR SW

E

SW

I

UK

U

SA

Changes in population structure by number of adults w orking

Changes in taxes and transfers

Changes in market-income poverty

Changes in poverty rates

 

Note:

 Data are based on a shift-share analysis applied to population living in households with a head of working age, 

broken by work attachment of household members (i.e. distinguishing between households with no workers, with one 
adult working, and with two or more adults working). The sum of the three components (shown ah bars) is equal to the 
total change in poverty rate (shown with a dark "diamond"). Exact years are those specified in the note to Table 1. 

Source:

 Calculations based on the Calculations from OECD questionnaire on distribution of household incomes. 

4. 

Poverty and inequality among children and households with children 

35. 

Poverty is a special concern when it affects individuals who cannot be held responsible for their 

situation and who are especially vulnerable to its consequences. These considerations have led to the 
explicit formulation of policy "targets" for child poverty in some OECD countries (

e.g. 

the United 

Kingdom) and to greater attention paid to children in the social policy agendas of most of them.

41

 This 

                                                      

39

  

In Germany, reforms introduced in 1996 increased child benefits and social minima (

Steuerfreies 

Existenzminimum)

.  

40

  

When looking at the household structure of the population (number of children and of adults present in the 
same households, i.e. four groups), this analysis also suggests that changes in population structure 
contributed to increase poverty in Australia, Germany, Netherlands, Sweden and the United Kingdom 
(mainly because of more persons living alone) and to reduce it (marginally) in the United States. 

41

  

In the United Kingdom, a quantified target has been set to reduce the number of children living in low-
income households by a quarter by 2004-05 as a contribution to the broader target of halving child poverty 
by 2010 and eradicating it by 2020. In Canada, the commitment was to "seek to eliminate child poverty" 
but no definition or indicator was agreed. The "New Zealand's Agenda for Children" (June 2002) embodies 
a commitment to eliminate child poverty. Among other EU countries, the setting of targets for child 
poverty is explicit in the National Action Plan of Greece, while other countries (e.g. Germany) have set 
targets in areas that may have an important impact on child poverty (such as cutting the number of youth 
not having obtained vocational qualifications by half by 2010).  

 DELSA/ELSA/WD/SEM(2005)1 

 

33

section looks at the recent record of OECD countries with respect to poverty among children and their 
families. It should be stressed, however, that results described below are critically shaped by how we adjust 
household income for household size (using the square root of household size): other assumptions about 
household equivalence of scale, for example, may significantly reduce the number of children that are 
counted as poor in any country. 

4.1. 

Levels and trends in relative income and poverty  

36. 

On average, across 23 OECD countries, around 12% of all children fell below the (50%) poverty 

threshold in 2000, an increase of close to 1 point relative to the level of the mid-1990s. Child poverty rates 
are especially low in the Nordic countries, where fewer than 4% of all children are poor, followed by 
France, Switzerland and the Czech Republic, with rates of around 7%. Child poverty is high in Mexico, 
Turkey and the United States, where it exceeds 20%, but also in Ireland, Italy, New Zealand, Portugal and 
the United Kingdom, where it is above 15% (Figure 16). Austria and New Zealand experienced significant 
increases in child poverty in the second half of the 1990s, while Italy recorded a significant decline. 

37. 

In most countries, relative poverty rates among children are also higher than for the entire 

population (Figure 16), but with much variation across countries: in the Nordic countries and Belgium, 
child poverty rates are between ½ and 

⅔

 of the overall rate, while in the Czech Republic, the Netherlands, 

Hungary and New Zealand child poverty rates exceed overall poverty by 50% or more. These differences 
suggest that specific factors increase risks of poverty for children in some OECD countries.

42

 

Figure 16. 

Relative poverty rates for children and the entire population 

0%

5%

10%

15%

20%

25%

30%

AU

S

AU

T

BE

L

CA

N

CZ

E

D

EN

FI

N

FR

A

GE

R

GR

C

HUN IR

L

IT

A

JP

N

LU

X

ME

X

NL

D

NZ

L

NO

R

POL PO

R

SP

A

SW

E

SW

I

TU

R

UK

US

OE

C

D (2

4)

child poverty rate

overall poverty rate

Mid-1990s

2000

 

Note:

 Exact years are those specified in the note to Table 1.

 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

4.2. 

The influence of household structure, mothers employment and benefit systems 

38. 

While several factors contribute to child poverty, three of the most important relate to the 

characteristics of the households where children live, to the employment status of their mothers, and to the 
role of taxes and transfers in reducing poverty risks. With reference to the first factor, the equivalised 
                                                      

42

  

On average, children constitute around a quarter of the poor population, with this share being 15% or less 
in Nordic countries, and 35% or more in the Czech Republic, Mexico, New Zealand, Poland, Turkey and 
the United Kingdom. 

DELSA/ELSA/WD/SEM(2005)1 

 

34

disposable income of households with children (and a head of working age) is, on average, slightly above 
80% of that of households with no children: the relative income of households with children is highest in 
the Nordic countries and Belgium, and lowest in Mexico and, to a lesser extent, Australia, New Zealand, 
Portugal and Turkey (Figure 17). Households with children generally have higher rates of relative poverty 
than those without children, and this difference increased over the past two decades.

43

 Among households 

with children, relative poverty rates are highest among single parents — at 40% or more in Australia, 
Canada, New Zealand, the United Kingdom and the United States, and over 50% in Ireland, Japan, Spain 
and Turkey — and, during the past 15 years, this increased the most in France, the Netherlands, New 
Zealand and the United Kingdom (while they it in some Nordic countries).  

Figure 17. 

Relative disposable income of households with children, 

2000

0.0

0.2

0.4

0.6

0.8

1.0

1.2

M

E

X

T

U

R

S

W

I

N

Z

L

P

O

R

A

U

S

U

K

G

L

U

X

U

S

A

A

U

T

IR

L

G

E

R

J

P

N

N

L

D

C

Z

E

C

A

N

S

P

A

H

U

N

F

R

A

IT

A

P

OL

GR

C

S

W

E

N

O

R

F

IN

D

E

N

B

E

L

OECD average

 

Note:

 Relative to households without children. Countries are ranked in increasing order of relative income. Data for Belgium and 

Spain refer to 1995. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

39. 

In many countries, however, it is not living in single-parent households 

per se 

that increases 

poverty, but rather the employment status of the parent. On average, the poverty rate for single parents (at 
32%) is three times higher than for all families with children; however, among those where the single 
parent is jobless, the poverty rate reaches 57% (while it falls to 21% among those where the parent is 
employed). Having an employment therefore reduces poverty risks among single parents by more than 
60%, although Greece, Japan and Turkey are notable exceptions. In several countries – notably the 
Australia, Italy Norway and Sweden – the poverty rate among single parents with a job is not that much 
different from the overall rate for families with children (Figure 18). Having a job also reduces the 
probability of falling into poverty for couples with children (by almost ¾ in the case of couples where both 
parents work, relative to those where only one parent does). Because of these patterns, OECD countries 
with higher employment rates among mothers also experience lower rates of child poverty (Figure 19). 

                                                      

43

  

On average, across 20 OECD countries, relative poverty rates of households with children were 17% 
higher than those of households without children in the mid-1980s, 22% higher in the mid-1990s, and 27% 
higher in 2000. 

 DELSA/ELSA/WD/SEM(2005)1 

 

35

Figure 18. 

Relative poverty rates in households with children and single-parent households, 2000 

 

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

AUS

AUT

BEL

CA

N

CZ

E

DE

N

FI

N

FRA

GE

R

G

RC

IRL

ITA

JP

N

LU

X

M

EX

NL

D

NZ

L

NO

R

PO

L

PO

R

SPA SWE

SW

I

TUR

UK

G

US

A

O

ECD (

26

)

Families with children

Non-working single parents

Working single parents

 

Note:

 Poverty thresholds at 50% of median income for the entire population. Data for Belgium and Spain refer to 1995. 

Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

Figure 19. 

Poverty among children and employment rates among mothers, 2000 

 

AS

AT

CA

CZ

DK

FI

FR

GE

GR

IR

IT

NL

NZ

NO

PT

SW

CH

UK

US

LX

0%

4%

8%

12%

16%

20%

24%

40%

60%

80%

100%

Employment rates of mothers

C

h

ild

 p

o

v

e

rty

 r

a

te

s

  

Note:

 Employment rates among women aged 25 to 54 with one and two or more children aged 15 or less (16 in the 

case of New Zealand and Sweden). Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes and OECD (2002), Employment Outlook, Paris. 

40. 

Changes in household composition — in terms of both household types and work attachment of 

different households — have influenced trends in poverty rates for households with children. Table 4 
compares relative poverty rates for households with children in the mid-1990s and 2000 (1

st

 and 2

nd

 

columns) in each country with those that would have prevailed had household structure remained as in 
1985 (4

th

 and 5

th

 columns); rates under "unchanged population structure" refer to relative poverty rates 

under assumptions of constant shares of single and couple families, as well as of constant shares in 
households with no, one and two or more workers. In most countries, differences between actual and re-
weighted rates are less than one percentage point in 1995, and below 2 points in 2000. Changes in 
household structure since the mid-1980s (3

rd

 and 6

th

 columns) have worsened trends in relative poverty for 

households with children in ten countries: Australia, Denmark and Norway (dampening the decline) as 
well as in the Czech Republic, Finland, France, Japan, Germany, New Zealand and the United Kingdom 

DELSA/ELSA/WD/SEM(2005)1 

 

36

(accentuating the increase). In the remaining 11 countries, changes in household structure have tended to 
smooth changes in relative poverty among households with children. 

Table 5.  Poverty rates in households with children under different household structure, mid-1990s and 2000 

 

Mid 1990s

2000

Point changes since 

mid-1980s

Mid 1990s

2000

Point changes since 

mid-1980s

Australia

9.4

10.2

-3.3

8.9

9.1

-4.4

Austria

5.5

11.5

7.0

7.0

13.5

8.9

Canada

11.0

11.5

-7.2

11.6

14.0

-4.7

Czech Republic

4.2

5.6

3.5

3.2

2.5

0.5

Denmark

1.5

2.1

-0.8

1.3

1.8

-1.2

Finland

1.9

3.3

0.9

1.8

2.1

-0.3

France

6.7

6.7

0.5

6.5

6.0

-0.2

Germany

8.6

10.4

4.5

6.7

8.1

2.2

Greece

11.1

11.1

-0.2

12.4

12.7

1.4

Italy

17.1

14.3

4.1

16.0

14.5

4.3

Japan

11.1

12.9

2.7

10.6

12.2

2.0

Luxembourg

7.2

6.9

1.4

6.6

7.4

2.0

Mexico

21.8

21.3

0.7

22.8

22.7

2.1

Netherlands

7.6

7.6

4.6

7.5

9.4

6.4

New Zealand

10.3

13.6

5.8

7.6

11.6

3.8

Norway

3.6

2.9

-0.3

2.8

2.2

-0.9

Portugal

12.6

13.1

3.4

14.0

16.2

6.6

Spain

11.5

..

-3.8

13.5

..

-1.8

Sweden

2.2

3.2

0.7

2.3

3.5

1.0

United Kingdom

14.6

13.6

5.5

13.7

13.0

4.9

United States

18.7

18.4

-2.6

20.4

21.8

0.8

Actual population structure

Unchanged population structure

 

Notes:

 Poverty thresholds at 50% of the median income for the entire population. Re-weighted poverty rates are calculated by holding 

constant the shares of five household groups (single parents with and without work, two or more adult households with children with 
no worker, one and two or more workers) at the mid-1980s level (1990 level in the Czech Republic and Portugal). Patterns highlighted 
are only proximate, as they do not allow for changes in poverty thresholds as household structure varies. Exact years are those 
specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

41. 

The last factor impacting on poverty among households with children is the tax and benefit 

system. While taxes and transfers reduce the extent of relative income poverty among all households, their 
effect is significantly lower in the case of households with children. On average, across 18 OECD 
countries, taxes and transfers lift out of market-income poverty more than half of the persons in households 
without children, but only 44% of those in households with children (Figure 20); the effects of taxes and 
transfers in reducing poverty among households with children is especially low in Japan, Italy and 
Portugal. Indeed, while tax and benefit systems in all OECD countries provide preferential advantages to 
households with children, these advantages are smaller than estimates of the higher household costs of 
larger families that are implicit in the elasticity used in this (and most others) studies to equivalised 
household income.

44

 

                                                      

44

  

The impact of taxes and transfers in reducing poverty also varies across different types of households with 
children. In most OECD countries, taxes and transfers have the largest poverty-reducing effect on 
households with children without work. Changes in patterns of support since the mid-1990s are limited. 

 DELSA/ELSA/WD/SEM(2005)1 

 

37

Figure 20. 

Poverty rates before and after taxes and transfers, households with and without children, 2000 

0%

5%

10%

15%

20%

25%

30%

35%

A

U

S

C

A

N

C

Z

E

D

E

N

F

IN

F

R

A

GE

R

IR

L

IT

A

JP

N

N

L

D

N

Z

L

N

O

R

P

O

R

S

W

E

S

W

I

U

K

G

U

S

A

OE

C

D

 (

1

8

)

Effect of taxes and transfers in
reducing poverty

Poverty rate, disposable income

Households with children

Households without children

 

Note:

 The "light" bar is the poverty rate at the level of disposable income, the "dark" bar is the arithmetic difference 

between the poverty rates at the levels of market and disposable income. Exact years are those specified in the note to 
Table 1. 

Source: 

Calculations from OECD questionnaire on distribution of household incomes. 

5. 

Income adequacy in old age: effects of pension reforms on the retirement-age population 

42. 

Most persons of retirement age (above 65) have withdrawn from the labour market, and depend 

on pensions and capital income for much of their daily living. Differences in the structure of their income, 
as compared to persons of working age, alter the mix of factors that most influence income inequality and 
relative poverty among those belonging to this age group. 

5.1. 

Levels and trends in relative income and poverty among the elderly 

43. 

Recent trends in income distribution and poverty among the elderly need to be described against 

the backdrop of longer-term trends towards significant improvement in their economic situation. Past 
OECD studies have highlighted steady gains in the relative incomes of prime-aged and elderly persons – 
especially those around retirement-age – in all OECD countries, as well as declines of their relative poverty 
rates – both in absolute terms and relative to other age groups (e.g. OECD 1998, OECD 2001, Förster and 
Pearson 2002). Historically, income inequality among the elderly population has also tended to be lower 
than among the population of working-age, and to decrease – or to increase by less – over time.  

44. 

Changes since the mid-1990s suggest some departures from these long-term patterns. Annex 

Table A.6 presents data on the income of individuals by age, relative to that of the entire population, and 
changes in that profile since the mid-1990s and mid-1980s.

45

 While the shape of equivalised disposable 

income by age of individuals is well established – higher relative income until an age of 41 to 50, followed 
by steady declines in later years – changes for the OECD average since 1995 suggest that: 

•

 

Relative incomes of youths (18 to 25) continued to fall as in the previous decade, but at a much 
lower pace. Children (aged less than 18) experienced small increases in their relative income, as 
compared to virtual stability in the previous decade. 

                                                      

45

  

These data on relative incomes take into account changes in population shares: an increase in the share of 
the elderly (with lower income) will depress overall income and suggest an increase in their relative 
position that only reflects their higher weight in the population. To avoid this potential bias, Annex Table 
A.6 uses a constant population structure at the base year to describe changes in relative income by age 
groups. 

DELSA/ELSA/WD/SEM(2005)1 

 

38

•

 

Prime-age adults (aged 41 to 50) experienced significant losses in relative income in the second 
half of the 1990s, which exceeded the gains of the preceding decade.  

•

 

Persons in later working-life (51 to 65) experienced further improvements in their relative 
incomes. 

•

 

Elderly persons (66 to 75) recorded small declines in their relative income, which contrast with 
significant gains in the previous decades, while the relative position of the very elderly (76 and 
over) was broadly stable.  

45. 

Declines in the relative income of persons aged 66 to 75 over the second half of the 1990s 

occurred in about half of the countries reviewed, and were particularly evident in Canada, France, 
Hungary, Luxembourg and Sweden (between 5 and 7 percentage points). In most of these countries, these 
declines followed improvements recorded over the 1980s. Recent trends are more diverse for the very 
elderly: these experienced substantial increases in relative incomes (6 points or more) in Austria, Greece 
and Turkey, but large falls in Finland, Ireland, Italy, Poland and Sweden.

46

 A more generalised decrease in 

relative incomes is experienced by adults aged 41 to 50, with the exception of Finland, Greece, Japan and 
Portugal. “Quasi-replacement rates” for younger senior citizens, 

i.e.

 income levels of those aged 66-75 

relative to that of persons aged 51 to 65, are — in most OECD countries — between 70 and 80%, and close 
to 90% in Austria, Poland and Turkey (Figure 21); and they fell in a majority of OECD countries — in 
particular in Canada, Finland, France, Hungary, Luxembourg and Sweden — in the second half of the 
1990s. 

Figure 21. 

Quasi-replacement rates for persons aged 66 to 75 

0

10

20

30

40

50

60

70

80

90

100

AU

S

AU

T

BE

L

CA

N

C

ZE

D

E

N

FI

N

FR

A

GE

R

GR

C

HU

N

IR

L

IT

A

JP

N

LU

X

M

E

X

N

LD NZ

L

NO

R

P

OL

PO

R

S

PA

SW

E

SW

I

TU

R

UK US

A

OE

C

D 

(1

7)

mid-80s

2000

mid-90s

 

Note:

 Quasi-replacement rates are defined as the mean disposable income of persons aged 66 to 75, relative to the 

mean disposable income of persons aged 51 to 65. For calculating relative income changes, population shares have 
been kept constant at the value recorded at the beginning of the period (mid-1980s, except for Czech Republic, 
Hungary, Poland and Portugal, where this is 1990). Data for Germany refer to old Länder. The OECD average refers to 
the average of 17 countries for which information is available in all three years (except Mexico and Turkey). Exact 
years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes 

                                                      

46

  

Some of the limits in using "current income" for assessing well-being of the elderly are discussed in Box 3. 

 DELSA/ELSA/WD/SEM(2005)1 

 

39

46. 

Differences across countries in the relative income of elderly people reflect differences in both 

family structures and in social protection systems. The role of family structures is highlighted by looking at 
households with a head of retirement-age (Figure 22). Individuals living in these households represent 
roughly 16% of the total population across the 24 countries shown, and their share exceeds by around 8% 
that of all elderly individuals.

47

 Beyond its size, there is much variation in family types. While in Greece, 

Mexico, Portugal and Turkey more than 80% of persons belonging to households with a retirement-age 
head live in households with two or more adults (i.e. couples of two elderly persons, or households 
composed of different generations), this proportion is less than 60% in Nordic countries. On average, 
around a third of all persons living in households with an elderly head in 2000 were living alone, and most 
often this person was not working. Elderly persons living alone are predominantly women, reflecting 
higher life-expectancy in old age and lower probability of entering a new union after separation or death of 
their partner. Across the OECD countries covered, only one out of four persons living in households with 
an older head was in employment. There is, however, significant variation across countries in this share, 
which ranges from less than 7% in the Czech Republic, Finland, Poland and Switzerland, to 40% or more 
in Greece, Ireland, Japan, Mexico, Portugal, Turkey and the United States.

48

 These differences are likely to 

mainly reflect the frequency of "work" (often part-time and occasional) following retirement from a 
person's main career, rather then delayed retirement from a full-time job until later ages.

49

 

Figure 22. 

Family structure among individuals living in households with an older head, 2000 

0%

20%

40%

60%

80%

100%

A

U

S

A

U

T

CAN

C

ZE

DE

N

FI

N

FR

A

GER G

R

C

H

UN

IR

L

IT

A

JP

N

LU

X

M

E

X

NL

D

NZ

L

N

O

R

PO

R

PO

L

S

W

E

SW

I

TU

R

UK

US

A

O

E

C

D 

(2

5)

Single, working

Single, not working

Two adults, both working

Two adults, one working

Two adults, both not working

 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

                                                      

47

  

This reflects the existence, within households with a retirement-age head, of younger spouses and — less 
often — their children. The ratio between persons living in households with a retirement-age head and the 
elderly population varies widely across countries, between values of 80% or less in Hungary and Poland — 
where many elderly live with their offspring — and 130% or more in Ireland and Portugal — where many 
teenagers and young adults continue to live in the parental home. These differences across countries may 
also reflect differences in the definition of "household heads" used in various surveys. 

48

  

This share can reflect different situations, as family members with jobs can be either younger persons 
living in a multi-generational household, or elderly workers themselves. The first pattern is more common 
in Mexico and Southern Europe, the second in Anglo-Saxon countries. 

49

  

The "work status" of different household types is defined on the basis of the presence or absence of 
earnings and self-employment income, rather than on the "self-reported" perceptions of survey respondents 
(with the exception of Germany). 

DELSA/ELSA/WD/SEM(2005)1 

 

40

47. 

Differences across countries in the structure of income of the elderly are also important. Public 

transfers (mainly old-age pensions) and capital income represent the largest components of the disposable 
income of the elderly representing, respectively, about two thirds and close to 30% of disposable income 
across the 17 countries for which information is available on all income sources (Annex Table 8). There is 
some evidence of a reduction in the share of public transfers in total income in about half the OECD 
countries. In several countries, the tax burden on elderly people also declined, in particular in Denmark, the 
Netherlands and New Zealand. The share of earnings (including self-employment income) in total income 
of the elderly remained stable on average, while increasing significantly in Ireland and New Zealand an 
decreasing in the Czech Republic, Hungary, Poland, Portugal and, in particular, Japan. The share of capital 
income slightly increased, in particular in some European countries. 

48. 

Changes in family types and income structure have affected both income inequality and relative 

poverty among the elderly. In a majority of countries (Greece, Ireland, Japan, Mexico, Portugal, 
Switzerland, Turkey and the United States are exceptions), income inequality among the elderly (Gini 
coefficients) remains lower than among the population of working age;

50

 and was broadly constant (on 

average) both in the decade from the mid-1980s to the mid-1990s and in the second half of the 1990s 
(although it increased slightly in Denmark, Finland, Ireland and New Zealand, Figure 23). Their relative 
poverty rate, at around 14% in 2000 across the 23 countries for which longer-trend data are available, 
decreased by 1 ½ points between the mid-1980s and mid-1990s and remained broadly stable during the 
past 5 years

51

 (Figure 24) — although, when excluding Mexico and Turkey, it increased by almost 1 

percentage point over the second half of the 1990s. However, cross-county averages hide great diversity of 
experience, with almost as many countries experiencing a decline in old-age poverty as those witnessing 
increases. In Ireland, the large increase in relative poverty rates among the elderly reflected strong growth 
in median income in real terms, and the failure of elderly income to increase at the same pace, rather than 
lower economic conditions among the elderly 

per se

. Moreover, assessing resources in retirement through 

monetary income may exaggerate poverty risks for the elderly, as it does not take into account lower work-
related expenditures and housing costs following retirement, and higher asset holding among the elderly 
(Box 3). 

                                                      

50

  

This has not always been the case: in the mid-1970s, inequality among the elderly was higher than among 
the working-age population in nine of the ten countries for which information is available; and in the mid-
1980s, this concerned 12 of the 20 countries. 

51

  

Among the elderly, those aged 76 and above have a much higher poverty risk than those aged 66 to 75, in 
almost all OECD countries (see Annex Table A.7) 

 DELSA/ELSA/WD/SEM(2005)1 

 

41

Box 3. Housing costs and poverty outcomes 

 

Most international comparisons of inequality and poverty among the elderly – including those presented in 

this paper – are based on current income. The capacity of households to respond to needs, however, depends on both 
income and expenditure. Housing costs, in particular, are the largest single component of household spending, and are 
a significant determinant of household ability to make ends meet. Government policies affect housing costs through 
both provision of subsidised housing to low-income groups, and through tax advantages to encourage home 
ownership. Some OECD countries (e.g. the United Kingdom) have translated the recognition of the importance of 
housing costs into the adoption of measures of poverty, both before and after housing costs.* 

 

As patterns of home ownership differ both across demographic groups (being higher among the elderly than 

among persons of working age) and countries (higher in Anglo-Saxon and Southern European countries than in Nordic 
and Continental European countries), housing costs significantly affect comparisons of the extent of poverty. An 
indication of the impact of housing costs is provided by the figure below, which uses matched data from income and 
expenditure surveys in Australia and Finland around the mid-1990s (Ritalkallio, 2003). This poverty measures relies on 
thresholds defined both before and after 

actual

 housing costs (expenditures for heating, electricity, water, repairs and 

maintenance, mortgage repayments and interest costs): in the case of the “after-housing-costs” poverty, the threshold 
is defined as 50% of (average) equivalised disposable income less the (average) equivalent rent paid by renters. 
Panel 

a

, which shows poverty rates and gaps (i.e. the shortfall of the disposable income of the poor from the poverty 

line), shows that consideration of actual housing costs reduces poverty rates but increases poverty gaps in both 
Australia and Finland, significantly narrowing differences between the two countries. Panel 

b

, which shows poverty 

rates among individuals of different ages, shows that consideration of housing costs lower poverty rates among the 
elderly, to a point where their poverty risk is lower than that of youths. 

Poverty rates

0%

5%

10%

15%

20%

Australia

Finland

Before  housing costs

After housing costs

Poverty gaps

0%

20%

40%

60%

80%

Australia

Finland

Before  housing costs

After housing costs

 

Australia

0%

10%

20%

30%

40%

0-2

4

25

-34

35

-4

4

45

-54

55

-6

4

65

+

Before Housing
costs
After housing costs

Finland

0%

10%

20%

30%

40%

0-2

4

25

-34

35

-4

4

45

-54

55

-6

4

65

+

Before Housing
costs
After housing costs

 

 

 

Source: Ritakallio (2003). 

* Consideration of housing costs in the measurement of poverty raises the difficult issue of how to measure support provided by 
governments through housing provided at subsidised rents. Changes in the form of government support (i.e. from social housing to 
housing cash benefits), as implemented in several OECD countries over the 1980s and 1990s, could distort assessment of trends in 
poverty (i.e. increases in cash income that do not correspond to improvements in the situation of low-income families

). 

DELSA/ELSA/WD/SEM(2005)1 

 

42

Figure 23. 

Gini coefficient of income inequality among the elderly 

0.00

0.20

0.40

0.60

AU

S

AU

T

BE

L

C

AN CZ

E

DE

N

FI

N

FR

A

GE

R

GR

C

HU

N

IR

L

IT

A

JP

N

LU

X

ME

X

N

LD

N

ZL

N

OR PO

L

PO

R

SP

A

SW

E

SW

I

TU

R

UK USA

OE

C

D

 (2

3)

mid-70s

mid-80s

mid-90s

2000

 

Note:

 Mid-1990s refer to early 1990s for Czech Republic, Hungary and Portugal. OECD (23) excludes Belgium, 

Poland, Spain and Switzerland. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

Figure 24. 

Relative poverty rates among the elderly 

0%

10%

20%

30%

40%

AU

S

AU

T

BE

L

CA

N

C

ZE

DE

N

FIN FR

A

G

ER

GR

C

H

U

N

IRL

IT

A

JP

N

LU

X

ME

X

NLD N

ZL

NO

R

PO

L

PO

R

SP

A

SW

E

SW

I

TU

R

UK US

A

O

ECD

 (2

3)

Mid-1970s

Mid-1980s

Mid-1990s

2000

 

Note:

 The poverty thresholds are set at 50% of the median income for the entire population. Elderly refer to the 

population aged 66 and above. Mid-1990s refer to early 1990s for Czech Republic, Hungary and Portugal. OECD (23) 
excludes Belgium, Poland, Spain and Switzerland. Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

49. 

Relative income poverty among the elderly tends to be concentrated among the very old and 

those living alone. Elderly people living alone and not working are, in all countries, at greater risk of 
poverty than other elderly, and this risk increased in the second half of the 1990s in Australia, Denmark, 
Finland, Germany, Sweden, the United Kingdom and the United States. Most of these are women, often 
widows with limited or no own pension entitlements. As a result of changes in both poverty risks and in 
the shares of the various household types, elderly persons living alone and not working make up the largest 
share of the poor in many OECD countries (Figure 25). However – in Canada, Greece, Ireland, Japan, 
Poland, Portugal, the United Kingdom and the United States – many poor elderly are both living with a 
partner and in a household with earnings. Conversely, in all countries, persons living alone and working, 
and those where both adults work, represent a small proportion of the poor. 

 DELSA/ELSA/WD/SEM(2005)1 

 

43

Figure 25. 

Structure of poverty among persons living in households with a retirement-age head 

0%

5%

10%

15%

20%

25%

30%

AU

S

AU

T

CA

N

CZ

E

D

EN

FI

N

FR

A

GE

R

GR

C

IR

L

IT

A

JP

N

LU

X

ME

X

NL

D

NZ

L

N

OR PO

L

PO

R

SP

A

SW

E

SW

I

TU

R

UK

US

OE

C

D

 (2

1)

Tw o adults, no one w orking
Tw o adults, one w orking
Tw o adults, both w orking

Single, not w orking
Single, w orking

2000

mid-90s

 

Note:

 The height of each bar represents the poverty rate (using a 50% threshold) of persons living in households with a 

retirement-age head. Data for Germany refer to old Länder. "Two adults" refer to households with two or more adults. 
Exact years are those specified in the note to Table 1. 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

5.2. 

Public pension systems and their impacts on the elderly population 

50. 

Given their weight in the disposable income of elderly people, public pensions play a major role 

in shaping income adequacy and poverty risks for this group of the population. When considered together, 
public transfers and taxes reduced inequality and poverty among the elderly in 2000 by more than they do 
with respect to the population of working-age. However, in a majority of countries this effect weakened 
over the second half of the 1990s (with the exceptions of the Czech Republic, France, Italy, Portugal and 
Sweden with respect to income inequality; and the same countries except Sweden but including the 
Netherlands and Norway with respect to poverty). 

51. 

Outcomes in terms of relative poverty among the elderly are affected by several features of 

public pension systems. The amount of spending on old-age pensions (public and private mandatory 
spending), however, does little by itself to influence poverty among the retirement-age population. 
Figure 26 highlights no relationship across countries (panel 

a

): in fact, some of the countries with higher 

spending on old-age pensions (e.g. Italy, France and Germany) experience higher poverty rates among the 
elderly than countries with much lower spending levels. This lack of association between pension spending 
and poverty outcomes reflects the importance of earnings-related pensions, and differences in the ceilings 
that are applied to high earnings. Indeed, where pension benefits increase in line with previous earnings, 
they may have a regressive impact on income distribution and relative poverty among the elderly. 

52. 

Other features of pension systems are likely to matter more for poverty outcomes among the 

elderly than aggregate spending, although the co-existence in a point in time of different rules applying to 
various groups of persons make it difficult to disentangle their importance.

52

 Among the features that are 

                                                      

52

  

Among these pension parameters is whether benefits are indexed to prices, earnings, or some combination 
of the two. In order to control expenditures, several OECD countries moved over the 1990s from wage to 
price indexation, a move which may tend to increase relative poverty among the elderly over time. To 

DELSA/ELSA/WD/SEM(2005)1 

 

44

most obviously related to poverty outcomes are pension floors provided by public pension and welfare 
systems. OECD countries vary significantly in the tools they use to minimise poverty risks among the 
elderly: some rely on “minimum pensions”, limited to persons with past contributory records, others on 
"basic" pensions, provided to all elderly citizens irrespectively of past contributions (but often subject to 
residence and means tests), and others yet on the general social assistance schemes applying to the entire 
population. In general, countries where pension floors – expressed as a ratio of the poverty line – are 
higher tend to display lower relative poverty rates among the elderly (panel b); the relation is however 
weak, as other characteristics — 

e.g.

 the extent to which the elderly share in the resources of extended 

households — impact on relative poverty in old-age. 

Figure 26. 

Relative poverty among the elderly and pension systems 

Poverty among the elderly and pension spending 

Poverty among the elderly and pension floors

US

UK

SZ

SW

PG

PL

NW

NZ

NL

LX

JP

IT

IR

GR

GE

FR

FN

DK

CZ

CA

AT

AL

0

10

20

30

40

0

5

10

15

20

Pension spending

Pov

e

rt

y

 rat

e

 am

ong t

he el

de

rl

y

US

UK

SZ

SW

PG

PL

NW

NZ

NL

LX

JP

IT

IR

GR

GE

FR

FN

DK

CZ

CA

AT

AL

0

10

20

30

40

20

40

60

80

100

120

140

Pension floor

Pov

e

rt

y

 rat

e

 am

ong t

he el

de

rl

y

 

Note:

 Public and mandatory private social spending for old-age and survivor benefits, as a share of GDP. Pension 

floors, expressed as a percentage of the 50% median income threshold, refer to the levels of “basic” or “targeted” 
pensions in first-tier pension systems of OECD countries.  

Source:

 OECD income distribution questionnaire, social expenditure and pension-monitoring databases. 

53. 

Despite the lack of a significant association between pension spending and poverty across 

countries, changes in the generosity of public transfers and taxes have played the largest role in shaping 
changes in poverty risks among the elderly within individual countries. Figure 27 applies the same shift-
share analysis described in Section 3 to changes (over the second half of the 1990s) in relative poverty 
rates among persons living in households with a retirement-age head, broken down by whether individuals 
are living alone or with other adults, and by the work status of household members. It suggests that, in all 
countries where changes in relative poverty rates are significant, they have been driven by changes in taxes 
and public transfers received by this group. Changes in the population structure have generally been minor, 
while changes in market income have tended to reduce risks of poverty in Australia, Canada and Finland, 
and to increase it in Japan and the United States. As in the case of the working-age population, a positive 
contribution of taxes and public transfers to higher relative poverty among the elderly may reflect increases 
in real benefits that lag those of median income, rather than declines in the real value of benefits. 

                                                                                                                                                                             

offset this effect, some countries have introduced specific measures to protect those more exposed to 
poverty risks (e.g. the Minimum Income Guarantee in the United Kingdom). 

 DELSA/ELSA/WD/SEM(2005)1 

 

45

Figure 27. 

Changes in relative poverty rates among households with a retirement age head by 

components, mid-1990 to 2000 

-8%

-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

AU

S

CA

N

D

EN

FI

N

FR

A

G

ER

IT

A

JPN

N

LD

NZ

L

NO

R

PO

R

SWE

UK

US

A

Changes in population structure
Changes in taxes and transfers
Changes in market-income poverty
Changes in poverty rates

 

Note:

 Data are based on a shift-share analysis applied to population living in households with a head of retirement-

age, broken by work attachment of household members (i.e. distinguishing between households with no workers, with 
one adult working, and with two or more adults working). The sum of the three components (shown as bars) is equal to 
the total change in poverty rate (shown as a "diamond"). Exact years are those specified in the note to Table 1. 

Source:

 Calculations based on the Calculations from OECD questionnaire on distribution of household incomes. 

5.3. 

Distributive patterns of public transfers and private capital income 

54. 

The distribution of public transfers (mainly old-age pensions

53

) and capital income (which 

includes private and occupational pensions) differs in important ways.

54

 Because of these differences, 

                                                      

53

  

On average, about 90% of public transfers going to the elderly are made up by old-age pensions, with 
somewhat lower shares in the Anglo-Saxon countries and close to 100% in Continental European 
countries. These differences, however, may partly reflect differences in classification of income transfers 
across countries. 

54

  

The measurement of income from occupational and private pension schemes raises difficult issues. 
According to the recommendations of the “Expert Group on Household Income Statistics” (the Canberra 
Group), current transfers received by households (“payments and receipts which are made without a 
matching 

quid pro quo 

in the period in which they are paid/received”) should include “employment-related 

pensions and other insurance benefits paid from private employer’s schemes” and “government schemes 
run entirely for benefit of government employees”, while excluding “lump-sum retirement payouts” 
(recorded as capital transfers) and “benefits from private insurance schemes where... participation in the 
scheme is entirely at the discretion of the contributor” (the latter being recorded as either non-life 
insurance, outside the scope of income, or as property income, when akin to payments from an annuity or 
similar investment instrument). In practice, it is unclear to what extent these conventions are strictly 
followed. As a result, because of the wide variety of private pension arrangements in OECD countries, 
benefits from private pension funds that are substantially similar may be classified differently by various 
countries. For the purpose of this analysis, private and occupational pensions are recorded among “capital 
incomes", although there are exceptions (Austria, the Czech Republic, Hungary, Italy and Mexico; in these 
countries, however, private pensions remain relatively under-developed). 

DELSA/ELSA/WD/SEM(2005)1 

 

46

reforms aimed at increasing the diversification of income sources in retirement may have distributive 
implications. One way to assess distributive patterns of public transfers and capital income is through 
“pseudo-Lorenz” curves. Figure 28 show these curves for public transfers (black dotted line) and capital 
income (black continuous line) among the retirement-age population, as well as that for disposable income 
of the working-age population (dashed line). The latter displays a familiar shape: on average, around 3% of 
disposable incomes accrue to working-age individuals in the lowest decile, 8% to those in the two lowest 
deciles, and around one fourth to persons in the highest decile of the working-age population. 

55. 

The extent to which pseudo-Lorenz curves for public transfers depart from the pattern for 

disposable income of working-age persons varies across countries. In countries where public pension 
systems are mainly earnings-related, as in the case of France, the (dotted) line for public transfers follows 
closely the (dashed) line for disposable incomes of persons of working age. In Canada, Ireland, the 
Netherlands, New Zealand and the United Kingdom, on the other side, pseudo-Lorenz curves for public 
transfers are close to the 45° line, as all persons receive a similar public pension. In Finland, public 
pensions are distributed progressively (the pseudo-Lorenz curve lies above the 45°line), as the share that 
goes to individuals in the lowest deciles exceeds that accruing to those at the top of the distribution. In 
Australia and Ireland, the middle deciles seem to profit relatively more from public pensions than both the 
poor and the rich. Despite these differences, however, public pensions are – on average and in all countries 
– more equally distributed than disposable income of the working-age population. 

56. 

Private capital income of the elderly is far more unequally distributed than public transfers. In 

Italy, for instance, more than 80% of this income source goes to the top 20 % of the retirement-age 
population and this percentage exceeds 65% in the Czech Republic, Greece, Luxembourg and New 
Zealand. On average, a little over 10% of private capital income accrues to the poorer half of elderly 
people, while more than 40% accrues to the top decile. In general, over the past five years, the share of 
private pensions and capital incomes in total incomes of the elderly increased in about half of the countries 
under review. In the case of Finland, this increase took place about equally across all income groups; in 
other countries, however, this increase affected primarily richer and middle income groups (

e.g.

 Denmark, 

Germany, Luxembourg, Ireland, Italy, Japan and the United Kingdom, Annex Table A.8). 

57. 

Evidence on the distributive pattern of public and private income sources underscores the extent 

to which resources of older people continue to be highly differentiated across the income ladder. On 
average, across the OECD countries reviewed in this paper, public transfers still account for almost 

all

 of 

the disposable income of the bottom quintile of the elderly population (Figure 29) and close to 80% of the 
incomes of the middle 60% of the distribution. Only individuals in the top quintile of the distribution enjoy 
a ‘balanced’ mix of income streams, where public transfers, private capital income, and earnings contribute 
about equally. Country differences around this average, however, are significant. In Finland, for example, 
the share of occupational pensions – which are included here among capital income – constitutes some 
70% of the disposable income of the elderly, as compared to only 20% for public pensions, and are 
distributed much more equally than in other countries. This reflects the more institutional role of such 
pensions in the Finnish retirement system – and their management by social security institutions. 

 DELSA/ELSA/WD/SEM(2005)1 

 

47

Figure 28. 

Distributive shape of public transfers and private capital income to the elderly and of disposable 

income of the working-age population, 2000 

OECD Average

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

public transfers

private capital

net income (work/age)

Australia

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Canada

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Hungary

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Ireland

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Italy

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

France

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Finland

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Denmark

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Germany

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Greece

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Czech Republic

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

 

DELSA/ELSA/WD/SEM(2005)1 

 

48

Figure 28. 

Distributive shape of public transfers and private capital income to the elderly and of 

disposable income of the working-age population, 2000 (cont.) 

Japan

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

public transfers

private capital

net income (work/age)

Netherlands

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

New Zealand

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Norway

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Poland

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Portugal

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Switzerland

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Sweden

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

United Kingdom

0

10

20

30

40

50

60

70

80

90

100

United States

0

10

20

30

40

50

60

70

80

90

100

Luxembourg

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

 DELSA/ELSA/WD/SEM(2005)1 

 

49

Figure 29. 

Income composition among the older population by income groups, OECD average 2000 

-30%

0%

30%

60%

90%

120%

all income groups

bottom 20% 

middle 60%

top 20%

Earnings

Private capital income

Public transfers

Taxes

Earnings

Taxes

Public 
transfers

Private, 
capital 
income

 

Source:

 Calculations from OECD questionnaire on distribution of household incomes. 

DELSA/ELSA/WD/SEM(2005)1 

 

50

BIBLIOGRAPHY 

Atkinson A., L. Rainwater and T. Smeeding (1995), 

Income Distribution in OECD Countries. Evidence 

from the Luxembourg Income Study, 

OECD, Paris. 

Atkinson, A. B. (1998), “Social Exclusion, Poverty and Unemployment”, in Atkinson A. B. and J. Hills 

(eds.), 

Exclusion, Employment and Opportunity, 

Centre for Analysis of Social Exclusion, London 

School of Economics, Case Paper No. 4, London.  

Atkinson A. B. (2002), "Income Inequality in OECD Countries: Data and Explanations", CESifo Working 

Paper No. 881. 

Bradshaw J. and N. Finch (2003), “Overlaps in Dimensions of Poverty”, 

Journal of Social Policy, 

32,4, 

Cambridge. 

Brewer M., A. Goodman, M. Myck, J, Shaw and A. Shephard (2004), "Poverty and Inequality in Britain: 

2004", The Institute For Fiscal Studies, Commentary No. 96, London. 

Burniaux J.M., T.-T. Dang, D. Fore, M.F. Förster, M. Mira d'Ercole and H. Oxley (1998), "Income 

Distribution and Poverty in Selected OECD Countries", 

OECD Economics Department Working 

Paper, 

No 189, March. 

Caner A. and E. N. Wolff (2004), "Asset Poverty in the United States", 

Public Policy Brief 

No. 76, The 

Levy Economic Institute of Bard College, Annandale-on-Hudson. 

Census Bureau (2003), “Supplemental Measures of Material Well-being : Expenditures, Consumption, and 

Poverty, 1998 and 2001”, 

Current Population Reports, 

Special Studies, September. 

Cohen-Solan M. and M. Leliévre (2003), « Niveau de vie et risqué de pauvreté parmi les retraités des pays 

européens », 

Études et Résultats, 

Direction de la Recherche des Études de l’Évaluation et des 

Statistiques, Ministère des Affaires sociales, du Travail et de la Solidarité, January, Paris. 

Förster, M (1994), “Measurement of Low Incomes and Poverty in a Perspective of International 

Comparisons”. 

OECD Labour Market and Social Policy Occasional Paper,

 No. 14. Paris. 

Förster, M. and M. Pearson (2002), “Income Distribution and Poverty in the OECD Area: Trends and 

Driving Forces”, 

OECD Economic Studies

, No.34, Paris. 

Förster, M., F. Mass and B. Marin 

eds.

 (2003), “Understanding Social Inclusion in a Larger Europe: An 

Open Debate”. EUROSOCIAL 71/03, European Centre for Social Welfare Policy and Research, 
Vienna. 

Freeman R. B. (2001), “The Rising Tide Lifts….?”, National Bureau of Economic Research Working 

Paper 8155, March, Cambridge. 

Gallie D. and S. Paugman (2002), 

Social Precarity and Social Integration, 

European Commission, 

Employment & social affairs, October, Brussels. 

 DELSA/ELSA/WD/SEM(2005)1 

 

51

Greenstein R. and I. Shapiro (2003), “The New, Definitive CBO Data on Income and Tax Trends”, Center 

on Budget and Policy Priorities, September, Washington D.C. 

Goodman A., M. Myck and A. Shephard (2003), “Sharing in the Nation’s Prosperity? Pensioner Poverty in 

Britain”, Institute for Fiscal Studies, London. 

Krueger A. and F. Perri (2002), “Does Income Inequality Lead to Consumption Inequality?", National 

Bureau of Economic Research Working Paper No. 9202, Cambridge. 

OECD (1998), 

Maintaining Prosperity in an Ageing Society, 

OECD, Paris. 

OECD (2001), 

Ageing and Income, 

OECD, Paris. 

OECD (2002), 

OECD Employment Outlook 2002, 

OECD, Paris. 

OECD (2003), "More and Better Jobs? Aggregate Performance During the Past Decade", 

OECD 

Employment Outlook, 

OECD, Paris. 

Siminski, P. Saunders, S. Waseem and B. Bradbury (2003), "Reviewing the Intertemporal Consistency of 

ABS Household Income Data with External Aggregates", 

Australian Economic Review

, Vol. 33, 

Issue 3, September.  

Ritakallio T.M. (2003), "The Importance of Housing Costs in Cross-National Comparisons of Welfare 

(State) Outcomes", Department of Social Policy, University of Turku, Turku.  

Teekens, R. and A. Zaidi (1990), “Relative and absolute poverty in the European Community”, 

Analysing 

Poverty in the European Community

, EUROSTAT, Luxemburg. 

UNICEF (2000), “A League Table of Child Poverty in Rich Countries”, Innocenti Report Card, Issue 

No. 1, June, Florence. 

Welniak E. J. (2003), "Measuring Household Income Inequality Using the CPS", 

Special Studies in 

Federal Tax Statistics, 

Annual Meetings of the American Statistical Association. 

DELSA/ELSA/WD/SEM(2005)1 

 

52

ANNEX 1. CHARACTERISTICS OF THE DATA USED IN THE ANALYSIS 

58. 

This annex describes the main features of the data used in this paper: both those that are common 

to the datasets of different countries and those that differ across countries and that are likely to distort 
comparisons. The basic concept underlying the data and indicators presented in this paper is that of 

household disposable income

. To account for possible scale economies in consumption, household income 

is "equivalised" using the square root of household size.

55

 Information is presented for various breakdowns 

of individuals and households: age of individuals, in the first case; age of the household head (below and 
above 65), presence of children (persons aged below 18), presence of other adults, and work status of 
household members (allowing to distinguish between households with zero, one, and two or more 
workers), in the second. 

59. 

Because of the emphasis, in this paper, on changes

 

in income inequality and poverty, an effort 

was made to improve data comparability over time for individual countries. To that effect, in cases of 
major changes in national survey methods (e.g. Sweden and Canada in 1995)

56

, data have been collected 

on both the “old” and “new” bases, so as to allow chain-linking of various indicators. The use of a common 
questionnaire and methodology (e.g. in terms of equivalence of scales, income components, and poverty 
thresholds) also allows better comparisons of levels

 

of the different variables across countries. However, 

the national data used in this paper differ in certain aspects that escape “standardisation” across countries, 
and this may affect cross-country comparisons. Detailed characteristics of country data are shown in 
Annex Table 1. Some of the main features that may affect comparisons across countries and time include 
the following: 

•

 

Differences in the definition of households

. For most countries, households refer to a group of 

people living in the same house and having common provisions for essential items. In some 
countries, however, children above a given age are considered as a separate household unit (e.g. 
Sweden until the mid-1990s), even if living in their parents’ home. More restrictive definitions of 
“household” will tend to reduce household size and equivalised income (and increase poverty 
rates) relative to other countries. 

•

 

Period over which income is assessed

. Data generally refer to income in the year preceding the 

interview, with the exception of Austria (date relate to monthly income) and Spain (data relate to 
quarterly income). Even for countries where annual income data are shown, however, income 
may be assessed over a shorter reference period and then converted to an “annual equivalent” 
(e.g., data refer to weekly income in the United Kingdom, while in Australia the reference period 
differs across income sources). Countries using shorter reference periods to measure income will 
generally display higher poverty rates because of the greater volatility of weekly income and 
higher probability of recording periods of “temporary” income shortfalls (i.e. poverty rates are 
likely to be higher than would have been found had income been recorded on an annual basis). 

                                                      

55

  

This implies that, to keep economic well-being unchanged, household income needs to increase by 41% 
when a second member joins the household, by a further 32% for a third one, and by 26% for the fourth.  

56

  

In the case of Canada, the 

Survey of Labour and Income Dynamics 

is used in place of the 

Survey of 

Consumer Finances 

starting in 1995. In the case of Sweden, the definition of households changed in 1995.  

 DELSA/ELSA/WD/SEM(2005)1 

 

53

•

 

Gross and net income.

 All income components are generally reported on a “gross” basis, i.e. 

before deduction of direct and payroll taxes (social security contributions) paid by individuals 
and households. Exceptions are Austria, the Czech Republic (for 2000), Greece, Hungary, 
Mexico, Poland, Spain and Turkey, where income components are recorded on a “net” basis (i.e. 
information on taxes paid is not available). Even for countries where taxes are separately 
identified, however, there are differences in the way these are computed, with some countries 
relying on data as reported by respondents (e.g. Japan, France), others on values from tax records 
(e.g. Belgium, Denmark), and others still on values “imputed” though microsimulation models 
applied to individual records (e.g. Italy, New Zealand, Portugal). In the latter case, cross-country 
differences in the details and assumptions used (e.g. with respect to tax evasion) may affect the 
comparability of results.

57

 

•

 

Income components. 

The data generally distinguish between earnings (broken down into the 

earnings of the household head, of the spouse and of other household member)

58

; self-

employment income; capital income (rents, dividends and interest); and current transfers received 
by households.

59

 Capital income is generally limited to income paid in cash (however, in the case 

of Denmark, Germany and Turkey, imputed rents of home-owners are included). Current 
transfers refer to cash transfers paid by government to individuals and households. Because of the 
exclusion of in-kind transfers, changes in the nature of government support (e.g. from provision 
of social housing at subsidised rates to housing benefits paid in cash) may distort results. Private 
transfers are generally included with capital incomes, with a few exceptions

60

. In some cases (e.g. 

the United States), microsimulation models have been used to impute values of some cash 
benefits (Earned Income Tax Credit, food stamps and housing benefits) that are not recorded in 
surveys. 

•

 

Recording of private pensions.

 There are large differences across countries in terms of the nature 

and institutional arrangements governing private pensions. These differences relate both to their 
mandatory or voluntary character, and to the nature of the agencies that are responsible for their 
management and administration (i.e. in some cases, they may be part of the social security 
administration, while in others they may be fully private). Also, private pensions are sometimes 
not separately identified in the household surveys of some countries. Because of these 
differences, private pensions that are substantially similar may be recorded differently across 
countries: as part of “capital income" in most countries, or as part of "public transfers” in Austria, 
the Czech Republic, Germany (prior to 2001)

61

, Hungary, Italy and Mexico (in these countries, 

however, private pensions are very small). No attempt has been made to correct for these 
discrepancies in classifications. 

                                                      

57

  

For example, because of the complexities of the German tax system, only standard deductions are 
considered by the German micro-simulation model used to generate the data presented in this paper. 

58

  

This breakdown of earnings, however, is not available in the case of Hungary, Norway and a few other 
countries. 

59

  

Negative income components are generally bottom coded to zero. 

60

  

In Sweden, for instance, part of alimonies are included in current government transfers as, for a number of 
families, there is a public intermediary between the parents. 

61

  

In Germany, private pensions are included in "capital income" only in 2001; for previous years, when 
private pensions were not separately identified in the survey, they are included in "public transfers". 

DELSA/ELSA/WD/SEM(2005)1 

 

54

Annex Table A.1. National sources and data adjustments 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Australia 

Household 
Expenditure 
Survey 

1984 
1989 
1994 
1999 

June to June, except 
1984 (calendar year)  

About 8,900 
households and 
78% response rate   

Persons living together 
in a private dwelling and 
having common 
provision for food and 
other essentials of living 

Current income 
from wages and 
salaries and 
government 
transfers. Annual 
income from other 
sources is pro-rated 
to a weekly 
equivalent 

Personal income 
taxes were 
collected until 1993-
94 and imputed 
thereafter 

Negative income 
bottom coded to 
zero. 
Features of the 
1975/76 survey 
skew the 
distribution away 
from the bottom 
end, distorting 
analysis of changes 
over time.  

Austria 

Micro 
census 

1983 
1993 
1999 

 

67% for income 
questions 

 Net 

income. 

Incomes are 
monthly averages. 
Income data 
exclude capital 
incomes and self-
employment 
incomes (if the self-
employed person is 
the household 
head) 

All income data are 
collected net of 
taxes and social 
security 
contributions and 
those are not 
imputed. 

Income 
components asked 
on individual level. 
Imputation of non-
response (1993, 
1999).  

Canada 

Survey of 
Consumer 
Finances 
 
Survey of 
Labour and 
Income 
Dynamics 

1975 
1985 
1995  
 
1995 
2000 

Income over the full 
calendar year 

About 30,000 
households and 
85% response rate   

A person, or group of 
persons, residing in a 
dwelling 

Market income and 
government 
benefits, net of 
income taxes 

Amounts received 
through some 
government 
transfers derived 
from other sources. 
Survey data on 
taxes are complete 
and do not require 
imputation 

Income items which 
were coded as non-
response in SLID 
were set to zero 

Czech 
Republic 

Micro 
census 
 
 

1992 
1996 
2002 

 About 

38,000 

dwellings and 76% 
response rate  

Private households 

Annual disposable 
income in each 
year. For 1992 no 
information on 
"taxes" is available 

 

No imputation, no 
negative incomes. 

 

DELSA/ELSA/WD/SEM(2005)1 

 

55

 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Denmark 

The Danish 
Law Model 
System 

1983 
1994 
2000 

Annual income  

About 170,000 
persons. For all 
these persons, 
income data are 
based on registers.   

Couples include both 
married and cohabitating 
partners. Children above 
17 living at home are 
considered as separate 
households  

Disposable income 
net of personal 
taxes and 
contributions to 
private pension 
schemes 

Data are derived 
from several tax 
and benefits 
registers (the 
Danish Law Model 
is not a survey). 

Negative incomes 
are set to zero. 
Payments from 
private pension 
schemes are 
included in capital 
income 

Finland 

Household 
Budget 
Survey 
Finnish 
Income 
Distribution 
Survey 
 

1976 
 
 
1986 
1995 
2000 

 Around 

13,000 

households and 
75% response rate   

Persons living in private 
households 

 

 

 

France 

Family 
Budget 
Survey  

1984 
1989 
1994 
2000 
 

Annual income if the 
12 moths preceding 
the survey 

Around 10,000 
households and 
70% response rate 

Persons living in the 
same housing unit 

Values for individual 
income components 
are aggregated into 
total income 

Income, housing 
and property taxes 
as declared in the 
survey. Social 
security 
contributions paid 
by workers are 
excluded. Capital 
income in 2000 
estimated by 
applying an 
average rate of 
return to survey-
measure of asset 
holdings 
 

Negative incomes 
are replaced with 
values over the 
three preceding 
years. Missing data 
are imputed. 

Germany 

Socio-
Economic 
Panel 

1984 
1989 
1994 
2001 (old 
länder) 
 
1994 
2001 (all 
länder) 
 
 

Annual income in the 
year preceding the 
survey 

Around 13,000 
households, initial 
response rate over 
50%, cross-
sectional response 
rate over 95% 

People living together 
and sharing their income 

Self-employment 
income is included 
in "earnings", 
occupational 
pensions in “current 
transfers”, private 
pensions in “capital 
income” 

Direct taxes and 
social-security 
contributions paid 
by workers are 
imputed from micro-
simulation models 

Income below the 
social minimum of 
DM 5000 per year 
is excluded. 

DELSA/ELSA/WD/SEM(2005)1 

 

56

 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Greece 

Household 
Budget 
Survey 

1974 
1988 
1994 
1999 

 96% 

93% 
86% 
84% 

Private households  

All incomes in cash, 
net of taxes and 
social insurance 
contributions 

 Missing 

incomes 

- 

households that did 
not provide income 
information - 
excluded from the 
sample 
 

Hungary 

Hungarian 
Household 
Panel 
 
Household 
Monitor 
Survey 
 

1991 
1995 
 
 
2000 

From April of the year 
in question to 
following March  

About 2,000 
households and 
67% response rate   

Private households  

Incomes in cash, 
net of taxes and 
social insurance 
contributions 

 No 

negative 

incomes. Missing 
incomes excluded 
in 1992, partly 
replaced by 
imputed values in 
1996 and 2001 

Ireland 

Survey of 
Income 
Distribution 
Living in 
Ireland 
Survey 

1987 
1994 
2000 

Current weekly 
income  

About 3,500 
households and 
69% response rate   

Persons living together, 
sharing budget 
arrangements, and 
meeting at least once 
per week for meals. 
Persons temporarily 
absent and living in 
collective households 
also included 
 

Income excluding 
non-monetary 
components 

 

 

Italy  

Bank of Italy 
Survey of 
Household 
Income and 
Wealth

  

1984 
1991 
1993 
1995 
2000 

Annual income  

About 8,000 
households and 
38% response rate   

Persons living in the 
same dwelling and 
contributing part of their 
income to the household 

Disposable income 
Income from 
financial assets (not 
available in 1984), 
“gifts” and family 
benefits excluded in 
all years. 

Gross income data 
based on a micro-
simulation model to 
estimate income 
taxes and social 
security 
contributions paid 
by workers 

Micro-simulation 
models used for 
1995 and 2000 
differ slightly from 
that used for 
previous years. 
Private transfers 
and pensions 
(minor items in 
Italy) are included in 
"public transfers" 
 

 

DELSA/ELSA/WD/SEM(2005)1 

 

57

 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Japan 

Comprehen
sive Survey 
of Living 
Condition of 
the People 
on Health 
and Welfare 

1985 
1995 
2000 

Annual income in the 
year preceding the 
survey  

About 32,000 
households and 
80% response rate   
 

Persons sharing the 
same housing unit and 
livelihood.  
Data exclude 
households headed by a 
person aged less than 
17, and all individuals 
whose age is not 
recorded 
 

Gross income 
All income items as 
reported in the 
survey 

Negative 
disposable income 
allowed, negative 
values for income 
components set to 
zero 

Persons with 
income three times 
larger than the 
standard deviation 
excluded (1.6% of 
all persons in 1995 
and 1.3% in 2000) 

Luxembourg 

Panel 
Socio-
Economiqu
e Liewen zu 
Lëtzebuerg 

1986/87 
1996 
2001 

Annual income 

About 2,300 
households and 
57% response rate 

 

All types of incomes 
in cash, net of taxes 
and social 
insurance 
contributions 

 

Include all private 
households in 
which at least one 
person belongs to 
national social 
security system 
(around 97% of the 
population). 
Negative incomes 
set to zero 

Mexico 

Survey of 
Household 
Income and 
Expenditure 

1984 
1994 
2002 

Income in the 3

rd

 

quarter of each year. 

About 20,000 
households and 
85% response rate 
in 2002 

Persons normally 
sharing a housing unit 
and having common 
expenditure for food 

Quarterly cash 
income net of direct 
taxes and soc. 
security 
contributions. 
Income items as 
reported in the 
survey 

Private pensions 
cannot be 
separately identified 
and are included in 
"public transfers" 

 

Netherlands 

Income 
Panel 
Survey 

1977 
1985 
1990 
1995 
2000 

 About 

82,000 

households and 
100% response rate   
(data from tax 
registers) 
 

Persons living at the 
same dwelling and 
running a common 
budget 

  Data 

exclude 

persons with zero 
or negative 
disposable 
household income 

New Zealand 

Household 
Economic 
Survey 

1986 
1991 
1996 
2001 

June to June in all 
years except 2001 
(June to March 
period) 

About 2,800 
households and 
73% response rate   

Persons sharing a 
private dwelling and 
normally spending four 
or more nights a week in 
it 

Disposable income 
All receipts received 
regularly or of a 
recurring nature 

Direct taxes and 
social security 
contributions paid 
by households 
imputed through 
microsimulation 
models 

Missing incomes 
are treated as zeros 

DELSA/ELSA/WD/SEM(2005)1 

 

58

 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Norway 

The Income 
Distribution 
Survey 

1986 
1995 
2000 

Calendar year 

About 

13,000 

households and 
75% response rate   

All individuals in the 
same dwelling having 
common housekeeping 

Annual disposable 
income. All income 
data collected from 
registers 

 

No missing data, 
negative income set 
to zero. Non-
respondents 
included in sample 
with missing data 
replaced by data 
from registers 
 

Poland 

Consortium 
for 
Household 
Economic 
Research 
Panel 
Database 
 

1995 
2000 

 About 

7,700 

households and 
100% response rate  

 Annual 

disposable 

income 

 Missing 

values 

imputed or set to 
zero 

Portugal 

Household 
Budget 
Survey  

1980 
1990 
1995 
2000 
 

Income in the year 
preceding the 
interview  

About 10,000 
households and 
response rate close 
to 100% in all years 

Persons living in the 
same dwelling 

Gross income, 
excluding all non-
monetary 
components 

 

 

Spain 

Continuous 
survey of 
household 
budgets 

 

1985 
1990 
1995 

Income in the 2

nd

 

quarter of each year 

About 3,200 
households and 
90% response rate 
in 1995 

Persons sharing a 
common budget 

Quarterly 
disposable income 

 All 

income 

components are on 
a net basis  

Sweden 

Income 
Distribution 
Survey 

1975 
1983 
1991 
1995 
2000 

Calendar year 

About 14,500 
households and 
75% response rate. 
Data based on tax 
registers, 
complemented with 
survey data.  
 

All individuals living 
together and sharing 
household resources. 

Annual disposable 
income. All income 
data collected from 
tax records 

 No 

missing 

incomes, negative 
incomes included, 
households with 
negative disposable 
incomes deleted. 

Switzerland 

Income and 
Consumption 
Survey 
 

1998 
2000 
2001 

 About 

3,700 

households and 
35% response rate   

 

Monthly gross and 
net income 

 No 

negative 

incomes, missing 
incomes imputed 

 

DELSA/ELSA/WD/SEM(2005)1 

 

59

 

Country Survey-

source 

Year to 

which 

income 

refers 

Period over which 

income is assessed 

Sample size and 

response rate 

most recent year 

Definition of 

households 

Recorded income 

Integration of 

survey data 

Other data 

features 

Turkey 

Household 
Income and 
Consumption 
Survey 
 

1987 
1994 
2002 

 

 

 

 

 

 

United 
Kingdom 

Family 
Expenditure 
Survey 

1975 
1985 
1991 
1995 
2000 

Income at the time of 
the interview for most 
items (over the 
previous 12 months 
for capital and self-
employment income) 
 

About 10,000 
households and 
60% response rate   

Persons living in the 
same dwelling 

Weekly gross 
income 

 Missing 

values 

excluded, negative 
values included 

United States 

Current 
Population 
Survey 

1974 
1984 
1995 
2000 

Year preceding the 
March interview 

About 50,000 
households and 
95% response rate   

Persons occupying a 
housing unit. 

Gross annual 
income 

Model-based 
estimates of taxes 
paid by each 
household and in-
kind government 
benefits added to 
survey data of 
gross annual 
income 

Negative values 
allowed when below 
$10 

DELSA/ELSA/WD/SEM(2005)1 

 

60

ANNEX 2. POVERTY THRESHOLDS USED IN THE ANALYSIS 

60. 

The thresholds used in this paper to measure relative poverty are based on percentages (50% and 

60%) of the median equivalised disposable income of all individuals, where household disposable income 
is equivalised using the square root of household size. These thresholds, for the latest available year, are 
shown in Annex Table 2. For example, for the United States, median equivalised disposable income of all 
individuals in 2000 is USD 23,954 per year (in current prices), and the poverty threshold (at 50% of the 
median) is USD 11,977 (second column). To ease interpretation, thresholds can also be expressed on a 
“non-equivalised” basis (i.e. with no adjustment for household size). A person living alone (whose 
“equivalised” and “non-equivalised” income are identical) will be considered as (relative) “poor” when his 
disposable income is less than USD 11,977 (column 3), while a household composed of two adults and one 
child will be counted as poor when its total household income is less than USD 20,745 (column 5). It 
should be noted, however, that these “non-equivalised” thresholds differ from those that would be obtained 
by applying a different elasticity for household size to micro data (

e.g.

 if all individuals had been ranked by 

their household income, with no adjustment for household size, median income and poverty thresholds 
would have differed from those shown below). Annex Table 2 also shows (in the last column) values of the 
poverty thresholds for a singe adult as a percentage of the take-home pay of an average production worker; 
in most OECD countries, poverty thresholds for a single person are in a range between 40% and 60% of 
the take-home pay of an average production worker, although they are higher (Hungary) and lower 
(Australia, Japan, Mexico, New Zealand, Poland and Turkey) in some countries.  

Annex Table 2. Values of the thresholds used in this paper for measuring 

relative poverty at half of median disposable income 

Single adult

Single adult with 

one child

Two adults with one 

child

Two adults with two 

children

A ustralia

1999

10,617

                    

10,617

             

15,015

                 

18,389

                   

21,234

                          

36%

A ustria

1999

104,972

                  

104,972

           

148,453

               

181,817

                 

209,944

                        

47%

C anada

2000

13,019

                    

13,019

             

18,412

                 

22,550

                   

26,039

                          

49%

C zech R ep.

2000

63,025

                    

63,025

             

89,131

                 

109,163

                 

126,051

                        

46%

D enm ark

2000

83,391

                    

83,391

             

117,933

               

144,438

                 

166,783

                        

53%

Finland

2000

49,733

                    

49,733

             

70,333

                 

86,139

                   

99,465

                          

49%

France

2000

48,284

                    

48,284

             

68,284

                 

83,631

                   

96,568

                          

48%

G erm any

2001

14,998

                    

14,998

             

21,210

                 

25,977

                   

29,996

                          

40%

G reece

1999

1,359,057

               

1,359,057

        

1,921,997

            

2,353,956

              

2,718,114

                     

49%

H ungary

2000

361,892

                  

361,892

           

511,792

               

626,815

                 

723,783

                        

63%

Ireland

2000

6,668

                      

6,668

               

9,429

                   

11,549

                   

13,335

                          

48%

Italy

2000

11,601

                    

11,601

             

16,406

                 

20,093

                   

23,201

                          

41%

Japan

2000

138

                         

138

                  

195

                      

239

                       

276

                               

38%

Luxem bourg

2001

552,877

                  

552,877

           

781,887

               

957,612

                 

1,105,755

                     

60%

M exico

2002

13,050

                    

13,050

             

18,455

                 

22,603

                   

26,100

                          

23%

N etherlads

2000

20,325

                    

20,325

             

28,743

                 

35,203

                   

40,649

                          

51%

N ew Zealand

2001

10,208

                    

10,208

             

14,436

                 

17,681

                   

20,416

                          

33%

N orway

2000

99,701

                    

99,701

             

140,999

               

172,687

                 

199,402

                        

52%

P oland

2000

5,740

                      

5,740

               

8,117

                   

9,941

                    

11,479

                          

37%

P ortugal

2000

718,005

                  

718,005

           

1,015,412

            

1,243,621

              

1,436,009

                     

57%

S weden

2000

78,833

                    

78,833

             

111,486

               

136,542

                 

157,665

                        

52%

S witzerland

2001

22,384

                    

22,384

             

31,656

                 

38,770

                   

44,768

                          

45%

Turkey

2002

1,468,727

               

1,468,727

        

2,077,094

            

2,543,910

              

2,937,454

                     

21%

U nited Kingdom

2000

5,981

                      

5,981

               

8,459

                   

10,360

                   

11,962

                          

43%

U nited States

2000

11,977

                    

11,977

             

16,938

                 

20,745

                   

23,954

                          

52%

Latest year

Poverty thresholds, non-equivalised disposable household incom e

50%  of nom inal 

equivalised 

disposable 

household incom e, 

nat. curr.

Poverty threshold for a 

single adult relative to 

take-hom e pay of an 

average production 

worker

 

Note:

 Data refer to annual disposable income. Values, as reported in country questionnaires, for the most recent year are expressed 

in prices of the base year. For the purpose of this table, these values have been adjusted in line with changes in the consumer price 
index. “Equivalised” disposable income is household disposable income divided by household size at the power 0.5.