INCOME DISTRIBUTION IN OECD COUNTRIES

Evidence from the Luxembourg Income Study

Prepared by Anthony B. Atkinson,

Lee Rainwater and

Timothy M. Smeeding

ORGANISATION FOR ECONOMIC CO~OPERATION AND DEVELOPMENT

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT

Pursuant to Article 1 of the Convention signed in Paris on 14th December 1960, and which came into force on 30th September 1961, the Organisation for Economic Co-operation and Development (OECD) shall promote policies designed:

to achieve the highest sustainable economic growth and employment and a rising standard of living in Member countries, while maintaining financial stability, and thus to contribute to the development of the world economy; to contribute to sound economic expansion in Member as well as non-member countries in the process of economic development; and to contribute to the expansion of world trade on a multilateral, non-discriminatory basis in accordance with international obligations.

The original Member countries of the OECD are Austria, Belgium, Canada, Denmark, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States. The following countries became Members subsequently through accession at the dates indicated hereafter: Japan (28th April 1964), Finland (28th January 1969), Australia (7th June 1971), New Zealand (29th May 1973) and Mexico (18th May 1994). The Commission of the European Communities takes part in the work of the OECD (Article 13 of the OECD Convention).

Publie en fran~ais sous le titre:

LA DISTRIBUTION DES REVENUS DANS LES PAYS DE L'OCDE

THE OPINIONS EXPRESSED AND ARGUMENTS EMPLOYED IN THIS PUBLICATION ARE THE SOLE RESPONSIBILITY OF THE AUTHORS AND DO NOT NECESSARILY REFLECT THOSE OF THE OECD OR OF THE GOVERNMENTS OF ITS MEMBER COUNTRIES.

© OECD 1995 Applications for permission to reproduce or translate all or part

of this publication should be made to: Head of Publications Service, OECD

2, rue Andre-Pascal, 75775 PARIS CEDEX 16, France.

FOREWORD

The distribution of income between households is commonly raised as an economic policy issue. The extent to which economic rewards are concentrated in particular households and less available in others is a frequent concern in both political and economic debate.

This study presents a comparison of income distribution in OECD countries. Using, for the most part, data which have been gathered in the database known as the Luxemburg Income Study (LIS), it compares the extent to which income is dispersed between households on a standardised basis, adjusting for the size of households (through the use of "equivalence scales") and for international differences in real incomes (by using national median household incomes as the reference point). To the extent that the data permit, the separate contributions of property and employment incomes and the effects of direct taxes and cash transfers are analysed. In addition, (non standardised) national studies are used to summarise national trends in income distribution over the last two decades.

This study was commissioned in 1991 by the then OECD Economics and Statistics Department and by the Education, Employment, Labour and Social Affairs Directorate from the Centre d'Etudes de Populations, de Pauvrete et de Politiques Socio-economiques of the International Network for Studies in Technology, Environment, Alternatives, Development (CEPSIINSTEAD) at which the LIS database has been compiled. It has been funded by additional grants from the authorities of OECD Member countries.

This report has been prepared by Anthony B. Atkinson (Nuffield College, Oxford), Lee Rainwater (Harvard University) and Timothy M. Smeeding (The Maxwell School, Syracuse University and LIS), all associates of CEPS/INSTEAD, and finalised in May 1995. It reflects their opinions and does not necessarily represent the views of the OECD. It is published on the responsibility of the Secretary-General of the OECD.

3

ACKNOWLEDGEMENTS

The authors would like to thank Richard Randell and Koen Vleminckx of the LIS for the great deal of work they contributed to this report. This includes compiling data and many technical details which underlie this report, as well as co-ordinating the production of the final manuscript. We thank Derek Blades, Betty Duskin, Henry Ergas, Michael Forster, Kjell Jansson, Peter Scherer, D. Verger and the members of the Advisory Board (Jean-Etienne Chapron, Ingemar Eriksson, Alois Guger, Gordon Harris, J.T.M. van Laanen, and Aino Salomliki) for their comments on the first two drafts. Professors Rainwater and Smeeding would also like to express their appreciation to the Russell Sage Foundation and to the Center for Advanced Study in the Behavioural Sciences which supported their work in the final stages of this project.

Many thanks are also due to the following individuals at their respective institutions: Holly Sutherland of Cambridge University; Cheri Minton of Harvard University; Deborah Bailey, Barbara Butrica, Esther Gray, Gina Husak and Deborah Milne of Syracuse University; Karen Gardiner of the London School of Economics; Caroline de Tombeur and Uwe Warner of the Luxembourg Income Study; and John Apruzzese (editor for the final version ofthis report).

We would also like to thank all those from national statistical offices, research agencies, and universities who contributed various statistical and research documents. Please note that these individuals are not responsible for our use of materials and comments which they supplied. We hope this list is complete:

Rolf Aarberge, Nick Adkin, Anders Bjorklund, John Bowdler, Bruce Bradbury, Andrea Brandolini, R. Brulard, Bill Callaghan, Bea Cantillon, Ian Castles, David Ellwood, Jon Epland, John Evans, Jean Louis Faure, J.P. Gaudement, Eystein Gjelsvik, Peter Gottschalk, Francis Green, Bjorn Gustafsson, Richard Hauser, Reino Hjerppe, J. H. M. Kok, Stephen Jenkins, C. Malherbe, Magda Mercader, Mikael Miiller, Brian Murphy, Brian Nolan, Antonio Perandones Garcia, Judy Raymond, Stein Ringen, Carlos Rodrigues, Peter Saunders, D. Slaats, Hans Steiner, Christian Valenduc, J. van Sinderen, and Michael Wolfson.

4

TABLE OF CONTENTS

Chapter 1

INTRODUCTION

1.1 Structure of the volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.2 A note on terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Chapter 2

METHODOLOGY

2.1 The definition of income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2 Unit and population definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.3 Adjustment for household size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.4 Presentation of results and measures of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Chapter 3

DATA QUALITY AND CONSISTENCY

3.1 The LIS database 25

3.2 Comparability of LIS datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.3 Definition of income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.4 Cash and non-cash benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5 Taxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.6 Comparison of reported income data with external aggregates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.7 Procedures adopted in use of LIS data in this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Chapter4

BASIC RESULTS

4.1 Distribution of income in seventeen OECD countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.2 Trends over time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.3 Sensitivity of estimates to methods employed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Chapter 5

COMPARISONS WITH NATIONAL STUDIES

5.1 Earlier studies of OECD countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

5.2 Evidence from national studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.3 Conclusions from national studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5

Chapter6

ANALYSIS OF COUNTRY DIFFERENCES: PRIMARY AND MARKET INCOMES

6.1 Level and trend in primary income distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.2 Level and trend in market income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

6.3 Relative income shares of primary and market incomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.4 Changing shares of income : husbands, wives, self-employment and property income . . . . . . . . . . . . . . . . . . . . . 93

6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

Chapter 7

TAXES AND TRANSFERS

7.1 The distribution of equivalent income levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.2 The distribution of taxes on income and social constributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

7.3 The distribution of transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

7.4 The role of transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

ChapterS

FUTURE RESEARCH PRIORITIES AND NEXT STEPS

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

1.

2.

3.

4.

5.

6.

7.

Appendices

LIS staff 123

Unit and head definitional details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Measures of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

Data type, quality and consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

Correspondence of micro-income data measures with SNA categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

Quality of income data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

Background tables on income sources and demographic characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

6

LIST OF TABLES

Chapter 2

2.1 Definition of income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 Family size exponents in different equivalence scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Chapter 3

3.1a OECD countries and LIS: country and year of data entry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1b Survey name and sponsor/administrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2a Standardised unemployment rates 1979-90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2b Rates of growth of real GDP 1979-90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2c Rates of increase in consumer prices 1979-90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.3 Types of survey data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.4 Correspondence between United Nations Guidelines and LIS microdata-based concepts of income . . . . . . . . . . 31 3.5 Expenditures for health, education and social welfare among the non-aged as a percentage of GDP in OECD

countries in 1987 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.6 Employees' social security contributions and personal income tax (including standard tax reliefs) at the income

level of an average production worker in 1987 as a percentage of gross earnings . . . . . . . . . . . . . . . . . . . . . . . . 33 3.7 Quality of income data: ratio of survey estimates to adjusted national accounts estimates (in per cent) . . . . . . . . 34

Chapter4

4.1 Summary of income distribution in OECD countries: percentiles of median . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.2 Summary of income distribution in OECD countries: cumulative proportions below percentiles of median . . . . 42 4.3 Summary of income distribution in OECD countries: cumulative decile shares . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.4 Measures of inequality in OECD countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.5 Trend over time in income distribution in OECD countries 1979-88: percentiles of median . . . . . . . . . . . . . . . . 47 4.6 Trend over time in income distribution in OECD countries: cumulative proportions below percentiles of median 48 4.7 Trend over time in income distribution in OECD countries: cumulative decile shares . . . . . . . . . . . . . . . . . . . . . 49 4.8 Trend over time in income distribution in OECD countries: summary measures of inequality . . . . . . . . . . . . . . . 49 4.9 Sensitivity of results to changes in equivalence scale: percentiles of median . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4.10 Sensitivity of results to changes in equivalence scale: cumulative decile shares . . . . . . . . . . . . . . . . . . . . . . . . . 53 4.11 Sensitivity of results to changes in weighting (persons versus households): percentiles of median . . . . . . . . . . . 55 4.12 Sensitivity of results to changes in weighting (persons versus households): cumulative decile shares . . . . . . . . . 57

5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9

~-'5.10

5.11 5.12 5.13 5.14 5.15 5.16 5.17 5.18 5.19

Chapter 5

International comparison by Sawyer (1976) ................................................... . International comparison by van Ginneken and Park (1984) of household distribution of income per head .... . National studies for a selection of OECD countries ............................................. . Income distribution in Australia 1981/82-1989/90 .............................................. . Earnings distribution in Austria 1970-87 Income distribution in Belgium 1985-92 Income distribution in Canada 1971-83 Income distribution in Finland 1966-85 Income distribution in France 1970-79 ....................................................... . Measures of inequality in France 1970-84 ..................................................... . Gini coefficient of inequality in Germany 1950-85 .............................................. . Income distribution in Germany 1973-90 ..................................................... . Income distribution in Ireland 1973-87 ....................................................... . Income distribution in Italy 1967-89 ........................................................ . Income distribution in Italy 1977-91 ........................................................ . Income distribution in Japan 1980-91 ........................................................ . Income distribution in the Netherlands 1981-89 ................................................ . Income distribution in Norway 1982-90: estimates by Central Statistical Bureau ....................... . Income distribution in Norway 1970-86: estimates by Ringen (1991) ................................ .

7

60 61 62 63 63 64 64 65 66 66 67 67 68 69 69 70 71 71 72

5.20 Income distribution in Portugal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 5.21 Income distribution in Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 5.22 Income distribution in Sweden 1975-90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 5.23 Income distribution in the United Kingdom 1977-90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 5.24 Income distribution in the United Kingdom 1979-1988/89 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 5.25 Income distribution in the United Kingdom 1968/69-1984/85 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 5.26 Income distribution in the United States 1967-91: income shares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 5.27 Income distribution in the United States adjusted for household size and for taxes/benefits 1973-90 . . . . . . . . . 78 5.28 Income distribution in the United States: average income as a percentage of poverty threshold . . . . . . . . . . . . . 79 5.29 Income distribution in the United States: relative incomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 5.30 Income distribution in the United States: estimates based on population census . . . . . . . . . . . . . . . . . . . . . . . . . 79

Chapter6

6.1 Percentage population shares for subtotal analyses of primary and market income, total households, and prime age households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6.2 Summary of level and trend in primary income distribution for households with heads aged 25 to 54 . . . . . . . . 84 6.3 Summary measures of level and trend in primary income distribution for prime age households with heads aged

25 to 54 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 6.4 Summary of level and trend in primary income distribution for all households . . . . . . . . . . . . . . . . . . . . . . . . . . 86 6.5 Summary measures of level and trend in primary income distribution for all households . . . . . . . . . . . . . . . . . . 87 6.6 Summary of primary income distribution for all households: cumulative decile shares . . . . . . . . . . . . . . . . . . . . 87 6.7 Summary of level and trend in market income distribution for all households . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 6.8 Summary measures of level and trend in market income distribution for all households . . . . . . . . . . . . . . . . . . . 90 6.9 Summary of market income distribution for all households: cumulative decile shares . . . . . . . . . . . . . . . . . . . . . 91

Chapter 7

7.1 Percentage distribution of low and modest, middle and high income groups by age . . . . . . . . . . . . . . . . . . . . . . 104 7.2 Percent low income by age group ............................................................ 104 7.3 Distribution of taxes by quintile and average tax as a percent of median equivalent income . . . . . . . . . . . . . . . . 105 7.4 Average taxes as a percent of median equivalent income by quintile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 7.5 Distribution of transfers by quintile and average transfers as a percent of median equivalent income . . . . . . . . . 107 7.6 Average transfers as a percent of median equivalent income by quintile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 7.7 Average transfers as a percent of median equivalent income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

A2.1 A4.1 A4.2 A4.3 A4.4 A4.5 A5.1 A6.1 A6.2 A6.3 A6.4

A6.5 A6.6 A6.7 A6.8 A6.9 A6.10 A6.11 A6.12

Appendices Household definition Primary objective of the survey ............................................................ . Sampling frame ........................................................................ . Response rates and LIS sample sizes ........................................................ . Data transformations .................................................................... . Descriptives of LIS second wave datasets with negative disposable income (DPI) ...................... . Correspondence of micro-income data measures with SNA categories ............................... . Comparison of Australian income survey with National Accounts aggregates, 1981/82 .................. . Comparison of Australian income survey with National Accounts aggregates, 1985/86 .................. . Comparison with Department of Social Security expenditure aggregates ............................. . Comparison of distribution of net taxable income in Belgium: tax statistics versus estimates on the basis of CSP-survey data ........................................................................ . Comparison of SCF estimates in Canada to adjusted personal income in National Accounts 1987 .......... . Comparison of survey estimates to National Accounts in Finland 1987 .............................. . Comparison of French survey estimates to National Accounts control totals 1984 ...................... . Comparison of German survey estimates to National Accounts control totals 1983 ..................... . Distribution of tax units over ranges of "total income" in Ireland ................................... . Expenditure on social welfare schemes in Ireland ............................................... . Comparison of Bank of Italy survey estimates to National Accounts control totals 1989 ................. . Comparison of FES estimates to National Accounts control totals in the United Kingdom 1977 ............ .

8

124 138 138 139 139 140 141 143 144 144

145 146 147 148 149 150 150 151 152

A6.13 Comparison of March Current Population Survey estimates of aggregate income with independent estimates 1987 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

A6.14 Non-response rates by category of income in March Current Population Survey (CPS) . . . . . . . . . . . . . . . . . . . 154 A 7.1 Percentage of income sources of gross income by decile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

Percentage of income sources of gross income by income level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Age and earning status distribution for total sample and persons with low and modest incomes . . . . . . . . . . . . . 161 Family type distribution for total sample and persons with low and modest incomes . . . . . . . . . . . . . . . . . . . . . 163

A7.2 A7.3 A7.4

LIST OF FIGURES

Chapter 2 2.1 Lorenz curves 22

Chapter4 4.1 United States and Sweden deciles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.2 Percentage relative to 80% and 120% median . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.3 Lorenz comparisons (based on decile points) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.4 Relative inequality in different countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Chapter6

6.1 Lorenz comparisons of primary income for all households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 6.2 Relative inequality rankings based on primary income in different countries . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 6.3 Lorenz comparisons of market income for all households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 6.4 Relative inequality rankings based on market income in different countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 6.5 Aggregate shares of primary income in the mid-1980s for households with heads aged 25 to 54 . . . . . . . . . . . . 94 6.6 Aggregate shares of primary income for all households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 6. 7 Aggregate shares of market income for all households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 6.8 Shares of primary income in the mid-1980s for households with heads aged 25 to 54: husbands, wives,

other earnings and self-employment income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 6.9 Shares of primary income in the mid-1980s for all households: husbands, wives, other earnings

and self-employment income ............................................................... 100 6.10 Shares of market income in the mid-1980s for all households: wages, self-employment income

and other private income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Chapter 7

7.1 Low-income rate of the non-elderly by transfers as a percentage of median equivalent income . . . . . . . . . . . . . . 108 7.2 Low-income rate of the elderly by transfers as a percentage of median equivalent income . . . . . . . . . . . . . . . . . 110

9

Chapter I

INTRODUCTION

Personal income distribution, relative low incomes, and income inequality in general, are increasingly recognised as important economic and social policy matters in OECD countries - especially as they become more interdependent. Comparing different countries' experiences is presently the focus of many studies. Growing economic interdependence and interest in cross-national distribution research did not come about by accident. Comparing cross-national data on income distribution delineates country similarities and differences over time and helps clarify how market and demographic forces and public policy affect the relative economic status of various groups.

Conclusions drawn from these comparisons nevertheless rely heavily on the underlying quality and comparability of income data. Accurate and accessible data resources are essential to serious cross-national policy analyses in the "survey sciences". This study aims to compare income distribution in selected OECD countries using microdata from the Luxembourg Income Study (LIS) database and national studies. The LIS database, described in Chapter 3 and Appendix 1, allows a comparison of cross-national income distributions in a unified household income database environment created explicitly for this purpose. Relying on the LIS in conjunction with OECD country studies permits us to avoid some of the problems of previous studies in this area, notably Sawyer's (1976), which is referred to on several occasions.

This study is particularly interesting at this time because it analyses the current economic, social and political forces that have widened income distribution in OECD countries, inquiring how long this trend has existed and what countries experienced (Taylor, 1992; Green, Henley and Tsakalotos, 1992).

Since the late 1970s, considerable changes of important elements have influenced income distribution in OECD countries. This resulted in part from broad economic developments in output, price level, industrial structure, and numbers of employed and unemployed, but also from employment structure changes, in age, sex and occupation, and the age and family composition of populations. Some of these changes occurred naturally without deliberation, while others were prompted by government policies.

Among the most important factors are the following:

differential experiences with macroeconomic performance across nations and continents, particularly in the 1980s; widening inequality in earnings distribution in several OECD countries (e.g. Gottschalk and Joyce, 1991; 1992; Davis, 1992); general declining male labour force participation and increasing early retirement juxtaposed with increasing female labour force participation and rising female wages relative to male wages in most OECD countries (Atkinson and Rein, 1993; Blau and Kahn, 1992); general population ageing has created, and will likely continue to create, severe budgetary pressures on governments' social retirement and health care finance systems (OECD, 1988); other large scale demographic changes such as the growth of single parenthood with serious implications for distributional outcomes (OECD, 1991); changing nature of taxation and income transfer policy and its distributive outcomes. These include a lowering of the top income tax brackets in several nations, and reduction in benefit levels and coverage in response to the budgetary pressures noted above; and a world-wide increase in capital income returns and changes in the distribution of ownership of assets, including programmes of privatisation.

This list nevertheless lacks many other possible elements such as immigration, technological change, and the transformation of Eastern Europe.

Fully documenting and analysing the distributional implications of these changes would be a rather difficult task. Mechanisms linking macroeconomic variables to personal income are complex for numerous reasons

11

including the existence of intervening institutions. National income categories do not map directly into those received by households. The corporate sector, the role of financial intermediaries, the significant ownership of assets by pension funds and assurance companies, all mean that changes in interest payments, profits and rent have implications for personal incomes which are not immediately apparent. The State is the single most important intermediary. It is therefore difficult to trace through the consequences of a privatisation programme, with the proceeds used to reduce income taxation.

Accounting for the distributional impact of these different factors is currently impossible. This report aims simply to outline the main empirical findings regarding personal income distribution in OECD countries and the changes in the mid-1980s. It is therefore important to establish how coherent a picture country studies (see Chapter 5) and the LIS database present of the evolution of income distribution. The LIS database, the comer­stone of this work, has significant limitations. It provides cross-sectional data from only a subset of OECD countries, relating to varying time periods in different countries. LIS data contain information on sources of money income and direct taxes only and do not follow the same families over time as do some country longitudinal household income panel datasets (e.g. U.S. Panel Study of Income Dynamics, German Socio-Economic Panel); or present complementary data on consumption or wealth.

LIS data were collected by different countries for different purposes at different times. Careful work by LIS staff improved data comparability, but some basic differences remain in areas such as survey design, income accounting units, and income definitions which this report purposely highlights along with possible inconsistencies and potential biases which may arise.

1.1 Structure of the volume

Chapter 2 presents the report's methodology, including income and unit definitions, adjustment factors for family size and measures of income inequality. Chapter 3 outlines the LIS dataset and coverage of Member and non-Member countries and examines data quality and consistency. It compares LIS dataset quality to one another, and to other benchmarks, notably income data with national accounts aggregates, and reviews adjustments to country data by LIS staff, changes in the nature and quality of country datasets over time, and additional standardizations made for this project.

The project's conclusions appear in the final four chapters. Chapter 4 analyses the basic results on the distribution of disposable incomes, both level and trend, as compiled from LIS datasets. Chapter 5 compares LIS results with others, in particular those from Sawyer's study and various OECD country studies. These chapters also provide evidence for countries not presently covered by LIS, including Austria, Japan, New Zealand, Portugal and Spain.

Chapter 6 analyses country differences in primary and market incomes: earnings, self-employment, and property income - definitions and use of these terms are explained later. The chapter considers the effects of earnings fluctuations for men, women and others, including separate analyses for prime age heads of households. Chapter 7 illustrates the arithmetic impact of direct taxes and transfers on gross income, and Chapter 8 summarises the main findings and comments on the needs for future research.

1.2 A note on terminology

Income distribution is an interstitial subject, lying in theoretical terms between micro-economics and macro­economics, and in statistical terms between national accounts statistics and household surveys. Each field has its own terminology. This report uses conventional micro-data terminology and relates it to terms used in other fields. Chapter 2 in particular discusses the different definitions of "income", and their relation with the United Nations Provisional Guidelines on Statistics of the Distribution of Income, Consumption and Accumulation of Households, Study M 61 (United Nations, 1977).

12

Chapter 2

METHODOLOGY

Issues of methodology are central to this investigation. Any comparison of income int;quality in two or more countries is only comprehensible by referring to the methods employed. Published comparisons are frequently based on different source types or distributions defined in different ways. Distribution of annual household income per capita in country A cannot be directly compared with that of weekly family income unadjusted for family size in country B, nor is it possible to eliminate all differences in definition. It is therefore necessary to understand how differences affect findings regarding inequality. Comparing data from LIS datasets with those in countries studies (Chapter 5), often based on a different methodology, allows for a preliminary assessment of the results.

'fhe major definition issues in this chapter are: definition of income (Section 2.1); unit and population definition (Section 2.2); adjustment for household size (Section 2.3); and presentation of results and measures of inequality (Section 2.4).

2.1 The definition of income

The present discussion deals with distribution of money income. Income is only one of several dimensions of economic well-being, and ideally requires a multidimensional approach. Two families with similar money incomes could have very different levels of economic well-being; and two families judged to have similar levels of economic well-being may have very different levels of money income. Well-being and income may diverge when families are able to finance consumption out of dissaving, however this report deals only with income, and not consumption. The difference between income and consumption is particularly relevant when considering a person's whole lifetime, which is discussed later. Well-being can depend on asset levels which provide additional security to income, or hours of work which may be a matter of choice. However, due to limited available information, this report treats income flows only.

Like the United Nations M 61 Provisional Guidelines (United Nations, 1977), this report gives priority to measuring income distribution. Income is the single most accessible indicator of economic well-being for the residents of any given country at any point in time. It is measurable, meaningful, and concrete, and annual or sub­annual money income microdata are regularly available for a range of OECD countries. Income is comparable across and within groups. Calculating income's central tendency (means and medians) and variability (size distribution of income and poverty rates) provides a rich background to gauge levels of economic well-being for individual households and demographically distinct groups of households, and to measure changes in the distribution of economic well-being over time.

The definition of income used here is derived from several components which are essential for a complete understanding of the sources of inequality. The left-hand column of Table 2.1 provides the terminology of this report. Income is defined in six stage:

1) wage and salary income. Stage 2 adds income from self-employment, to form 2) 'primary income (excluding property income). These represent the incomes received from direct participation in production. Stage 3 adds property income

received (but does not subtract interest income paid): 3) primary income (including property income). This does not include capital gains or losses. Adding private cash and near-cash income (such as occupational

pensions, alimony and child support, but not subtracting payments made) yields: 4) market income.

13

Heading (terminology used in text)

1. Wage and salary income (before deduction of direct taxes or employee social security contributions)

2. Primary income (excluding property income) (before deduction of direct taxes or employee social security contributions)

3. Primary income (including property income) (before deduction of direct taxes or employee social security contributions)

4. Market income

5. Gross income (before deduction of direct taxes or employee social security contributions)

6. Disposable income

Notes:

Table 2.1 Definition of Income

United Nations Provisional Guidelines

Compensation of employees includes employers' contributions to social security and similar schemes

I + gross entrepreneurial income + income from producers' co-operatives = Primary income

2 + property income includes rent, dividends, interest and imputed rents of owner-occupied housing

3 + social security benefits + pensions/annuities + other current transfers (alimony and child support) = Total household income

5 - direct taxes - social security contributions - pension fund contributions = Total available household income

Definition Adopted Here

Compensation of employees excludes employers' contributions to social security and similar schemes includes sick pay paid by government

I + gross self-employment income

2 + realised property income excludes imputed rents of owner-occupied housing

3 + occupational pensions" + other cash income&

4 + social insurance cash transfersc + universal cash transfersd + social assistance'

5 - direct taxes - social security contributions

a Occupational pensions include all pensions paid from non-social retirement schemes including employer-based pensions for private sector workers and public employees.

b Other cash income includes regular private transfers, alimony and child support benefits, other sources of regular cash income, not classified above. c Social insurance transfers include: accident or short-term disability pay, long-term disability pay, social retirement benefits (old age and survivors),

unemployment pay, maternity allowances, military or veteran's benefits, other social insurance. d Universal cash transfers include child and/or family allowances if paid directly by governments. Universal cash transfers paid as refundable income tax credits

are counted as negative amounts in the income tax of some countries. e Social assistance includes all income-tested and means-tested benefits, both cash and near-cash.

The addition of all regular public transfers (such as social retirement pensions, family allowances and unemployment compensation) gives in tum

5) gross income.

Appendix 7 presents the percentage of distribution of four income sources (Primary Income, Other Private Income, Social Transfers and Taxes) for each country by deciles of disposable income (Table A7.1) and groups defined in terms of percentages of median income (Table A7.2). It is important to emphasize that the boundary between market and gross income may in some contexts be somewhat artificial. A country providing old age income via compulsory membership in private pension schemes experiences less of an increase between stages 4 and 5 than a country where pensions are state provided. The balance between market and gross income in the United Kingdom, which allows people to contract out of the state pension scheme, depends on the proportion choosing this option.

Finally, deducting personal income tax and social security contributions yields:

6) disposable income.

Definitions and terminology vary in this field. People approach it from different perspectives and work in different traditions. Income is a crucial variable for macro-economic statistics, as it is for distributional analysis, which are treated separately by different people and in different agencies. Bridging the gap is important. The United Nations made a major contribution in 1977 when drawing up the Provisional Guidelines, which outlines a structure for distribution statistics relating to national accounts, covering such topics as the definition and classification of statistical units and incomes. 1

1. Subsequent work in this area has been sparse with the exception of a recent important piece by Reich (1991) which points out several unresolved conundrums.

14

This work adheres as close as possible to the United Nations Provisional Guidelines, though certain differences remain, which are identified in italics in Table 2.1. As shown in the second column, the Provisional Guidelines use different names - "total household income" for gross income and "total available household income" for disposable income, and "primary income" excludes property income.2 Here the Provisional Guidelines are ignored because the gross/disposable terminology is now well-established in income distribution literature.

Column 3 of Table 2.1 illustrates the difference in principle between the definitions adopted here and those in the Provisional Guidelines. The most important concern the exclusion of employers' contributions (to state or occupational transfer schemes) from wages and salaries, and that of imputed rents on owner-occupied houses, which leads to the question of money income versus income in-kind.

Money income

This study focuses on money income and generally excludes non-cash income in the form of housing, medical care, education, transport and related types of income in-kind (e.g. net subsidies). The principal reasons for this exclusion are unavailability of data and difficulty estimating cash equivalent value. However, many countries designate "near-cash" subsidies those which are fungible and clearly measured in currency terms, such as food stamps in the United States, housing allowances, and student scholarships in Germany. These forms of income are in principle connected to a particular expenditure - food, rent and tuition - though in general spending is loosely restricted. These subsidies, termed "near cash" income, are treated here as disposable income.

The two most significant income in-kind omitted are the imputed rental value of owner-occupied housing and the value of home production, particularly food produced for own consumption, and own household services. Because of large differences in home ownership across countries and among groups within countries, imputed rent may have a substantial effect on the distribution of economic well-being. The first, and most difficult, way to measure imputed rent is to estimate the cash rental value of a housing unit and compare it to the actual cost of home ownership, including depreciation, upkeep, and interest paid on mortgage loans. The difference between these is imputed rent which is incalculable for most LIS countries.

An alternative method is to assign some rate of return to net home equity. But the concept of income here includes only actual, realised and regular returns on wealth, and not unrealised returns or one-time realised capital gains from sale of assets. To assign an imputed return to one type of asset, housing, and not others would be inequitable. The absence of realised capital-gains income for most countries forces the LIS to exclude them from income. Net gains or losses from buying or selling assets such as houses, stocks, and bonds are also excluded.

Although imputed rent on owner-occupied housing is omitted for all countries, this may affect income distribution comparisons. Countries differ in owner-occupation and scarcity of housing and land. Imputation can affect the relative degree of income inequality.

These datasets and studies concentrate largely on the world's richest and most market-oriented economies where production for own consumption, particularly food and shelter, is far less common than in developing countries. With the exception of Spain, Portugal, and Ireland, excluding production for own consumption will not cause large variations in estimates of well-being. There are however important country differences between married women's market and non-market work patterns, which will affect the results because market work for wages reduces time for household service production, and households where most or all adults work in the market tend to use services such as house cleaning, prepared food, and child care. Measures of net "earnings capacity" would correct some of these differences, but are beyond the scope of this report (see Garfinkel and Haveman, 1977; Haveman and Buron, 1993).

Measurements of tax collection and public service provisions included here are incomplete. Traditional public goods such as infrastructure (roads, buildings, inforriiation services), defence, and protection services (police, fire), are excluded along with much of health, education and other in-kind subsidies. This report also counts only direct taxes such as national or federal income taxes and payroll taxes paid by employees and the self­employed, and omits indirect taxes such as value-added taxes, sales taxes, and property taxes. It does not claim to measure the overall net effect of the government budget, taxing and spending, on household income (O'Higgins and Ruggles, 1981).

Time period

Because of data availability, traditional accounting periods and public policy concerns, this study employs the annual income accounting convention. Most direct taxes, such as income and payroll taxes, use an annual

2. It should be noted that in the revised System of National Account definitions "primary income" includes net property income received.

15

accounting framework. Public benefit programmes, particularly those aimed at the poor, are often based on shorter accounting periods, the week or month, but expenditures for such programmes are recorded in annual amounts. Data are recorded annually for all but a small number of LIS countries.

Adopting an annual assessment period has obvious implications for measuring inequality, especially when considering alternatives. The shortest realistic measurement period is the week (or the month for monthly paid employees) which is the basis for United Kingdom data. For many weekly income is subject to considerable variation: in any week an employee may work a short shift or receive a special bonus. Weekly incomes may vary with the weather, as for a window-cleaner, or with the season, as for a shopkeeper. As a result of these short-run fluctuations, the distribution of weekly income appears more unequal than income measured over a longer period such as a year.

This analysis seems to indicate that a longer assessment period reduces the observed degree of dispersion. It is arguably for this reason that the study of annual income creates a misleading picture. Some propose changing to a lifetime income concept, which could generate quite different results. There are at all times people with low incomes who will later have higher incomes (e.g. students or apprentices), or who in the past had higher incomes (e.g. pensioners). Lifetime incomes will, by the same argument, generally be more equally distributed than current incomes. However, not only do conceptual problems arise when defining lifetime incomes [e.g. the choice of discount rate(s)], but such measures of income are rarely available and are subject to great discrepancies when estimating. The future income of today's children is only measurable through numerous assumptions about future levels of human capital (education and experience), the returns to that capital, and the type and amount of work effort by individuals. Though the annual accounting framework may be somewhat erroneous in measuring long­term income, it is still a more accurate and objective measure of economic status than lifetime income estimates (Palmer, Smeeding, and Jencks, 1988). Consequently, the report does not consider lifetime income measurement further, but does highlight life-cycle differences by the degree of dispersion within and across age groups.

2.2 Unit and population definition

The unit of analysis is an important issue often disregarded in income distribution studies. It is evident that the ultimate source of concern is the welfare of the individual. But the individual is not the appropriate unit of analysis. If the individual were the basic unit of analysis, a substantial number of individuals would emerge with virtually no recorded income, notably children and spouses working at home. They may nevertheless be enjoying a high standard of living as a result of income sharing with parents/spouses. That most married couples share incomes (perhaps not equally), and most children are supported by their parents, affirms there is a very important degree of income-sharing. If the extent of intra-family transfers were known with reasonable accuracy, it would be possible to add them to the income of the wife and children, but this would still not allow for the economies of scale which arise from living together as a unit.

Considering a wider unit than the individual raises major issues of definition. The main choices are listed below:

- common residence, where a household comprises a common dwelling and some sharing of common house-keeping -this typically constitutes the most extensive unit of analysis; common spending, where the spending unit makes spending decisions to a considerable degree in common -this may cover people who have no family relationship; blood or marital relationship, where members of the family unit are related by marriage/cohabitation or by blood relations; and

- dependence, where the unit includes a single person or couple plus any dependent children - this constitutes the inner family.

Criteria are typically applied cumulatively, so that the family unit refers to family members residing in the same household. The choice between these definitions is important in cross-country comparisons for two reasons:

- the pattern of household or family formation may differ across countries; and - the available data may relate to different definitions.

Adopting a wider unit, such as household rather than inner family, generally reduces the degree of income dispersion, basically averaging the incomes of the different members. The same distribution of individual incomes in two countries may lead to lower observed inequality in country A because people tend to live in larger units or because data refer to households rather than families. Even where the same definition is adopted, it may have different significance. For example, the age when children become independent may vary greatly across countries due to differences in national school systems, custom, and economic factors. The same may apply to comparisons over time within a country.

16

Two key issues help choose between two different units of assessment: income sharing and economies of scale. Three unrelated people sharing the same living quarters can be treated as three single person families or one three person household. The former ignores any sharing among the three persons and any economies of scale (e.g. in using household durables). The latter (household) definition accounts for economies of scale but also lumps the three individual incomes together and assumes they are equally shared. When omitting individual needs, three unrelated persons living together as one household are empirically no different than single mothers with two children, or married couples with one child. Unfortunately, little is actually known about intra-unit sharing of income (Jenkins, 1991); the topic of economies of scale is discussed in conjunction with equivalence scales.

This study relies on the household as the primary unit of analysis for several reasons. First, it is the most common unit among LIS datasets. Second, it is the unit of choice for most researchers in this area, and therefore increases the comparability of this study with others. The household concept can also evade the issue of unmarried couples, now a common phenomenon in all countries. But national statistical offices treat these units very differently. In Sweden, Norway, and the Netherlands unmarried couples living together are treated as married couples, with no differentiation between household and family. In other countries, such as the United States, Canada, and Australia, they are treated as two single person families, but as one household. The household definition therefore considers such couples as the same for all countries but the family definition does not.

A single consistent definition of households is nevertheless impossible for all LIS national datasets. "Households" in Sweden are defined as one or two adults with or without children under 18. Those over 18 are treated as single person families (or households) and cannot be linked to their actual household or family living situation. In Italy and in one Canadian file, only the family definition is available; in all other countries the "household" definition is reasonably consistent (see Appendix 2 and Table A2.1). It is possible in many countries to use several unit definitions and compare their effects on the results. Appendix 7, Tables A7.3 and A7.4 provide the distribution of samples in each country by family type, age group, and earnings status.

In addition to the key question of the unit of analysis, a number of other definitional issues arise.

Population coverage

Population in national surveys is generally determined by a population sampling frame. Countries either use electoral registers, decennial population censuses, or postal address registers, and most include documented aliens, and exclude recent and unregistered immigrants. Full-time military, the homeless and the institutionalised are generally excluded. The institutionalised includes nursing homes residents (the aged and disabled), other long­term sanitarium and hospital patients, and prisoners. The remaining civilian, non-institutionalised population represents 96 per cent or more of the total. In no country is a major part of the population excluded.

Separately identifying minorities is difficult since they represented only a small fraction of the population in the mid-1980s. Besides African-Americans and Hispanics in the United States and the Turkish minority in Germany, purposely oversampled in the dataset for easy identification, minorities are very hard to identify in other datasets and are not separately considered here (see Palmer, Smeeding and Torrey, 1988, on this issue). Regional differences, which are of particular interest in Italy and in geographically larger countries such as Canada and United States, are also ignored.

Definition of household head

Considering males as household heads in this study is a matter of comparability, not of choice. While most surveys allow the household to designate the head, not all do so. LIS uses the following rule when comparing datasets designating a male household head:

For each country survey in the LIS database the head of household, if the person designated head is male, is also designated as the household head by LIS, and no checks are made to determine who that person is. In those cases when a female is the designated household head in the original survey, checks are performed to see if there is a male spouse present. If so, the male person is designated by LIS as the household head.

For a more complete description of the household definition in all countries, the identification of household members, and how the unit head was designated in each of the original country surveys, see Appendix 2.

Age categorisation

Classifying subgroups of populations and household heads and members according to age leaves several options. The LIS defines children as 18 years or younger, prime age adults as between 25 and 54, and the elderly as 65 or older. School attendance and tradition in countries such as France and Italy make it customary for children to live with their parents until age 25 or beyond. Income tax systems in these countries make allowances for such persons as dependents. All countries must decide how to classify 18-25 year olds, particularly those who board at

17

school. In some countries, a married person cannot be a "child" - unmarried 17 year olds are children; married 17 year olds are not. In others, children are defined according to rules determining tax dependents. Depending on the country, oldest age cut-offs for tax dependents can differ between ages 18, 21, or 25, and in others children can be any age - the survey respondent decides. Comparing the sensitivity of these definitions is allowed in some countries (see Appendix 2).

The incomes of 55-64 and 18-25 year olds pose similar problems. Males generally retire before age 65, voluntarily (or involuntarily) in most OECD countries, reducing their money incomes to increase leisure time. Early retirement can create a wide variance in earnings and other money income sources, and no simple definition of retirement exists in OECD countries. Many 55-64 year old with retirement income continue to work, some receive pension income, others do not. Retirement does not necessarily mean receipt of pension income in these age ranges (Smeeding, 1982). This report considers the entire income distribution first, and then separately prime age, 25-54 year olds, where retirement and school attendance are both less likely to independently affect labour supply and earnings variability.

2.3 Adjustment for household size

Since households differ in size and in composition, it is necessary to adjust income to account for differences in need. Equivalence scales were designed to accomplish this. Total household income divided by the equivalent number of adults measures household well-being. Traditionally, studies have used the household income per capita (or per member) measure to adjust total incomes according to the number of persons in the household. But such an adjustment ignores economies of scale in household consumption relating to size and other differences in needs among household members. Additional influential family characteristics include region, location, and particularly age of adults and children. This project considers only how family size, the most common factor, affects measures of poverty and inequality through various equivalence scales used in the OECD countries studied. Size is crucial because it is always used in equivalence adjustments, is often the only factor in equivalence scales, and is given the greatest weight in the few scales adding other considerations.

Equivalence scales currently used in policy or discussed in scholarly literature greatly vary in their emphasis of increments to family size in the calculation of need. One extreme viewpoint ignores size and deals only with disposable income, without adjusting income to account for size. At the other extreme, per capita income ignores economies of scale in producing and consuming household goods and services. Current equivalence scales rather evenly cover the range between these two extremes. Most plausible assumptions about economies of scale in family consumption are apparently extant in policy and academic consideration of inequality issues.

This diversity of equivalence scales is evident in comparative studies of the distribution of income and well­being. In country studies, choosing equivalence scales is often foreclosed by conventional usage or public policy practice. Which scale should be used in cross-national comparisons? And how does the choice of scale affect the results of the analysis? Buhmann, Rainwater, Schmaus and Smeeding (1988) discovered that the choice of equivalence scales definitely affects the ranking of countries. This part of the report updates this analysis and outlines the basis for selecting the equivalence scales in this project.

Equivalence scales are generally presented as income amounts, or ratios of amounts, needed by families of different size and/or structure. If a one-person family needs one unit of income to maintain a given level of living, a two-person family is said to need 1. 7 units, and a three-person family 2.2 units. In the range of studies considered here, however, a single parameter can closely approximate the specific amounts or ratios in the equivalence scales: the family size elasticity of need. This report assumes that economic well-being (W), or "adjusted" income, equates disposable income (D) and size (S) in the following way:

W=DfSE (1)

The equivalence elasticity, E, varies between 0 and 1; the larger it is, the smaller are the economies of scale assumed by the equivalence scale. (For further discussion, see Coulter, Cowell and Jenkins, 1992.)

Table 2.2 summarises the results by fitting a linear regression based on equation (1) to a wide range of equivalence scales. The equation fitted takes logarithms from both sides to explain the logarithm of the scale, as a function of a constant, and the logarithm of family size. Few scales, particularly those based on regression analyses of survey data, actually specify this equation- most do not try to adapt this kind of mathematical relation. Some state a simple rule of thumb -for example, additional adults after the first have weights of 0. 7, and additional children 0.5. Others incorporate diminishing weights for each additional person. Some scales are phrased in terms of age rather than number of children. Despite these differences, equation (1) summarises quite well the relation between need and size, although two alternative scales with the same value of E may generate different results (as is likely the case when they are differentiated in dimensions other than family size).

18

Line

1

2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55

56

Table 2.2 Family Size Exponents in Different Equivalence Scales A. Individual Scales

Value ofE

Survey Scales Expert Scales

Type of Scale SUBJ CONS PROG STAT

Unadjusted Family Income E = 0

IEQ- France 0.12 IEQ - Belgium 0.17 IEQ - United Kingdom 0.18 MIQ- U.S. Dubnoff data 0.18 MIQ - U.S. ISDP 0.21 IEQ - Netherlands 0.22 Necessities- U.S. 1960-61 0.23 MIQ- U.S. Gallup 0.23 IEQ - Switzerland 0.26 IEQ - Germany 0.27 IEQ - Denmark 0.27 IEQ corrected - Netherlands 0.29 IEQ - Ireland 0.32 PIE- U.S. 0.33 Dutch Poverty 0.35 MIQ corrected - Netherlands 0.36 Expenditures- U.S. 1960-61 0.37 Expenditures- U.S. 1972-73 0.38 Food- U.S. 1960-61 0.47 France 0.57 Swedish Poverty 0.54 Australian Poverty 0.55 Austria (Pro-Kopf-Haushaltseinkommen) 0.79 Austria (Katholischer Familienverband) 0.81 Austria (Tyrol Benefits) 0.71 Austria (Oberosterreich Benefits) 0.75 Austria (Niederosterreich Benefits) 0.78 Swiss Poverty 0.56 U.S. Official Poverty 0.56 Canadian Official LICOs 0.56 Expenditures - Switzerland 0.57 British Poverty 0.59 German Poverty 0.67 Australian Bureau of Statistics 0.50 Australia Henderson (head not working) 0.51 Australia Henderson (head working) 0.58 European Poverty Line 3, LIS 0.70 Jenkins/O'Higgins 0.72 U.S. Bureau of Labor Statistics 0.72 OECD Social Indicators 0.73 European Poverty Line 1 0.84 Finland (TASKU Scale) 0.65 Germany 0.88 Italy (Sarpellon) 0.70 Italy (Ministero del Lavoro) 0.56 Italy (2nd Report of Poverty Commission) 0.72 Italy (ISPE) 0.76 Italy (CNEL) 0.61 Ireland (Supplementary Welfare Allowance) 0.59 Luxembourg 0.70 Netherlands 0.70 Sweden 0.74 UK (McC!ements before housing costs) 0.78 UK (McClements after housing costs) 0.91

Per Capita Income E = 1.0

19

Correlation with Ln (Size)

1.00 1.00 1.00

1.00 1.00 1.00

1.00 1.00 0.98

1.00 1.00 0.98 1.00 0.95 0.99 1.00 n.a. 1.00 1.00 0.87 0.91 0.91 0.99 0.87 1.00 0.98 1.00 0.98 0.99 1.00 0.96 0.98 0.99 1.00 1.00 1.00 1.00 1.00 0.97 0.89 1.00 0.99 1.00 1.00 0.98 0.94 0.98 0.88 0.97 0.82 0.81

Line Type of Scale

Minimum Value Maximum Value Median Value Mean Value

B. Summary Statistics

Survey Scales

SUBJ I CONS

0.12 0.36 0.25 0.24

0.23 0.76 0.57 0.55

Value ofE

Expert Scales

PROG I 0.35 0.81 0.59 0.62

STAT

0.50 0.91 0.72 0.69

Notes: IEQ - Income Evaluation Question; MIQ - Minimum Income Question; PIE - Public Income Evaluation. Summary statistics do not include lines I and 56.

Sources by line: 2,3,4,10,11,12,14: 5: 6: 7, 13, 16, 17: 8, 20: 9, 15: 18,40: 19: 21: 22: 23: 24, 25, 26, 27, 28: 29: 30: 31: 32: 33: 34: 35, 36,37: 38:

39, 41,42: 43: 44: 45: 46: 47: 48: 49: 50: 51: 52: 53: 54,55:

Van Praag, Hagenaars and van Weerden [ 1982] Rainwater [1987] Danziger, Vander Gaag, Taussig and Smolensky [1984] Kapteyn, Kooreman and Willense [ 1987] Watts [1967] Rainwater [1974] Lazear and Michael [1980] Van der Gaag and Smolensky [1982] D. Verger Wahlstrom [1987] Henderson [1975] [ 1990]. Supplied by Herr Wolf. Buhmann [1988] U.S. Department of Commerce [ 1987] Statistics Canada [ 1987] OECD, "Social Indicators" [ 1982] Scale Ramprakash [ 1986] Dobroschke-Kohn [1987] Supplied by B. Bradbury Hauser and Nouvertne [1980]; also used in initial LIS research papers, see Smeeding, Rainwater and O'Higgins [1988] Jenkins and O'Higgins [1987] Supplied by A. Salomaki Supplied by I. Fischer Sarpellon [1982] Ministero del Lavoro [1983] Carbonaro [ 1985] DiBiase, DiMarco and DiNicola [1993] Rossi [1993]. Supplied by B. Whelan Supplied by CEPS, Luxembourg Supplied by the Netherlands Ministry of Social Affairs Supplied by the Swedish Central Statistical Office Department of Social Security [1993], p 124.

The results in Table 2.2 span almost the entire range from no adjustment to per capita adjustment- extending from 0.12 for a scale developed from the van Praag Income Evaluation Question (IEQ) in France to 0.91 for the equivalence scale by McClements (1978).

Four types of scales were developed: two using experts' knowledge, and two empirically from analyses of survey data. Social science analysts who develop expert scales using various of materials are usually responsive to policy and precedent considerations. Scales may explicitly affirm how need varies by family size, as in the US Poverty Line, or implicitly establish amounts payable by a transfer programme, as in the Income Support scheme (and associated housing supplements) in the United Kingdom. Two somewhat different goals of expert scales are therefore apparent:

Expert Statistical (STAT) Scales in this case are developed only for statistical purposes to count persons below or above a given standard of living, for example minimum adequacy. The Bureau of Labor Statistics family budgets are a good example, or the scales used in OECD Social Indicators or by the European Commission to count low income population.

Expert Programme (PROG) The second type of expert scale focuses on defining benefits for social programmes - the Swedish "base amount" is an example of a scale used to calculate benefits under social protection programmes. The US

20

poverty line was initially developed for statistical purposes but now also serves as a guide to the adequacy of programme benefits.

Survey-based scales presenting a second general approach employ multivariate analyses of either consumption expenditures or respondents' assessment of the adequacy of income in terms of some particular target (making ends meet, not being poor, having a very good income, etc.).

Consumption (CONS) This aims to measure utility indirectly through revealed preferences of consumer spending constrained by disposable income.

Subjective (SUBJ) The goal here is to measure directly the utility associated with particular income levels for families of given characteristics. These scales address three important topics: Income Evaluation Question (IEQ), Minimum Income Question (MIQ), and Public Income Evaluation (PIE).

Table 2.2 illustrates the various forms of different kinds of scale. Expert scales usually have the highest elasticities, with a median of 0.72 for statistical, and 0.59 for programme oriented. Consumer expenditure scales have a median of 0.57, and subjective scales only 0.25, therefore a family of 4 persons is treated as equivalent to 1.4 adults on the basis of a value of 0.24, but 2.7 adults with a value of 0.72.

Measures of scale used for this study are S1, or per capita household income, and S0·5• These were chosen because adjustment for need is important and per capita household income is a well recognised international standard which will increase the comparative usefulness of our study for current and future projects. The S0·5

measure provides a good contrast between per capita income S 1 and the case of no adjustment SO, which is also considered when conducting sensitivity tests. Of the 54 estimates in Table 2.2, fourteen range from 0.4 to 0.59. Many country studies (Chapter 5) also use SO.

Population weighting

One final technical issue regarding income distribution estimates is the choice of population weights. Interview units in survey data are drawn from the survey population at large (see Appendix 4, Table A4.2 for sampling frames in LIS countries). Each type of unit is then weighted inversely to their probability of selection. Household incomes are then multiplied by this household weight to produce representative estimates for all households in the country (region or other grouping) as a whole. Household incomes for survey interviewees are therefore "weighted" to equal total household income.

Economists and public policy analysts are however most often concerned with the economic well-being of individuals and not with the well-being of households per se. Re-weighting household income by the number of people in each unit to produce so-called "person weights" facilitates this adjustment. A six-person unit "counts" six times as much as a one-person unit. Person weighting produces an estimate of the overall distribution of income among individuals in the population.

The issue of weighting is particularly important when examined with adjustments for household size. It makes sense to treat each household as a single unit (i.e. to apply household weights) if no adjustment is made (scale SO). This is supported by the wide use of this measure by countries in their own analyses of inequality. Where household income is divided by the number of members to give per capita income, person weights seem appropriate. This report uses person weights in the following, but examines the sensitivity of the results.

2.4 Presentation of results and measures of inequality

This report presents in a number of ways the comparisons of the degree of inequality over time, across countries, and across income measures within countries. It aims to provide a thorough overview of income distribution before judging a country's rank or relative level of inequality.

The first method of presentation expresses the percentiles of distribution as percentages of the median. In the United States dataset for 1986, for example, the median equivalent income per adult, applying the scale S0·5, was $13,364 per year. For a family of 4, this corresponded to a total income of $26,728, since the equivalence scale is simply the square root of 4. The equivalent income per adult at the lower quartile, 25 per cent up from the bottom, was $8,240, so that the twenty-fifth percentile, expressed as a percentage of the median, was 61.7 per cent. This is shown as P25.

The bottom decile was $4,639, giving a percentage of 34.7 per cent, shown as P10 and the top decile was $27,540, giving a percentage of 206.1 per cent. The ratio of the top to bottom decile is referred to as the decile ratio. The movements in these indicators show the changes in the relative distribution of income. It is of course

21

possible for P10 to be falling, but for the bottom decile to be rising in ter ms of purchasing power (if the median is rising sufficiently).

Comparisons of income distributions are more frequently based on the cumulative distribution of income compared to the cumulative distribution of persons in households (i.e. the familiar Lorenz curve). Figure 2.1 shows Lorenz curves for two hypothetical countries. Along the horizontal axis are shown cumulative percentages of the population: the bottom 10 per cent, the bottom 20 per cent (which includes the bottom 10 per cent), the bottom 30 per cent and so on. If all incomes were equal, then the bottom 10 per cent would receive 10 per cent of total income, and the diagonal line in Figure 2.1 would be the "line of identical incomes". In fact the bottom 10 per cent receive less than 10 per cent of total income, so that the Lorenz curve lies below the diagonal. As long as the Lorenz curve for one country (Country A in Figure 2.1) remains above (or touches) the other (Country B), it is possible to draw conclusions from the Lorenz curves about the degree of inequality (Atkinson, 1970). The Lorenz curve is a relative measure, concerned with shares of total income, and ignores differences in total income. The share of the bottom 20 per cent in total income may be 6 per cent in country B, compared with 10 per cent in country A, but if total income is twice as high in country B, then it may still be better off despite having a smaller share. Since this report concentrates wholly on relative measures of inequality, judgements concerning tradeoffs between relative and absolute levels of well-being are beyond its scope.

In practice, Lorenz curves do not necessarily resemble those in Figure 2.1 since they may intersect. When they intersect, ambiguous conclusions will necessarily arise - unless inequality is weighted at different points of the distribution. Comparing Lorenz curves also provides no quantitative measure of the extent of differences in inequality. Are the differences large or small? (This is a separate question from that of the statistical significance of the differences.)

The standard approach to these questions is to use a summary measure of inequality. Two such measures are used here:

the Gini coefficient; and - the Atkinson inequality index with coefficients of 0.5 and 1.0.

Figure 2.1 Lorenz Curves

Lorenz Curves

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

22

Cumulative %of total income

These measures reduce the degree of inequality to a single number. The Gini coefficient for example is equal to the area between the Lorenz curve and the diagonal expressed as proportion of the whole triangle. It is alternatively equal to the expected average difference in incomes, relative to the mean, between any two persons drawn at random from the population. As indicated by Atkinson (1970), Sen (1973), and Luethi (1981), this means that all summary measures imply some a priori value judgements about the distribution itself. The Gini coefficient is most sensitive to inequality changes around the median. Weighting is made explicit in the Atkinson index coefficient, where the sensitivity to changes in the lower part of the income distribution rises with the value of the coefficient (only two values are used here, but others could be employed). The issue of sensitivity arises, not just at the level of social judgements but also in terms of the robustness of the results regarding errors in data. Measures emphasizing the extremes, such as the coefficient of variation (not used here), may be highly sensitive to the inclusion or exclusion of outliers.

In presenting the results, the distribution is usually considered as a whole (relative percentiles and Lorenz curves), rather than a single summary measure, making individual judgements possible from the whole distribution. Relative income positions -low income, modest income, middle and high income- are also explored. Political discontent has emerged in recent years in many industrialised countries due to a perceived notion that income inequality was increasing and the middle class shrinking (Taylor, 1992; Duncan, Smeeding and Rodgers, 1993). Concerns about these trends can be empirically examined in this report. Household income distributions are examined as in earlier studies in four adjusted income groupings:

low income (adjusted incomes below 0.5 times median income); modest income (adjusted incomes between 0.5 and 0.7 times median income); middle income (adjusted incomes between 0.7 and 1.5 times median income); and high (adjusted incomes above 1.5 times median income).

23

Chapter3

DATA QUALITY AND CONSISTENCY

The value of this study depends crucially on the quality of the underlying survey data. The availability of microdata and extensive technical documentation allow for a more complete examination of this issue than in the past. Section 3.1 introduces the LIS dataset, explains the procedures for dealing with those OECD countries not covered by the dataset, and draws attention to some of the major limitations of a LIS-based approach. Describes the LIS dataset in more detail, examining in particular how much the information for different countries is comparable. As explained in Chapter 2, the definition of income is key to relating the study of income distribution to that of macroeconomy. Section 3.3 examines the relation between the United Nations M 61 Provisional Guidelines (the reference point of this study) and data from LIS datasets.

An important means of assessing the quality of the survey data on income distribution is to compare implied aggregates with other estimates of aggregate income based on national accounts or other sources, such as administrative records. This is the subject of Section 3.1, which compiles evidence about the comparisons made in nine of the countries studied. This is apparently the first occasion where evidence is presented on a comparative basis for a wide range of countries. The conclusions are to some extent reassuring, but also underline the problems associated in particular with survey measurement of self-employment income, property income and certain means­tested transfers. Section 3.3 describes the main procedures adopted in LIS data to allow for differential non­response, income item non-response, income under-reporting, and top- and bottom-coding. Much technical detail is relegated to Appendices 4, 5 and 6, with brief explanations in the text and notes to the tables.

3.1 The LIS database

The Luxembourg Income Study (LIS) project began in 1983 under the joint sponsorship of the government of Luxembourg and the Centre for Population, Poverty, and Policy Studies (CEPS) in Walferdange. It is now funded on a continuing basis by CEPS/INSTEAD and by the National Science Foundations of its member countries. The main objective of the LIS project has been to create a database containing social and economic data collected in household surveys from different countries.

Since its inception in 1983, the experiment has grown into a co-operative research project with membership of countries in Europe and North America, as well as Australia. The database contains information for some 25 countries for one year or more from 1960 to 1989. Surveys were added in 1993 to represent the early 1990s. Extensive documentation on the technical aspects of survey data, and the social institutions of income provision in member countries will soon be available to users. This technical documentation underlies many of the appendices to this report.

The OECD has made use of LIS data on several occasions, appearing in reports on reforming public pensions (1987), lone parents (1991) and occupational pensions (1992). Current OECD research in progress uses the LIS database for projects related to poverty and material deprivation, low earnings and related issues.

Table 3.1a contains a list of OECD countries for which income data were available for this study, covering altogether the majority of OECD Member countries.

There are three groups of countries covered in the study:

- Group A: Six major OECD countries which have made their data available to the LIS. For four of these countries we have data at two different dates, and for the remaining two we have data for the

25

Table 3.la OECD Countries and LIS: Country and Year of Data Entry

Canada (CN) France (FR) Germany (GE) Italy (IT) United Kingdom (UK) United States (US)

Australia (AS) Austria (OS) Belgium (BE) Ireland (IR) Luxembourg (LX) Norway (NW) Sweden (SW) Switzerland (SZ)

Finland Netherlands New Zealand

A. Major OECD Countries

1981, 1987 1979, 1984 1984 1986 1979, 1986 1979, 1986

B. Other OECD Countries Where LIS Data are Used

1981-82, 1985-86 1987 1985, 1988 1987 1985 1979, 1986 1981, 1987 1982

C. Other OECD Countries Supplying Similar Information

1987, 1990 1983, 1987 1983-84, 1987-88

mid-1980s. The level and trend in inequality are extensively analysed in Chapters 4, 6 and 7. In Chapter 5, the results are compared to other national studies within each country.

-Group B: Eight other OECD countries for which we used data in the LIS. For a number of these we have data on two years. They will be included in Chapters 4, 6 and 7. Where possible, comparisons are made with national studies in Chapter 5.

- Group C: Three other OECD countries for which results have been made available on a similar basis. These include the Netherlands, which is a member of the LIS, and Finland which is joining the LIS as this report is being finalised.

Table 3.lb shows for reference purposes the official names of the surveys in the LIS dataset (i.e. the 14 countries in categories A and B), and the responsible agencies.

The LIS project aims to increase the degree of cross-national comparability, though complete comparability is impossible, even if LIS surveys were conducted in each country. Comparability is a matter of degree; achieving an acceptably high level is the only reasonable goal. Major limitations are therefore important to note at the outset.

First, all comparisons are made in terms of intra-country relative measures and do not seek absolute comparisons of levels of well-being in different countries. Adjusting country currencies to common measures is possible (e.g. purchasing power parities), but they vary by source and time period. And to convert all country money incomes to common terms implicitly assumes that money incomes (and taxes, and cash transfers) capture the same fraction of total national income in each country. This is violated to the extent that countries spend different amounts for non-cash versus cash benefits (e.g. health care), or rely more or less heavily on indirect taxes, price subsidies or other elements of redistribution such as tax expenditures. This violation will occur even if the measures of money income are identical in all countries.

Second, the unavailability of certain microdatasets creates issues of interior versus exterior comparability. Having countries raw data files, questionnaires, codebooks, editing procedures and other technical details, makes it possible to measure the extent of comparability among datasets, and to make sensitivity comparisons, impose identical restrictions on other countries' data in order to assess the extent of bias due to non-comparability. It is doubtless much less satisfactory to process data by sending a set of specifications to another party which create the output. Having some modicum of control over datasets is therefore preferable.

Third, using the available data and, for example, information covering only two years makes it difficult to recognise general trends. Changes are hard to monitor when not occurring at a steady rate. It is also apparent from Table 3.la that the timing of LIS surveys differs, and even if all surveys were available for the same year, timing differences of business cycles across countries would still create a hindrance. Dates when data are drawn correspond to different aggregate economic conditions (e.g. inflation, unemployment, economic growth). Economic growth, consumer price change and unemployment are each liable to affect the outcome of analyses.

26

AS81; AS85

OS87 BE85; BE88

CN81; CN87

FR79; FR84

GE84

IR87

IT86

LX85

NW79,NW86

SW81; SW87

SZ82

UK79; UK86

US79; US86

Table 3.lb Survey Name and Sponsor/Administrator

Australian Income and Housing Survey Australian Bureau of Statistics Austrian Microcensus Panel Survey of the Centre for Social Policy, Centre de Politique Sociale/Centrum voor Sociaal Beleid (University of Antwerp, UFSIA) Survey of Consumer Finances Household Surveys Division, Statistics Canada French Survey of Income from Income Tax (Enquete sur Ies Revenus Fiscaux) Institut National de Ia Statistique et des Etudes Economiques (INSEE) German Socio-Economic Panel Study (Das sozio-okonomische Panel) Deutsches Institut fiir Wirtschaftsforschung (DIW) Survey of Income Distribution, Poverty and Usage of State Services Economic and Social Research Institute Bank of Italy Income Survey (Indagine Campionaria sui Bilanci Delle Famiglie) Luxembourg Household Panel Study (Panel Socio-Economique "Liewen zu Letzeburg") Centre d'Etudes de Populations, de Pauvrete et de Politiques Socio­Economiques (CEPS) Income and Property Distribution Survey (Inntekts- og Formuesundersokelsen) Central Bureau of Statistics Income Distribution Survey (InkomstfOrdelningsundersokningen) Statistika Centralbyran Swiss Income and Wealth Survey (Schweizerische Einkommens- und Vermogensstichprobe) Volkswirtschaftliches Institut, Universitat Bern Family Expenditure Survey Central Statistical Office March Current Population Survey Bureau of the Census

Economic growth and unemployment affect the level and trend in earnings, and inflation can impact the real value of public income transfer benefits which are not indexed. The interaction between indexation schedules and survey timing is also an issue.

These major economic factors are identified in Table 3.2a to c, with highlighted figures indicating data years in Table 3.la. Output grew strongly for four countries in 1979, but in a number of countries the early 1980s saw little or no growth: 0 in Sweden in 1981 and -0.9 in Switzerland in 1982. By the mid-1980s growth was strengthened for those countries with inequality data (except Belgium). The variation in growth rates is clearly illustrated by Belgium, whose first growth rate was 0.8 per cent (1985) and second was 5.0 per cent (1988), and by Finland, whose first growth rate was 3.3 per cent (1987) and second 0 per cent (1990).

The annual rate of consumer price inflation declined in every country from the first to the second period (except Finland, first observed in 1987). In most cases this decline was from a double to a single digit level. Consumer prices actually fell in the Netherlands in 1987. Price changes were typically accompanied by increased unemployment (except in Belgium, Finland and Sweden). In countries with two observations of income equality, unemployment increased substantially in France and the United Kingdom, whereas the United States and Canada experienced only mild unemployment increases over the period. Because both the level and trend in unemployment can affect the absence or presence of earnings and their levels, unemployment rates are quite important in the analyses which follow.

Demographic change was also occurring in these countries. While birth rates were fairly stable, several studies indicate that lone or single parent families as a percentage of all families with children increased by 15 to 25 per cent (Ennisch, 1987; Blundell and Walker, 1988). Still, lone parent families make up 26 per cent of all families with children in the United States and 23 per cent in Sweden, far outdistancing the 12 to 14 per cent estimates in France, Germany, Canada, and the United Kingdom. While single parent families are not separated in the following studies, their relative growth may still have some impact on results. Population ageing and retirement age did not change substantially enough to outweigh economic and other policy changes over this period.

27

Table 3.2a Standardised Unemployment Rates 1979-90

Country 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Australia 6.2 6.0 5.7 7.1 9.9 8.9 8.2 8.0 8.0 7.2 6.1 6.9 Austria* 1.8 1.6 2.2 3.1 3.7 3.8 3.6 3.1 3.8 3.6 3.1 3.2 Belgium 8.2 8.8 10.8 12.6 12.1 12.1 11.3 11.2 11.0 9.7 8.0 7.2 Canada 7.4 7.4 7.5 10.9 11.8 11.2 10.4 9.5 8.8 7.7 7.5 8.1 Finland 5.9 4.6 4.8 5.3 5.4 5.2 5.0 5.3 5.0 4.5 3.4 3.4 France 5.8 6.2 7.4 8.1 8.3 9.7 10.2 10.4 10.5 10.0 9.4 8.9 Germany 3.2 2.9 4.2 5.9 7.7 7.1 7.1 6.4 6.2 6.2 5.6 4.8 Ireland n/a n/a n/a n/a 14.0 15.5 17.0 17.1 16.7 16.2 14.7 13.4 Italy 7.6 7.5 7.8 8.4 8.8 9.4 9.6 10.5 10.9 11.0 10.9 10.3 Luxembourg* 0.7 0.7 1.0 1.3 1.6 1.7 1.6 1.4 1.6 1.4 1.3 1.1 Netherlands 5.4 6.0 8.5 11.4 12.0 ll.8 10.6 9.9 9.6 9.2 8.3 7.5 New Zealand n/a n/a n/a n/a nla n/a n/a 4.0 4.1 5.6 7.1 7.7 Norway 2.0 1.6 2.0 2.6 3.4 3.1 2.6 2.0 2.1 3.2 4.9 5.2 Portugal n/a n/a n/a n/a 7.9 8.4 8.5 8.5 7.0 5.7 5.0 4.6 Spain 8.4 11.1 13.8 15.6 17.0 19.7 21.1 20.8 20.1 19.1 16.9 15.9 Sweden 2.1 2.0 2.5 3.2 3.5 3.1 2.8 2.7 1.9 1.6 1.4 1.5 Switzerland* 0.3 0.2 0.2 0.4 0.8 1.0 0.8 0.7 0.6 0.6 0.6 0.6 United Kingdom 5.0 6.4 9.8 11.3 12.4 11.7 11.2 11.2 10.3 8.6 7.2 6.8 United States 5.8 7.0 7.5 9.5 9.5 7.4 7.1 6.9 6.1 5.4 5.2 5.4

Notes: * "Commonly used definitions" rather than "standardised unemployment rates." There are breaks in the series: Belgium (after 1982), Finland (after 1981), Germany (after 1983), the Netherlands (after 1982), Norway (after 1979), Sweden (after 1986), and the United Kingdom (after 1983).

Source: All except those marked* from OECD Economic Outlook, December 1993, Table Al9, those marked with* from OECD Economic Outlook, December 1992, Table Rl9.

Table 3.2b Rates of Growth of Real GDP 1979-90

Country 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Australia 4.0 2.5 3.4 -0.1 0.7 7.5 4.8 1.9 4.4 4.4 4.6 1.4 Austria 4.7 2.9 -0.3 1.1 2.0 1.4 2.5 1.2 1.7 4.1 3.8 4.6 Belgium 2.2 4.1 -0.9 1.5 0.5 2.1 0.8 1.6 2.1 5.0 3.9 3.3 Canada 3.9 1.5 3.7 -3.2 3.2 6.3 4.7 3.3 4.2 5.0 2.4 -0.2 Finland 7.3 5.3 1.6 3.6 3.0 3.1 3.3 2.8 3.3 5.4 5.4 0.0 France 3.2 1.6 1.2 2.5 0.7 1.3 1.9 2.5 2.3 4.5 4.3 2.5 Germany 4.2 1.0 0.1 -0.9 1.8 2.8 2.0 2.3 1.5 3.7 3.6 5.7 Ireland 3.1 3.1 3.3 2.3 -0.2 4.4 3.1 -0.4 4.6 4.2 6.5 9.1 Italy 5.8 4.1 0.6 0.2 1.0 2.7 2.6 2.9 3.1 4.1 2.9 2.1 Luxembourg 2.3 0.8 -0.6 1.1 3.0 6.2 2.9 4.8 2.9 5.7 6.7 3.2 Netherlands 2.4 0.9 -0.7 -1.5 1.4 3.1 2.6 2.0 0.9 2.6 4.7 4.1 New Zealand 1.5 0.4 4.7 3.4 1.0 8.6 1.2 0.6 -2.2 3.0 -0.7 0.5 Norway 5.1 4.2 0.9 0.3 4.6 5.7 5.3 4.2 2.1 -0.5 0.6 1.7 Sweden 3.8 1.7 0.0 1.0 1.8 4.0 1.9 2.3 3.1 2.3 2.4 1.4 Switzerland 2.4 4.4 1.4 -0.9 1.0 1.8 3.7 2.9 2.0 2.9 3.9 2.3 United Kingdom 2.8 -1.9 -1.2 1.6 3.6 2.3 3.8 4.1 4.8 4.4 2.1 0.5 United States 2.5 -0.5 1.8 -2.2 3.9 6.2 3.2 2.9 3.1 3.9 2.5 1.2

Source: OECD Economic Outlook, December 1993, Table AI.

3.2 Comparability of LIS datasets

Perhaps the most important issue of comparability lies with the relative quality and consistency of LIS datasets themselves. The types of survey data used by the LIS are not uniform in nature, purpose or objective. The lowest common denominator the LIS requires is the existence of a substantial level of detail concerning income sources and totals. The surveys themselves are quite diverse, as illustrated in the upper part of Table 3.3 (see Appendix 4 Table A4.1 for more detail). Some surveys are designed first and foremost to collect income data; others are derived from income tax records; and still others come from special supplements to labour force surveys. Some LIS datasets are based on income questions taken from expenditure surveys (as in the case of the United Kingdom); others are separate waves of longitudinal household panel data (e.g. Germany); and still others are taken, at least in part, directly from government administrative data.

28

Table 3.2c Rates of Increase in Consumer Prices 1979-90

Country 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988

Australia 9.1 9.8 10.1 11.2 10.1 3.9 6.7 9.1 8.5 7.3 Austria 3.7 6.3 6.8 5.4 3.3 5.7 3.2 1.7 1.4 1.9 Belgium 4.5 6.7 7.1 8.7 7.7 6.3 4.9 1.3 1.6 1.2 Canada 9.1 10.2 12.4 10.8 5.8 4.3 4.0 4.2 4.4 4.0 Finland 7.5 11.6 12.0 9.6 8.3 7.1 5.9 2.9 4.1 5.1 France 10.8 13.6 13.4 11.8 9.6 7.4 5.8 2.7 3.1 2.7 Germany 4.1 5.5 6.3 5.3 3.3 2.4 2.2 -0.1 0.2 1.3 Ireland 13.2 18.3 20.4 17.1 10.5 8.6 5.5 3.8 3.1 2.1 Italy 15.7 21.1 18.7 16.3 15.0 10.6 8.6 6.1 4.6 5.0 Luxembourg 4.5 6.3 8.1 9.4 8.7 5.6 4.1 0.3 -0.1 1.4 Netherlands 4.2 6.5 6.7 5.9 2.7 3.3 2.3 0.1 -0.7 0.7 New Zealand 13.7 17.2 15.4 16.2 7.3 6.2 15.4 13.2 15.8 6.4 Norway 4.8 10.9 13.7 11.3 8.4 6.3 5.7 7.2 8.7 6.7 Portugal 23.9 16.6 20.0 22.4 25.5 28.8 19.6 11.8 9.4 9.7 Spain 15.7 15.6 14.5 14.4 12.2 11.3 8.8 8.8 5.2 4.8 Sweden 7.2 13.7 12.1 8.6 8.9 8.0 7.4 4.2 4.2 5.8 Switzerland 3.6 4.0 6.5 5.6 3.0 2.9 3.4 0.8 1.4 1.9 United Kingdom 13.4 18.0 11.9 8.6 4.6 5.0 6.1 3.4 4.1 4.9 United States 11.3 13.5 10.3 6.1 3.2 4.3 3.5 1.9 3.7 4.1

Source: OECD Economic Outlook, December 1993, Table Al5.

2 3 4 5

Table 3.3 Types of Survey Data

Income or Living Standard Survey"

Combination of survey and administrative records Income Tax Recordsb Panel study Labor Force Survey Supplement:' Expenditure Survey!

A. Type of Survey

Netherlands, Australia, Canada, Ireland, Italy, Switzerland Finland, Sweden

France, Norway Belgium, Germany, Luxembourg United States, Austria United Kingdom

B. Differential Income Data Quality: A Conceptual Breakdown

Row Income Concept Difference

"True Income"

2 Administrative Record Income

3 Tax Reported Income

4 Edited Survey Income"

5 Reported Survey Income

Black Economy' Tax Evasion and Avoidance' Reporting Error• Item Non-response;

Sweden, Finland

Norway, France

Australia, United States, United Kingdom, Germany, Luxembourg, Canada, Belgium, Italy, Ireland Netherlands, Switzerland

a Survey primarily aimed at necessary living standards or income. Secondary aims may include other items such as wealth, expenditure, earnings, home ownership, finances, etc.

b Survey basis is from income tax records. Additional imputations are made for non-taxed income sources and related issues. In Finland, additional information is obtained from interviews.

c Primary survey objective is labour force participation, employment, unemployment, etc., special supplement provides income data.

d Primary purpose of survey is expenditure data, but monthly/weekly income information is also gathered. Black economy consists of net income from illegal activities.

f Tax evasion refers to legal sources of income which are not reported to income tax authorities, while tax avoidance refers to use of legal means of reducing tax liabilities.

g Reporting error refers to the difference between the amount of income reported on a survey and the amount actually received.

h Edited survey income refers to survey income which has been adjusted for item non-response (see Appendix 4 for additional detail). Item non-response refers to the failure of a respondent to report the amount of income received from a specific income source.

29

1989 1990

7.5 7.3 2.6 3.3 3.1 3.4 5.0 4.8 6.6 6.1 3.6 3.4 2.8 2.7 4.1 3.3 6.6 6.1 3.4 3.7 1.1 2.5 5.7 6.1 4.6 4.1

12.6 13.4 6.8 6.7 6.4 10.5 3.2 5.4 7.8 9.5 4.8 5.4

0 I

The lower part of Table 3.3 presents a reasonable way to envision how these differences are likely to affect the quality of income data. Five conceptual levels of income reporting are suggested and the approximate level at which each LIS country dataset lies. Income reporting in the upper rows is considered more complete than in lower rows. This is not intended to be more than indicative. There are studies of income distribution which lie between rows 1 and 2, such as those combining administrative data with survey information, as in the case of Sweden, Finland or the Blue Book series in the United Kingdom (see Ramprakash, 1975).

Up the rows from bottom to top, Table 3.3 begins with the amount of income actually reported by the population, excluding entire non-interviews but leaving partial or "item" non-response intact (row 5), as is the case in the Dutch and Swiss surveys. The Dutch and Swiss however make extensive imputations for some types of income (e.g. social security, child allowances). Their income data is perhaps more congruous with the next level, which is edited income (row 4) whereby all item non-responses are corrected. These adjustments may take many forms, including "hot-deck" imputation (e.g. the United States Census Bureau technique), where there is imputation of a value taken from the most recent (on the tape) respondent with the same characteristics as the non­respondent, or "cold deck" imputation, whereby the imputation is taken from a matrix which computes the average value of all respondents and assigns the average value to the non-respondent (see Appendix 4).

Row 3 refers to the amount of income recorded in information from tax records. Norwegian and French data are at this level. Table 3.3 shows that incomes for tax purposes are more reliably reported than survey incomes, which may be true for some but not all countries [for evidence of this in the United States, see Radner (1983)]. Tax-based surveys may also suffer from omissions of certain types of non-taxable income or non-taxpayers, in addition to tax evasion and tax avoidance. Row 2 raises gross incomes to the total amount recorded by some administrative intermediary, based on totals drawn from national income accounts or administrative records of government agencies. Swedish data are mainly drawn from such records. Differences between the top row, true income, and the administrative amounts usually arise from amounts of income which in principle are recorded in the national accounts, but are not readily allocated to individual households. This largely includes the underground, informal, or "shadow" economy.

3.3 Definition of income

One concern shared by economists and statisticians is the correspondence between macroeconomic aggregate income data- such as those contained in the SNA (Systems of National Accounts) and microeconomic survey­based income data. Although the United Nations Guidelines (1977) provides a valuable link, Chapter 2 explains that it was necessary to modify both the terminology and definitions of income categories in order to bridge the gap to income distribution.

The relation between the UN Guidelines and information in the LIS datasets is summarised in Table 3.4. The left hand column delineates the categories in the UN Guidelines (more detailed than in Table 2.1). Entries on the right are based on information supplied by the LIS contacts in each Member country. In each country, matrices were created to compare the UN Guidelines definition with micro-survey concepts. Appendix 5 presents the results in detail and Table 3.4 annotates major discrepancies such as in Sweden where realised capital gains on the sale of property are included in property income.

In certain cases such as the exclusion of employers' contributions to social security and similar schemes' and of imputed rents on owner-occupied housing, the UN Guidelines (see Chapter 2) are not followed. In other cases, the survey data cover the same items while the method of estimation may differ in a systematic way from that typically applied in national accounts. Entrepreneurial or self-employment income in micro-surveys usually refers to an "income tax" definition of self-employment income. Differences in depreciation allowances for fixed capital and in inventory accounting affect the micro survey and macro estimates in most countries.

Another example of a systematic difference is excluding lump-sum transfer payments for example at retirement under pension schemes or Christmas bonuses for those on social security. LIS datasets generally include cash social security benefits and other cash transfers (e.g. social assistance) paid by govemments.2

Occupational pensions, life insurance and/or annuity benefits received regularly are counted as occupational pensions. But one time or lump-sum transfers, whether state or private, are not consistently reported and therefore not included in basic income definitions. Similarly, non-recurring lump-sum payments such as realised capital gains, lottery winnings, inheritances, and/or insurance settlements are not always counted as income, although they are recorded in the LIS dataset where available and may be combined with other items.

1. Non-mandatory employer contributions such as employer contributions to voluntary employee health or life insurance policies in the United States or Canada (which are covered by employer and/or employee payroll taxes in other countries) are not counted.

2. In countries such as Sweden where sick pay is a government benefit, it has been added back into wages and salaries.

30

Table 3.4 Correspondence Between United Nations Guidelines and LIS Microdata-based Concepts of Income

United Nations Guidelines

1. Primary Income Gross of Consumption of Fixed Capital a. Compensation of employees

i. wages and salaries

ii. employers' contributions to social security and similar schemes

b. Income of members from producers' cooperatives

c. Gross entrepreneurial income of unincorporated enterprises, including withdrawals from entrepreneurial income of quasi-corporate enterprises

2. Property Income Received a. Imputed rent of owner­

occupied dwellings b. Interest

c. Dividends

d. Rent, royalties, patents, copyrights, etc.

3. Current Transfers and Other Benefits Received a. Social security benefits b. Pensions and life insurance - annuity

benefits

c. Other current transfers

4. Direct Taxes Paid

5. Social Security and Pension Fund Contributions a. Social security contributions

b. Pension fund contributions

Comments

Australia (1981 and 1985): no earnings in-kind; (1981) wages received from own limited liability company included with self-employment income. Austria: net income only; no earnings in-kind. Finland: includes wages and salaries paid to owners of limited liability companies. Ireland: not clear. Netherlands: not clear if in-kind earnings included (e.g. use of company car not included). Sweden: includes income in kind. United States: includes salaries of persons owning incorporated business but no income in-kind. Not included. Australia: not applicable, no employer contributions to social security.

Australia: included under gross entrepreneurial income. France: included under gross entrepreneurial income. Germany: included under gross entrepreneurial income. Ireland: not clear. Italy: included under gross entrepreneurial income. Netherlands: included under wages and salaries. Sweden: included in gross entrepreneurial income. Australia (1985): excludes wages and salaries paid to owners of limited liability companies (recorded under wages and salaries). Finland: net of consumption of fixed capital, and excludes wages and salaries paid to owners of limited liability companies (recorded under wages and salaries). France: net of consumption of fixed capital. Italy: net of income tax, social contributions, and consumption of fixed capital. Sweden: net of consumption of fixed capital. United States: net after expenses and consumption of fixed capital.

Not included.

Germany: interest and dividends not separated. Italy: after tax only. Sweden: interest and dividends not separated, and capital gains (amount defined in the taxation rules) from selling property are included in property income. Australia (1981): included under rent, royalties, etc. France: 30 per cent of incomes from stocks and shares not included in tax return, so not included. Germany: interest and dividends not separated. Italy: after tax only. Australia (1981): includes dividends. Australia (1985): only rents collected. Germany (1984): only rents and royalties available. Italy: only rents available. Sweden: net rent after expenses. United States: rent is net rent after expenses.

Australia: annuities available from other surveys specifically, classified on pensions. Austria: only social retirement benefits and public sector pensions. Germany (1984): only old age pensions. Ireland: only occupational pensions. Life insurance benefits included or counted as lump-sum income in some nations. Varies by country and type

Property taxes available in some surveys. For instance: France: land taxes and car taxes not included. Germany (1984): property and wealth taxes not available. Netherlands: imputed. Sweden: wealth tax included.

Generally available and included. Australia: not applicable. Austria: could be estimated. Employee contributions generally counted in wages and salaries

31

3.4 Cash and non-cash benefits

Different countries rely on different mixes of cash and non-cash benefits to provide income support, access to necessities (health care, food, housing) and human capital investment (education). The money income concept in this report includes only public cash and near cash benefits. That country differences develop due to preferences for cash versus non-cash transfer is to be expected. A similar type of difference may occur if countries rely on employers to provide some types of benefits (e.g. occupational pensions or health insurance for workers in the United States), while governments provide others (e.g. health insurance in most other countries). Though no systematic evidence exists of employers versus government public benefits, investigating the mix of cash versus non-cash benefits is still possible in OECD countries.

The OECD Social Data Bank allows comparisons of public cash subsidies (social welfare expenditures) with in-kind subsidies (public health and education benefits) as a percentage of GDP across countries in 1987 (Table 3.5).This reveals that non-cash benefits tend to be much more uniform than cash benefits. Education benefits vary from 4.0 to 6.9 per cent of GDP, health benefits range from 3.9 to 7.7 per cent of GDP excluding Turkey (1.4 per cent). In contrast, cash social welfare subsidies run from 3.3 per cent in Australia to 14.1 per cent in Sweden. The relative uniformity of non-cash benefits also extends to the likely number and type of beneficiaries. Basic public education and health benefits are much more equally distributed across the population at large than cash benefits, which are more likely to benefit the poor, disabled, sick and unemployed.

In-kind benefits tend to be a small (large) share of total social transfers relative to cash benefits in countries with small (large) shares of GDP spent on the latter. Because of this pattern, the exclusion of non-cash benefits are not expected to greatly affect the level or trend in cash income inequality found here.3

Table 3.5 Expenditures for Health, Education and Social Welfare Among the Non-aged as a Percentage of GDP in OECD Countriesa in 1987

Country In-Kind Cash Social Total In-Kind Welfare As Per Cent of

Total

Education Health

Australia 4.9 5.5 3.3 13.7 75.9 Austria 5.9 5.6 na na na Belgium 5.2 6.2 10.0 21.2 47.2 Canada 6.9 6.5 6.7 20.1 66.7 Finland 5.8 5.8 13.1 24.7 47.0 France 5.4 6.5 8.7 20.6 57.8 Germany 4.0 6.4 8.1 18.5 56.2 Irelandb 6.2 6.5 8.6 21.3 59.6 Italy 4.4 5.7 3.5 13.6 74.2 Japan 4.5 4.9 na na na Luxembourg na 6.7 7.7 na na Netherlands 6.9 6.1 12.1 25.1 51.8 New Zealand 5.1 6.0 4.0 15.1 73.5 Norway 6.2 7.2 11.4 24.8 54.0 Portugal 4.0 3.9 4.6 12.5 63.2 Spainb 4.6 4.5 4.9 14.0 65.0 Sweden 6.8 7.7 14.1 28.6 50.7 Switzerland 5.0 5.1 na na na Turkey na 1.4 na na na United Kingdom 4.9 5.2 7.7 17.8 56.7 United States 4.8 4.7 3.5 13.0 73.0

a Excludes Denmark, Iceland and Greece. b 1980 value. Source: OECD Social Data Bank.

3. In a recent LIS-based study covering Australia, Canada, Germany, the Netherlands, Sweden, the United Kingdom, and the United States, Smeeding, Saunders et al. (1993) found that imputation of health and education benefits had an equalising impact in all countries, but did not affect the inequality ranking of countries when cash and non-cash rankings were compared to those based on the same disposable cash income ranking used in this report.

32

3.5 Taxes

Table 3.6 Employees' Social Security Contributions and Personal Income Tax (Including Standard Tax Reliefs) at the Income Level

of an Average Production Worker in 1987 as a Percentage of Gross Earnings•

Country Single People 1\vo-Child Families

Employee Social Personal Total Employee Social Personal Security Income Tax Security Income

Contributions Tax• Contribution Tax•

Australia 1.3 22.4 23.7 1.3 17.8 Austria 16.4 9.5 25.9 16.4 6.4 Belgium 12.1 23.5 35.6 12.1 15.6 Canada 4.1 21.3 25.4 4.1 12.2 Finland 3.4 30.8 34.2 3.1 25.2 France 16.7 6.8 23.5 16.5 -Germany 17.1 18.6 35.7 17.1 8.6 Greece 13.3 3.5 16.8 13.3 1.8 Ireland 7.8 27.9 35.7 7.8 17.9 Italy 8.8 18.4 27.2 8.8 15.2 Japan 7.0 8.5 15.5 7.0 2.7 Luxembourg 12.2 14.0 26.2 12.2 1.0 Netherlands 25.5 11.9 37.4 25.5 8.9 New Zealand - 23.6 23.6 - 20.5 Norway 10.9 22.7 33.6 10.4 15.2 Portugal 11.0 5.1 16.1 11.0 4.0 Spain 6.0 14.1 20.1 6.0 10.0 Sweden - 36.6 36.6 - 35.0 Switzerland 10.3 10.8 21.1 10.3 6.0 Turkey 9.4 22.4 31.8 9.4 22.4 United Kingdom 9.0 20.3 29.3 9.0 16.5 United States 7.1 20.0 27.1 7.1 13.3

a Excluding Denmark and Iceland.

Total Tax

19.1 22.8 27.7 16.3 28.3 16.5 25.7 15.1 25.7 24.0

9.7 13.2 34.4 20.5 25.6 15.0 16.0 35.0 16.3 31.8 25.5 20.4

b The figures shown in this table do not take into account the effects of non-standard expense-related reliefs, which are found in all countries' income tax systems. In some countries the inclusion of these reliefs would substantially reduce the average rates of income tax shown in this table.

Source: OECD (!990), Tables l, 3.

Personal property, wealth and church taxes are generally not counted as direct taxes (though Swedish data include wealth tax). Since it is not possible to separately estimate the proportion of property taxes on renters (due to the uncertain incidence of the property tax), property taxes on owner-occupied homes (or other durables) are not subtracted, and church taxes are considered voluntary uses of income (user charges).

Because of varying degrees of reliance on employer and employee social security contributions and different mixes of personal, business, earnings, income, property and goods (expenditure, VAT, sales) taxes in OECD countries, the manner in which taxes are collected is liable to affect the results of analyses. Here only personal income taxes (including standard tax reliefs and subsidies such as child allowances) and employee payroll taxes are taken into account. The effect of these taxes is illustrated in a recent report presenting the level of employee social security contributions and personal income taxes relative to the gross earnings of an average production worker in 1987 (OECD, 1990). Both the total burden of these two taxes and each separate portion of taxes vary considerably as a percentage of the gross earnings for a production worker. Two cases are presented in Table 3.6: one for a single worker, the other for a worker with two children.

Employee social security contributions range from zero in Sweden and New Zealand (where all such taxes are paid by the employer) to over 25 per cent in the Netherlands. As expected, very little variance exists in employee payroll taxes between the single earner case and the two-child family case. Personal income taxes vary from 5.1 per cent in Portugal to over 36 per cent in Sweden for singles, and from zero in France to 35.0 per cent in Sweden for two-child families. While countries with larger cash social welfare benefits (Table 3.5) tend to have heavier employee social security and income tax burdens, there is still quite a bit of residual variance across countries. For instance, Ireland has a relatively heavy income and employee tax burden, but only average cash benefits. No evidence is presented on the structure (progessivity) of personal income or employee social security taxes, though they too vary across countries.

33

Country data vary from these generally applied rules, as is illustrated in Table 3.4. Earnings in-kind are not covered in a number of countries; property income is sometimes recorded net of tax; salaries of unincorporated businesses owners may be reported as salary; and coverage of "other current transfers" varies.

Differences in LIS datasets should be kept in mind throughout the analysis. In certain cases, for example, they do not affect the total disposable income, but only the division into component variables. Reporting property income net of tax (as in Italy) should not, in principle, influence the measurement of disposable income. And including the salary paid to business owners in wages and salaries should not impact the total income recorded. However, the reporting of earnings net of tax (in France, Italy, Luxembourg, Belgium) will affect comparisons with countries that report earnings gross of tax. The wide range of possible choices in defining income and other variables also means that researchers using the same underlying dataset may arrive at different estimates of income distribution. This should be considered when comparing the results of this report with others (see Chapter 5) using the same datasets. They may be more accurate reflections of national positions than LIS definitions, which were designed to cater to the needs of international comparability.

3.6 Comparison of reported income data with external aggregates

Similarities and differences in the quality of reported income amounts are important in survey measurement. What can be learned about the overall quality of income data from comparisons with national accounts and other external sources is an important question for the LIS. The results for seven countries are summarised in Table 3.7, with more complete coverage in Appendix 6.

Three points should be made before comparing reported income amounts. First, national income accounts or administrative data may not always be superior to survey data in some countries. National accounts aggregates are themselves estimates whose reliability is the subject of much literature. In the United Kingdom, for example, Maurice (1968) grades different variables according to their reliability - self-employment income is placed in

Table 3.7 Quality of Income Data: Ratio of Survey Estimates to Adjusted National Accounts Estimates (in Per Cent)

Country and Year

United United Australia Canada Finland Germany Italy Kingdom States

Income Item (1981-82) (1985-86) (1981) (1987) (1987) (1983) (1989) (1987) (1979) (1987)

Wages and Salaries 92.2 100.6 101.6 100.0 101.5 108.8 106.9 93.7 97.4 99.4 Self-Employment Income 124.91 83.7 78.2 90.4 73.4 36.32 53.1 75.7 84.2 78.5 Property Income 50.7 66.7 60.5 47.7 82.5 78.4 50.6 45.1 55.2 Occupational Pension Income 85.4 74.5 81.5 81.6 Government transfers 75.4 66.4 77.5 75.5 90.6 50.6 74.3 90.9 82.8 86.9 Total (All Income)' 83.0 81.7 92.4 90.1 93.54 76.9 80.6 89.0 89.0 89.2

Notes: I. In 1981-82 (but not 1985-86) wages received by persons from their own limited liability company have been grouped with self-employment income,

whereas the convention followed by the national accounts is to classify this income as wages and salaries.

Sources:

2. Includes property income. 3. Based on sum of items presented above only. Some income amounts, e.g. alimony and child support or private transfers, have no administration

data to which the survey data can be compared. 4. Sum of items above.

Australia:

Canada 1981:

Canada 1987:

Finland: Germany: Italy: United Kingdom:

United States 1979: United States 1987:

comparisons between Unit Record File and National Accounts provided by Bruce Bradbury of Social Policy Research Centre, University of New South Wales. survey data from Survey of Consumer Finances for 1981; comparisons from unpublished tabulations based on family income data provided by Statistics Canada. survey data from Survey Consumer Finances for 1987; comparisons from SCF/National Account Reconciliation, Statistics Canada, Household Survey Division. provided by Aino Salomaki of the Government Institute for Economic Research, Helsinki. comparison of SOEP Transfer Survey and National Accounts, from Kassella and Hochmuth (1989), Table 14. Brandolini (1993). data from Family Expenditure Survey for 1977; comparisons as reported by Atkinson and Micklewright (1983) using, in part, methodology developed by Ramprakash (1975). survey data from the Current Population Survey for 1979, as reported in U.S. Bureau of the Census (1981, Table A-2). survey data from Current Population Survey for 1987 as reported in U.S. Bureau of the Census (1991, Table A-1, page 74).

34

range B, with a 90 per cent confidence interval of +1- 3-10 per cent. In the case of property income, she comments that:

"like all figures obtained as residues, the estimate of personal income from rent, dividends and net interest cannot be regarded as accurate" (1968, p 103).

Second, administrative data may need adjusting to produce estimates for comparable income concepts and populations before comparing it to survey data (or tax data). For example, national accounts may include households together with non-profit organisations. It may be necessary to subtract the interest income received by tennis clubs or charities, or income received by households not in the survey population (e.g. non-residents, the deceased, and the institutionalised).

As put by the United States Bureau of the Census,

"Deriving independent estimates of aggregate income for purposes of evaluating the survey data is difficult. The survey and administrative sources use different definitions, cover different universes, and are based on concepts that are not exactly the same. Therefore, adjustments to the administrative sources must be made to help correct for these inconsistencies and arrive at a valid independent estimate that can be used to make fair and accurate estimates of the quality of the survey estimates ( ... ). [In the United States, these adjustements] attempt to 1) remove income that is received by the institutional population, the deceased, and persons not residing in the United States at the time of interview, 2) remove any components of income that are received as 'in-kind' payments or benefits, and 3) remove any lump-sum or one-time payments, withdrawals, etc." (US Department of Commerce, 1991, p. 215).

In the United Kingdom, adjustments are necessary for differences in timing. In the case of self-employment income, survey figures relate to "profit in the most recent 12 months for which figures are available". Evidence from the Royal Commission on the Distribution of Wealth (1979, Appendix D) showed that the average lag from the mid-point of the accounting year was 15 months. According to Atkinson and Micklewright (1983, Table 3), adjustment for timing differences raises the survey estimate in 1977 from 50.9 per cent of national accounts to 58.7 per cent. National accounts estimates also differ in their definition of profits, so that applying the same adjustment as Ramprakash (1975) raises the survey figure to 75.7 per cent. Adjustments are also made in other categories. For example, national accounts estimates of occupational pensions include lump-sum payments and refund of contributions. Atkinson and Micklewright show (1983, Table 4) that the 1977 comparable estimate of occupational pensions is only £3,334 million or 54.9 per cent of the total national accounts figure (£6,070 million) once these adjustments are made.

Third, it is important when comparing income amounts to bear in mind that differences between income aggregates may arise from different sources: varying non-response to the survey (for example, a low response rate from high income groups may cause understated investment income); item non-response by households taking part; or inaccurate reporting by respondents. For the last case, Table 3.7 illustrates that the United States total wages and salaries reported in the US survey are 99.4 per cent of the adjusted administrative amount in 1987. This does not mean that all individuals reported 99.4 per cent of their true wages and salaries. This is an average based on some individuals who have over-reported or under-reported their incomes (see Appendix 6). Multiplying reported amounts by the reciprocal of the percentage reported is not the appropriate way to make an adjustment. A direct record-for-record comparison is needed for further discussion. Under-recording may appear as failure to report an income source, but it may be indistinguishable from genuine zero entries. Overall readings of data quality, such as those in Table 3.7, do not provide all the ingredients necessary to adjust microdata for reporting errors (see Radner, 1983).

Table 3.7 shows information for seven countries on the ratio of survey to administrative estimates for each of five categories of income. In three cases, there are comparisons for two years. This information is based on the work of others which may not be fully comparable. The LIS decided not to make comparisons in this context because it would require a great deal of research (see Smeeding, 1982).

Total or "all income" constitute about 90 per cent of comparable national income totals for four of the seven countries: Canada, Finland, the United Kingdom and the United States. Other countries registered an aggregate shortfall of some 20 per cent. This is generally re-assuring because in some cases at least part of the difference is explained by the fact that the totals are not fully comparable: for example, that the institutional population in some countries is still included in national accounts totals, or that lump-sum payments from national accounts were impossible to exclude.

In specific categories, estimates often differ by some non-trivial degree. The following discusses the individual items by applying these broad classifications:

"good" : "moderate" : "questionable" :

90 per cent or more 70 - 89 per cent less than 70 per cent.

35

- Wage and salary income is generally well reported, with figures close to or above 100 per cent. The lowest figure- that for Australia in 1981-82 -illustrates the definitional problems which may still remain. Wages received in 1981-82 (but not 1985-86) by persons from their own limited liability company are grouped with self-employment income, whereas national accounts classify this income as wages and salaries. This may account for shortfalls in wages and salaries in 1981-82. The total is essentially the same in the two sources in 1985-86 (see Table A6.2).

- Self-employment income reporting differs substantially across surveys, though Atkinson and Micklewright (1983) deem it hard to specify the meaning of "self-employment income" either in surveys or in the national accounts. In the case of Germany, the figure is combined with that for property income, which tends to have a lower rate of coverage. Apart from Italy, the accuracy of other countries is classified as "moderate"; the estimate can never be described as "good" (aside from Australia in 1981-82 for the reason explained above).

- Incomplete reporting of property income plagues virtually all types of income surveys. Because of its highly skewed pro-rich distribution by income and age, this differential reporting problem must be carefully noted. For instance, adjustments in the United States for non-reporting all types of income among the elderly in the 1973 Current Population Survey, based on a record-for-record match with several sources of administrative data, indicate that the overall incomes of the elderly would increase by 37 per cent, if accurately reported, compared to about 9 per cent for the population as a whole (Radner, 1983). This variance was mostly due to property income non-reporting among the high-income elderly, as well as under-sampling.

- The accuracy of government transfer income reporting differs across countries, with only Finland and the United Kingdom classified as "good". The pro-poor nature of transfer income demands careful attention to differences which may exist across types of transfer benefits, because means-tested benefits are typically poorly reported. While the US overall transfer income reporting rate is 82.8 per cent, means tested benefits are only about 75 per cent reported (Smeeding, 1982, Appendix F). In Canada, only about 50 per cent of social assistance, provincial income supplements, and provincial tax credits, which are largely means-tested, are reported. Problems may arise when identifying the type of benefit. Though the United Kingdom monthly data generally report overall transfer income better than other annual income­based surveys, evidence of misreporting of transfer income by transfer type does exist (Atkinson and Micklewright, 1983).

It is also possible to compare government transfers with administrative data, which may be a major source for national accounts. The check is therefore not necessarily wholly independent from that already considered, though information may be available from administrative sources on the number of recipients as well as the average amounts. In the case of Australia, for example, Income Survey estimates of selected aggregate income support benefits received during the year were compared with Department of Social Security (DSS) expenditure statistics. Although excluding the institutionalised population from the scope of the Income Survey makes these two sources of data not entirely compatible, some problems of under-reporting are apparent. Income from Unemployment Benefits, for example, is apparently under-reported by 30 per cent. Similar rates of under-reporting apply to supporting parents benefits and widows pensions. The exclusion of the institutionalised will mainly affect the comparison with age and invalid pension expenditure. Similar rates of under-recording apply to the current data on numbers of persons in receipt of the different pensions or benefits. However, it is not known whether the discrepancy is mainly a result of non-reporting of income receipts by respondents, or under-sampling of the pensioner/beneficiary population.

The United Kingdom data (Atkinson and Micklewright, 1983, Table 6) appear to accurately assess the number of beneficiaries in the Family Expenditure Survey, except for Supplementary Benefit where the numbers in 1977 were 87.1 per cent of the administrative total. It is presumed that a sizeable part of this shortfall arises because of respondent failure to separate Supplementary Benefit from other benefits which are payable simultaneously. In the case of Ireland, comparisons show that the aggregate level of transfers reported in the survey was about 90-95 per cent of actual social security expenditures and, despite some misclassification of widow for old age pensions, its distribution in social welfare schemes converges with official statistics.

Another source of comparisons is with the distribution of income indicated by tax statistics. In the case of Belgium, a research group at the University of Louvain (HIVA) estimated the distribution of net-taxable incomes on the basis of survey-data, and compared this with official statistics. It appears that both lower and higher incomes are over-represented. In the case of Ireland, the level and distribution of wage income is close to that reported by the Revenue Commissioners for the nearest fiscal year; the main problems concern self-employment and investment income.

Such differences should be carefully noted when comparing relative incomes across countries. Due to relatively better reporting of property income and relatively worse reporting of transfer income, overall measures

36

of income inequality in Canada may be more unequal than in other countries - on account of relative income data quality alone. Still, our major aggregates tend to be reasonably consistently reported across these countries.

3.7 Procedures adopted in use of LIS data in this study

It is rare that survey data are presented in a completely raw form. Even where little reference is made to data adjustments, as in the UK Family Expenditure Survey, item non-response typically leads to using imputation methods to complete at least some of the data gaps. The data used here have been supplied in a form where a number of such adjustments have already been made. The comparisons made in the previous section are in general conducted after such adjustments have been made.

LIS data have undergone further adjustments. In a few cases LIS has, under the supervision of country officials, changed data files after their arrival. These include removing several large negative incomes caused by income tax losses in the Switzerland file (essentially a country-imposed bottom code). A rather different LIS decision was to subsample the Australian, Canadian and US data tapes to a size more consistent with the other countries and more manageable for the remote access system. In this case, every third (fifth) record was selected for Australia and Canada (the United States) and every selected record had its household weight multiplied by a factor of three (five).

Particular reference should pe made to top and bottom codin~- the process of arbitrarily assigning a maximum (top) or minimum (bottom) level to an income type. In some countries, survey takers protect the confidentiality of their respondents by top or bottom coding incomes, both income aggregates (such as gross incoJlle or disposable income) and individual income components (such as earnings), or other variables (e.g. top coding age at 80 or 85). Indeed many components of income can be negative or zero. That countries differ in their practice of top and bottom coding necessitates considering the procedure they applied.

The results provided here are not always sensitive to top and bottom values. This applies to the percentile break points: none of the percentile points in the income distributions (e.g. 95th or 90th percentile) are affected by top coding. But, some summary measures of inequality, such as the Theil index, are sensitive to top coding, while others, such as the Atkinson index, are sensitive to bottom coding. The commonly used Gini index is perhaps less influenced, but a recent paper by Fichtenbaum and Shahidi (1988) indicates that the Gini coefficient may still be sensitive to top-coding practices.4 Adjusting top and bottom incomes affect the cumulative income shares, and hence the Lorenz curves.

Because the impact of top coding is unknown in most of LIS surveys (i.e. the aggregate shares of income or maximum income values without top coding) no adjustments were done for top coding. A small number of experiments were done doubling the maximum value of the top income in two countries. This experiment produced less than a .001 change in the calculated Gini. The Theil measure is not implemented.

In the main results of chapter 4, no adjustment is made for bottom coding. Since some datasets already come to LIS bottom coded (e.g. no value less than zero or a minimum set at some small positive value), this affects the degree of comparability of, in particular, the decile shares; and this qualification should be borne in mind by the reader. For the reasons indicated above, we have however in the calculations of the Gini and Atkinson coefficients bottom coded the datasets to correspond to one per cent of the mean disposable income. In Chapter 6, dealing with components of income where zero or negative values are more common, the following procedure was adopted:

- For any dataset with a zero or positive minimum value for an income concept (including primary income, market income, disposable income, etc.), array these incomes from lowest to highest creating 40 separate income groupings (quantiles). Calculate the mean value for the bottom 2.5 per cent (quantile) of the population rated by income using only datasets which have zero or positive minimum values for a given income concept. Divide this mean value by the upper bound of the bottom quantile to get a minimum ratio of mean to upper bound. Set equal to this value the mean income of the bottom 2.5 per cent for all other datasets.

For instance, this procedure produced minimum ratio of mean values to upper bounds for the bottom quantile of the primary and market income distributions which were between 39 and 44 per cent of the upper bound of the bottom quantile. Experiments with this methodology using 40, 50 or 60 per cent of the upper bound of the bottom

4. According to the Fichtenbaum and Shahidi (1988) paper, the U.S. dataset used in this study had a top-code in 1979 which could create up to a 4.1 percent difference in the top coded Gini vs. a Gini which estimated an upper tail of the distribution using a Pareto distribution. This problem was lessened in 1985 when the top code was tripled in the U.S. dataset. Yet some bias may remain.

37

quantile as a mean value, or calculating the average value of the mean income in the bottom quintile for datasets with zero or positive values, produced no appreciable difference in the Atkinson coefficient when compared to this report's chosen methodology.

3.8 Conclusion

The aim of this study has been to improve the comparability of income distribution estimates across countries, but this chapter ends by emphasising three points:

Full comparability is impossible and differences will always remain between data for different countries or in the interpretation of data in the social and economic contexts of different countries; Adopting a common set of definitions means that estimates for any one country may be less satisfactory if that country were isolated, since in a comparative study one is compelled to use the most common practice rather than the optimum in each case; and Standardizing on a common approach does not necessarily imply comparability: the same definitions may be applied in two countries but with different consequences (e.g. the choice of equivalence scale may affect the relative extent of low incomes in two countries, because one has a higher proportion of large households).

38

Chapter4

BASIC RESULTS

This chapter has a first look at the results, concentrating on the distribution of disposable income. Section 4.1 begins by examining the distribution for the most recent available year in each of seventeen OECD countries. These results relate to the period 1984-88 (with one exception). Changes over time are the subject of Section 4.2, which illustrates how distribution has changed in the 1980s in eleven countries. As explained in Chapter 2, income distribution can be measured in a number of different ways - Section 4.3 examines the sensitivity of the estimates to the methods employed.

4.1 Distribution of income in seventeen OECD countries

This section describes the evidence from the LIS dataset about the distribution of disposable income in OECD countries in the mid to late 1980s (although the results for Switzerland relate to 1982). The estimates allow a comparison of the degree of income inequality in the different countries, though full comparability is impossible. Though this report's comparability is stronger than others which lacked access to the micro-data, sources still differ in many respects. The comparisons may similarly be sensitive to the methods employed. This section concentrates on disposable income per equivalent adult, using an "intermediate" scale of household size to the power of a half, and weights each household by the number of persons. Section 4.3 compares the results to those obtained using per capita incomes, or without adjusting for household size, and examines the sensitivity of estimates to the use of household rather than person weights.

The findings depend in part on the method of presentation. In order to allow readers to consider the distributions in different ways, and draw their own conclusions, Tables 4.1 to 4.4 present four different ways of looking at the evidence. Data include all sources of income for all countries but one: Austrian data exclude self­employment income. Tables 4.1, 4.2 and 4.3 include a second panel presenting results for Austria and six other countries with disposable income excluding self-employment income. In broad comparisons, the exclusion of self­employment income does not seem to greatly affect the results. Because the Austrian income base is not the same as other countries, it is not included in rankings that summarise the whole distribution (such as Table 4.4). Tables A 7.1 and A 7.2 provide information on the pattern of different sources of income at different percentiles of the distribution, and at low, middle and high income levels.

Percentiles of the distribution

The first method of presentation (Table 4.1) expresses the percentiles of the distribution as percentages of the median. For example, in the United States dataset for 1986, the median equivalent income per adult was $13,364 per year. For a family of 4, this corresponded to a total income of $26,728, since the equivalence scale used is the square root of the household size- in this case, the square root of 4, i.e. 2 (in that year the average official poverty threshold for a family of 4 in the United States was $11,203). The equivalent income per adult at the lower quartile, 25 per cent up from the bottom, was $8,240, so that the twenty-fifth percentile, expressed as a percentage of the median, was 61.7 per cent. This is shown as P25 in Table 4.L The bottom decile was $4,639, giving a percentage of 34.7 per cent, and the top decile was $27,540, giving a percentage of 206.1 per cent. The ratio of the top to bottom decile, referred to as the decile ratio, is shown in the final column.

39

Table 4.1 Summary of Income Distribution in OECD Countries: Percentiles of Median

PIO P, P, P,o P, p ,JPIO

A. Disposable Income Per Equivalent Adult

Australia 1985 46.5 66.4 142.1 186.5 218.5 4.01 Belgium 1988 58.5 74.5 128.8 163.2 190.8 2.79 Canada 1987 45.8 68.5 137.5 184.2 218.0 4.02 France 1984 55.4 72.1 139.7 192.8 233.5 3.48 Germany 1984 56.9 75.0 132.7 170.8 201.7 3.00 Ireland 1987 49.5 66.7 150.9 209.2 252.2 4.23 Italy 1986 48.9 68.8 145.0 197.9 233.8 4.05 Luxembourg 1985 58.5 75.1 132.7 184.0 228.1 3.15 Norway 1986 55.3 76.0 128.7 162.2 187.3 2.93 Sweden 1987 55.6 75.6 125.1 151.5 170.4 2.72 Switzerland 1982 53.9 73.6 134.3 185.1 244.6 3.43 United Kingdom 1986 51.1 67.6 144.6 194.1 232.1 3.79 United States 1986 34.7 61.7 149.6 206.1 247.3 5.94

Finland 1987 58.9 76.5 125.5 152.7 173.6 2.59 Netherlands 1987 61.5 75.7 135.0 175.0 206.4 2.85 New Zealand 1987/8 53.6 186.6 3.48

B. Disposable Income (Excluding Self-employment Income) Per Equivalent Adult

Austria 1987 56.3 75.9 129.9 162.5 186.7 2.89

France 1984 57.6 73.5 137.5 184.6 223.1 3.20 Germany 1984 57.2 75.0 131.6 167.9 192.3 2.94 Italy 1986 50.4 69.3 142.9 187.8 224.3 3.72 Sweden 1987 57.8 77.0 124.1 149.3 167.9 2.58 United Kingdom 1986 53.1 68.6 145.6 194.7 226.9 3.67 United States 1986 34.7 61.1 148.7 203.7 242.7 5.87

Notes: The equivalence scale employed in these calculations is the square root of the household size:

household size scale I person 1.00 2 persons 1.41 3 persons I. 73 4 persons 2.00 5 persons 2.24 6 persons 2.45

The households are weighted by the number of household members. Source: 13 countries above line in Panel A, and Austria, from the LIS; Finland data from the income distribution statistics

database of Statistics Finland; the Netherlands: data supplied by J. T. M. van Laanen of the Department of Statistics on Income and Consumption, the Hague (based on the same data as those contained in the LIS database but applying a different set of weights); New Zealand: data supplied by Phillipa O'Brien, Department of Statistics, New Zealand.

The decile ratio in the United States is 5.94, the largest value recorded in the last column of Table 4.1. From Table 4.1, the lower part of the distribution of equivalent disposable income does appear to be different in the United States. The bottom decile is only slightly over a third of the median, compared with around 45 per cent in Australia and Canada, and values in excess of 55 per cent irt Belgium, Finland, France, Germany, Luxembourg, Norway, and Sweden. In the Netherlands, the bottom decile is over 60 per cent of the median.

Differences are less marked at the top. The top decile is over 190 per cent of the median in France, Italy, the United Kingdom, the United States and Ireland. As far as P95 is concerned, the values for France, Italy, Luxembourg, Switzerland, the United Kingdom, the United States and Ireland all range from 225 per cent upwards. The distribution at the top is noticeably less unequal in Finland and Sweden, with Germany and the Netherlands having an intermediate position. The overall pattern of the distribution is compared for Sweden and the United States in Figure 4.1, which shows the decile points expressed as percentages of the median. These two countries represent two ends of the spectrum in terms of the recorded degree of inequality. The decile points in Sweden are higher to the left of the median (100 per cent) and lower to the right.

Estimates excluding self-employment income suggest that, for comparable countries, the percentiles are slightly closer to the median. For example, the decile ratio for Sweden is 2.58 compared with 2.72 when self­employment income is included. The Austrian distribution appears, on this basis, quite close to that in Germany, with top percentiles somewhat lower.

40

Figure 4.1 United States and Sweden Deciles

- United States -Sweden

Low and high incomes

An alternative way to express the same information often adopted in international comparisons is to ask what proportion of the population is below specified percentages of the median. How many people are for instance living in households with disposable incomes below half the median? Table 4.2 shows the cumulative percentages below different percentages of the median. If, following the European Commission, 50 per cent of average income is taken as a standard of "low income", and if "average" is measured by the median, then the United States stands out with 18.4 per cent of the population living in households below 50 per cent of the national median. The next highest figures are those for Australia and Canada (around 12 per cent). The lowest figures are recorded in Belgium, Finland and Luxembourg, followed by Austria (part B of the table), Germany, France, Norway and Sweden. The proportion with "modest" incomes between 50 and 70 per cent of the median is lowest in the United States, and largest in the United Kingdom.

At the other end of the scale, "well-to-do" may be defined as those living in households with equivalent income above 1.5 times the median. In the United States in 1986, this amounted to some $40,000 a year for a family of four. The proportion above 150 per cent of the median is lowest in Sweden and Finland (only 10.5 per cent and 11.0 per cent of the population respectively). The proportions in Australia, Ireland, Italy, the United Kingdom, and the United States are more than double that in Sweden.

One interesting feature revealed by Table 4.2, and brought out graphically for a selection of three countries in Figure 4.2, is the extent of concentration around the middle of the distribution. For this purpose, the middle ranges from 80 to 120 per cent. In the United States, the United Kingdom, Italy, Ireland, and Australia, about a quarter of the population are within this central range. In Canada, France, and Switzerland, the proportion is around a third. In Germany and Luxembourg it is higher, and in the Scandinavian countries some 40 per cent are within 20 percentage points of the median. The shape of the distribution is different from one country to another, as is illustrated in Figure 4.2. France has around one third below 80 per cent of the median, one third above 120 per cent, and one third in the middle group. The United States shows more in both the lower and upper groups and Sweden fewer.

41

Table 4.2 Summary of Income Distribution in OECD Countries:

Cumulative Proportions Below Percentiles of Median

50 60 70 80 100 120 !50 200

A. Disposable Income Per Equivalent Adult

Australia 1985 12.3 20.3 27.5 34.5 50.0 63.4 78.5 92.6 Belgium 1988 4.7 11.1 19.8 29.9 50.0 68.8 85.9 96.3 Canada 1987 12.2 18.2 26.2 34.3 50.0 65.3 80.5 92.7 France 1984 7.5 13.2 22.8 32.1 50.0 65.0 79.0 91.1 Germany 1984 6.5 12.6 21.5 30.2 50.0 66.2 83.4 94.6 Ireland 1987 10.7 19.9 27.5 36.0 50.0 61.4 74.8 88.7 Italy 1986 10.5 17.4 27.2 35.2 50.0 62.7 76.7 90.4 Luxembourg 1985 5.4 11.2 21.0 29.6 50.0 67.1 82.6 94.9 Norway 1986 7.3 12.8 20.1 29.0 50.0 68.2 85.2 95.5 Sweden 1987 7.6 12.6 20.0 28.6 50.0 70.9 89.5 98.0 Switzerland 1982 8.0 13.8 21.7 31.3 50.0 66.4 81.6 91.7 United Kingdom 1986 9.1 17.6 27.1 35.2 50.0 62.7 77.2 91.2 United States 1986 18.4 24.1 30.3 36.5 50.0 61.7 75.1 89.0

Finland 1987 5.0 10.7 18.7 28.5 50.0 70.5 89.0 98.0

B. Disposable Income (Excluding Self-employment Income) Per Equivalent Adult

Austria 1987 6.7 11.9 20.4 29.6 50.0 68.3 86.0 96.3 France 1984 6.3 11.6 21.6 31.4 50.0 65.7 80.1 92.3 Germany 1984 6.6 12.4 21.3 29.1 50.0 66.1 84.3 95.9 Italy 1986 10.0 16.1 26.0 35.4 50.0 63.3 77.8 91.9 Sweden 1987 6.9 11.5 18.7 27.8 50.0 71.8 90.3 98.4 United Kingdom 1986 7.7 16.0 26.4 34.7 50.0 62.1 77.3 91.3 United States 1986 18.6 24.4 30.5 36.7 50.0 61.5 75.5 89.4

Notes: see Table 4.1. Source: LIS.

Lorenz curves

A third form of presentation is in terms of shares of total income, which are the ingredients for the conventional Lorenz curve. Table 4.3 shows cumulative shares by decile groups. This picture is partly different because the mean income is taken as the basis for expressing relative position, rather than the median. There are quite substantial differences between the ratio of the mean and the median in different countries. In the Scandinavian countries, the mean is close to the median, whereas in the United States the mean is 13 per cent higher than the median, and in Switzerland the mean is some 20 per cent higher than the median.

Countries with a higher ratio of mean to median are normally expected to exhibit more inequality in Table 4.3 than in the earlier results. Switzerland has the same share for the bottom decile as Canada, whereas in Table 4.1 the bottom decile is some 8 percentage points higher. On the other hand, this is not the only factor at work. The share of the bottom decile group depends not only on the income level demarcating this group (the first decile point) but also on the distribution of incomes below this point.

The estimates in Table 4.3 excluding self-employment income suggest that, for comparable countries, the cumulative shares are slightly larger, though in some cases (such as the United States) the differences are slight. In the case of Austria, which only appears in part B of the table, the Lorenz curve lies above that for Germany, with a maximum difference of 1 percentage point. The Lorenz curve for Austria starts above that for Sweden, and then crosses around the lower quartile. In view of the difference of definition, Austria is not included in the subsequent comparisons, although among the 7 countries in part B of Table 4.3, its Lorenz curve is not dominated by any other country, and the share of the bottom 20 per cent is the highest among these 7 countries.

The Lorenz curve allows for an unambiguous comparison of the relative distributions in cases where curves do not intersect, since the same ranking will be reached with a wide range of summary measures of inequality (Atkinson, 1970, Cowell, 1977). This requires that, for all x, the share of the botto x per ceilt'ih-country A is greater than that in country B. In such a situation, the distribution in country A is "Lorenz superio.t:Jto that in country B. Errors surrounding the estimates of the distributions must be considered w en judging such comparisons. It would be possible to calculate the sampling errors associated with the Lorenz curve, and require that one curve be significantly different from another at a specified level of confidence. This focuses on sampling

42

Figure 4.2 Percentage relative to 80% and 120% median

lllill!lil Total population - below 120% of median

- below 80% of median

United States France Sweden

errors and excludes other non-sampling errors which may be quantitatively more important. Differences in definition may lead to sizeable differences in measured shares, as discussed later in this chapter, and there are other variations not considered. 1 Calculating sampling errors is not given priority, though it certainly warrants fuller attention.

A simpler approach considers that only differences greater than 1 percentage point between cumulative shares are "significant". It is insufficient that the share of the bottom 40 per cent in France be 21.8 per cent, compared with 21.5 per cent in Canada (example from Table 4.3). Comparisons are done on the basis of the shares of the decile groups, starting at 10 per cent and ending at 90 per cent, or the shares of the bottom 10 per cent, the bottom 20 per cent, and so on up to 90 per cent. Therefore, the results are not unduly influenced by what happens at the tails of the distribution.

Figure 4.3 summarises the possible Lorenz comparisons on this basis. The grid should read: a + indicates that for the country shown in the row, we have Lorenz superiority over the country shown in the relevant column, with a difference exceeding 1 percentage point for at least one cumulative decile group; a - indicates Lorenz inferiority with a difference exceeding 1 percentage point for at least one cumulative decile group.

The first conclusion is that in more than half the cases an unambiguous comparison is possible. Of the 120 possible pair-wise comparisons of the 16 countries, in 89 cases one country is Lorenz superior, and the

I. Just to give one example, the following figures for the Netherlands 1987 illustrate the differences which can emerge from different sets of weights. The figures show the cumulative shares of decile groups: Figures supplied by the Central Bureau of Statistics: 4.1 10.1 16.9 24.5 33.0 42.5 53.2 65.3 79.4 Figures calculated from LIS dataset: 3.6 9.6 16.3 23.8 32.1 41.6 52.3 64.6 79.2

43

Table 4.3 Summary of Income Distribution in OECD Countries: Cumulative Decile Shares

SIO szo s3o s4o sso s60 s1o sso s9o s9s

A. Disposable Income Per Equivalent Adult

Australia 1985 2.9 7.7 13.7 21.0 29.4 39.0 50.2 63.0 78.3 87.3 Belgium 1988 4.2 10.2 17.1 25.0 33.8 43.5 54.3 66.4 80.3 88.4 Canada 1987 2.8 7.8 14.1 21.5 30.1 39.8 50.7 63.3 78.4 87.5 France 1984 3.0 8.3 14.6 21.8 29.9 39.1 49.5 61.6 76.3 85.5 Germany 1984 4.0 9.8 16.6 24.2 32.9 42.5 53.2 65.3 79.4 87.8 Ireland 1987 2.5 7.1 12.6 19.3 27.1 36.3 47.0 59.6 75.1 84.7 Italy 1986 3.1 8.0 13.9 20.7 28.7 38.0 48.7 61.2 76.2 85.4 Luxembourg 1985 4.3 10.2 17.1 24.8 33.5 43.1 53.9 66.0 80.4 88.8 Norway 1986 3.9 9.8 16.9 24.9 33.9 43.7 54.6 66.7 80.6 88.7 Sweden 1987 3.3 9.5 16.9 25.3 34.6 44.8 55.9 68.2 81.9 89.7 Switzerland 1982 2.8 8.0 14.1 21.0 29.0 37.8 47.7 58.9 72.5 81.3 United Kingdom 1986 2.5 7.5 13.5 20.5 28.7 38.2 49.1 61.8 77.1 86.4 United States 1986 1.9 5.7 11.2 18.0 26.2 35.7 46.9 60.2 76.3 86.2

Finland 1987 4.5 10.8 18.1 26.4 35.6 45.6 56.6 68.6 82.2 90.0 Netherlands 1987 4.1 10.1 16.9 24.5 33.0 42.5 53.2 65.3 79.4 87.8 New Zealand 1988 3.2 8.5 14.7 21.9 30.2 39.9 51.0 63.9 79.1 n.a.

B. Disposable Income (Excluding Self-employment Income) Per Equivalent Adult

Austria 1987 4.1 10.1 17.2 25.4 34.4 44.2 54.8 67.2 81.1 91.8 France 1984 3.4 9.1 15.6 22.9 31.3 40.6 51.1 63.3 77.9 91.1 Germany 1984 4.1 9.9 16.8 24.7 33.5 43.2 54.1 66.4 80.5 91.7 Italy 1986 3.3 8.4 14.6 21.7 30.2 39.6 50.5 63.2 78.3 90.9 Sweden 1987 3.5 9.8 17.4 25.8 35.2 45.4 56.5 68.6 82.2 92.3 United Kingdom 1986 2.9 8.1 14.1 21.2 29.4 38.8 49.8 62.5 77.9 90.9 United States 1986 2.0 5.8 11.3 18.1 26.3 35.9 47.2 60.4 76.5 90.2

Source and notes: see Table 4.1.

Figure 4.3 Lorenz Comparisons (Based on Decile Points)

BE CN FR GE IR IT LX NW sw sz UK us Fl NL NZ

AS + + + BE + + + + + + + + + + CN + + + FR + + + GE + + + + + + IR

IT

LX + + + + NW + + + + sw + + + + sz UK + us Fl + + NL +

44

difference at some point exceeds 1 percentage point. Figure 4.4 summarises the comparisons in terms of a Hasse diagram, which indicates the cases where definite conclusions can be drawn about the rankings of different countries. Countries towards the top of the diagram have lower levels of inequality, and a traceable line downwards from country A to country B implies that the Lorenz curve for country A is superior to that of country B (the difference at some point exceeds 1 percentage point).

Figure 4.4 illustrates that the Lorenz curve for Finland is superior to those for all other countries. Below Finland, Lorenz curves for Belgium, Luxembourg, Norway and Sweden are not inferior to any other country. They are either less than 1 percentage point apart (e.g. Belgium and Luxembourg) or are not comparable without ambiguity: the Lorenz curve for Belgium starts above that for Sweden, but falls below at the third decile.

Australia, Canada, and New Zealand form a group in the middle. The Lorenz curve for the last of these dominates those for France and Italy whose curves intersect with those for Australia and Canada. The curve for France starts off slightly (less than 1 percentage point) above that for Canada and then falls below around the middle of the distribution. Lorenz curves for Australia and Canada lie inside that for the United Kingdom, which

Figure 4.4 Relative Inequality in Different Countries

Belgium

I

Germany Netherlands Luxembourg

New Zealand

I France

r-------~~~~------~

Italy

Switzerland Ireland

Finland

Australia

-

United Kingdom

United States

45

Norway Sweden

Canada

I

intersects those for France and Italy. The United Kingdom and France (but not Italy) in tum dominate the United States, and France dominates Ireland and Switzerland.

These results relate in each case to a single year- but the situation in a particular year may be atypical in some significant respect. A relatively high proportion of Ireland's population works in agriculture- the reference year (1986 for farm incomes) was marked by a decline in agricultural income (see Chapter 5). The relative ranking of Ireland in another year might well be different.

Summary measures of inequality

Using summary measures of inequality to analyse intersections of Lorenz curves yields definite rankings and serves to quantify the extent of inequality. At the same time, where Lorenz curves cross, different summary measures may yield different rankings, and so the choice of measure may be important. Two such measures are described in Chapter 3, and one of these (the Atkinson index) contains a parameter which allows explicitly for different distributional judgements. Table 4.4 presents different measures of inequality, and (as with Table 4.8) draws no distinction between countries "above the line" from the LIS and those using data from other sources.

Table 4.4 lists countries in order of their Gini coefficients, which are perhaps the single most popular index (the ~a~~ e for New Zealand). On this basis, the Scandinavian countries, Belgium and Luxembourg cle~ lylve the lowest asured inequality, followed by Germany and the Netherlands.

~i coefficien , provides a quantification of the extent of inequality. As interpreted by Sen (1976), it is possible tOC'atcu ate the "equivalent" level of national income taking account of the "cost" of inequality as (100-G) per cent of national income. It may be inferred from Table 4.4, other things equal, that inequality reduces national income by between a fifth (in Finland) and a third~Switzerland, the United States and Ireland). The "other things equal" ignores the possibility that any measure to reduce inequality may affect total national income; the focus here is on the equity dimension. Or, average income would have to be higher by 25 per cent in Finland and 50 per cent in the United States to be equivalent to an equal distribution of income.

The quantitative measure depends on a social judgement about the importance of inequality. This is made explicit in the equally distributed equivalent measure (Atkinson, 1970), where the parameter applied can be varied to account for different distributional judgements (e.g. Sawyer, 1976). Table 4.4 shows the results for the values of 0.5 and 1.0, giving the "cost" of inequality as a proportion of average income. The ranking is generally similar to that of the Gini coefficient, but not identical (and the magnitude of the cost is different). Those countries moving up the ranking, compared with the Gini coefficient, are shown in italics.

Table 4.4 Measures of Inequality" in OECD Countries

Country Year Gini Atkinson 0.5 Atkinson 1.0

Finland 1987 20.7 3.6 7.5 Sweden 1987 22.0 4.6 10.3 Norway 1986 23.4 4.6 9.5 Belgium 1988 23.5 4.9 10.3 Luxembourg 1985 23.8 4.6 9.2 Germany 1984 25.0 5.2 10.1 Netherlands 1987 26.8 n/a n/a Canada 1987 28.9 7.0 14.6 Australia 1985 29.5 7.5 15.8 France 1984 29.6 7.7 16.0 United Kingdom 1986 30.4 8.2 18.1 Italy 1986 31.0 8.0 15.3 Switzerland 1982 32.3 9.9 18.4 Ireland 1987 33.0 9.3 18.8 United States 1986 34.1 9.9 21.2

a For definitions of inequality measures, see Appendix 3. Source: see Table 4.1.

46

4.2 Trends over time

How the distribution of income has been changing over time is of particular interest. For Australia, Belgium, Canada, Finland, France, the Netherlands, New Zealand, Norway, Sweden, the United Kingdom, and the United States, information exists for two dates. These cover a range of countries, with regard to both their intrinsic features and their degree of income inequality.

The two data points correspond in most, but not all, cases to the end of the 1970s/beginning of the 1980s, and to the middle of the 1980s (1984-88). The difference in years between observations are:

3 years 4 years 5 years 6 years 7 years

Finland and Belgium Australia and the Netherlands France and New Zealand Canada and Sweden Norway, the United Kingdom and the United States

Two important points should be considered when analysing the results: differing gaps and years covered (e.g 1979-84 for France, and 1985-88 for Belgium). Moreover, as emphasised in Chapter 3, the results should be interpreted carefully in view of the differing macro-economic climate at different dates and in different countries. The use of two points in time may also be misleading due to changing emphasis on redistributive policy, as in the Netherlands.

Percentiles of the distribution

In considering changes over time, the report adopts the same methods of presenting the information as in Section 4.1. Table 4.5 shows the changes over time in terms of percentiles expressed relative to the median. Little significance should be attached to small changes over time. That the bottom decile in Australia rose from 46 per cent of the median to 46.5 per cent in table 4.5 is of little importance.

To begin with, the decile ratio increased in Finland, the Netherlands, Norway, Sweden, the United Kingdom and the United States; remained essentially unchanged in Australia, Belgium, France and Canada; and fell slightly in New Zealand. This more or less summarises the picture that will emerge throughout this section. The majority of countries covered here show a rise in inequality, particularly the United States and the United Kingdom, but this was not universal. Country experiences are diverse.

Table 4.5 Trend Over Time in Income Distribution in OECD Countries 1979-88: Percentiles of Median

P,o P, P, P,. P, P.JP,o

Australia 1981 46.0 68.3 141.9 186.3 216.4 4.05 1985 46.5 66.4 142.1 186.5 218.5 4.01

Belgium 1985 59.3 74.7 128.7 162.5 187.2 2.74 1988 58.5 74.5 128.8 163.2 190.8 2.79

Canada 1981 44.9 69.3 138.3 182.7 211.5 4.07 1987 45.8 68.5 137.5 184.2 218.0 4.02

France 1979 53.6 72.5 138.4 186.5 232.3 3.48 1984 55.4 72.1 139.7 192.8 233.5 3.48

Norway 1979 57.0 76.7 126.6 158.1 181.9 2.77 1986 55.3 76.0 128.7 162.2 187.3 2.93

Sweden 1981 61.5 79.2 124.4 150.9 167.0 2.45 1987 55.6 75.6 125.1 151.5 170.4 2.72

United Kingdom 1979 50.9 70.4 138.5 179.7 208.9 3.53 1986 51.1 67.6 144.6 194.1 232.1 3.79

United States 1979 38.1 64.5 141.8 187.6 221.9 4.93 1986 34.7 61.7 149.6 206.1 247.3 5.94

Finland 1987 58.9 76.5 125.5 152.7 173.6 2.59 1990 57.0 76.4 126.2 156.2 178.5 2.74

Netherlands 1983 64.8 77.2 135.5 176.1 208.1 2.72 1987 61.5 75.7 135.0 175.0 206.4 2.85

New Zealand 1983/4 53.2 189.6 3.56 1987/8 53.6 186.6 3.48

Source and notes: see Table 4. I.

47

It would however be misleading to group countries together and make broad generalisations. For instance, in considering the top decile relative to the median, little change is found in the Netherlands and Sweden, where the decline in the relative position of the bottom groups is mainly responsible for the rise in inequality. The rise is however marked in the United Kingdom, the United States and in France. The pattern of change has its distinctive features in each country.

Low and high incomes

The proportions of the population with less than specified percentages of the median are shown in Table 4.6. Countries "below the line" are not included in this table. The proportion below half the median changed slightly in Australia, Canada and France. There was no change in the United Kingdom according to these estimates.2 In the United States, the proportion below this low income cut-off rose by some 1 3/4 percentage points, and in Norway and Sweden the increase was somewhat larger (starting from a much lower base).

The top of the distribution showed little change in the proportion above one and a half times median income for Australia, Canada, and Sweden. In the United Kingdom, the proportion rose from 19.4 per cent to 22.8 per cent, and in the United States from 21.4 per cent to 24.9 per cent.

The previous section drew attention to the concentration around the middle (from 80 to 120 per cent of the median) in Sweden (and Norway). This concentration was even more marked in 1981, with 44.8 per cent of the Swedish population grouped in this middle range. The 1980s saw a decline in the middle income class in both Sweden and the United States (both fell by some 2.5 percentage points). The difference was that the movement edged towards the lower range in Sweden (below 80 per cent of the median), but towards the upper range in the United States (above 120 per cent of the median).

Table 4.6 Trend Over Time in Income Distribution in OECD Countries: Cumulative Proportions Below Percentiles of Median

50 60 70 80 100 120 150 200

Australia 1981 12.5 19.4 26.1 34.1 50.0 63.5 78.3 92.7 1985 12.3 20.3 27.5 34.5 50.0 63.4 78.5 92.6

Canada 1981 12.6 19.1 25.5 33.4 50.0 64.8 80.1 93.5 1987 12.2 18.2 26.2 34.3 50.0 65.3 80.5 92.7

France 1979 8.2 14.6 22.7 31.7 50.0 64.3 80.3 91.9 1984 7.5 13.2 22.8 32.1 50.0 65.0 79.0 91.1

Norway 1979 5.0 12.0 19.0 28.2 50.0 69.9 87.3 96.9 1986 7.3 12.8 20.1 29.0 50.0 68.2 85.2 95.5

Sweden 1981 5.4 9.2 16.0 26.1 50.0 70.9 89.7 98.7 1987 7.6 12.6 20.0 28.6 50.0 70.9 89.5 98.0

United Kingdom 1979 9.2 17.3 24.8 33.3 50.0 64.8 80.6 93.9 1986 9.1 17.6 27.1 35.2 50.0 62.7 77.2 91.2

United States 1979 16.6 22.3 28.6 35.1 50.0 63.2 78.6 92.5 1986 18.4 24.1 30.3 36.5 50.0 61.7 75.1 89.0

Source and notes: see Table 4.1.

Changes in the Lorenz curves are illustrated in Table 4. 7. In eight of the eleven countries, the maximum shift in the Lorenz curve is less than 1 percentage point. (Differences in time periods should be considered when comparing such absolute differences.) In five of these eight countries, the Lorenz curves shifted outward, and in the remaining three cases (Australia, Canada and France) the Lorenz curves cross.

Lorenz curves shift outwards at some point by more than 1 percentage point for all deciles in the case of the United Kingdom, from the second deeile upwards in Sweden, and from the third decile in the United States.

2. In making any comparison with the estimates made in the official Households Below Average Income study, it should be borne in mind that we are here looking at 50% of the median, rather than the mean, as used in the official study.

48

Table 4.7 Trend Over Time in Income Distribution in OECD Countries: Cumulative Decile Shares

s,. s,. s,. s,. s,. s60 s,. s,. s,. s,

Australia 1981 2.8 7.7 13.9 21.3 29.8 39.6 50.7 63.7 79.0 88.1 1985 2.9 7.7 13.7 21.0 29.4 39.0 50.2 63.0 78.3 87.3

Belgium 1985 4.2 10.3 17.3 25.1 34.0 43.9 54.8 66.9 80.9 89.1 1988 4.2 10.2 17.1 25.0 33.8 43.5 54.3 66.4 80.3 88.4

Canada 1981 2.7 7.6 14.0 21.5 30.1 39.9 50.9 63.7 78.8 87.8 1987 2.8 7.8 14.1 21.5 30.1 39.8 50.7 63.3 78.4 87.5

France 1979 3.1 8.4 14.6 21.9 30.0 39.2 49.7 61.6 76.0 84.9 1984 3.0 8.3 14.6 21.8 29.9 39.1 49.5 61.6 76.3 85.5

Norway 1979 4.1 10.2 17.4 25.6 34.6 44.4 55.2 67.2 80.9 88.9 1986 3.9 9.8 16.9 24.9 33.9 43.7 54.6 66.7 80.6 88.7

Sweden 1981 4.0 10.6 18.3 26.7 36.0 46.1 57.2 69.2 82.9 90.6 1987 3.3 9.5 16.9 25.3 34.6 44.8 55.9 68.2 81.9 89.7

United Kingdom 1979 3.5 8.7 15.1 22.6 31.1 40.8 51.8 64.4 79.2 88.0 1986 2.5 7.5 13.5 20.5 28.7 38.2 49.1 61.8 77.1 86.4

United States 1979 2.1 6.4 12.4 19.7 28.4 38.3 49.6 62.7 78.3 87.6 1986 1.9 5.7 11.2 18.0 26.2 35.7 46.9 60.2 76.3 86.2

Finland 1987 4.5 10.8 18.1 26.4 35.6 45.6 56.6 68.6 82.2 90.0 1990 4.3 10.5 17.7 26.0 35.1 45.1 56.0 68.1 81.8 89.7

Netherlands 1983 4.4 10.6 17.4 25.0 33.4 42.8 53.3 65.3 79.4 87.8 1987 4.1 10.1 16.9 24.5 33.0 42.5 53.2 65.3 79.4 87.8

New Zealand 1984 3.3 8.7 14.9 22.2 30.5 40.1 51.2 64.1 79.2 n.a. 1988 3.2 8.5 14.7 21.9 30.2 39.9 51.0 63.9 79.1 n.a.

Source and notes: see Table 4.1.

Summary measures of inequality

Table 4.8 illustrates changes over time in the summary measures for ten countries (results for New Zealand being unavailable). There was virtually no change in Canada and France, though the small differences illustrate how diverse summary measures may move in various directions, with the Gini coefficient and the Atkinson (1.0) measure moving in opposite directions. Changes of around 1 percentage point in the Gini coefficient are observed in Finland, Norway, Belgium and Australia. Sweden and the Netherlands experienced a 2 percentage point increase in the Gini coefficient over a shorter period, and in the United Kingdom and the United States there was a more than 3 percentage point increase.

Table 4.8 Trend Over Time in Income Distribution in OECD Countries: Summary Measures of Inequality<'

Country Year Gini Atkinson 0.5 Atkinsoni.O

Finland 1987 20.7 3.6 7.5 1990 21.5 3.8 8.2

Sweden 1981 19.9 3.6 7.8 1987 22.0 4.6 10.3

Norway 1979 22.2 4.5 9.5 1986 23.4 4.6 9.5

Belgium 1985 22.8 4.4 9.0 1988 23.5 4.9 10.3

Netherlands 1983 24.7 n/a n/a 1987 26.8 n/a n/a

Canada 1981 28.6 7.0 15.0 1987 28.9 7.0 14.6

Australia 1981 28.7 7.1 15.3 1985 29.5 7.5 15.8

France 1979 29.7 7.7 15.0 1984 29.6 7.7 16.0

United Kingdom 1979 27.0 6.1 12.7 1986 30.4 8.2 18.1

United States 1979 30.9 8.4 19.0 1986 34.1 9.9 21.2

a For definitions of inequality measures, see Appendix 3. Source: see Table 4.1.

49

Table 4.9 Sensitivity of Results to Changes in Equivalence Scale: Percentiles of Median

Pw P, P, p9{) P, P.JPw

Australia 1981 no adj 38.1 64.8 141.7 190.1 222.6 4.99 scale= 0.5 46.0 68.3 141.9 186.3 216.4 4.05 per capita 47.6 70.9 152.1 218.5 264.1 4.59

1985 no adj 38.3 62.4 139.9 186.2 219.1 4.86 scale= 0.5 46.5 66.4 142.1 186.5 218.5 4.01 per capita 48.0 71.0 153.0 224.0 272.0 4.67

Belgium 1985 no adj 49.0 70.6 134.0 173.6 198.4 3.54 scale= 0.5 59.3 74.7 128.7 162.5 187.2 2.74 per capita 56.6 73.5 132.4 175.9 205.9 3.11

1988 no adj 47.9 70.3 133.7 172.2 200.8 3.60 scale= 0.5 58.5 74.5 128.8 163.2 190.8 2.79 per capita 56.9 73.8 134.2 174.1 201.6 3.06

Canada 1981 no adj 37.7 64.6 138.4 182.0 214.8 4.83 scale= 0.5 44.9 69.3 138.3 182.7 211.5 4.07 per capita 45.4 69.3 148.0 210.8 259.4 4.64

1987 no adj 37.9 63.6 140.4 184.8 224.6 4.88 scale= 0.5 45.8 68.5 137.5 184.2 218.0 4.02 per capita 45.9 69.6 145.9 210.2 256.3 4.58

France 1979 no adj 44.4 68.4 137.9 185.2 227.6 4.17 scale= 0.5 53.6 72.5 138.4 186.5 232.3 3.48 per capita 48.9 69.6 146.2 209.5 263.9 4.29

1984 no adj 45.9 68.8 139.4 193.7 238.6 4.22 scale= 0.5 55.4 72.1 139.7 192.8 233.5 3.48 per capita 47.5 68.2 144.9 206.1 253.9 4.33

Germany 1984 no adj 45.8 68.8 134.1 178.6 213.8 3.90 scale= 0.5 56.9 75.0 132.7 170.8 201.7 3.00 per capita 54.3 73.3 141.9 193.2 230.9 3.56

Ireland 1987 no adj 42.0 62.5 149.7 206.7 252.4 4.92 scale= 0.5 49.5 66.7 150.9 209.2 252.2 4.23 per capita 44.4 64.1 148.1 215.8 289.1 4.86

Italy 1986 no adj 45.2 67.8 148.7 205.0 251.3 4.53 scale= 0.5 48.9 68.8 145.0 197.9 233.8 4.05 per capita 47.1 66.9 147.9 205.3 245.2 4.35

Luxembourg 1985 no adj 51.4 71.4 138.3 189.3 216.5 3.68 scale= 0.5 58.5 75.1 132.7 184.0 228.1 3.15 per capita 56.1 73.5 138.6 186.4 226.5 3.32

Norway 1979 no adj 43.1 69.9 127.0 152.5 174.8 3.54 scale= 0.5 57.0 76.7 126.6 158.1 181.9 2.77 per capita 59.2 76.2 140.0 196.2 232.2 3.32

1986 no adj 40.4 68.9 132.4 173.1 205.5 4.28 scale= 0.5 55.3 76.0 128.7 162.2 187.3 2.93 per capita 59.9 75.8 134.6 175.4 208.3 2.93

50

Table 4.9 (contd) Sensitivity of Results to Changes in Equivalence Scale: Percentiles of Median

PIO P, P, p90 P, P.JP,o

Sweden 1981 no adj 45.9 66.0 138.4 164.8 185.1 3.59 scale= 0.5 61.5 79.2 124.4 150.9 167.0 2.45 per capita 56.9 76.0 133.5 168.4 185.9 2.96

1987 no adj 39.0 61.8 136.7 164.5 183.7 4.22 scale= 0.5 55.6 75.6 125.1 151.5 170.4 2.72 per capita 58.2 77.1 136.6 171.7 194.2 2.95

Switzerland 1982 no adj 41.2 69.2 137.9 194.2 261.2 4.71 scale= 0.5 53.0 73.6 134.3 185.1 244.6 3.43 per capita 51.6 68.0 153.1 217.7 276.0 4.22

United Kingdom 1979 no adj 39.8 67.3 139.2 183.8 220.1 4.62 scale= 0.5 50.9 70.4 138.5 179.7 208.9 3.53 per capita 54.2 72.4 142.9 197.3 237.6 3.64

1986 no adj 41.2 62.9 146.7 204.7 240.2 4.97 scale= 0.5 5l.l 67.6 144.6 194.1 232.1 3.79 per capita 45.6 68.5 144.7 203.9 247.9 4.47

United States 1979 no adj 31.9 59.9 142.7 191.7 224.6 6.01 scale= 0.5 38.1 64.5 141.8 187.6 221.9 4.93 per capita 37.4 63.9 150.3 217.7 264.6 5.82

1986 no adj 31.8 58.0 149.2 210.1 252.1 6.62 scale= 0.5 34.7 61.7 149.6 206.1 247.3 5.94 per capita 33.2 60.7 157.3 232.0 288.4 6.99

Finland 1987 no adj 42.1 67.0 129.6 161.9 184.4 3.8 scale= 0.5 58.9 76.5 125.5 152.7 173.6 2.6 per capita 60.4 77.6 130.0 165.3 190.9 2.7

1990 no adj 39.9 65.0 130.2 162.0 186.4 4.1 scale= 0.5 57.0 76.4 126.2 156.2 178.5 2.7 per capita 62.5 78.0 132.3 171.8 200.5 2.7

Netherlands 1983 no adj 56.0 75.1 135.9 183.0 222.6 3.3 scale= 0.5 64.8 77.2 135.5 176.1 208.1 2.7 per capita 57.9 73.6 141.3 196.7 236.4 3.4

1987 no adj 50.8 72.9 134.4 181.7 216.9 3.6 scale= 0.5 61.5 75.7 135.0 175.0 206.4 2.8 per capita 55.8 72.5 140.6 192.8 228.3 3.5

New Zealand 1984 no adj 47.9 183.9 3.8 scale= 0.5 53.2 189.6 3.6 per capita 46.8 214.0 4.6

1988 no adj 46.8 187.1 4.0 scale= 0.5 53.6 186.6 3.5 per capita 48.3 214.7 4.5

Note: The households are weighted by the number of household members. Source: See Table 4.1.

51

4.3 Sensitivity of estimates to methods employed

This section examines the sensitivity of the results described above to assumptions made about the treatment of households of different sizes and the use of person or household weights.

Equivalence scale

For the reasons discussed in Chapter 2, the calculations of equivalent income used in this report are based on a very simple scale that divides household income by the square root of the number of household members. The scale is N°·5 which is referred to as the 0.5 scale. Tables 4.9 and 4.10 illustrate the effect of two alternative - extreme- assumptions: that no adjustment is made for household size, and that household income is divided by the number of household members (a per capita scale).

It is hardly surprising that such changes in methods change the results, since they amount to very different treatment of households of different types. A household which on the basis of total household income is comfortably above the median may find itself below the median on a different basis. That the change in method affects both the evaluation of the individual household position and the median or mean income should be kept in mind. A single person household has an income of $X whatever scale is chosen, but $X becomes a different fraction of the median. It is also not surprising if the impact of a change in equivalence scale affects different countries differently, since it depends on the joint distribution of household size and income. There are different patterns of household structure in the countries studied.

The first conclusion that appears to emerge from Tables 4.9 and 4.10 is that measured inequality tends to be higher with either of the alternative assumptions. Viewing them along a spectrum, inequality tends to fall when moving from "no adjustment" to the 0.5 scale, and then rise when moving to a per capita basis. In Table 4.9, the decile ratio is systematically higher on a "no adjustment" basis, in some cases by more than 1.0. In Table 4.10, the share of the bottom 20 per cent is lower by typically 1 to 2 percentage points. The use of a per capita scale leads to a higher value of the decile ratio than the 0.5 scale (except in the case of Norway 1986), although for all countries except France, the Netherlands (in 1983), New Zealand and the United States (in 1986), it is lower than with no adjustment.

The differences are important in the context of international comparisons. A difference of approximately 1.0 in the decile ratio is of the magnitude of that between Germany and Canada. Sweden on a "no adjustment" basis has a decile ratio in 1987 of 4.22, which compares unfavourably with those calculated using a 0.5 scale around the same date in Canada, France, Germany, Italy, Luxembourg, the Netherlands, and the United Kingdom. If in the case of the United Kingdom in Table 4.10, the share of the bottom 20 per cent in 1979 which was calculated with no adjustment for household size were compared with that for 1986 with the 0.5 scale, then the share would appear to have increased - whereas on a consistent basis there is a decrease.

Adopting a consistent basis across countries and across years is therefore important. But does the choice of method affect the conclusions drawn using a consistent basis? In the case of trends over time, Section 4.2 showed that there was little apparent change in Canada and France on the basis of the 0.5 scale, but the same conclusion does seem to hold for no adjustment and per capita calculations. The decile ratio in Table 4.9 is essentially the same in both years, and the Lorenz curves in Table 4.10 differ by less than 1 percentage point.

Among the countries showing a rise in inequality, results are confirmed for the United States in both cases. For the United Kingdom, the change in the shares of bottom income groups is less marked on a no adjustment basis and more marked on a per capita basis: the share of the bottom 20 per cent in the United Kingdom fell by either 0.5, 1.2 or 1.6 percentage points, depending on the method of adjusting for household size. The findings for Sweden demonstrate the potential sensitivity to changes in definition. The rise in inequality is less marked on a per capita basis than on the 0.5 scale. The outward shift in the Lorenz curve between 1981 and 1987 is less than 1 percentage point on a per capita basis, and the bottom decile rises relative to the median, so that the decile ratio is effectively unchanged.

As far as comparisons across countries are concerned, moving to one of the alternative methods would definitely affect the rankings, although the broad pattern remains the same. On a no adjustment basis, ranking according to the decile ratio would change, compared to that with the 0.5 scale, with Sweden falling, and Switzerland and France changing positions, as do Germany and Norway, and the United Kingdom and Italy. On a per capita basis, compared with the 0.5 scale, Italy would jump ahead of the United Kingdom.

52

Table 4.10 Sensitivity of Results to Changes in Equivalence Scale: Cumulative Decile Shares

Calculations based on person weights

s,. 520 s,. s,. s,. s .. s1• s,. s,. s,

Australia 1981 no adj 2.1 6.4 12.3 19.7 28.3 38.1 49.4 62.4 78.0 87.4 scale= 0.5 2.8 7.7 13.9 21.3 29.8 39.6 50.7 63.7 79.0 88.1 per capita 2.6 7.4 13.3 20.2 27.9 36.9 47.6 60.3 76.3 86.2

1985 no adj 2.2 6.4 12.1 19.4 28.0 37.9 49.1 62.1 77.5 86.8 scale= 0.5 2.9 7.7 13.7 21.0 29.4 39.0 50.2 63.0 78.3 87.3 per capita 2.6 7.3 13.1 19.8 27.5 36.4 46.7 59.4 75.4 85.5

Belgium 1985 no adj 3.3 8.6 15.2 22.9 31.7 41.6 52.7 65.3 80.0 88.6 scale= 0.5 4.2 10.3 17.3 25.1 34.0 43.9 54.8 66.9 80.9 89.1 per capita 3.9 9.6 16.3 24.0 32.6 42.2 52.9 65.0 79.3 87.8

1988 no adj 3.3 8.5 15.1 22.7 31.5 41.4 52.5 65.0 79.5 88.1 scale= 0.5 4.2 10.2 17.1 25.0 33.8 43.5 54.3 66.4 80.3 88.4 per capita 3.9 9.6 16.2 23.9 32.5 42.0 52.6 64.7 78.9 87.3

Canada 1981 no adj 2.2 6.6 12.5 19.8 28.4 38.8 50.1 63.0 78.3 87.5 scale= 0.5 2.7 7.6 14.0 21.5 30.1 39.9 50.9 63.7 78.8 87.8 per capita 2.5 7.1 13.0 19.8 27.7 36.8 47.4 60.0 75.7 85.5

1987 no ad 2.3 6.6 12.5 19.8 28.4 38.2 49.5 62.4 77.7 87.1 scale= 0.5 2.8 7.8 14.1 21.5 30.1 39.8 50.7 63.3 78.4 87.5 per capita 2.7 7.4 13.3 20.3 28.2 37.3 47.8 60.3 75.8 85.7

France 1979 no adj 2.5 7.2 13.2 20.4 28.6 38.1 48.7 61.0 75.6 84.5 scale= 0.5 3.1 8.4 14.6 21.9 30.0 39.2 49.7 61.6 76.0 84.9 per capita 2.7 7.4 13.1 19.8 27.5 36.5 46.7 58.9 74.0 83.7

1984 no adj 2.4 7.1 13.1 20.3 28.5 37.8 48.5 60.8 75.8 85.1 scale= 0.5 3.0 8.3 14.6 21.8 29.9 39.1 49.5 61.6 76.3 85.5 per capita 2.5 7.2 12.9 19.7 27.6 36.6 47.1 59.4 74.6 84.2

Germany 1984 no adj 3.0 8.1 14.4 21.9 30.5 40.3 51.2 63.6 78.3 87.1 scale= 0.5 4.0 9.8 16.6 24.2 32.9 42.5 53.2 65.3 79.4 87.8 per capita 3.8 9.0 15.3 22.5 30.7 39.9 50.5 62.7 77.4 86.4

Ireland 1987 no adj 1.9 6.2 11.5 18.1 26.0 35.3 46.2 59.1 74.8 84.7 scale= 0.5 2.5 7.1 12.6 19.3 27.1 36.3 47.0 59.6 75.1 84.7 per capita 2.1 6.3 11.6 18.0 25.6 34.4 44.6 56.9 72.6 82.7

Italy 1986 no adj 2.8 7.3 13.0 19.7 27.5 36.6 47.3 59.9 75.3 84.8 scale= 0.5 3.1 8.0 13.9 20.7 28.7 38.0 48.7 61.2 76.2 85.4 per capita 2.9 7.5 13.1 19.9 27.8 36.9 47.6 60.1 75.3 84.8

Luxembourg 1985 no adj 3.4 8.8 15.2 22.7 31.2 40.8 51.6 64.0 79.0 88.2 scale= 0.5 4.3 10.2 17.1 24.8 33.5 43.1 53.9 66.0 80.4 88.8 per capita 4.0 9.4 15.8 23.2 31.5 40.8 51.4 63.7 78.4 87.4

Norway 1979 no adj 2.9 8.1 14.9 23.1 32.4 42.7 54.0 66.5 80.4 88.4 scale= 0.5 4.1 10.2 17.4 25.6 34.6 44.4 55.2 67.2 80.9 88.9 per capita 3.7 9.4 16.0 23.4 31.6 40.7 51.1 63.2 78.1 87.2

1986 no adj 2.6 7.4 13.8 21.6 30.6 40.5 51.6 64.1 78.8 87.6 scale= 0.5 3.9 9.8 16.9 24.9 33.9 43.7 54.6 66.7 80.6 88.7 per capita 4.2 10.1 16.9 24.5 33.0 42.5 53.1 65.2 79.4 87.7

53

Table 4.10 (contd) Sensitivity of Results to Changes in Equivalence Scale: Cumulative Decile Shares

Calculations based on person weights

Sw s'" S"' s" s,. s60 s" s" s .. s.,

Sweden 1981 no adj 2.9 8.0 14.3 21.8 30.6 41.0 53.0 66.3 81.2 89.5 scale= 0.5 4.0 10.6 18.3 26.7 36.0 46.1 57.2 69.2 82.9 90.6 per capita 3.7 9.7 16.8 24.8 33.7 43.5 54.6 67.1 81.6 89.9

1987 no adj 2.3 6.9 12.9 20.3 29.2 39.9 52.0 65.5 80.6 89.1 scale= 0.5 3.3 9.5 16.9 25.3 34.6 44.8 55.9 68.2 81.9 89.7 per capita 3.3 9.4 16.4 24.3 33.1 42.7 53.7 66.2 80.7 89.0

Switzerland 1982 no adj 2.0 6.3 12.1 19.1 27.0 35.8 45.9 57.5 71.5 80.7 scale= 0.5 2.8 8.0 14.1 21.0 29.0 37.8 47.7 58.9 72.5 81.3 per capita 2.6 7.1 12.4 18.6 25.9 34.3 44.3 56.3 71.3 80.7

United Kingdom 1979 no adj 2.6 7.0 13.1 20.6 29.2 39.0 50.1 62.9 78.2 87.3 scale= 0.5 3.5 8.7 15.1 22.6 31.1 40.8 51.8 64.4 79.2 88.0 per capita 3.3 8.6 14.9 22.0 30.2 39.3 49.8 62.2 77.4 86.7

1986 no adj 2.2 6.5 12.1 18.9 27.1 36.7 47.7 60.7 76.5 86.2 scale= 0.5 2.5 7.5 13.5 20.5 28.7 38.2 49.1 61.8 77.1 86.4 per capita 2.3 7.0 12.9 19.8 27.8 37.0 47.6 60.1 75.3 85.0

United States 1979 no adj 1.7 5.7 11.3 18.3 26.9 36.9 48.5 61.9 77.8 87.5 scale= 0.5 2.1 6.4 12.4 19.7 28.4 38.3 49.6 62.7 78.3 87.6 per capita 1.9 5.9 11.3 18.0 25.9 35.1 46.0 58.9 74.9 85.2

1986 no adj 1.7 5.3 10.4 17.1 25.3 35.0 46.2 59.5 75.9 86.1 scale= 0.5 1.9 5.7 11.2 18.0 26.2 35.7 46.9 60.2 76.3 86.2 per capita 1.6 5.1 10.1 16.4 24.0 33.0 43.9 56.9 73.5 84.1

Finland 1987 no adj 3.0 8.0 14.5 22.5 31.8 42.1 53.5 66.3 80.8 89.2 scale= 0.5 4.5 10.8 18.1 26.4 35.6 45.6 56.6 68.6 82.2 90.0 per capita 4.5 10.6 17.8 25.8 34.6 44.2 54.9 66.9 80.8 88.9

1990 no adj 2.8 7.6 14.0 21.9 31.2 41.6 53.1 65.9 80.5 89.0 scale= 0.5 4.3 10.5 17.7 26.0 35.1 45.1 56.0 68.1 81.8 89.7 per capita 4.6 10.8 17.8 25.6 34.2 43.7 54.3 66.3 80.3 88.7

Netherlands 1983 no adj 3.6 9.1 15.7 23.2 31.6 41.0 51.5 63.5 77.9 86.8 scale= 0.5 4.4 10.6 17.4 25.0 33.4 42.8 53.3 65.3 79.4 87.8 per capita 3.8 9.2 15.4 22.6 30.7 39.8 50.2 62.3 77.2 86.4

1987 no adj 3.1 8.4 14.9 22.5 31.0 40.6 51.3 63.5 78.1 87.0 scale= 0.5 4.1 10.1 16.9 24.5 33.0 42.5 53.2 65.3 79.4 87.8 per capita 3.6 8.9 15.2 22.4 30.5 39.8 50.4 62.7 77.7 86.7

Note: The households are weighted by the number of household members. Source: See Table 4.1.

Household weights

The results in this chapter weight each household according to the number of members. This weighting procedure is adopted in a number of countries. There are other countries, however, which give each household a weight of 1. Among the national studies cited in Chapter 5, this applies to France, Italy, Japan, the Netherlands, and the United Kingdom (Economic Trends estimates). It is therefore of interest to examine the sensitivity of the findings to different choices with respect to weighting.

54

Table 4.11 Sensitivity of Results to Changes in Weighting (Persons Versus Households): Percentiles of Median

PIO p25 P, p90 P, P.JPIO

Australia 1981 person 46.0 68.3 141.9 186.3 216.4 4.05 household 45.3 62.0 144.5 190.2 221.1 4.20

1985 person 46.5 66.4 142.1 186.5 218.5 4.01 household 46.0 61.0 147.0 192.0 224.0 4.17

Belgium 1985 person 59.3 74.7 128.7 162.5 187.2 2.74 household 59.7 74.4 131.6 167.5 194.9 2.80

1988 person 58.5 74.5 128.8 163.2 190.8 2.79 household 57.8 73.6 133.3 169.6 200.0 2.93

Canada 1981 person 44.9 69.3 138.3 182.7 211.5 4.07 household 41.1 64.0 142.3 189.2 220.5 4.60

1987 person 45.8 68.5 137.5 184.2 218.0 4.02 household 42.7 65.7 142.1 191.4 226.6 4.48

France 1979 person 53.6 72.5 138.4 186.5 232.3 4.02 household 50.9 70.4 140.1 189.5 235.7 3.72

1984 person 55.4 72.1 139.7 192.8 233.5 3.48 household 55.0 72.2 141.5 195.9 238.6 3.56

Germany 1984 person 56.9 75.0 132.7 170.8 201.7 3.00 household 53.8 72.8 134.0 173.8 206.0 3.23

Ireland 1987 person 49.5 66.7 150.9 209.2 252.2 4.23 household 50.8 64.9 156.1 219.8 269.7 4.32

Italy 1986 person 48.9 68.8 145.0 197.9 233.8 4.05 household 49.0 68.4 146.5 198.6 238.1 4.06

Luxembourg 1985 person 58.5 75.1 132.7 184.0 228.1 3.15 household 55.7 73.4 133.1 173.4 202.6 3.11

Norway 1979 person 57.0 76.7 126.6 158.1 181.9 2.77 household 52.2 70.4 130.5 164.1 188.1 3.15

1986 person 55.3 76.0 128.7 162.2 187.3 2.93 household 49.6 69.6 132.3 167.3 193.8 3.37

Sweden 1981 person 61.5 79.2 124.4 150.9 167.0 2.45 household 60.6 78.5 126.7 155.4 172.6 2.56

1987 person 55.6 75.6 125.1 151.5 170.4 2.72 household 51.6 72.1 128.8 159.0 180.8 3.08

Switzerland 1982 person 53.9 73.6 134.3 185.1 244.6 3.43 household 46.2 69.6 134.7 181.9 239.2 3.94

55

Table 4.11 ( contd) Sensitivity of Results to Changes in Weighting (Persons Versus Households): Percentiles of Median

PIO p25 P, p90 P, P.JP,o

United Kingdom 1979 person 50.9 70.4 138.5 179.7 208.9 3.53 household 50.2 65.1 144.1 187.9 220.4 3.74

1986 person 51.1 67.6 144.6 194.1 232.1 3.79 household 53.8 68.6 149.1 203.0 246.8 3.78

United States 1979 person 38.1 64.5 141.8 187.6 221.9 4.93 household 35.3 60.9 145.0 195.1 232.6 5.53

1986 person 34.7 61.7 149.6 206.1 247.3 5.94 household 33.5 58.9 151.8 209.8 252.9 6.26

Finland 1987 person 58.9 76.5 125.5 152.7 173.6 2.6 household 55.6 73.0 129.2 159.4 181.9 2.9

1990 person 57.0 76.4 126.2 156.2 178.5 2.7 household 53.5 70.8 131.2 163.8 189.3 3.1

Netherlands 1983 person 64.8 77.2 135.5 176.1 208.1 2.7 household 64.0 76.3 135.5 176.1 207.3 2.8

1987 person 61.5 75.7 135.0 175.0 206.4 3.4 household 60.1 74.3 136.2 175.8 207.5 3.5

New Zealand 1984 person 53.2 189.6 3.6 household 53.3 192.9 3.6

1988 person 53.6 186.6 3.5 household 53.4 188.5 3.5

Source and notes: see Table 4.1.

Tables 4.11 and 4.12 show the effect of treating each household as 1 unit (household weights) in place of giving it a weight equal to the number of members (person weights). The effect in almost all cases is to move the Lorenz curve outwards over at least part of the range, the only exception being Italy. The difference is generally small with respect to the Lorenz curves, but larger in certain cases, for example the share of the bottom 20 per cent falls by 1 percentage point or more in Norway (1979), Switzerland and New Zealand (1984).

56

Table 4.12 Sensitivity of Results to Changes in Weighting (Persons Versus Households): Cumulative Decile Shares

SIO s,. s,. s,. s,. s .. s,. s,. s .. s,

Australia 1981 person 2.8 7.7 13.9 21.3 29.8 39.6 50.7 63.7 79.0 88.1 household 2.6 7.2 12.8 19.9 28.3 38.2 49.6 62.8 78.5 87.8

1985 person 2.9 7.7 13.7 21.0 29.4 39.0 50.2 63.0 78.3 87.3 household 2.8 7.3 12.8 19.6 27.9 37.6 49.0 62.1 77.8 87.0

Belgium 1985 person 4.2 10.3 17.3 25.1 34.0 43.9 54.8 66.9 80.9 89.1 household 4.0 10.0 16.9 24.7 33.4 43.1 54.0 66.2 80.3 88.7

1988 person 4.2 10.2 17.1 25.0 33.8 43.5 54.3 66.4 80.3 88.4 household 3.9 9.7 16.5 24.1 32.7 42.4 53.2 65.4 79.4 87.7

Canada 1981 person 2.7 7.6 14.0 21.5 30.1 39.9 50.9 63.7 78.8 87.8 household 2.4 6.9 12.7 20.0 28.4 38.2 49.4 62.5 78.0 87.2

1987 person 2.8 7.8 14.1 21.5 30.1 39.8 50.7 63.3 78.4 87.5 household 2.6 7.2 13.2 20.2 28.6 38.3 49.4 62.2 77.6 86.9

France 1979 person 3.1 8.4 14.6 21.9 30.0 39.2 49.7 61.6 76.0 84.9 household 3.0 8.0 14.1 21.2 29.3 38.5 49.1 61.2 75.8 84.8

1984 person 3.0 8.3 14.6 21.8 29.9 39.1 49.5 61.6 76.3 85.5 household 2.8 8.1 14.3 21.4 29.5 38.6 49.0 61.1 75.9 85.1

Germany 1984 person 4.0 9.8 16.6 24.2 32.9 42.5 53.2 65.3 79.4 87.8 household 3.8 9.3 15.8 23.4 31.9 41.5 52.3 64.4 78.6 87.1

Ireland 1987 person 2.5 7.1 12.6 19.3 27.1 36.3 47.0 59.6 75.1 84.7 household 2.4 6.8 12.2 18.6 26.1 35.1 45.7 58.4 74.2 84.2

Italy 1986 person 3.1 8.0 13.9 20.7 28.7 38.0 48.7 61.2 76.2 85.4 household 3.2 8.1 13.9 20.8 28.8 38.0 48.7 61.2 76.2 85.5

Luxembourg 1985 person 4.3 10.2 17.1 24.8 33.5 43.1 53.9 66.0 80.4 88.8 household 4.0 9.8 16.6 24.3 32.9 42.5 53.3 65.5 79.9 88.4

Norway 1979 person 4.1 10.2 17.4 25.6 34.6 44.4 55.2 67.2 80.9 88.9 household 3.6 9.1 15.8 23.7 32.6 42.6 53.7 66.1 80.3 88.6

1986 person 3.9 9.8 16.9 24.9 33.9 43.7 54.6 66.7 80.6 88.7 household 3.6 8.9 15.4 23.2 32.1 42.0 53.1 65.5 79.9 88.3

Sweden 1981 person 4.0 10.6 18.3 26.7 36.0 46.1 57.2 69.2 82.9 90.6 household 3.7 10.2 17.8 26.2 35.3 45.4 56.5 68.7 82.5 90.4

1987 person 3.3 9.5 16.9 25.3 34.6 44.8 55.9 68.2 81.9 89.7 household 2.8 8.6 15.6 23.7 32.9 43.1 54.3 66.8 81.0 89.2

Switzerland 1982 person 2.8 8.0 14.1 21.0 29.0 37.8 47.7 58.9 72.5 81.3 household 2.3 7.0 12.9 19.9 28.0 37.0 47.2 58.7 72.5 81.2

57

Table 4.12 (contd) Sensitivity of Results to Changes in Weighting (Persons Versus Households): Cumulative Decile Shares

s,. s,. s,. s,. s,. s60 s,. s,. s,. s,

United Kingdom 1979 person 3.5 8.7 15.1 22.6 31.1 40.8 51.8 64.4 79.2 88.0 household 3.5 8.5 14.3 21.4 29.7 39.3 50.4 63.3 78.5 87.5

1986 person 2.5 7.5 13.5 20.5 28.7 38.2 49.1 61.8 77.1 86.4 household 2.9 8.0 13.8 20.5 28.4 37.6 48.3 61.0 76.4 85.8

United States 1979 person 2.1 6.4 12.4 19.7 28.4 38.3 49.6 62.7 78.3 87.6 household 1.8 5.9 11.4 18.5 26.9 36.8 48.2 61.4 77.4 87.0

1986 person 1.9 5.7 11.2 18.0 26.2 35.7 46.9 60.2 76.3 86.2 household 1.7 5.4 10.5 17.2 25.2 34.8 46.0 59.3 75.6 85.6

Finland 1987 person 4.5 10.8 18.1 26.4 35.6 45.6 56.6 68.6 82.2 90.0 household 4.1 10.0 16.9 24.9 33.9 43.9 54.9 67.3 81.3 89.3

1990 person 4.3 10.5 17.7 26.0 35.1 45.1 56.0 68.1 81.8 89.7 household 4.0 9.6 16.3 24.1 33.1 43.0 54.1 66.6 80.8 89.1

Netherlands 1983 person 4.4 10.6 17.4 25.0 33.4 42.8 53.3 65.3 79.4 87.8 household 4.2 10.2 17.0 24.6 33.0 42.4 53.0 65.0 79.1 87.5

1987 person 4.1 10.1 16.9 24.5 33.0 42.5 53.2 65.3 79.4 87.8 household 3.8 9.6 16.2 23.7 32.2 41.7 52.4 64.6 78.9 87.4

New Zealand 1984 person 3.3 8.7 14.9 22.2 30.5 40.1 51.2 64.1 79.2 household 3.5 7.7 14.7 21.8 30.1 39.7 50.8 63.8 79.2

1988 person 3.2 8.5 14.7 21.9 30.2 39.9 51.0 63.9 79.1 household 3.4 8.6 14.6 21.5 29.7 39.4 50.6 63.7 78.9

Source and notes: see Table 4.1.

4.4 Conclusion

This chapter presented the developments in the distribution of disposable income for seventeen OECD countries in the late 1970s and 1980s. While bearing in mind the problems of comparability which remain, the evidence clearly depicts the Scandinavian countries, Benelux and Germany as exhibiting the least relative inequality. The highest levels of relative inequality are recorded in the United States, Switzerland and Ireland. Where inequality is comparable at different dates, there was a rise in the 1980s for the majority of countries studied, but this was not universal. Increases in measured income inequality were largest in the Netherlands, Sweden, and particularly the United Kingdom and the United States.

58

ChapterS

COMPARISONS WITH NATIONAL STUDIES

This chapter sets the results of Chapter 4 in the context of national studies of income inequality and extends the coverage of OECD countries. This review of the evidence available from other studies is not a comprehensive survey of all published material. The purpose is rather to build a bridge between the LIS dataset, which emphasizes raising the degree of comparability of data employed, and the much more disparate national studies, which for good reason have employed a wide variety of sources and definitions. Since this study seeks to compare the LIS dataset with other approaches, it does not refer to other studies based on the LIS dataset, such as Smeeding, Rainwater and O'Higgins (1990), and Mitchell (1991).

In view of the differences in definitions, sources, and timing, the results from the national studies are not necessarily expected to show the same level of inequality as found in Chapter 4. Nor need the trends be the same since dissimilarities may have a different impact at different dates. The trends in the national studies are however of particular interest because the estimates typically cover a longer time period and include more observations.

Section 5.1 begins by discussing the earlier comparison of OECD countries by Sawyer (1976) and that published by the International Labour Office by van Ginneken and Park (1984). Section 5.2, which forms the bulk of this chapter, considers the evidence available from national studies for a selection of OECD countries, including in addition to those covered in Chapter 4, Japan, Portugal and Spain. The discussion is organised by country with broad conclusions drawn out in Section 5.3.

5.1 Earlier studies of OECD countries

The most widely known earlier study of income distribution in OECD countries is by Sawyer (1976). Table 5.1 summarises the main findings for the size distribution of post-tax income for 12 OECD countries around 1970. The countries are ranked in order of the Gini coefficient (highest at the top) with two other measures of inequality shown for comparison. The countries fall into three main groups, distinguished by the solid horizontal lines in the table (these correspond to differences in the Gini coefficient of more than 2.5 percentage points):

- France, Italy, Germany and the United States; - Spain, Canada and the Netherlands; and - the United Kingdom, Japan, Australia, Norway and Sweden.

This grouping is rather surprising because it does not correspond to what one might expect knowing the features of these societies.

The Sawyer study met with lively reactions, notably from the French Government, which published a reply (Begue, 1976). There are indeed a number of serious questions about the degree of comparability:

- Data are derived from different types of source. In the majority of cases, the source is a household survey, such as the U.S. Current Population Survey, but in other cases data are based on tax records (France, the Netherlands and Norway) or a synthesis of different sources (Germany). Some indication of the consequences are provided by Sawyer's additional memorandum items for Germany (which replaces the synthetic estimate by one from a household income and expenditure survey) and the United Kingdom (which replaces the expenditure survey figure by a synthetic estimate). (The synthetic estimate for the United Kingdom is that usually known as the "Blue Book" estimate, which combines information from the tax records with household survey data and other information. This estimate is discussed further below.)

59

Table 5.1 International Comparison by Sawyer (1976)

Country Year Gini Charnpernowne Share of % % bottom20%

France 1970 41.4 27.6 4.3

Italy 1969 39.8 24.8 5.1 Germany 1973 38.3 22.0 6.5

United States 1972 38.1 24.8 4.5

Spain 1973-4 35.5 20.1 6.0 Canada 1969 35.4 22.0 6.0 Netherlands 1967 35.4 19.2 6.5

United Kingdom 1973 31.8 16.6 6.3 Japan 1969 31.6 15.6 7.9

Australia 1966-7 31.2 15.7 6.6

Norway 1970 30.7 16.2 6.3 Sweden 1972 30.2 15.8 6.6

Note: The Charnpemowne index is an alternative inequality measure. Estimated by applying average tax rates to pre-tax data

Source: Sawyer ( 1976), Tables 4 and 6.

Germany United Kingdom

Source: Sawyer (1976), Table 6.

Gini coefficients

Synthetic estimate

38.3 33.5

Comments

* Based on tax records

Synthesis of different sources

*

Based on tax records Relates to tax unit rather than household

* * Excludes agricultural, forestry workers and fishermen. Only households in urban areas Based on tax records

Survey estimate

31.2 31.8

This provides a rather different picture of the relative income inequality in the two countries. The difference in the Gini coefficient is reduced from 6.5 percentage points in Table 5.1 to 4.8 percentage points if synthetic estimates are taken in both cases, and actually reversed for survey estimates. Data do not cover the whole population in two cases (Japan and Australia). The exclusions may be expected to reduce the recorded degree of inequality, and may be one reason why they are found in the lowest inequality group. Sawyer did not have access to the original micro-data, and was in some cases obliged to make aggregate adjustments, particularly in going from pre-tax income to post-tax income (the countries marked by an* in the Comments column). As described by Sawyer, "One of these distributions had to be estimated from the other by utilising data on the average amount of tax paid by each income class ( ... )inequality tends to be underestimated since households have not been ranked by the derived income concept" (1976, p. 12). In the case of the United States, for example, Sawyer used the published tables from the Current Population Survey as follows: "The income accruing to each income class is not reported and has been estimated by logarithmic interpolation. The post-tax income distribution has been estimated by applying the average tax rates by income class reported in the Statistical Abstract of the United States and the social security contribution rates in force at the time" (1976, p. 32). It does not seem likely that this method can produce very accurate estimates of the distribution of disposable income, since it makes no allowance for the variation in taxes paid for a given gross income. The distributions relate to household income, but in the main figures no adjustment is made for differences in household size. As described in the previous chapter, the "no adjustment" figures can indicate a different degree of inequality and a different ranking of countries.

60

It is therefore not surprising if the figures in this report differ from those in the earlier Sawyer study. There is not only a difference of some ten years in the period covered, but the methods applied are significantly different. Though LIS estimates are not superior in all cases, they are closer to being on a consistent basis across countries.

A more recent comparative study by van Ginneken and Park (1984) for the ILO and World Bank1 pays considerable attention to the problems of generating internationally comparable estimates (this was indeed the title). The authors concentrated on estimates which in principle were based on the same income concepts and income units, and where possible the original survey data were adjusted in line with the national accounts. At the same time, the study did not generally have access to the original micro-data, and in a number of cases the estimates are based on adjusting tabulated data. For example, the per capita distribution for the United Kingdom is calculated from a table cross-classifying households by household size.

The results of van Ginneken and Park cover 23 developed and developing countries, including seven OECD Member countries. The findings for these countries are shown in Table 5.2, where again the countries are classified in order of the Gini coefficient (highest at the top). The basis for classifying households is per capita income- one reason why these results differ from those of Sawyer. There are also differences in the timing and in data sources for some countries. The results correspond better to Chapter 4 in that there is less of a gap between France and other countries and inequality in Sweden is distinctly lower.

Table 5.2 International Comparison by van Ginneken and Park (1984) of Household Distribution of Income Per Head

Country Year Gini Share of Share of Comments % top20% bottom 20%

France 1975 35 44.2 7.9 Based on tax records Spain 1973-74 34 41.8 7.2 * Germany 1974 32 41.6 8.4 Synthesis of different sources Ireland 1973 29 39.7 9.0 United Kingdom 1979 27 37.4 9.5 * Denmark 1976 27 37.0 9.5 Sweden 1979 22 32.0 9.5 *

Relates to family unit rather than household

Notes: * Includes imputed rent of owner-occupied houses. Unless otherwise indicated, data from household surveys.

Source: van Ginneken and Park (1984), Table 2.

5.2 Evidence from national studies

Countries considered in this section are listed in Table 5.3, together with the years covered. The studies are classified according to their origins. Those without an asterisk were carried out in an official statistical agency or in conjunction with the agency (although this distinction is not always easy to draw). Those with a dollar sign make use of data sources different from those in Chapter 4, or a number of sources. Each country is discussed in tum below.

Australia

Sources for the national estimates by Saunders (1993) shown in Table 5.4 are the same as for the LIS dataset though different methods are applied, including a different equivalence scale, and different definitions of income unit and treatment of negative incomes. Comparing Table 5.4 with estimates for 1981 in Table 4.7 suggests that the latter gives a higher degree of inequality: the Lorenz curve from S20 to S70 being between 1 and 2 percentage points further from the diagonal.

1. Reference may also be made to the study by Stark (1977), who assembled a wide variety of income distribution estimates for nine countries: Australia, Canada, France, Ireland, Japan, Sweden, the United Kingdom, the United States and Germany. Recent surveys which have helped us include Gardiner (1993) and Green, Henley and Tsakalotos (1992).

61

Table 5.3 National Studies for a Selection of OECD Countries

Australia *Saunders (1993) 1981/2, 89/90

Austria Guger (1989a) 1970, 76, 81 and 87

Belgium $ Valenduc (1991) 1983-1986 *Cantillon et a! (1994) 1985, 1988 and 1992

Canada Wolfson (1986) 1971, 75, 79, 82, 83

Finland $ Uusitalo (1989) 1966, 71, 76, 81, 85

France Sandoval (1989) 1970, 75 and 79 Canceill and Villeneuve (1990) 1979, 1984 $ Assemat and Glaude (1989)

Germany *$ DIW Guger (1989) 1950, 55, 60, 64, 68, 70, 73, 75, 78,80,83, 84,85

Ireland

Italy

Japan

Netherlands

Norway

Portugal

Spain

Sweden

United Kingdom

United States

Notes:

*$ EVS Hauser and Becker (1993) 1973, 78 and 83 * Socio-Economic Panel Hauser and Becker (1993) 1983, 85, 87 and 90

$Callan and Nolan (1993) 1973, 80 and 87

Brandolini (1993) 1967,68,69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79,80, 81,82, 83,84,86,87,89

Family Income and Expenditure Survey 1980,85,89,91

Central Bureau of Statistics 1981, 83, 85, 87, 88 and 89

*Ringen 1970, 73, 76, 79, 82 and 86

*Rodrigues (1993) 1980 and 90

*Mercader 1980/1 National Institute of Statistics (INE) 1990/1

Jansson (1990) Gustafsson and Palmer (1993) 1975, 78, 80-90

Economic Trends 1977-92 Households Below Average Income 1979, 81, 87 and 88/89 $Blue Book (estimates exist also for earlier years) 1968/9-78/9, 81/2, 84/5.

Annual CPS data 1967-1991 Adjusted estimates 1973-90 Trends in Relative Incomes 1964, 69, 74, 79, 84 and 89 *$ Decennial Population Census 1950, 60, 70 and 80

With the exception of those studies marked with *, the estimates are the work of staff of the central bureau of statistics or central bank (Italy).

$ Using a different data source from that in the LIS dataset. estimates exist also for earlier years.

62

Estimates in Table 5.4 cover the 1980s, and indicate a rise in inequality, the Gini coefficient increasing by 2 percentage points. Saunders comments that the increase is "considerable over what is a relatively short period" (1993, p. 19). The upper half of the Lorenz curve shifts outwards by 1.4 percentage points, which corresponds to broadly the same annual rate of shift as the 0.5 to 0.7 percentage points recorded over the period 1981-82 to 1985-86 in Chapter 4.

Austria

1981182 1989/90

Table 5.4 Income Distribution in Australia 1981/82-1989/90 Cumulative decile shares of total income % and Gini coefficient

s,o s,o s,o s40 s,o s60 s,o s,o

3.2 8.7 15.3 22.9 31.5 41.1 51.9 64.4 3.0 8.3 14.7 22.0 30.3 39.7 50.5 63.0

s90 Gini

79.2 27 77.8 29

Note: Distribution among persons of equivalent disposable family income; the equivalence scale is that embodied in the Henderson poverty line.

Source: Saunders (1993), Table 5.

Data for Austria are available in the LIS dataset, but are used only to a limited extent in Chapter 4, since self­employment income is not covered and are therefore not comparable with estimates for other countries.

Data in Table 5.5 on the distribution of earnings for the period 1970 to 1987 show a rise in inequality between 1970 and 1976 but little apparent change from 1976 to 1987.

Belgium

1970 1976 1981 1987

Note:

Table 5.5 Earnings Distribution in Austria 1970-87 Cumulative quintile shares of total income % and Gini coefficient

s,o

6.8 6.6 6.6 6.9

21.1 20.2 20.2 20.4

s60

39.8 38.3 38.1 38.2

s,o

63.4 61.7 61.4 61.2

The data relate to the distribution among employees of individual gross earnings from employment, For the limitations of the data, see Guger (1989a, p. 190).

Source: Guger (1989a), Table 5.

Gini

29.3 31.2 31.4 31.6

The source for the national estimates by Cantillon et al. (1994) in the upper part of Table 5.6 is the same for the LIS dataset, though different methods are applied. Cantillon et al. use the OECD Social Indicators scale, which markedly gives more weight to larger families (Table 2.2 shows a value ofE of 0.73). Comparing these estimates with results in Table 4.3 for 1988 shows that the cumulative decile shares here are rather higher up to S40, and then somewhat lower. The difference is not large: for example only 0. 7 percentage points for the share of the top 20 per cent. ·

The estimates of Cantillon et al. suggest that there was a modest increase in inequality over the period 1985 to 1992, the Gini coefficient increasing by slightly over 1 percentage point. The share of the bottom 20 per cent fell from 10.7 per cent in 1985 to 10.4 per cent in 1992

As in a number of countries, income tax records provide an alternative source of information. However, this has both advantages and disadvantages - the latter including the fact that statistics do not cover those not subject to taxation (some 15 per cent of the population). The exclusion of an important section of the population has a major impact on the estimated degree of inequality, as is demonstrated by the calculation of Valenduc (1989) of the Gini coefficient for disposable income:

63

Household surveys

1985 1988 1992

Tax records

1983 1984 1985 1986

Table 5.6 Income Distribution in Belgium 1985-92 Cumulative decile shares of total income % and Gini coefficient

SIO

4.6 4.6 4.4

2.4 2.1 2.1 2.1

s"

10.7 10.5 10.4

s,. s ..

17.6 25.4 17.3 25.0 17.2 25.0

s, s .. s,. s,.

34.1 43.8 54.6 66.6 33.6 43.2 53.8 65.9 33.7 43.3 54.1 66.2

s,. Gini

80.5 22.5 79.6 23.4 80.2 23.7

79.3 27.4 78.9 27.4 78.8 27.9 78.5 28.3

Note: Distribution among persons of disposable household income; in the case of the household survey estimates, it is expressed per equivalent adult using the OECD Social Indicators scale.

Source: Household survey from Cantillon eta/. (1994), Table 30; tax records from Valenduc (1991).

Taxable households (as in Table 5.6)

All households

27.4 per cent

32.7 per cent

Both of these estimates indicate a substantially greater degree of inequality than that found in Chapter 4. This may be due to the fact that no adjustment is made for family size, and that certain transfer income is excluded, but it is also possible that the household survey data may understate high incomes.

Tax record data indicate a modest rise in inequality between 1983 and 1986. However, a change in tax legislation in 1984 meant that certain investment income no longer appeared. It has been suggested that the relative stability hides a more marked rise in the inequality of professional earnings (Valenduc, 1994).

Canada

The source for the national estimates shown in Table 5.7 is the same as in the LIS dataset, though slightly different years of the annual Survey of Consumer Finances have been employed in the study by Wolfson (1986). There are also small differences between the LIS dataset and those statistics produced internally by Statistics Canada (for example, in the imputation for taxes). Data for 1982, taken as the nearest year to 1981 (used in Chapter 4), show a large difference in the estimated degrees of inequality. According to Wolfson (1986), the Gini coefficient was 38 per cent, compared with 29 per cent in Table 4.8; the share of the top 20 per cent was 41.6 per cent compared with 36.3 per cent, and the share of the bottom 20 per cent was only 4.6 per cent compared with 7.6 per cent.

Table 5.7 Income Distribution in Canada 1971-83 Cumulative shares of total income % and Gini coefficient

s,. s .. s .. s,. s,. s, Gini

1971 3.4 13.8 31.7 57.3 74.1 84.3 39.8 1975 3.8 14.5 32.3 57.9 74.7 85.0 38.8 1979 4.2 15.0 32.8 58.4 75.2 85.5 38.0 1982 4.6 15.7 33.3 58.4 74.9 85.2 37.5 1983 4.6 15.3 32.6 57.7 74.5 84.8 38.2

Note: The data relate to the distribution of total disposable income per family unit. Source: Wolfson (1986), Table 3.

Differences in definition may account for part of the differences in the findings. The Wolfson results are not adjusted for family size, and are not weighted by the number of people. According to Wolfson's alternative calculations, applying an equivalence scale to pre-tax income (he does not apply the scale to after-tax income) would reduce the statistics as follows:

64

Change in percentage points Gini -3.8 Share of top 20% -2.1 Share of bottom 20% + 1.5

[from Wolfson (1986), comparing lines (a) and (d) in Table 3]

According to LIS results (Table 4.12), weighting by persons increases the share of the bottom 20 per cent by significantly less than 1 percentage point and reduces the share of the top 20 per cent by significantly more. These could therefore explain part, but not all, of the difference.

A further important difference is that the Wolfson data adopt a definition of the family unit, consisting of unattached individuals and parent(s) living with never-married children (Wolfson, 1986, p. 338). This is a narrower unit than the household, and means that multiple person households may appear as two or more family units. As a result, the recorded degree of inequality should be higher.

The trend is of particular interest. In Chapter 4, measured income distribution seems to show little change between 1981 and 1987. The Wolfson figures suggest the same for the period 1979-83, and the overall conclusion that he reaches for this period as a whole is that "the shape of the distribution of total money income has not changed appreciably in the last two decades" (1986, p. 352).

Finland

Estimates reported in Table 5.8 are based on household budget surveys, prepared in standardised form by the Central Statistical Office. Uusitalo's comparison (1989, p. 29) with the income distribution statistics suggests that both the rich and the poor are under-represented in the budget surveys, but that the difference is not large- around 1 or 1.5 percentage points in the Gini coefficient.

Table 5.8 Income Distribution in Finland 1966-85 Cumulative decile shares of total income % and Gini coefficient

SIO s,. s,. s,. s,. s .. s,. s,. s .. Gini

1966 3.1 7.8 13.5 20.3 28.2 37.4 48.0 60.4 75.6 31.8 1971 3.7 9.1 15.5 22.9 31.3 40.8 51.5 63.8 78.3 27.0 1976 4.5 10.7 17.9 26.0 35.0 44.9 55.8 67.9 81.6 21.6 1981 4.4 10.8 18.2 26.5 35.6 45.6 56.6 68.7 82.3 20.6 1985 4.8 11.3 18.8 27.1 36.2 46.1 56.9 68.8 82.1 20.0

Note: Distribution among persons of disposable household income per equivalent adult. Source: Uusitalo (1989), Table 5.4.

When comparing estimates for 1985 in Table 5.8 with those for 1987 in Chapter 4, it is important to remember that the equivalence scale is different. Uusitalo uses the "OECD scale" in the results cited, which gives a weight of 0.7 to the second adult and 0.5 per child. The results are in fact close: the Gini coefficient is 20.0 compared with 20.7 in Chapter 4.

The trends are particularly interesting. Estimates in Table 5.8 show a rapid reduction in measured inequality between the 1960s and the 1970s: the Gini coefficient fell by ten percentage points in a decade. The share of the top 20 per cent fell from 40 per cent to 32 per cent. The subsequent fall was much less substantial: about 1-1.5 percentage points in both cases. The slight fall in inequality between 1981 and 1985 in Table 5.8 and the slight rise between 1987 and 1990 in Tables 4.7 and 4.8 may indicate that there was little overall change in the 1980s.

The comparison of income distribution in Finland with that in Sweden is discussed below in the sub-section on Sweden.

France

Two major sources of evidence about the distribution of income exist in France. The first is the Enquete sur les Revenus Fiscaux, referred to as ERF, based on income tax declarations. Inquiries have been carried out periodically since 1956, with the most recent in 1970, 1975, 1979 and 1984. The ERF proceeds by drawing a representative sample of the population for which the income tax declarations are supplied by authorities. The

65

Table 5.9 Income Distribution in France 1970-79 Cumulative decile shares of total income %

SIO s" s,. s40 s,. s .. s,. s,. s,.

1970 0.9 3.6 7.2 12.9 19.8 28.3 39.1 51.8 68.0 1975 1.2 4.3 8.9 14.8 21.9 30.5 40.9 53.6 69.8 1979 1.4 4.8 9.6 15.7 22.9 31.6 42.0 54.6 70.5

Note: The original figures for 1970 do not add up to 100% on account of rounding. Source: Calculated from Sandoval (1989), Table 3.1.

Table 5.10 Measures of Inequality in France 1970-84

1970 1975 1979 1984

Gini 39.8 38.4 36.4 37.2 Gini * * 44.4 * 41.8 * 40.4 Kuznets 28.5 27.4 25.7 25.9 Theil 27.6 26.3 23.8 23.5 Atkinson 0.25 6.7 6.3 5.7 5.6 Atkinson 1.50 34.3 34.9 31.3 31.6

Note: Except* covers only households where the head is an employee or is retired (18 million out of20 million households); * covers whole population.

Source: Except * from Canceill and Villeneuve (1990), p. 71. *from Sandoval (1989), Table 3.3.

majority of non-taxable income (family benefits, minimum-vieillesse) is incorporated by an imputation procedure. This source is used in studies of income inequality in France (see Tables 5.9 and 5.10)- these are the data lodged in the Luxembourg Income Study (and hence used in Chapter 4). The second source of evidence in France is the household budget survey, the Enquete sur les Budgets Familiau:x, referred to as EBF. The EBF is conducted periodically - the 1984-85 survey is referred to here. Information is obtained by interview on expenditure, income and other variables.

The two sources are compared below. For the moment, this section concentrates on comparing the results using the ERF as shown in Tables 5.9 and 5.10- particularly the results of Sandoval for 1979, which cover the entire population (other results in Table 5.10 are limited to households where the head is an employee or retired) and may be compared with those in Table 4.7. The decile shares in Table 5.9 are much lower: that of the bottom 20 per cent for example is 4.8 per cent, compared with 8.4 per cent in Table 4.7. The Gini coefficient in Table 5.10 is 40.4 per cent, compared with 29.5 per cent in Table 4.8. Again differences in definition are important. The estimates of Sandoval are not adjusted for household size and are not weighted by the number of persons, although Table 4.12 does not suggest that the latter made a great deal of difference.

It was suggested in Chapter 4 that the distribution had not changed much between 1979 and 1984. The same conclusion appears in the national studies. According to Canceill and Villeneuve (1990), there was a slight rise in the Gini coefficient and a scarcely perceptible fall in the Theil and Atkinson (parameter = 0.25) indices. This "quasi statu quo" (the title of the article by Canceill and Villeneuve) may be contrasted with the more noticeable reduction in inequality between 1970 and 1979.

The existence of two sources raises the question of their relative reliability. The ERF and the EBF have been compared by Assemat and Glaude (1989) and Assemat (1989). Their calculations show that 8.9 per cent of households were below 50 per cent of the median, and 14.0 per cent below 60 per cent of the median, in the ERF in 1984 (the comparable figures in Table 4.2 aie 7.5 per cent and 13.2 per cent), but the estimates based on the EBF are 10.1 per cent and 17.2 per cent. Assemat and Glaude ( 1989) make use of the fact that part of the sample from the 1984 ERF consisted of the households in the sample for the 1984-85 EBF, which allowed a matching of households in the two surveys. This in tum permitted an examination of differential non-response and the comparability of the income figures reported, which suggests that the proportion of low income is understated in the EBF. A second important difference concerns the reporting of income-related social security benefits. The ERF imputes receipt of minimum pension (minimum vieillesse) in all cases of apparent eligibility, taking no account of incomplete take-up, which causes the proportion with low incomes to be underestimated in the ERF.

To sum up, the two sources of data in France have their relative advantages and disadvantages. The fiscal data (ERF) understate the low income population, which however may be overstated in the budget survey (EBF).

66

Germany

Section 5.1 made reference to the synthetic estimates of the distribution of income in Germany, combining information from a variety of sources. There is no reason to expect these to generate the same results as estimates in Chapter 4, which are based solely on household survey information. Moreover, the synthetic estimate shown in Table 5.11 relates to the household distribution of disposable income not adjusted for household size. On these accounts, one would expect the measured degree of inequality to be greater than in Table 4.4, but a sizeable part of the difference is likely to be attributed to differences in source. In fact, the Gini coefficient in Table 5.11 is 33.4 per cent in 1984, compared with 25.1 per cent in Table 4.4.

Alternative estimates from household survey data are shown in Table 5.12, based on the work of Hauser and Becker (1993). Two estimates are given for 1983: the first from the Income and Expenditure Survey and the second from the Socio-Economic Panel Study, which is the source used in Chapter 4. The distribution is based on a different equivalence scale from that used in Chapter 4, and excludes households with a non-German head (Hauser and Becker note that the latter makes only a very slight difference to inequality measures). The two distributions for 1983 are close and differ little from that reported for 1984 in Chapter 4.

Trends over time illustrate that the DIW synthetic estimates in Table 5.11 indicate a decline in inequality over the post-war period as a whole. The more detailed position is summarised by Guger as follows:

"The general picture which emerges is that in the 1950s and early 1960s the dispersion of income narrowed somewhat but widened again in the second half of the 1960s. From 1970, when the dispersion in income was nearly as high as in 1950, to the early 1980s, the development of quintile shares and Gini coefficients indicates a greater shift towards equality" (Guger, 1989, p. 66).

Estimates based on household survey data (Table 5.12) indicate that there was virtually no change in inequality between 1973 and 1983, but that inequality was slightly greater in 1990 than in 1983. As described by Hauser and Becker, there is a "very moderate trend to increasing inequality" (1993, p. 16).

1950 1955 1960 1964 1968 1970 1973 1975 1978 1980 1983 1984 1985

Table 5.11 Gini Coefficient of Inequality in Germany 1950-85 Synthetic estimates by DIW

DIW

39.6 38.4 38.0 38.0 38.7 39.2 37.0 36.6 36.4 36.6 33.9 33.4 35.2

Note: DIW is a synthetic estimate combining information from a variety of sources. Source: DIW from Guger (1989}, Chart I.

1973 1978 1983

1983 1985 1987 1990

Note: Source:

Table 5.12 Income Distribution in Germany 1973-90 Cumulative shares of total income (per cent) and Gini coefficient

s,. s,. s .. s,.

Expenditure Survey 10.4 24.5 42.0 64.1 !0.4 24.55 42.1 64.1 10.1 24.3 42.0 64.4

Socio-Economic Panel 10.1 24.3 42.2 64.8 9.8 23.8 41.6 64.3

10.1 24.4 42.2 64.6 9.9 23.9 41.5 64.0

The data relate to the individual distribution of equivalent disposable income per household. Hauser and Becker (1993}, Table 7.

67

Gini

25.4 25.4 25.5

25.0 25.95 25.2 26.0

Ireland

Evidence about the distribution of income is available from the Household Budget Survey (HBS), which has been carried out at seven-year intervals: 1973, 1980 and 1987. Data for 1973 were used by Nolan (1978). Figures for all three years, taken from Callan and Nolan (1993), are shown in Table 5.13. The last of the surveys was carried out in the same year as the ESRI survey available in the LIS dataset, and used in Chapter 4, but the results may be different:

"One important difference between the two surveys is that fieldwork for the ESRI one was carried out in the first half of the year and the farm income information obtained was retrospective and referred to 1986, which was a particularly bad year. The HBS, on the other hand, was carried out through 1987 and used farm accounts for that year, when farm incomes were on average 25 per cent higher" (Callan and Nolan, 1993, p. 8).

The results in Table 5.13 are also unadjusted for household size and have household weights. Comparing them with the "no adjustment" figures in Table 4.10, one can see that they are fairly close: for example, the share of the bottom 20 per cent is 5.9 per cent, compared with 6.2 per cent in Table 4.10.

Estimates in Table 5.13 suggest that there was a modest reduction in inequality between 1973 and 1980: the Lorenz curve moves upwards, but the difference in no case exceeds 1 percentage point. Between 1980 and 1987, the Lorenz curve shifts upwards more markedly at the bottom, but the curves cross at the median, before crossing again around the upper quartile.

Table 5.13 Income Distribution in Ireland 1973-87 Cumulative decile shares of total income % and Gini coefficient

s,. s,. s,. s •• s,. s60 s,. s,. s,. Gini

1973 1.7 5.0 10.0 16.5 24.3 33.5 44.4 57.4 73.6 36.7 1980 1.7 5.2 10.3 16.9 24.8 34.1 45.1 58.1 74.3 36.0 1987 2.2 5.9 10.9 17.2 24.8 34.0 45.0 58.4 75.0 35.2

Note: Distribution among households of disposable total household income. Source: Callan and Nolan (1993), Tables 3 and 4.

Italy

Estimates for Italy in Table 5.14 are based on the same data source used in Chapter 4 (the Bank of Italy survey), but they are different in definition from the results of the main tables in that no adjustment is made for family size and each family receives a weight of one. It may be noted that the results for 1986 (below) approximate the "no adjustment" estimates in Table 4.10. Estimates given by Brandolini and Sestito (1994, Table 2b) applying an equivalence scale with an exponent of 0.5, as in Chapter 4, but with each family weighted as 1, are very close to the household weighted figures in Table 4.12. As found in Chapter 4, the change in weighting appears to make little difference on the results for Italy: Brandolini and Sestito report that the mean value of the Gini coefficient over the period 1977-91 is 31.1 per cent on a household basis and 30.7 per cent on a personal basis.

The Bank of Italy survey has undergone substantial changes over time which are documented in Brandolini (1993). In addition to the change in sampling frame, shown as a break in Table 5.15, the definition of income has also changed. One should therefore be careful when drawing conclusions about trends over time.

The evolution of the distribution of income has been the subject of a number of studies; particular reference should be made to the reports of CNEL (Rossi, 1993 and 1994). According to Brandolini and Cannari (1992), there was a period of reduction in inequality from the middle of the 1970s to the beginning of the 1980s. This is evident in Table 5.14 where the Lorenz curve for 1983 (first estimate) lies clearly above those for the 1970s. Brandolini and Cannari suggest that subsequently in the 1980s inequality first rose and then fell after 1987 (although they note that inequality in 1987 is perhaps slightly overestimated). Brandolini and Sestito (1994) fit a regression equation to explain the Gini coefficient over the period 1977 to 1991 which suggests that there is a significant downward trend, but that inequality increases when the economy is growing faster. As they note, the latter conclusion counters those typically found in the United States, the United Kingdom and Canada.

68

Japan

1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

1983 1984 1986 1987 1989

Notes:

Source:

1977 1978 1979 1980 1981 1982 1983 1984 1986 1987 1989 1991

Notes:

Table 5.14 Income Distribution in Italy 1967-89 Cumulative decile shares of total income %

SIO s,. s,. s .. s,. s .. s,. s,. s,.

1.6 4.8 9.6 15.7 23.3 32.1 42.0 54.0 69.5 1.7 5.0 9.8 15.8 23.2 31.7 41.7 54.2 70.7 1.8 5.4 10.3 16.4 23.6 32.2 42.2 54.2 69.8 1.9 5.1 10.0 16.1 23.6 32.5 42.8 55.4 71.5 1.6 4.8 9.5 15.6 23.4 31.7 42.3 54.6 70.7 1.8 5.2 10.0 16.3 23.8 32.6 43.0 55.6 71.6 1.7 4.9 9.5 15.4 22.4 30.6 40.7 52.4 67.3 1.8 5.1 9.8 15.3 22.4 30.8 41.0 53.1 68.1 2.0 5.7 10.6 16.6 23.9 32.5 42.6 54.7 70.0 2.4 6.5 12.0 18.6 26.2 35.0 45.4 57.9 73.4 2.3 6.2 11.4 17.7 25.1 33.8 44.1 56.5 72.1 2.4 6.6 12.0 18.4 25.9 34.8 45.3 57.7 72.8 2.2 6.0 11.1 17.3 24.7 33.5 44.0 56.4 72.0 2.4 6.5 11.7 17.9 25.2 33.8 44.0 55.8 70.2 2.8 7.4 13.1 19.8 27.6 36.6 47.1 59.5 74.7 2.9 7.5 13.2 20.0 27.9 37.1 47.8 60.5 76.1 2.7 7.2 12.8 19.4 27.1 36.1 46.6 59.3 75.0

2.6 7.0 12.6 19.2 26.9 35.9 46.5 59.2 74.8 2.7 7.0 12.6 19.2 26.9 36.0 46.7 59.3 74.8 2.7 7.0 12.5 19.1 26.7 35.7 46.3 58.9 74.4 2.7 6.9 12.1 18.4 25.8 34.6 45.2 57.8 73.6 2.7 7.0 12.5 19.0 26.7 35.8 46.5 59.1 74.8

Data relate to the family distribution of equivalent family after-tax incomes. Prior to 1983 sample drawn from the electoral register; after 1983 the sample was drawn from the registry office records. Brandolini (1993}, Table B.7.

Table 5.15 Income Distribution in Italy 1977-91 Gini coefficient %

Gini

34.9 33.4 33.8 31.2 30.6 28.7 29.2 30.1 30.2 31.9 29.7 29.2

Data relate to the family distribution of equivalent family after-tax incomes. Prior to 1983 sample drawn from the electoral register; after 1983 the sample was drawn from the registry office records.

Source: Brandolini and Sestito ( 1994 ), Table 2a.

Evidence about the distribution of income in Japan in Table 5.16 is derived from the Family Income and Expenditure Survey, which does not cover the whole population. It excludes households engaged in agriculture, forestry or fishing, one-person households, foreigner households, households which manage restaurants, hotels or

69

Table 5.16 Income Distribution in Japan 1980-91 Percentiles of median

1980 1984 1985 1989 1991

PlO 53.5 50.5 49.5 50.9 49.7 P20 67.6 65.1 64.1 64.8 63.9 P30 79.4 77.2 75.8 76.6 76.0 P40 89.3 87.9 87.5 88.6 88.1 P60 112.8 112.8 113.2 113.1 114.9 P70 128.1 128.4 129.0 129.0 131.4 P80 148.9 148.6 150.1 151.3 152.6 P90 187.2 183.3 185.6 188.2 194.0 P90/Pl0 3.50 3.63 3.75 3.70 3.90

Cumulative decile shares (per cent)

1980 1984 1985 1989 1991

810 3.6 3.4 3.3 3.5 3.3 820 9.0 8.6 8.4 8.5 8.2 830 15.4 15.0 14.6 14.7 14.2 840 22.8 22.4 21.8 21.8 21.3 850 31.1 30.9 30.1 30.1 29.4 860 40.4 40.4 39.5 39.4 38.7 870 50.9 51.2 50.2 49.9 49.3 880 63.0 63.5 62.5 62.0 61.4 890 77.5 78.2 77.2 76.7 76.2

Note: Distribution of household disposable income (i.e. household weights and no adjustment for household size). Source: Family Income and Expenditure Survey.

similar establishments, households with 4 or more living-in employees, and households where the head is absent on long-term. For this reason alone, the estimates are not comparable with those for other OECD countries examined here. Comments on this comparison are therefore avoided. 2

Bearing in mind that trends over time may also be affected by the incomplete coverage of the population, one may note that the decile ratio in Table 5.16 appears to have increased over the 1980s. The bottom decile fell as a proportion of the median in the first half of the 1980s, and the top deciles rose relative to the median in the second half. At the same time, the changes are not dramatic, with the bottom 20 per cent losing less than 1 percentage point and the top 20 per cent gaining 1.5 percentage points. It should also be noted that even though data cover 5 years, the overall picture may be affected by cyclical variations in the degree of inequality.

The Netherlands

Data for the Netherlands in Table 5.17 are based on the same source used in Chapter 4, but with a different definition of income (the estimates in Table 5.17 deduct from net income interest paid, health care and life insurance premia, wealth tax payments, and alimony paid). There is also a difference in definition from Tables 4.1 to 4.8. The estimates in Table 5.17 use household weights and make no adjustment for household size. The significance of both these differences is visible in the following calculations of the Gini coefficient in 1987:

No adjustment Scale= 0.5 Per capita

(Source: Central Statistical Bureau).

Household weight Person weight

31.0 29.8 27.4 28.7

26.8 28.3

2. The different sources of income distribution data in Japan are discussed by Bauer and Mason (1992).

70

Table 5.17 Income Distribution in the Netherlands 1981-89 Percentiles of median

1981 1983 1985 1987 1988 1989

P!O 49.0 50.3 48.5 45.2 45.6 45.2 P20 66.0 66.8 65.1 62.1 61.8 61.5 P30 77.6 77.8 76.9 74.9 74.6 74.2 P40 88.8 88.9 88.6 87.5 87.2 87.0 P60 112.8 112.3 112.7 113.4 113.7 114.1 P70 127.9 127.8 127.2 129.2 129.3 130.2 P80 147.4 147.5 146.3 148.7 148.1 149.9 P90 181.1 !80.7 178.4 180.2 180.3 182.8 P90/P10 3.69 3.59 3.68 3.99 3.96 4.05

Cumulative decile shares (per cent)

1981 1983 1985 1987 1988 1989

S!O 2.7 3.0 2.9 2.5 2.6 2.6 S20 8.0 8.4 8.1 7.5 7.6 7.4 S30 14.5 15.0 14.6 13.8 13.9 13.5 S40 22.0 22.5 22.1 21.3 21.4 20.8 S50 30.5 31.0 30.7 29.9 30.0 29.3 S60 40.2 40.5 40.4 39.6 39.9 39.0 S70 51.1 51.3 51.3 50.7 51.1 50.1 S80 63.5 63.7 63.7 63.4 63.8 62.8 S90 78.2 78.3 78.4 78.3 78.7 77.8 Gini 28.3 27.8 28.1 29.4 29.0 29.6

Note: Distribution of household disposable income (i.e. household weights and no adjustment for household size). Source: Central Bureau of Statistics.

First, comparing the underlined estimate with that in Table 5.17 (29.4 per cent) shows that the difference in the definition of income served to reduce the measured degree of inequality in the national study compared with Chapter 4 estimates. Second, the use of a household weight/no-adjustment definition served to increase the measured inequality by a larger amount (compare the bold italic figure with that underlined).

The trend over time in Table 5.17 suggests that there was little change in the first half of the 1980s but a rise in inequality in the second half. The Gini coefficient increased by 1.5 percentage points in 4 years, and the decile ratio widened from 3.7 to 4.1.

Norway

The results obtained in Chapter 4 may be compared with those from two Norwegian studies using data based on the same source. Table 5.18 shows estimates produced by the Central Statistical Bureau for 1982, 1986, 1989 and 1990. The equivalence scale differs from those used in Chapter 4, being the "OECD" form, but one should not expect the results to be greatly different. Comparing the estimates for 1982 with those given in Chapter 4 for

1982 1986 1989 1990

Table 5.18 Income Distribution in Norway 1982-90: Estimates by Central Statistical Bureau

Cumulative decile shares of total income and Gini coefficient (per cent)

sw s,. s,. s,. s,. soo s,. s,.

4.2 10.3 17.3 25.2 34.0 43.7 54.4 66.3 4.5 10.6 17.7 25.7 34.6 44.3 55.0 67.0 4.1 10.1 17.1 25.0 33.7 43.2 53.7 65.4 4.1 10.2 17.2 25.2 34.0 43.7 54.4 66.2

s"' Gini

79.9 23.4 80.6 22.6 78.9 24.4 79.7 23.7

Note: Individual distribution of household disposable income divided by an equivalence scale of 0.7 for second household member and 0.5 for subsequent members.

Source: Central Bureau of Statistics.

71

1979 illustrates that the decile shares are not identical, but are quite close at the bottom. The Gini coefficient is 23.4 per cent, compared with 22.2 per cent in Chapter 4. The estimate for 1986 is close to that given in Chapter 4.

The second set of estimates by Ringen (1991), in Table 5.19, are substantially different. According to Ringen's estimates for 1986, the Gini coefficient is 33 per cent, compared with the figure of 22.6 per cent from estimates of the Central Statistical Bureau. The difference is not yet explainable.3

Both series shed light on trends over time. The trends in Table 5.19 have been summarised by Ringen (1991) as indicating a rise in inequality from 1970 to 1976 and a fall from 1976 to 1986, but as he says:

"The trends observed here are not strong- there is not much more inequality in 1976 than in 1970 and not much less in 1986 than in 1976" (1991, p. 7).

The estimates in Table 5.18 indicate that the Gini coefficient did increase by nearly 2 percentage points between 1986 and 1989, but the estimate for 1990 was little different from its value in 1982. There is no clear evidence of a steady upward trend in income inequality over the 1980s.

Portugal

1970 1973 1976 1979

1979 1982 1986

Notes:

Table 5.19 Income Distribution in Norway 1970-86: Estimates by Ringen (1991)

Cumulative decile shares of total income and Gini coefficient (per cent)

s,o s,o s,o s" s,o s60 s,o s,o s,o Gini

2.0 6.3 12.5 20.1 28.9 38.8 50.0 62.9 78.1 30.5 0.7 4.3 9.9 17.2 26.0 36.0 47.6 60.9 76.6 34.8 1.9 6.0 12.0 19.4 28.0 37.9 49.3 62.3 77.6 31.9 2.0 5.9 11.7 19.4 28.3 38.4 49.9 63.0 78.4 31.4

1.9 5.6 11.1 18.2 26.6 36.3 47.3 59.9 74.8 34.6 2.0 5.9 ll.5 18.7 27.2 36.9 47.9 60.5 75.5 33.4 1.9 5.7 11.3 18.6 27.3 37.2 48.5 61.4 76.7 33.0

In the data prior to 1979 (first estimate) depreciation provisions are deducted from self-employment income; after that date they are not deducted. Individual distribution of household disposable income divided by an equivalence scale of 0.7 for second household member and 0.5 for subsequent members. Original data do not always add to 100 per cent on account of rounding.

Source: Calculated from Ringen ( 1991 ), Table 2.

The evidence for Portugal in Table 5.20 differs from most of the other national studies in that these estimates were prepared by an individual researcher, using household survey data, to the same specification as Tables 4.1-4.8. While the data have not yet been fully incorporated into the LIS, it is interesting to compare them with findings for other countries in Chapter 4.

Compared with other countries, the decile ratio values would place Portugal similar to Ireland. The Gini coefficient in 1989/90 is similar to that in Italy.

The trend over the 1980s was towards less inequality, in that the Lorenz curve for 1989-90 is above that for 1980-81, but the difference is small, particularly at the top. The results of Gouveia and Tavares (1993), based on the same source, similarly show a reduction in relative inequality, using both income and expenditure variables and taking E = 0 or 1.0. On the latter basis, the per capita income distribution shows a fall in the Gini coefficient of 1 percentage point.

3. One possible reason for differences in income inequality estimates lies in the definition of disposable income. Ringen takes income recorded gross of deductions for tax purposes, which include all private interest payments. These are substantial: in 1982 interest on debt accounted for almost 7 per cent of gross income (Ringen, 1986, p 61). Interest is much more important for the upper income groups, representing only 1.6 percent of gross income for the bottom 20 per cent in 1982 but 22.0 per cent of gross income for the top 1 per cent (Ringen, 1986, p 61).

72

1980/81 1989/90

1980/81 1989/90

Table 5.20 Income Distribution in Portugal Percentiles of median and decile ratio

PIO P, P, p90

47.4 69.2 143.5 203.2 48.6 69.0 143.2 202.4

Cumulative decile shares of total income per cent

s,. s,. s,. s,. s,. s .. s,.

3.1 7.8 13.8 20.1 28.7 37.9 48.3 3.4 8.0 13.9 20.9 28.9 38.1 48.5

P, P.JPIO

252.7 4.29 248.5 4.16

s,. s90 Gini

60.6 75.8 32 60.8 75.8 31

Note: The results are for the distribution among persons of household disposable income adjusted by an equivalence scale equal to (household size)0·5•

Source: Rodrigues (1993).

This does not however account for the differences between Tables 5.20 and 5.21, since the Central Statistical Bureau estimates also refer to gross property income, and the impact of deducting interest paid on the Gini coefficient in 1986 is only to reduce it by 0.6 percentage points. [Source: Epland, 1992 (comparing Table 4 and Table 5); it may be noted that the share of the top decile group is reduced from 19.5 per cent to 18.7 per cent when interest paid is deducted.]

Spain

As in the case of Portugal, the first set of estimates in Table 5.21 was prepared by an individual researcher, using household survey data, to the same specification as Tables 4.1-4.8. The second set of estimates, for 1990-91 was supplied by the National Institute of Statistics (INE). While the data are not yet fully incorporated into the LIS, it is interesting to compare them with findings for other countries in Chapter 4.

The first conclusion from Tables 5.20 and 5.21 is that the distributions appear relatively similar in Portugal and Spain. In 1980-81, the decile ratio is 4.3 in Portugal and 4.4 in Spain. There is more inequality in Spain as the Lorenz curve lies below that of Portugal, but the difference is slight (nowhere larger than a 0.6 percentage point). In 1989-90, the Lorenz curves for the two countries cross, being higher for much of the range in Spain where the decile ratio is lower. The value of the Gini coefficient for Spain in Table 5.19 is close to that for Italy in Table 4.4.

In view of source differences between the two estimates for Spain, no attempt is made to draw conclusions about changes over time.

1980/81 1990/91

1980/81 1990/91

Table 5.21 Income Distribution in Spain Percentiles of median and decile ratio

P,o P, P, p"'

46.3 68.1 143.4 203.0 50.1 69.3 143.2 197.9

Cumulative decile shares of total income (per cent)

SIO s,. s,. s,. s,. s .. s,.

2.8 7.4 13.2 20.1 28.2 37.5 47.9 3.3 8.6 14.6 21.6 29.6 38.6 49.0

P, p .JPw

248.1 4.38 243.8 3.95

s,. s90 Gini

60.2 75.5 32.1 61.2 75.8 30.7

Note: The results are for the distribution among persons of household disposable income adjusted by an equivalence scale equal to (household size )0·5.

Sources: 1980/81 supplied by M. Mercader, DELTA, Paris. 1990/91 supplied by INE, Madrid.

73

Sweden

Data underlying estimates for Sweden in Table 5.22 are drawn from the same source as those in Chapter 4, but with a number of differences in method. The equivalence scale employed in Table 5.22 follows the social assistance norms recommended by the National Social Welfare Board, based on the number of adults and the number and ages of children, with a regional adjustment for housing costs. The definition of disposable income in Table 5.22 includes an allowance for the imputed rent on owner-occupied homes, equal to 2.5 per cent of the net asset value. These differences must be accounted for when considering the approximately 1.5 point difference in the Gini coefficients recorded for 1981 and 1987. In particular, attention should be drawn to the fact that Table 5.22 shows a relatively modest fall in the share of bottom income groups between 1981 and 1987: for example the share of the bottom 20 per cent fell from 11.3 per cent to 10.9 per cent, whereas Table 4.7 shows a fall from 10.6 per cent to 9.5 per cent. The possible role of the equivalence scale leading to this difference in findings is indicated by the fact that the per capita estimates in Table 4.10 show only a fall from 9.7 per cent in 1981 to 9.4 per cent in 1987, which is similar to that in Table 5.20.

Estimates in Table 5.22 show a decline in income inequality up to 1981, and then a reversal of that trend. In this respect the picture from the two years studied in Chapter 4 (1981 and 1987) is representative of that decade. The evidence of Gustafsson and Palmer (1993) indicates that there was a substantial increase in inequality in 1990 and 1991 (not shown in Table 5.22 in view of the discontinuity in the series).

There have been a number of interesting studies comparing the distribution of income in Sweden with that in other Nordic countries. Ringen estimates the family distribution (i.e. the unit of analysis is the family) of family equivalent disposable income in 1982 and concludes that:

"the distribution of disposable income is less inegalitarian in Sweden than in Norway" (1986, p. 43).

For example, the share of the top 20 per cent in disposable income is estimated to be 35.5 per cent in Norway, compared with 31.1 per cent in Sweden. These figures may be compared with those of 33.7 per cent for Norway in Table 5.18 and 31.2 per cent for Sweden in Table 5.22.

Comparisons of Sweden and Finland for the 1970s were done by Sandstrom (1980) and Nygard (1984). The following summary of their findings by Gustafsson and Uusitalo (1990) provides a salutary lesson:

"Their studies demonstrate the importance of a careful co-ordination of the data. When the distributions of disposable income as defined in the national data sets were compared, Sweden was shown to have a more egalitarian distribution than Finland. When differences in the estimates of imputed income from home-owner occupancy were removed, the difference disappeared or changed such that Finland was shown to have the more egalitarian distribution. When the differences in family definitions were removed, Sweden was again shown to be the more egalitarian" (1990, p. 76).

Table 5.22 Income Distribution in Sweden 1975-90 Cumulative decile shares of total income (per cent) and Gini coefficient

s,. s,. s,. s'" s,. s'" s,. s,. s,. Gini

1975 4.4 10.7 17.9 25.9 34.8 44.7 55.6 67.7 81.4 21.3 1978 4.6 11.2 18.6 26.8 35.8 45.7 56.6 68.7 82.3 20.0 1980 4.6 11.3 18.8 27.1 36.2 46.1 57.0 68.9 82.3 19.4 1981 4.4 11.3 19.1 27.5 36.6 46.5 57.3 69.1 82.5 19.1 1982 4.4 11.2 18.9 27.3 36.4 46.3 57.0 68.8 82.1 19.4 1983 4.4 11.2 18.9 27.3 36.4 46.3 57.1 68.9 82.3 19.4 1984 4.1 10.8 18.4 26.8 35.9 45.8 56.5 68.3 81.7 20.4 1985 4.4 11.1 18.6 26.9 35.9 45.6 56.2 67.9 81.3 20.5 1986 4.1 10.7 18.2 26.4 35.3 45.0 55.6 67.3 80.6 21.4 1987 4.3 10.9 18.4 26.7 35.8 45.6 56.3 68.1 81.1 20.5 1988 4.3 10.9 18.5 26.9 35.9 45.7 56.4 68.2 81.5 20.4 1989 4.3 10.9 18.3 26.5 35.5 45.3 56.0 67.8 81.2 21.0 1990 4.0 10.4 17.7 25.9 34.9 44.7 55.4 67.3 80.8 21.9

Note: Distribution among persons of disposable household income per equivalent adult, using the equivalence scale embodied in the social assistance norms.

Source: Gustafsson and Palmer (1993).

74

United Kingdom

The first set of results for the United Kingdom in Table 5.23 are based on the same source as Chapter 4 (the Family Expenditure Survey). They allow a direct comparison with Table 4.7 for 1979 and 1986. For 1979, the share of the bottom 20 per cent is a little higher in Table 5.23: 9.7 per cent compared with 8.7 per cent. But, when considering the rounding-off of published figures, the other quintile shares for 1979 in Table 4.7 appear to be virtually the same. For 1986, the share of the bottom 20 per cent is again higher in Table 5.23 (8.8 per cent compared with 7.5 per cent), but the differences are not large: the Gini coefficient is 31 per cent, compared with 30.4 per cent in Table 4.8.

The methods applied in obtaining the estimates in Table 5.23 differ from those in Chapter 4 in that they use a different equivalence scale, and the Economic Trends estimates use household weights. The estimates in Table 5.24, taken from the Households Below Average Income study, apply person weights- but the equivalence scale is again different from Chapter 4. The share of the bottom 20 per cent in 1979 is again slightly higher than in Table 4.7.

There are other differences in methods which may in part account for any differences in findings. These include the exclusion of observations (for example, those where there is an absent spouse, or where a member of the household has been self-employed for less than 2 months), the use of price indices to adjust income information, and the factors used to gross-up to population totals (see Department of Social Security, 1992). The Household Below Average Incomes estimates (apart from those for 1983 and 1985) use data from tax records (Survey of Personal Incomes) to adjust the top of the distribution.

The trend in inequality in Table 5.23 is clear: the Lorenz curve at the end of the 1980s is distinctly inferior to that in 1979. LIS estimates in Chapter 4 showed the share of the bottom 40 per cent to be falling by 2 percentage points between 1979 and 1986. The Economic trends estimates show a fall of 3 percentage points (in rounded figures) from 1979 to 1987; the Households Below Average Incomes figures show a fall of 2.7 percentage points. The Gini coefficient was shown in Table 4.8 as falling by 3.6 percentage points between 1979 and 1986. The Economic Trends estimates show a fall of 6 points between 1979 and 1987. [For further discussion of the trend in income inequality in the United Kingdom, see, among others, Atkinson (1993) and Gardiner (1993).]

Table 5.25 shows the results from a different source: the "Blue Book" estimates referred to in Section 5.1. These combine information from the tax records with household survey data and other information to produce a synthetic estimate. They also differ in being based on the narrower family unit; they are not adjusted for family size. The level of inequality (for example a Gini coefficient of 33.5 per cent in 1978-79) cannot therefore be directly compared with the earlier estimates (26.8 per cent for 1979 in Table 4.8). On the other hand, the trend in the Gini coefficient is not dissimilar: it fell by 2.5 points between 1978-79 and 1984-85 (the most recent estimate available on a Blue Book basis). It may be noted that this followed a period of slight decline from 1968-69.

Table 5.23 Income Distribution in the United Kingdom 1977-90 Cumulative decile shares of total income (per cent) and Gini coefficient

Economic Trends Estimates

s20 s"' s60 s,. Gini

1977 9.7 23 41 64 27 1978 9.8 24 42 65 26 1979 9.4 23 41 64 27 1980 9.2 22 40 63 28 1981 9.3 22 39 62 28 1982 9.5 23 40 63 28 1983 9.5 22 39 62 28 1984 9.6 23 40 63 28 1985 9.2 22 39 62 29 1986 8.8 22 39 62 31 1987 8.2 20 36 59 33 1988 7.6 19 35 58 35 1989 7.6 19 36 59 34 1990 7.0 18 34 57 36

Note: Household distribution of household equivalent income. Source: Economic Trends, March 1991, Table N, January 1993, p. 187, and January 1994, pp. 122-123.

75

Table 5.24 Income Distribution in the United Kingdom 1979-1988/89

1979 1981 1987 1988/89

1983 1985

Cumulative decile shares of total income (per cent)

Households Below Average Income Estimates

SIO s,o s,o s,o s,o soo

4.2 9.9 16.6 24.1 32.7 42.2 4.1 9.7 16.2 23.6 32.0 41.4 3.6 8.6 14.5 21.4 29.4 38.4 3.2 7.9 13.5 20.3 28.2 37.3

4.1 9.8 16.4 23.8 32.2 41.5 4.2 9.7 16.1 23.4 31.7 41.0

Notes: Individual distribution of household equivalent income before housing costs. Estimates for 1983 and 1985 based on earlier methodology and relate to Great Britain.

s,o s,o

53.0 65.2 52.1 64.3 48.9 61.0 47.8 60.0

64.1 63.8

s90

79.4 78.8 75.9 74.9

Source: Department of Social Security (except 1983 and 1985 from Households Below Average Income 1981-87, Table AI).

Table 5.25 Income Distribution in the United Kingdom 1968/69-1984/85 Cumulative decile shares of total income (per cent)

s,o s,o s,o s,o s,o soo s,o s,o s90 Gini

1968/69 6.6 11.9 18.5 26.6 36.3 47.8 60.9 76.4 33 1969/70 6.4 11.6 18.3 26.4 36.1 47.5 60.8 76.4 34 1970/71 6.6 11.8 18.3 26.1 35.6 46.8 60.1 76.0 34 1971/72 6.6 11.7 18.1 25.9 35.3 46.6 60.0 75.9 34 1972173 6.8 12.3 18.8 26.8 36.3 47.5 60.7 76.5 33 1973/74 3.2 7.4 12.8 19.2 27.0 36.5 47.7 60.9 76.4 33 1974/75 3.1 7.5 12.8 19.2 27.0 36.4 47.8 61.0 76.8 32 1975/76 3.2 7.6 13.0 19.5 27.4 37.1 48.5 61.9 77.7 32

1975/76 3.1 7.4 12.7 19.0 26.7 36.3 47.6 61.1 76.9 33 1976/77 3.0 7.5 12.6 19.2 26.9 36.2 47.5 60.8 76.8 33 1977/78 3.0 7.4 12.6 18.8 26.5 35.8 47.1 60.5 76.7 33 1978/79 2.9 7.0 12.1 18.5 26.2 35.5 46.8 60.3 76.6 33.5 1981182 2.4 6.4 11.6 17.9 25.2 34.0 44.8 58.0 74.4 36 1984/85 2.7 6.9 11.8 17.8 24.9 33.5 43.9 56.9 73.5 36

Note: The "new series" from 1975n6 adds back mortgage interest paid. Source: See Atkinson and Micklewright (1992), Tables Bll and BI2.

United States

In the United States there is considerable wealth of income data. Here the report concentrates primarily on data drawn from the same source as employed in Chapter 4 (the Current Population Survey) but which are customarily analysed in different ways. Table 5.26 reproduces the standard annual Bureau of the Census estimates which differ from those in Chapter 4 in that they:

relate to household income unadjusted for size; use household weights; relate to gross income, not deducting direct taxes; and do not include food stamps.

We should therefore expect the estimates in Table 5.26 to exhibit a greater degree of inequality than those in Tables 4.7 and 4.8. (Table 4.10 indicates that the no-adjustment (for household size) estimates of the decile shares were some 1-1.5 points lower (in the middle), and Table 4.12 that the use of household weights made a similar difference.)

76

------·-·-·--

Table 5.26 Income Distribution in the United States 1967-91: Income Shares Cumulative shares of total income (per cent) and Gini coefficient

s,. s,. s .. s,. s, Gini

1967 4.0 14.8 32.1 56.3 82.5 39.9 1968 4.2 15.3 32.8 57.2 83.4 38.8 1969 4.1 15.0 32.5 57.0 83.4 39.1 1970 4.1 14.9 32.3 56.8 83.4 39.4 1971 4.1 14.7 32.0 56.5 83.3 39.6 1972 4.1 14.6 31.7 56.2 83.0 40.1 1973 4.2 14.7 31.8 56.4 83.4 39.7 1974 4.3 14.9 31.9 56.5 83.5 39.5 1975 4.3 14.7 31.7 56.4 83.4 39.7 1976 4.3 14.6 31.6 56.3 83.4 39.8 1977 4.2 14.4 31.3 56.0 83.2 40.2 1978 4.2 14.4 31.3 56.0 83.2 40.2 1979 4.1 14.3 31.1 55.8 83.1 40.4 1980 4.2 14.4 31.2 56.0 83.5 40.3 1981 4.1 14.2 30.9 55.7 83.5 40.6 1982 4.0 14.0 30.5 55.0 83.0 41.2 1983 4.0 13.9 30.3 54.9 82.9 41.4 1984 4.0 13.9 30.2 54.8 82.9 41.5 1985 3.9 13.7 29.9 54.3 82.4 41.9 1986 3.8 13.5 29.7 54.0 82.0 42.5 1987 3.8 13.4 29.5 53.8 81.8 42.6 1988 3.8 13.4 29.4 53.7 81.7 42.7 1989 3.8 13.3 29.1 53.1 81.1 43.1 1990 3.9 13.5 29.4 53.4 81.4 42.8 1991 3.8 13.4 29.3 53.5 81.9 42.8

Note: Household distribution of household gross income (i.e. household weights and no adjustment for household size). Source: U.S. Department of Commerce (1992), Table B-3.

There is wide agreement that income inequality has increased in the United States. According to the 1994 Economic Report of the President,

"Starting some time in the late 1970s, income inequalities widened alarmingly in America" (1994, p. 25).

Table 5.26 suggests that this widening took place particularly in the first half of the 1980s. The distribution remained relatively stable from 1967 to 1979, with the Gini coefficient around 40 per cent. In the 1980s, it began to increase. The overall rise in the coefficient from 1979 to 1986 (2.1 percentage points) is less than that recorded in Table 4.8 (3.2 percentage points), but the direction of movement is similar, and there is no suggestion that 1986 (the second year studied in Chapter 4) is atypical.

The trend over time may be affected by the choice of definition, and a different picture may arise from that in Table 5.26 when adjusting for household size or when moving from gross to net income. Mayer and Jencks, for example, argue:

"that trends in inequality differ for families and households; that trends for households differ when we weight all persons equally instead of weighting all households equally; that trends for total household income differ from trends for per capita household income; that trends in the top half of the distribution do not always mirror trends in the bottom half of the distribution; and that the decennial Census tells a somewhat different story from the Current Population Survey (CPS) about the 1960s and 1970s" (1993, p. 123).

However, they immediately go on to say that:

"Nonetheless, all measures derived from the CPS indicate that income inequality grew in the early 1980s and remained unusually high in the late 1980s" (1993, p. 123).

Other recent U.S. studies (e.g. Karoly, 1993, and Danziger and Gottschalk, 1993) find this same pattern of rising inequality in U.S. incomes for 1979 through 1989 using several different income measures.

A number of the variations listed by Mayer and Jencks have been considered in Chapter 4 and are discussed briefly here. Table 5.27 shows estimates prepared by the Congressional Budget Office of the distribution adjusted

77

Table 5.27 Income Distribution in the United States Adjusted for Household Size and for Taxes/Benefits 1973-90

Cumulative quintile shares of total income

s,. s,. s .. s,.

A Unadjusted Gross income 1973 4.0 14.2 31.3 56.3 1979 3.9 13.9 30.5 55.6 1989 3.6 12.8 28.5 53.0 1990 3.5 12.7 28.4 52.9

B Divided by Gross income poverty threshold 1973 5.5 17.3 34.4 58.3 1979 5.1 16.7 34.0 58.3 1989 4.3 14.8 31.3 55.3 1990 4.3 14.9 31.4 55.6

c Per capita Gross income 1973 5.2 16.2 32.2 55.4 1979 4.9 15.9 32.2 55.8 1989 4.1 14.1 29.7 53.1 1990 4.1 14.1 29.7 53.1

D Divided by poverty threshold After taxes + food and housing benefits 1979 6.4 19.0 36.8 61.0 1989 5.6 17.0 34.0 58.0 1990 5.6 17.0 34.0 58.1

Note: Individual distribution of family equivalent (or per capita) income; see text for definition of equivalence scale. Source: Committee on Ways and Means, US House of Representatives, 1992 Green Book, p. 1356.

for family size and allowing for taxes and food/housing benefits. Adjusting for household size is done by dividing by the weighted average official poverty threshold, which varies with family size.4 As illustrated in Table 2.2 (line 30), this corresponds to a value of E of 0.56, which is close to the value of 0.5 used in Chapter 4 (for a comparison, see Ruggles, 1990, Table 4.4.). Adopting an equivalence scale shifts the Lorenz curve upwards, the shares of the bottom three quintile groups being larger and those of the top quintile groups smaller. The shift is less on a per capita basis. The after tax and benefit series, to be compared with panel B, shows a significant reduction (about 5 percentage points) in the share ofthe top 20 per cent. Other than housing benefits, the definition of income resembles that adopted in Chapter 4.

Table 5.28 provides a more comprehensive run of years for the series labelled B in Table 5.27, where the distribution is presented in terms of the average income, relative to the poverty line, of people in each quintile group.5 So that in 1973, the bottom 20 per cent received on average 93 per cent of the poverty threshold income (before tax). This percentage visibly fell sharply between 1979 and 1983; it has since recovered, but remains below its 1973 level. The next quintile group changed little over the period as a whole, while the upper three quintile groups experienced an increase in incomes relative to the poverty threshold.

Yet another representation is provided by the Bureau of the Census Trends in Relative Income series shown in Table 5.28, which presents the cumulative percentage below different proportions of the median. This series uses person weights and adjusts by the same equivalence scale as in Chapter 4 (i.e. it is different from the poverty threshold used in Tables 5.27 and 5.28). The results show the proportion with incomes below 50 per cent of the median rising from 20.0 per cent in 1979 to 21.8 per cent in 1984. The comparable estimates in Table 4.6 are 16.6 per cent and 18.4 per cent (in 1986). The longer term picture is somewhat different from Table 5.28 in that these figures suggest that the proportion with low incomes has risen since 1969: 32.9 per cent had less than three-

4. The official poverty threshold in the United State varies not only with family size but also by age of householder (under 65 years, or 65 and over) and number of family members under 18 years (see U.S. Department of Commerce, 1993, Table A, p. vii); and in earlier years it varied also by gender of head and location. The "weighted average thresholds" are calculated as the average, for a particular family size, of those for different numbers of children and for those under and over 65, weighted by the number in the category (in Table 5.27 the weights are those for 1989). In contrast, Table 5.28 uses the full range of variation in official poverty thresholds.

5. It should be noted that these estimates are not fully comparable with those in Table 5.28, since the equivalence scale now adjusts for age, gender and location of family head, as well as for family size (see previous footnote).

78

Table 5.28 Income Distribution in the United States: Average Income as a Percentage of Poverty Threshold

Quintile group Lowest Second Middle Fourth Top

1973 93 197 285 394 686 1975 86 188 276 384 667 1977 89 196 291 407 702 1979 92 209 309 431 736 1980 87 200 299 422 716 1981 82 193 294 418 737 1982 77 189 290 417 748 1983 76 187 292 423 757 1984 79 193 300 434 795 1985 81 197 305 441 807 1986 82 204 316 459 837 1987 82 206 322 466 843

1987 84 209 326 473 873 1988 85 208 325 472 860 1989 87 2!0 328 475 880 1990 83 203 318 464 855

Notes: Individual distribution of family equivalent gross income. See text for definition of equivalence scale. A revised method of processing the data was introduced in 1987.

Source: Committee on Ways and Means, US House of Representatives, 1992 Green Book, p. 1376.

Table 5.29 Income Distribution in the United States: Relative Incomes

1964 1969 1974 1979 1984 1989

Note: Source:

1950 1960 1970 1980

Cumulative percentage below different percentiles of the median

< 25 <50 < 75 < 100 < 125

6.7 19.2 33.7 50.0 64.3 5.5 17.9 32.9 50.0 65.0 5.5 18.7 34.0 50.0 64.9 6.7 20.0 34.5 50.0 63.8 6.2 21.8 35.8 50.0 62.4 8.3 22.1 36.2 50.0 62.7

Individual distribution of family equivalent gross income. US Department of Commerce (1991), Table A.

Table 5.30 Income Distribution in the United States: Estimates Based on Population Census

Cumulative shares of total income

s" s,o soo s,o

3.1 13.6 30.4 54.8 4.4 15.5 32.0 55.4 4.7 16.0 32.7 56.1 4.5 15.6 32.5 56.5

Note: Individual distribution of family equivalent gross income. Source: Committee on Ways and Means, US House of Representatives, 1992 Green Book, p. 1437.

< 200

88.3 89.1 89.0 88.1 85.9 85.3

s90

71.8 71.6 72.2 72.9

Overall average

331 320 337 355 345 345 344 347 360 366 379 384

393 390 396 385

< 300

96.9 97.2 97.2 96.7 95.8 95.2

s.,

82.8 82.1 82.5 83.3

quarters of the median in 1969 but 34.5 per cent in 1979 and 36.2 per cent in 1989. The picture one obtains of the trends may therefore depend on how data are analysed and on adjustments to data for differences in family size, taxes and weights.

79

In order to see the effect of other data sources, Table 5.30 presents the findings from the Census Public Use tapes by Danziger and Cancian for the individual distribution of gross equivalent income using the official poverty thresholds. This series suggests that in the 1970s the share of the bottom 20 per cent began to shrink after rising in the previous two decades. The results for this period are not inconsistent with those in other tables, but no information is provided about the 1980s (1990 Census data were unavailable at the time of this study).

Finally, the standard Census Bureau series on poverty measurement (using a different equivalence scale from that in Table 5.29, but the same as that in Tables 5.27 and 5.28) shows that after declining in the 1960s the poverty rate for persons in 1979 (11.7 per cent) was little different from that in 1969(12.1 per cent). The rate increased in the early 1980s reaching 15.2 per cent in 1983. The most recent estimates in this series put the poverty rate at 14.5 per cent in 1992, while the 1986 rate was 13.6 per cent (U.S. Department of Commerce, 1992a, Table 2). A different Census Bureau series, using the same poverty line but adjusting incomes for taxes and benefits, starts at a lower poverty rate for persons (8.9 per cent in 1979) and shows the same general trend, with these low-income rates rising to 12.7 per cent in 1983 and 11.4 per cent in 1991 (U.S. Department of Commerce, 1992b, Table H, line 14).

5.3 Conclusions from national studies

This review of national studies has served three main purposes. The first was to compare LIS results with those by other studies using similar data. In general, the results are broadly consistent with those reported in Chapter 4, with differences largely explained by differences in definitions, though in some cases (such as Norway) there is room for further investigation.

The second purpose was to extend the range of countries, introducing for example evidence for Austria, Japan, Portugal and Spain. In the former two cases, the coverage of data sources did not allow comparison with estimates for other countries, so that while the results for Austria and Japan are of considerable interest in their own right, conclusions cannot be drawn about the relative degree of inequality. Data for Portugal and Spain were more comparable in form, and the estimates suggested that their degree of inequality was broadly similar to that of Italy or Ireland.

The third purpose was to develop LIS analyses of trends over time. The national studies typically provide more comprehensive coverage of years, therefore allowing the two years considered in Chapter 4 to be placed in temporal context. The findings may be summarised as follows (while keeping in mind that any such summary risks over-simplification):

Australia Austria Belgium Canada Finland France Germany Ireland Italy Japan Netherlands Norway Portugal Sweden United Kingdom United States

Trend in income inequality over the 1980s

rise in inequality (1981/82-1989/90) little change 1981-87 (in earnings distribution) modest rise 1985-1992 little change in the early 1980s (1979-1983) little change in the early 1980s (1981-1985) little change in the early 1980s (1979-1984) modest rise 1983-1990 little change 1980-1987 downward trend, with cyclical behaviour (1980-1991) rise in inequality (1980-1991) little change in first half of 1980s, modest rise in inequality in second half little apparent trend 1982-1990 modest fall in inequality 1980-1989 rise in inequality 1980-1990 definite rise in inequality 1980-1990 rise in inequality 1980-1991

Of the sixteen countries, Belgium, the Netherlands and Germany show a "modest" rise in inequality (between a 1 and 2 percentage points increase in the Gini coefficient), and Australia, Japan, Sweden, the United Kingdom and the United States show a bigger rise, with the United Kingdom exhibiting by far the largest increase in income inequality in the 1980s.

80

Chapter6

ANALYSIS OF COUNTRY DIFFERENCES:

PRIMARY AND MARKET INCOMES

This chapter probes more deeply into LIS findings by examining the level and trend in income inequality across countries. In terms of the definitions introduced in Chapter 2, this chapter is concerned with primary income (excluding property income) and market income. Primary income (excluding property income) consists of wages and salaries and self-employment income, net of employer contributions for insurance and other benefits, but gross of employee contributions to such schemes. It forms the bulk of aggregate household income in all countries. For shorthand, this aggregate is referred to as "earnings." Earnings are the foundation of household economic well­being, particularly for the prime working age (25-to-54-year old household heads) population. Earned income not only meets the direct needs of workers and their families, it also provides a large fraction of the tax base which supports redistributive government social welfare programmes. Prior to government taxation and transfer, other sources of self-generated private market income also come into play. "Market income" includes primary income plus other sources of income which are not redistributive government transfers. These other sources include property income (interest, rents, and dividends received on a regular basis), occupational pensions from former employers, regular interhousehold cash transfers, and court ordered payments in the form of child support and alimony. Only households with non-zero levels of primary income or market income are counted here, including households with negative earnings (e.g. from self-employment).

The most complete level of comparability is reached across countries in terms of disposable income. However, treatment of subcomponents of disposable income vary across countries, thus hampering the potential level of comparability. Data for Austria exclude self-employment income and are not included here. In Sweden, to be consistent with other countries, reinserting sick pay into earnings was possible despite its administration and payment by the tax office. Other differences were not so easily resolved. Belgium, France, Italy and Luxembourg collect all of their information on a post-income tax and post-employee social security contribution basis. The level and trend in earnings inequality in these countries is therefore affected both by market driven patterns of wages and earnings and by tax policy. Furthermore, unemployment insurance is included in French earnings and cannot be separated. And, Finnish data were prepared by consultants and not strictly under LIS control. Analyses of earnings and market income in these countries are therefore treated as less comparable than those in others. In most of these analyses, they are included "below" the line and separated from other countries.

Another problem arises in countries such as Sweden, France and Germany where it is impossible to separate occupational pensions from social retirement. In this time period, social retirement forms the vast bulk of their total pension income (Pestieau, 1991). In these cases it was decided to treat all pensions as social retirement and hence include them as part of the overall tax and transfer scheme in Chapter 7. Separating the two income types was possible for most countries with extensive occupational pensions.

The measures employed here involve the same person-weights used in earlier chapters, though no adjustments were made for differences in household size or other characteristics. In terms of earlier terminology, this is equivalent to using an E coefficient of 0 as an equivalence factor where earnings (or market incomes) are the measure of household income. The only other departure in this chapter is the analysis of changes in the level and trend in primary income inequality among the prime working age group - households with 25-to-54-year-old heads. Differences in their earnings are less likely to be contaminated by changes in retirement patterns, school leaving and related demographic and social phenomena which affect the labour market attachment of other age groups and hence earnings and market income distributions. When considering all households, the indirect effects of government policy are noticeable on labour supply (e.g. retirement decisions, welfare programme effects), on

81

realised property income (e.g. income tax policy), and on related economic behaviours. But these primary and market income distributions are still essential building blocks toward understanding the distribution of income prior to direct government taxation and transfer policy.

It is important for the reader to note how the choices made for this chapter affect the interpretation of the results. Because the analyses treat households with any non-zero primary income, the results are affected by the choices of husbands and wives for full-time or part-time work, and by variation in hours of work. One cannot therefore draw conclusions about the dispersion or trend in wages or in full-year full-time salaries (see Gottschalk and Joyce, 1991 and 1992, for LIS evidence on this issue). To the extent that zero primary income is the result of individual choice (reflecting, for example, preferences for leisure over the life cycle), the sample studied here is self-selected and does not allow conclusions to be drawn about the population as a whole. There are a number of different ways in which data may be viewed, and the analysis here does not claim to provide more than a partial view.

The chapter contains three basic types of analyses. First, it examines summary measures of income distribution and change using various ratios of percentile break points in the size distribution, cumulative income shares and aggregate summary measures of inequality. The percentile break points are particularly intuitive and are less subject to top and bottom coding biases inherent in household income microdata. Second, it concentrates on the cumulative size distribution of incomes, Lorenz dominance, and on differences in aggregate income shares between the bottom 50 and top 10 per cent in each country and time period and for each group. Adjustments for negative incomes are made as outlined in Chapter 3. These comparisons give a broad idea of the concentrations of primary and market incomes in each country and how they are changing over time. Finally, it examines the differences and changes in the aggregate shares of primary incomes attributable to various groups: husbands', wives' and others' earnings, self-employment income, and also differences and changes in shares of aggregate market incomes (earnings of all household members, self-employment income and property income, and other private income). Each analysis investigates both the shares of the bottom 50 and top 10 per cent of the distribution and their composition. The reader should carefully note the income concepts employed and measures used here for they serve as the basis for the next chapter which concentrates on the effect of government taxes and transfers on income distribution.

Table 6.1 Percentage Population Shares for Subtotal Analyses of Primary and Market Income, Total Households, and Prime Age Households

Non-zero Primary Non-zero Primary Non-zero Market Income, Prime Age Income, All Income, All

Country/Year Households" Households Households

Australia, 1981 61.6 78.9 93.1 AUSTRALIA, 1987 65.4 81.4 93.8 SWITZERLAND, 1982 65.1 85.2 98.3 Canada, 1981 67.7 89.4 95.7 CANADA, 1987 68.1 87.7 95.6 GERMANY, 1984 62.6 81.1 94.7 IRELAND, 1987 60.2 81.2 88.4 Netherlands, 1983 60.9 73.1 83.3 NETHERLANDS, 1987 63.0 74.2 85.5 NORWAY, 1979 61.7 87.9 94.6 Sweden, 1981 58.8 83.5 95.9 SWEDEN, 1987 58.6 83.3 97.6 United Kingdom, 1979 62.7 82.4 94.3 UNITED KINGDOM, 1986 56.7 72.5 90.6 United States, 1979 64.6 87.4 95.2 UNITED STATES, 1986 65.6 85.9 94.6

Belgium, 1985 72.0 78.0 79.9 BELGIUM, 1988 69.7 76.8 78.9 FINLAND, 1987 71.0 98.0 98.8 France, 1979 67.5 87.0 91.6 FRANCE, 1984 67.2 84.0 90.6 ITALY, 1987 63.1 83.3 88.9 LUXEMBOURG, 1985 67.9 82.3 85.5

Overall Averageb 64.4 83.0 91.8

Notes: a Prime age households are headed by persons aged 25 to 54. b Overall average of countries in capitals.

82

It should be noted that, in contrast to Chapter 4, data are presented for all countries and years together, rather than separating analyses of the level, and then the trend, of inequality. In the tables which follow,1data for the most recent year which serves as the basis for comparisons of levels across countries are indicated by a capitalised country name. Data for earlier years, used in trend analyses, are indicated by the country name in normal type or separated from the more recent estimates. Data for Belgium, Finland, France, Italy, and Luxembourg are separated by a line to highlight their less than full comparability with the data for other countries studied here.

Because of differences in labour force behaviour, and in type and receipt of occupational pensions, child support and property income, the percentage of the entire population included in each country varies. Table 6.1 shows the fraction of the entire household population (where households with disposable income are 100 per cent) contained in each major subgrouping. Prime age households with non-zero primary income average about 64 per cent of all households, varying from below 60 per cent in the United Kingdom and Sweden to around 70 per cent in Belgium, Finland and Canada. While the United Kingdom monthly income reporting format may bias their estimates downward, and while the French inclusion of prime age unemployment insurance recipients tends to make their estimate higher than it would otherwise be, these estimates are not very different from other countries with more consistent data.

If all households with non-zero primary income are considered regardless of age, one would find that the coverage rises on average to 83 per cent, with the United Kingdom and the Netherlands at about 73 per cent (note the fall in the United Kingdom between 1979 and 1986) and Canada, France, Norway and the United States approaching 90 per cent.

Households with non-zero market incomes are on average 91 per cent of all units, ranging from 79 per cent in Belgium, and 86 per cent in Luxembourg and the Netherlands, to 98 per cent in Sweden and Switzerland. (It should be noted that these figures may be affected by the non-reporting of income such as that from property; it has been suggested that this may be partly responsible for the figure of 79 per cent in Belgium.) For most OECD countries, an overwhelming proportion of households have at least some income derived from market sources. On the other hand, there is evidence that reliance on non-market income sources in the form of government transfer payments is also substantial. While 83.0 per cent of households had non-zero primary incomes, only 74.7 per cent relied on primary income for at least half their total income. While 91.8 per cent of households had non-zero market incomes, only 79.2 per cent relied on market income for at least half their total income. Thus, while receipt of market income is high, on average more than a fifth of all households rely on the government as the major source of their income.

6.1 Level and trend in primary income distribution

The size distribution of primary income, typically and often interchangeably termed "earnings" in quantitative economic studies such as this, varies considerably across countries and over time. Household earnings are a package of adult and child earnings with the head of the household - the husband in two earner units by international convention - typically having the largest amount of such income. However, wives and other earners - typically children of all ages but also other relatives in some cases - have played an increasingly larger role in generating earned income in recent years. This issue is treated in greater detail later in the chapter. This section only examines aggregate earnings which may be affected by these patterns. Earnings also vary according to a country's industrial structure. The mix of wages and salaries vis-a-vis farm and non-farm (small business) self­employment income also vary widely by country. The issue of labour income mix is discussed later in this chapter as well.

Recent studies of trends in the size distribution of earned income indicate a trend toward greater inequality in a number of OECD countries, including several studies based on the LIS database (Green, Coder, Ryscavage, 1992; Gottschalk and Joyce, 1991, 1992). These studies point to three general explanations for changes in earned income inequality in advanced countries:

- Structural economic change or "deindustrialisation" by which the growth in personal and business service sectors is larger than that in the manufacturing sector; The growing internationalisation of economies whereby changing patterns of international trade are causing large swings in the composition of output and in patterns of specialisation of industries within countries; and

- Technical and technological change which transfers production processes and which generally reduces demand for low skill labour.

Significant differences exist in the changes in earnings dispersion across OECD countries and no one single explanation seems to dominate (OECD, 1993).

83

In those countries where earnings dispersion has increased significantly, it has been argued that such changes are secular and structural, not cyclical short-term phenomena. These changes have brought with them a general increase in the return to high quality human capital, particularly to higher education. In many countries such as the United States, the United Kingdom, and Canada, the gap between the earnings of tertiary school graduates and elementary and secondary school graduates has increased dramatically during the 1980s (Lorenz and Wagner, 1990; Levy and Murnane, 1992).

Cyclical economic phenomena are also important. LIS data employed here cover different years in different countries. Because the speed of economic expansion and/or contraction vary considerably across countries, even same year data would not translate into the same stage of the business cycle in each country. Unemployment rates, a primary macroeconomic indicator of labour market tightness and labour demand, vary significantly across countries and over time (unemployment rates can be found in Chapter 3, Table 3.2 earlier in this volume). These patterns should be considered when moving through the figures and charts which follow. Earlier analyses (Srneeding and Coder, 1993) have shown that the percentile points at the lower end of the primary income distribution are sensitive to cyclical factors while the upper end of the earnings distribution is much less sensitive. The reader should realise that no attempt was made to identify full-year full-time workers or wages (earnings per hour) in these analyses. Others (e.g. see Gottschalk and Joyce, 1992) have conducted such analyses using this same database. The results presented here differ in that they incorporate the effects of variation both in weeks worked per year and in part-time/full-time work. 1

The vehicles employed to summarise the level and trend in the earnings distribution are five percentile break points (lOth, 25th, 75th, 90th, and 95th) relative to the median, "top" relative to "bottom" percentile points (90/10 and 80/40), two summary measures of income concentration: Atkinson (using exponent of 0.5) and Gini, and a comparison of Lorenz curves via a Hasse diagram. The summary measures are explained in Chapter 2 and Appendix 3. The logic of the percentile break points and the ratios of these points to one another is straightforward. The points P10, P25, P75, P90, and P95, capture the dispersion of incomes relative to the median

Table 6.2 Summary of Level and Trend in Primary Income Distribution for Households with Heads Aged 25 to 54a

Percentile Ratios to the Median'

Country/Year PIO P, P, p90 P•s P,JP,o P,JP,o

Australia 1981 30.1 67.2 139.4 180.8 206.5 600.7 171.6 AUSTRALIA 1985-86 36.2 67.8 136.4 175.9 201.9 485.5 167.1 SWITZERLAND 1982 51.7 76.5 135.4 179.0 218.7 346.4 160.9 Canada 1981 31.4 63.5 137.3 175.8 199.4 559.3 173.4 CANADA 1987 27.3 62.3 139.9 181.7 207.9 665.1 177.7 GERMANY 1984 47.6 69.6 135.6 174.9 202.8 367.7 167.4 IRELAND 1987 27.0 62.6 155.8 225.7 273.1 835.4 207.7 Netherlands 1983 56.9 75.7 137.9 180.4 214.4 316.8 164.9 NETHERLANDS 1987 54.9 75.0 141.5 185.3 214.3 337.4 171.7 NORWAY 1979 42.9 72.4 129.9 155.6 175.4 362.8 152.9 Sweden 1981 33.9 67.9 135.7 165.6 187.5 488.3 164.4 SWEDEN 1987 28.1 63.0 133.5 165.1 186.3 588.4 168.0 United Kingdom 1979 44.1 70.9 137.2 179.2 203.2 406.2 167.9 UNITED KINGDOM 1986 32.2 65.3 142.9 192.8 220.4 598.2 179.9 United States 1979 27.7 60.0 142.4 190.1 220.1 686.3 179.3 UNITED STATES 1986 26.3 59.0 147.4 201.0 239.5 764.5 192.0

Belgium 1985 52.7 69.0 134.1 167.6 187.7 318.2 165.6 BELGIUM 1988 49.9 68.0 135.6 168.9 185.4 338.4 162.6 FINLAND 1987 43.7 66.7 134.7 178.0 210.7 406.9 166.4 France 1979 41.3 65.2 143.9 192.0 228.5 465.0 187.3 FRANCE 1984 40.0 62.4 146.3 200.6 238.9 500.9 187.3 ITALY 1986 47.6 72.8 142.4 183.2 204.2 384.6 180.1 LUXEMBOURG 1985 58.8 76.4 131.9 174.0 203.7 296.0 161.4

Overall Average• 40.9 68.0 139.3 182.8 212.2 485.2 173.5

a Primary incomes are unadjusted and recorded only for those with non-zero amounts. b Percentile ratios are the ratios of income levels dividing the specified percentiles of the distribution multiplied by I 00. c Overall average of countries in capitals.

1. The United Kingdom data, being based on weekly or monthly earnings, are not affected by unemployment spells, though the number of households with earnings is affected as mentioned above.

84

household (P50), while the ratios of P90 to PIO and P80 to P40 show the incomes of the highly ranked relative to those of the lowly paid. For trend analysis, widening inequality is captured by declines in relative percentiles below the median and/or increases in relative percentiles above the median. Higher ratios of top to bottom incomes illustrate rising inequality, as the higher percentile increases relative to the bottom percentile. All ratios are multiplied by 100 to present them in percentage terms. This section begins by examining the distribution of earnings among households with prime working age heads (ages 25 to 54), and then moves to distributions of earnings and market income for all units.

On average, earnings in prime age households at the lOth percentile are about 41 per cent of the median, while those at the 90th percentile are roughly 183 per cent of the median household. But there is a wide variance across countries (Table 6.2). At the bottom end of the distribution (PlO, P25), relative earnings are highest in smaller central European countries (the Netherlands and Switzerland) while they are the lowest in the United States, Ireland and Canada.2 These positions reverse at the upper end of the distribution where Ireland and the United States stand out as the countries with the relatively highest earnings levels at the 90th and 95th percentiles, followed by the United Kingdom, Switzerland, and the Netherlands at the 95th percentile. Ireland and the United States have the highest overall top to bottom percentile ratios, while the Netherlands, Norway and Switzerland have the least relative spread. In all countries, as expected, the 80/40 ratios are much more closely grouped than are the 90/10 figures. The least relative inequality appears in Luxembourg, but this could be very much affected by their income and employee payroll tax system.

Table 6.3 Summary Measures of Level and Trend in Primary Income Distribution for Prime Age Households with Heads Aged 25 to 54

Most Recent Period Earlier Period

Country/Year Gini Atkinson 0.5' Country/Year Gini Atkinson 0.5"

IRELAND, 1987 36.6 11.5 UNITED STATES, 1986 35.7 11.3 United States, 1979 32.8 9.9 CANADA, 1987 32.3 9.5 Canada, 1981 30.7 8.9 UNITED KINGDOM, 1986 32.2 9.1 United Kingdom, 1979 28.3 7.0 SWITZERLAND, 1982 30.7 8.4 SWEDEN, 1987 29.7 8.4 Sweden, 1981 28.4 7.5 AUSTRALIA, 1985/86 29.6 7.9 Australia, 1981/82 30.6 8.6 NETHERLANDS, 1987 28.6 6.9 Netherlands, 1983 27.9 6.6 GERMANY, 1984 27.6 6.5 NORWAY, 1979 25.4 6.2

FRANCE, 1984 34.9 10.3 France, 1979 35.5 11.4 ITALY, 1986 29.2 7.2 FINLAND, 1987 29.1 7.5 BELGIUM, 1988 24.9 5.0 Belgium, 1985 25.2 5.1 LUXEMBOURG, 1985 24.3 4.7

a Negative incomes were adjusted to a uniform bottom code (see Chapter 3).

The summary measures of inequality in Table 6.3 are consistent with the percentile discussions based on Table 6.2. Countries are ranked by the Gini index for the most recent observation period, though the Atkinson measure gives almost exactly the same ranking. Ireland and the United States stand out as having the most primary income inequality, followed by Canada and the United Kingdom. At the other end of the spectrum, Norway has the least unequal primary income distribution among the countries studied here. The remainder of the countries are closely bunched with Ginis from 30.7 per cent (Switzerland) to 27.6 per cent (Germany). In considering the high recorded level of inequality in Ireland in Tables 6.3 and 6.5 (below), and the high value of the P90/P10 ratio, readers should bear in mind that the farm income information obtained referred to 1986, which was a particularly bad year for agricultural incomes in Ireland.

2. Luxembourg and Belgium, which are on an after tax basis, have the relatively highest earnings at the lOth percentile. Sweden has a relatively low PlO ratio most likely due to the high prevalence of part-time work, differences due to the classification of single persons as separate households, and the reaction of employers' wage policies to high marginal income tax rates. Overall, Sweden has one of the most equal distributions of prime age earnings among full-year full-time earners among all LIS nations (Gottschalk and Joyce, 1992).

85

Table 6.4 Summary of Level and Trend in Primary Income Distribution for All Households•

Percentile Ratios to the Median'

CountryfYear P,. P, P, p90 P, P,JP,. P80/P40

Australia 1981 26.1 64.9 143.2 188.1 215.7 720.0 178.8 AUSTRALIA 1985-86 28.1 64.3 141.1 183.5 210.5 652.0 175.8 SWITZERLAND 1982 33.0 71.7 138.1 185.7 229.1 562.5 168.2 Canada 1981 23.5 59.3 142.4 184.4 209.6 786.1 183.3 CANADA 1987 21.4 56.8 146.4 193.9 223.9 907.8 189.0 GERMANY 1984 37.7 68.3 138.2 179.2 210.7 475.0 172.2 IRELAND 1987 22.2 58.0 161.6 235.7 286.0 1063.2 217.3 Netherlands 1983 50.6 75.0 140.0 185.2 219.6 366.3 169.9 NETHERLANDS 1987 51.0 72.8 141.1 185.7 217.2 364.2 176.4 NORWAY 1979 18.9 67.6 135.4 165.5 186.4 877.7 165.5 Sweden 1981 13.8 56.1 147.4 182.9 206.9 1329.5 186.9 SWEDEN 1987 9.8 52.0 151.1 190.7 215.9 1944.9 199.7 United Kingdom 1979 36.7 68.0 140.1 185.1 211.8 504.9 172.1 UNITED KINGDOM 1986 27.3 61.1 145.2 196.8 226.3 720.6 187.4 United States 1979 19.3 54.9 148.3 200.0 236.4 1037.6 195.9 UNITED STATES 1986 18.8 53.6 152.4 214.5 256.3 1138.5 203.1

FINLAND 1987 18.1 53.9 145.9 197.0 234.4 1087.3 194.7 France 1979 31.7 63.7 149.7 202.8 244.1 638.9 193.1 FRANCE 1984 32.4 60.2 150.0 210.0 252.6 648.4 199.3 ITALY 1986 39.8 71.8 149.2 193.9 221.0 487.5 187.0 Belgium 1985 52.7 69.9 137.3 173.9 195.6 329.9 171.8 BELGIUM 1988 50.0 68.2 134.7 168.2 186.4 336.4 163.2 LUXEMBOURG 1985 53.8 72.7 135.8 179.0 209.7 333.0 166.0

Overall Averagec 30.8 63.5 144.4 192.0 224.4 773.3 184.3

a Primary incomes are unadjusted and recorded only for those with non-zero amounts. b Percentile ratios are the ratios of income levels dividing the specified percentiles of the distribution multiplied by I 00. c Overall average of countries in capitals.

For several countries, there are observations at two dates. In the case of the United States, the United Kingdom and Canada, where there is a clear move towards increased disparities, the Gini coefficient increased by at least 1.5 percentage points. For the other countries, the picture is more mixed. While a trend towards increased dispersion is visible in the top to bottom percentile ratios (Table 6.2), where every country experiences a trend toward greater inequality in the P90/PIO ratio except for Australia, the Gini coefficient either rises by a small amount (Sweden and the Netherlands) or falls (Australia, France and Belgium). Again it should be emphasised that the distribution of wages for full-time, full-year workers, which can change over time in a different manner, is not the focus here. For evidence that wages became more unequal in Australia over this period, see Gottschalk and Joyce (1992).

Adding in non-prime age household heads naturally widens the distribution of primary income, even when restricting the analysis to households with non-zero levels of earnings (Table 6.4). At the bottom of the distribution, low earnings of college students and younger workers and part-time workers of all ages push down the P ratios. At the top end, highly paid older workers raise the average levels. But the countries at the extremes are not much changed- smaller European countries such as the Netherlands (and below the line, Belgium) tend to have relatively high earnings at the bottom end of the distribution (and also Switzerland and Germany at the P25 level), while Sweden, Norway, Finland, and the United States are among the lowest. The high prevalence of part­time work among Scandinavian women may account for a great deal of this pattern in Sweden, Norway, and Finland. The upper end of the distribution shows high P-ratios for Sweden, Finland, Ireland and the United States and relatively meager high earnings in Norway (and Belgium). Top to bottom ratios are lowest in the Netherlands (for the P90/PIO ratio), and highest in Sweden, the United States and Ireland (for the P90/P10 ratio).

The rankings of countries by Gini and Atkinson measures of primary income for all households (Table 6.5) are similar, but not identical to the rankings for prime age households (Table 6.3). Ireland and the United States have the most inequality, followed by Sweden. Canada and Switzerland are grouped at the next level. At the very bottom of the distribution, primary income is much more equally distributed in the Netherlands and Norway than in other countries (except for Luxembourg and Belgium, which are found below the line). For countries with two observations, the United States, Canada and the United Kingdom, which were among those with the highest levels of first period inequality, show a marked increase in summary inequality measures.

86

Table 6.5 Summary Measures of Level and Trend in Primary Income Distribution for All Households

Most Recent Period Earlier Period

Country/Year Gini Atkinson 0.5" Country/Year Gini Atkinson 0.5"

IRELAND, 1987 40.3 14.2 UNITED STATES, 1986 38.6 13.4 United States, 1979 35.9 11.8 SWEDEN, 1987 36.6 13.7 Sweden, 1981 34.6 11.9 CANADA , 1987 35.3 11.6 Canada, 1981 33.1 10.1 SWITZERLAND, 1982 35.1 11.4 UNITED KINGDOM, 1986 34.0 10.2 United Kingdom, 1979 30.4 8.3 AUSTRALIA, 1985/86 32.2 9.5 Australia, 1981182 32.3 9.6 NORWAY, 1979 30.5 9.5 GERMANY, 1984 29.9 7.7 NETHERLANDS, 1987 29.9 7.7 Netherlands, 1983 29.4 7.4

FRANCE, 1984 37.6 12.1 France, 1979 37.8 12.6 FINLAND, 1987 37.5 14.1 ITALY, 1986 31.5 8.4 LUXEMBOURG, 1985 25.8 5.3 BELGIUM, 1988 25.7 5.4 Belgium, 1985 26.0 5.5

a Negative incomes were adjusted to a uniform bottoin code (see Chapter 3).

Lorenz curves present yet another way to present cross-national differences in inequality. Here only the most recent data from each country are included to show the cumulative decile shares of income and the share of the top decile (Table 6.6). Unlike the percentile ratios, which rely on distance from the median to create comparisons, Lorenz measures are based on mean income differences. The picture of inequality which emerges from this comparison is somewhat different from that shown by the P-ratios in Table 6.4. Switzerland leads Ireland, followed by the United States in having the highest income share for the top decile. At the bottom of the distribution, the Netherlands has the highest income share in both the lowest decile and quintile (with Belgium and Luxembourg, below the line, second). Ireland, Sweden, the United States (and Finland below the line) have the lowest decile and quintile shares of primary income.

The Lorenz curves can be constructed from the cumulative decile shares. Figure 6.1 summarises the Lorenz comparisons which can be made on this basis. Belgium, Luxembourg, France, and Italy have been included, but they are in italics to indicate the limits on their comparability across countries. As a criterion of significance, the LIS employs a 1.0 percentage point difference.

An unambiguous comparison is possible in 54 of the 78 cases in Figure 6.1. The comparisons are summarised in a Hasse diagram (Figure 6.2). Luxembourg and Belgium (in italics) have the least unequal distributions as their Lorenz curves are superior to every other country's curves, although it should be noted that they are not fully comparable with the countries which are not in italics. Germany; the Netherlands, Italy, and Australia follow with some rankings among those possible (e.g. Italy is inferior to the Netherlands, and Australia to Germany). The rest of the countries are closely bunched, although Ireland and the United States are more often Lorenz inferior as

Table 6.6 Summary of Primary Income Distribution for All Households: Cumulative Decile Shares

10% 20% 30% 40% 50% 60% 70% 80% 90% top 10%

AUSTRALIA, 1985 1.6 6.1 12.1 19.3 27.8 37.6 48.8 61.6 77.0 23.0 CANADA, 1987 1.2 4.7 10.1 17.0 25.4 35.2 46.7 60.0 76.0 24.0 GERMANY, 1984 2.2 7.3 13.6 20.9 29.4 39.1 50.1 62.7 77.9 22.1 IRELAND, 1987 0.2 3.2 8.0 14.2 21.8 30.8 41.8 55.2 72.4 27.6 NETHERLANDS, 1987 2.8 8.1 14.4 21.6 29.7 38.8 49.5 61.9 76.7 23.3 SWEDEN, 1987 0.5 3.3 8.5 15.3 23.8 34.2 46.5 60.5 76.8 23.2 SWITZERLAND, 1982 1.7 6.3 12.4 19.4 27.2 36.1 46.1 57.8 71.9 28.1 UNITED KINGDOM, 1986 1.6 5.5 11.2 18.1 26.4 36.1 47.2 60.3 76.4 23.6 UNITED STATES, 1986 1.0 4.0 8.9 15.3 23.3 32.7 43.8 57.2 74.0 26.0

FINLAND, 1987 0.6 3.3 8.4 15.1 23.7 33.8 45.6 59.3 75.8 24.2 FRANCE, 1984 1.7 5.6 10.7 16.9 24.5 33.5 44.1 56.7 72.4 27.6 ITALY, 1986 2.5 7.4 13.5 20.4 28.4 37.6 48.5 61.1 76.2 23.8 LUXEMBOURG, 1985 3.8 9.5 16.1 23.5 32.0 41.5 52.3 64.6 79.5 20.5 BELGIUM, 1988 3.7 9.1 15.5 23.0 31.8 41.8 53.0 65.5 80.1 19.9

87

Figure 6.1 Lorenz Comparisons of Primary Income for All Households

BE CN FR GE IR IT LX NL SWE sz UK us

Australia, 1985 + ? + ? ? + ? + + Belgium, 1988 \ + + + + + = + + + + + Canada, 1987 \ ? + ? ? + France, 1984 \ + ? ? ? ?

Germany, 1984 \ + ? ? + + + + Ireland, 1987 \ ? ?

Italy, 1986 \ ? + ? + Luxembourg, 1985 \ + + + + + Netherlands, 1987 \ ? + + + Sweden, 1987 \ ? ? ?

Switzerland, 1982 \ ? ?

United Kingdom, 1986 \ +

The grid should be read as follows: + indicates that for the country shown in the row we have Lorenz superiority over the country shown in the relevant

column. - indicates Lorenz inferiority. ? indicates that the Lorenz curves cross, with a difference of more than 1 percentage point in at least one direction. = indicates that none of these conditions are satisfied. Source: LIS Database.

compared to most others. Finally, Ireland is Lorenz inferior to Canada and the United States, but not to Switzerland or Sweden.

It is very difficult to pinpoint the exact source of these patterns both in level and change in primary income inequality. There is no single explanation that fits all cases. Moreover, a combination of factors is likely to be at work in most countries. Early retirement, part-time work trends, increasing returns to education and other human capital variables, cyclical and structural economic change may all play a role in explaining these patterns, with relatively greater or lesser importance depending on the country under examination. A sorting out of these factors will be a task for future analysts.

6.2 Level and trend in market income

This section broadens the focus to capture changes in full market income. Here the effects of property income, occupational pensions and private transfers come into play. Forces helping to drive changes in the distribution of market income include population ageing (which affects property income levels and occupational pensions), the growth of single parent families (which evoke an increase in private transfers for child support and alimony) and changes in savings levels and patterns of asset holdings (which affect property income receipt) across countries. Also affecting the market income distribution in a selective but indirect fashion is the relative size and structure of government social retirement pensions and tax policy. Countries with high levels of taxation of income and wealth and extensive public pension systems are more likely to have low overall property income shares and relatively smaller occupational pension systems relative to countries which are otherwise situated. These factors will affect the results which follow.

In moving from primary to market incomes roughly 10 per cent of households was added to this sample (see Table 6.1). Most of these units have relatively small amounts of private transfers, property incomes or occupational pensions. At the other extreme, a few have very large amounts of property incomes. As a result, distribution of market income is more skewed than the distribution of primary income (Tables 6.7 and 6.8

88

Figure 6.2 Relative Inequality Rankings Based on Primary Income in Different Countries

Luxembourg Belgium

Netherlands Germany

~

Italy Australia

United Kingdom

Canada

France United States

I I

Switzerland Ireland Sweden

compared to Tables 6.4 and 6.5). Older households with only small amounts of property income enter the universe of households counted at the bottom of Table 6. 7, while those with significant asset holdings and/or occupational pensions push up the high income household count. In fact, because of such extreme cases, ratios of the 90th to the lOth percentile became too volatile to be an accurate indicator and hence, were dropped from Table 6.7. And the Gini measures in Table 6.8 jump by large amounts in some countries. As a result, the reader should not pay a great deal of attention to the PlO levels due to the entrance of many retiree and single parent households with only small amounts of property income, occupational pensions or child support.

It appears that after taking account of these qualifications there is some change among countries with the lowest values for PlO and P25 (compare Tables 6.4 and 6.7), with Sweden having by far the lowest relative value for market income in the P25 range in Table 6.7. The Netherlands and Norway tend to have the least unequal distributions (highest P25 and lowest P90, P95s), while Ireland, the United Kingdom and the United States tend to

89

Table 6.7 Summary of Level and Trend in Market Income Distribution for All Households"

Percentile Ratios to the Median'

Country/Year PIO P, P, P, P, P,/P,o

Australia 1981 9.5 59.2 147.1 198.7 231.3 189.1 AUSTRALIA 1985-86 8.1 52.7 146.5 195.0 228.2 193.2 SWITZERLAND 1982 14.3 59.4 141.2 193.2 238.7 178.1 Canada 1981 19.6 56.6 145.3 191.2 220.8 190.1 CANADA 1987 16.4 53.1 149.4 199.9 233.4 199.6 GERMANY 1983 8.1 50.0 146.0 192.0 223.1 185.9 GERMANY 1984 0.7 54.0 147.3 197.2 233.0 191.7 IRELAND 1987 12.2 53.3 167.5 248.4 304.5 230.6 Netherlands 1983 20.4 71.2 143.0 191.3 228.1 177.9 NETHERLANDS 1987 17.9 68.6 146.1 194.3 231.3 182.2 NORWAY 1979 10.1 61.4 138.3 170.3 192.4 172.8 Sweden 1981 4.7 34.8 157.9 200.0 227.9 217.7 SWEDEN 1987 3.8 27.1 169.5 216.7 246.9 237.8 United Kingdom 1979 7.6 57.4 145.1 193.7 222.3 185.3 UNITED KINGDOM 1986 3.3 41.9 160.0 223.0 260.5 220.0 United States 1979 13.9 50.0 149.9 208.0 247.3 205.1 UNITED STATES 1986 13.0 47.7 154.8 220.5 269.5 215.7

Belgium 1985 49.9 68.3 139.7 175.6 199.6 176.1 BELGIUM 1988 49.5 67.6 135.1 171.2 186.5 168.0 FINLAND 1987 17.5 53.4 146.0 198.8 237.4 195.6 France 1979 19.6 61.3 153.2 209.6 251.0 200.2 FRANCE 1984 15.3 57.5 155.1 220.9 269.0 209.6 ITALY 1986 17.2 66.7 150.0 201.1 232.2 192.7 LUXEMBOURG 1985 48.4 72.6 138.7 190.6 218.5 171.0

Overall Average' 17.1 59.1 159.4 215.5 253.7 209.6

a Market incomes are unadjusted and recorded only for those with non-zero amounts. b Percentile ratios are the ratios of income levels dividing the specified percentiles of the distribution multiplied by I 00. c Overall average of countries in capitals.

Table 6.8 Summary Measures of Level and Trend in Market Income Distribution for All Households

Most Recent Period Earlier Period

Country/Year Gini Atkinson 0.5' Country/Year Gini Atkinson 0.5'

IRELAND, 1987 46.1 20.0 SWEDEN, 1987 43.9 20.0 Sweden, 1981 41.1 17.7 UNITED KINGDOM, 1986 42.8 18.7 United Kingdom, 1979 36.5 13.8 UNITED STATES, 1986 41.1 15.5 United States, 1979 38.8 14.1 SWITZERLAND, 1982 40.7 15.6 GERMANY, 1984 39.5 17.7 AUSTRALIA, 1985/86 39.1 15.2 Australia, 1981/82 36.9 13.6 CANADA, 1987 37.4 13.0 Canada, 1981 35.0 11.4 NETHERLANDS, 1987 34.8 11.7 Netherlands, 1983 33.9 11.0 NORWAY, 1979 33.5 11.9

FRANCE, 1984 41.7 15.8 France, 1979 40.6 15.1 FINLAND, 1987 37.9 14.3 ITALY, 1986 36.1 12.3 LUXEMBOURG, 1985 28.0 6.5 BELGIUM, 1988 27.3 6.3 Belgium, 1985 27.5 6.7

a Negative incomes were adjusted to a uniform bottom code (see chapter 3).

90

Table 6.9 Summary of Market Income Distribution for All Households: Cumulative Decile Shares

10% 20% 30% 40% 50% 60% 70% 80% 90% top 10%

AUSTRALIA, 1985 0.4 3.1 8.1 14.9 23.2 32.9 44.2 57.5 73.6 26.4 CANADA, 1987 0.9 4.0 9.0 15.6 23.9 33.6 45.1 58.6 75.0 25.0 GERMANY, I 984 0.1 1.8 7.2 14.3 22.7 32.7 44.3 58.0 74.5 25.5 IRELAND, 1987 0.1 2.0 6.7 12.7 19.6 28.7 39.8 53.5 71.1 28.9 NETHERLANDS, 1987 0.9 5.0 11.1 18.2 26.4 35.8 46.8 59.7 75.4 24.6 SWEDEN, 1987 0.2 1.1 4.2 10.1 18.2 28.5 41.2 56.3 74.0 26.0 SWITZERLAND, 1982 0.8 3.7 9.0 15.7 23.5 32.4 42.6 54.6 69.3 30.7 UNITED KINGDOM, 1986 0.2 1.7 5.7 11.9 19.9 29.7 41.5 55.6 73.1 26.9 UNITED STATES, 1986 0.6 3.2 7.6 13.7 21.5 30.9 42.0 55.5 72.5 27.5

FINLAND 1987 0.6 3.3 8.4 15.1 23.7 33.8 45.6 59.3 75.8 24.2 FRANCE, 1984 0.7 4.0 8.7 14.6 22.0 30.8 41.4 54.0 70.0 30.0 ITALY, 1986 1.0 5.0 10.7 17.5 25.4 34.7 45.6 58.6 74.5 25.5 LUXEMBOURG, 1985 3.0 8.4 14.9 22.2 30.6 40.1 50.9 63.4 78.7 21.3 BELGIUM, 1988 3.2 8.4 14.7 22.2 30.9 40.9 52.0 64.4 79.2 20.8

have the highest ratios at the P90 and P95 levels. In these countries the top quintile of households with market incomes is at least two and one-half times as large as the median household in most countries. Looking to the ratio of the 80th to the 40th percentiles, Sweden has the highest ratio, with Luxembourg at the other end of the spectrum. The United States, the United Kingdom, and Ireland also have levels of 80-40 ratios in the 215 per cent and above range.

The summary measures of inequality (Table 6.8) reinforce these findings. Ireland, Sweden, and the United Kingdom have the most unequal distributions of primary income, with the United States and Switzerland having Ginis slightly below those and above the rest. However, as shown by the Atkinson index, Germany belongs in this same group. At the other end of the distribution, the countries with the least unequal market income distribution are Norway and the Netherlands (excluding Belgium and Luxembourg).

There are two observations for a number of countries. As measured by changes in the Gini ratio, the trend is toward greater overall inequality in market incomes in all countries with the exception of Belgium. Canada, Sweden, the United Kingdom, Australia and the United States all experienced a 2 percentage points or more increase in their Gini ratio. Changes in the P-ratio points in Table 6. 7 were more muted but still substantial in some cases, particularly in the United Kingdom, the United States and Sweden.

The Lorenz curve analyses (Table 6.9) of market income can be compared to the percentile points and the summary measures of inequality. The Swiss have the largest income share accruing to the top decile, followed by the French (though tax policy may have some effect here) and then by the Irish. Belgium and Luxembourg (below the line) appear at the other end of the spectrum with a top decile share of market income only two-thirds as large as the Switzerland share. Most other countries are bunched between 25 and 28 per cent shares for the top decile. At the other end of the distribution, Belgium and Luxembourg (below the line) have a share of 8.4 per cent for the bottom quintile group; above the line, the Netherlands with 5.0 has the highest share. The lowest share for the bottom quintile is found in Sweden, followed by Germany and the United Kingdom. The analysis of Lorenz dominance is shown in Figures 6.3 and 6.4 (where Finland is not included). Here 29 cells show ambiguous crossings (? in Figure 6.3) and 49 curves allow an unambiguous comparison. As a result, Belgium and Luxembourg (in italics) clearly sit at the top of the Hasse diagram in Figure 6.4, with the Netherlands and Italy corning next, followed by Canada. Other countries are too close to rank definitively, apart from the United States and Australia being Lorenz superior to Ireland, and Australia being superior to the United Kingdom.

6.3 Relative income shares of primary and market incomes

The most traditional way to view the concentration of household income is to examine the cumulative shares at various points in the distribution as shown above. While the Gini ratios presented earlier capture these differences in a single parameter, and while Tables 6.6 and 6.9 present numerical comparisons, it is also instructive to visually examine cumulative shares at selected points in the distribution. Presented here are figures showing the cumulative primary income and market income shares for the bottom 50 and top 10 per cent of households ranked by each income concept.

91

Figure 6.3 Lorenz Comparisons of Market Income for All Households

BE CN FR GE IR IT LX NL SWE sz UK us

Australia, 1985 ? ? + ? ? + ?

Belgium, 1988 \ + + + + + ? + + + + + Canada, 1987 \ + + + ? + ? + + France, 1984 \ ? ? ? ? ? ?

Germany, 1984 \ ? ? ? ? ?

Ireland, 1987 \ ? ? ?

Italy, 1986 \ ? + + + + Luxembourg, 1985 \ + + + + + Netherlands, 1987 \ + + + + Sweden, 1987 \ ? ? ?

Switzerland, 1982 \ ? ?

United Kingdom, 1986 \ ?

The grid should be read as follows: .

+ indicates that for the country shown in the row we have Lorenz superiority over the country shown in the relevant column. indicates Lorenz inferiority.

? indicates that the Lorenz curves cross, with a difference of more than 1 percentage point in at least one direction.

Source: LIS Database.

Figure 6.5 illustrates households having prime age heads, a group for whom cumulative shares have not been constructed until now. A somewhat different pattern emerges. The earnings share of the bottom half of the distribution was highest in the mid-1980s in Norway, Germany, the Netherlands, and Switzerland and lowest in Ireland and the United States. The countries where the earnings share of the top decile was greatest, were Ireland and Switzerland. Canada, Germany, Norway, and the Netherlands had the lowest shares accruing to the top decile. Belgium, France, Italy, and Luxembourg are to the right of the break in the figure to reflect their post-tax nature and other anomalies. In fact, while the Italy and France figures are not much different from those in other countries, Luxembourg and Belgium are clearly outliers, most likely reflecting the post-tax nature of the data they have made available.

Much the same patterns appear when adding in household heads of other ages (Figure 6.6). Norway, Germany, and the Netherlands have the highest primary income shares among the bottom half of households, while Ireland, Sweden, and the United States have the lowest shares accruing to the bottom half of the distribution. For all of these groups, the shares in Figure 6.6 are lower than those in Figure 6.5, reflecting lower average earnings for non-prime age groups. Switzerland and Ireland have the largest primary income shares for those in the top decile, while Norway and Germany have the least.

Figure 6.7 changes the variable under consideration to market income. The highest shares for the bottom 50 per cent are still found in the Netherlands and Norway, while the lowest shares for the top 10 per cent are found in Norway and Canada. Sweden, and to a lesser extent, Ireland and the United Kingdom, have the lowest shares for the bottom decile. Switzerland and Ireland have the highest market income shares amongst the top half of households. The countries with post-tax data, in particular Luxembourg and Belgium, are clear outliers here, having by far the highest bottom share and the lowest top decile share. France is closer to the other extreme as far as the top market income decile share is concerned.

92

Figure 6.4 Relative Inequality Rankings Based on Market Income

in Different Countries

Belgium

Netherlands

I" I"'\

Canada

I I United States

Australia

I I I

Switzerland Ireland I

United Kingdom

Sweden

Luxembourg

Italy

France Germany

6.4 Changing shares of income: husbands, wives, self-employment and property income

The cumulative shares of income among the top decile and bottom half of the distribution are rather antiseptic due to a lack of clear explanation of the patterns observed in Figures 6.5-6.7. In order to lend some life to these estimates, the most recent data year was taken for each income concept and the sources of income were examined for the cumulative income shares of the bottom 50 per cent and for the top I 0 per cent. For primary incomes, each country's totals were apportioned into four pieces: wages and salaries accruing to the husband (or head regardless of sex), to the wife (in husband-wife families), to other earners (who were neither head nor spouse), and to self­employment income regardless of its source. For market incomes, the total was divided into three segments: earnings (regardless of source), self-employment income, and other private income (property income, occupational pensions and private transfers). Though trend data are not shown separately, they will be discussed.

This section begins by examining earnings shares by source for prime aged households in the mid-1980s (Figure 6.8). Husbands' (or heads') wages and salaries constitute the bulk of total household earned income among the bottom 50 per cent of families in every country. While the range varies from 69 per cent in the United Kingdom to 92 per cent in Switzerland, every country relies on heads' earnings for at least two-thirds of primary

93

\0 .j::>.

*

Figure 6.5 Aggregate Shares of Primary Income* in the Mid-1980s for Households with Heads Aged 25-54 35

30

25

~201··········· ··~~· ~- Bottom50%

ilBll Top10% (j;

(J_ 15 10

5

0 L!l co CJ) ~

.~ l'! Ci5 :J <(

C\1 co CJ)

1:J c ~

~ .§ (f)

1'- '<t co co CJ) CJ) ~ ~

<1l >-1:J c <1l <1l c E <1l (j; (J

<9

1'-co ~ 1:J c <1l

~

1'­co CJ)

(f) 1:J c <1l

~ .r:: Q5 z

CJ) 1'-1'- co CJ) CJ) ~ ~

>- c <1l Q)

1!: 1:J Q)

0 :;; z (f)

co co CJ) ~

E 0

1:J Ol c

S2 1:J 2 ·c: ::J

co co ~ (f) Q)

1ii Ci5 1:J 2 ·c: ::J

. . Q) Ol ~ Q)

~

'<t co co co CJ) CJ) ~ ~

Q) >-" ~ c ~

L1..

L!l co CJ)

e> :J 0 .c E Q) X :J

....J

co co CJ)

E :J ·o Qi co

Primary income includes all earnings, i.e. wages and salaries of all household members plus farm and non-farm self-employment income before taxes and transfers. Overall average for ten countries: Bottom 50%: 28.7; Top 10%: 22.9.

c Ql ()

Q; 0...

1.0 Ul

Figure 6.6 Aggregate Shares of Primary Income* for All Households

35

30

25

20

15

10

5

0 11) C\1 1'- '<!" 1'- 1'- (J) 1'- <0 <0 . '<!" <0 LO co <X) <X) <X) <X) <X) <X) 1'- <X) <X) <X) . <X) co co co (J) (J) (J) (J) (J) (J) (J) (J) (J) (J) Ql (J) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Ol ~ >-Cll "0 Cll >- "0 (/) >- c E (/) ~ Ql

~ !:'!' E

c "0 c c "0 Cll Ql Ql Ql () :::J :::J

1i! Cll Cll Cll Cll c 2: "0 0

~ > c 0 ·c,

~ c E e! Cll Ql "0 ~ ~ .0 Q;

ti Cll ~ 0 ;;: Ol E CD ::J .ti () Q; - z c lL Ql

~ C/) ::;:: "0 X

'§ (!} .c :::J Q) "0 :m --'

C/) z 2 c ·c: :::> :::>

Primary income includes all earnings, i.e. wages and salaries of all household members plus farm and non-farm self-employment income before taxes and transfers. Overall average for ten countries: Bottom 50%: 26.3; Top 10%: 24.2.

income in the lower half of the distribution. These earnings alone make up nearly 77 per cent of aggregate income. Self-employment income accounts for at most 14 per cent of earnings in Australia (13 per cent in the United Kingdom) and much less in most other countries, averaging 7.7 per cent of the total. Wives' earnings are an important source of income for those in the bottom half of the distribution in Sweden (17 per cent), Canada (16 per cent) and in the United States (16 per cent) and average 11.6 per cent of earnings. In Switzerland and the Netherlands, wives' earnings are only 4 and 7 per cent of the total. Earnings of other household members are 6 per cent or less in all countries, averaging only 3.5 per cent of the total. Though four countries whose income is of a post-tax nature are separated in Figure 6.8, the component shares of primary incomes differ very little from other countries, with the exception of self-employment income in Italy (20 per cent of the total).3

In contrast, the make-up of aggregate primary incomes of the top 10 per cent of households relies much less on heads' wages and salaries than on other types of earnings. Heads' wages and salaries are at most 66 per cent of the total incomes of the top decile in Sweden and 60 per cent in the United States and Switzerland. They are half or less in Germany, and average 56.3 per cent of the aggregate. Wives' wages and salaries are largest in Sweden (31 per cent), Canada (23 per cent), and Australia (21 per cent), and average a full18.5 per cent of the total. Self-employment income is important in Switzerland (35 per cent), and also in the Netherlands (25 per cent) and Norway (25 per cent), but very small in Sweden (3 per cent). The average is 17.3 per cent. In Italy, self­employment income is 39 per cent of the top decile share, 4 percentage points more than husbands' earnings. There appears to be a much greater difference in the sources of income constituting the top decile as compared to the bottom half of household incomes in each country. Wives' wages and salaries, other income, and self­employment income grow both in average level and in variance, while husbands' wages and salaries shrink by almost 21 per cent when comparing the income sources of the bottom half of earners to those at the other end of the distribution.

The trend in almost all countries (not shown) was for wives' wages and salaries to play a larger role in later versus earlier years at both the top and the bottom of the income distribution. This most often came at the expense of a declining earnings share for heads and other earners. Self-employment income changed little between years in most countries. In the future, wives' earnings are expected to continue to play a growing role in family income purchases and in overall primary income inequality, as growing female labour force participation and experience, coupled with work oriented family policy (e.g. subsidised child care), and the breakdown of segmented labour markets for women tend to raise the level and share of women's earnings in the household income package.

Adding in non-prime-age heads (Figure 6.9) makes little difference in the findings for the composition of the incomes of the bottom 50 per cent in the prime-age group. Self-employment income increases marginally in importance (from 7.7 per cent to 8.2 per cent of the total), but heads' wages and salaries remain the most important component of earned income, averaging 7 4.0 per cent. The most significant change is in others' wages and salaries which grew by 2.9 percentage points- still only 6.4 per cent of the total. Differences at the top of the distribution are even more difficult to find. Patterns are generally much the same as in Table 6.7, though others' earnings increase by 1.6 percentage points. The trends toward a greater contribution of wives' wages and salaries at both ends of the distribution remain strong for this group as well.

Shifting focus to market income in Figure 6.10 demonstrates that wages and salaries dominate the composition of market incomes among those in the bottom half of the income distribution, accounting for 79.7 per cent of the total, varying from 69.3 per cent in the United Kingdom to over 85 per cent in Germany and the Netherlands. Self-employment income is 14 per cent (Ireland) or less, averaging 7.7 per cent of the total. Other income types (private transfers, property income and occupational pensions are over 20 per cent of total market income in the United Kingdom and averages 10.3 per cent of the total in the bottom half of the distribution). In the top decile of the income distribution, wages and salaries fall in importance to 76 per cent of the total, as does other income (7.7 per cent) while self-employment increases to 16.5 per cent, more than double its average at the bottom. Wage and salary incomes vary from 60 per cent in Switzerland to almost 92 per cent in Sweden. On the other hand, self-employment income makes up 28 per cent of market income in Switzerland. Other income has almost a 15 per cent share in Australia. On average, it is clear that other income is more important in the bottom 50 per cent of families than in the top 10 per cent, while self-employment income is the reverse. Wage and salary income is more important in the top 10 per cent of the distribution when compared to the bottom 50 per cent only in Sweden and in the United Kingdom.

3. We have excluded Belgium from Figures 6-8, 6-9 and 6-10 because self-employment income is not separately identifiable in this dataset.

96

Figure 6.7 Aggregate Shares of Market Income* for All Households

35

30

25

~ 20, ••••••• ..~ ••• •••• I - Bottom 50% _. Top10%

Q; \0 c.. .....:) 15

10

5

0

"' C\l .... .... .... .... "' .... CD CD . .... co "' co

"' "' "' "' "' "' .... "' "' "' .

"' co "' co

"' "' "' "' "' "' "' "' "' "' Q) "' ~ "' ~

~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 0> ~ >- "" "' "0 "' >- "0 rJl >- c E rJl ~ Q) ~

E c "0 c c "0 "' Q) Q) Q) (.) ::> ::>

~ "' "' "' "' c i:: "0 0

~ > c 0 ·a,

"0 ..0

w 1ij c E ~ "' 0 Q) 0> <1: ~ E

a:;

"' 1ij ;:: !D ::l ~ () Q) z c ll.. Q)

<1: .s:: (/) S2 "0 X

·~ CJ ::> a; "0 .l!l -' (/) z .l!l ·c:

·c: ::::> ::::>

Market income includes primary income plus property income, occupational pensions and private transfers such as child support, before taxes and transfers. ** Overall average for ten countries: Bottom 50%: 22.5; Top 10%: 27.3.

6.5 Conclusion

This brief tour through the level and trend in inequality in primary and market incomes has been instructive, even if in some cases data were not totally comparable. Inequality varies to some extent by overall level of national economic well-being, but not exclusively. In some comparisons, Ireland exhibited the largest degree of primary and market income inequality, yet the next closest in inequality of primary incomes was the United States. Ability to measure trends over time is limited by the amount of available information, and the picture is rather mixed, but a number of countries such as the United States and the United Kingdom appear to have experienced growing inequality in primary and market incomes during the 1980s. For market income, the summary measures of inequality show a sizeable increase in Australia, Canada, Sweden, the United Kingdom and the United States. Women's earnings and self-employment income play different roles depending on the country and the part of the income distribution under analysis. The report now attempts to assess the effect of government tax and transfer policy on country incomes.

98

Figure 6.8 Shares of Primary Income in the Mid-1980s for Households with Heads Aged 25-54:

Average•

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Husbands, Wives, Other Earnings and Self-Employment lncome1

Luxembourg

1985

~~~~~~~~~~~~~=~~~~~i~~~~=i~~~~~~ Italy 1986

France 1984

0% 20% 40% 60% 80% 100%

Bottom 50%:

- SelfEmp.

lal Other

H~,,,,;;l Wife

- Husband

Average for the first nine countries: Husband: 77.2% Wives: 11.6% Other: 3.5% Self-Employed: 17.7%

Average•

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Luxembourg

1985

~~~~~~~~~~~~~i~~~~~i~~~~~~~~~~~~ Italy 1986

France 1984

0% 20% 40% 60% 80% 100%

Top 10%:

- SelfEmp.

lii!llllllll Other

li?;;t";"l Wife

- Husband

Average for the first nine countries: Husband: 56. 3% Wives: 18.5% Other: 7.9% Self-Employed: 17.3% 1. Households with only one earner count earnings for heads only as husbands.

99

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Luxembourg 1985

Italy 1986

France 1984

Figure 6.9 Shares of Primary Income in the Mid-1980s for All Households: Husbands, Wives, Other Earnings and Self-Employment Income1

-IIIII c::::::J -

0% 20% 40% 60% 80% 100%

Average for the first nine countries: Husband: 74% Wives: 11.5% Other: 6.4% Self-Employed: 8.2%

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Luxembourg 1985

Italy 1986

France 1984

~--

0% 20%

--I ' I -40% 60% 80% 100%

Average for the first nine countries: Husband: 55.4% Wives: 17.7% Other: 9.5% Self-Employed: 17.4% 1. Households with only one earner count earnings for heads only as husbands.

100

Bottom 50%:

Self Emp.

Other

Wife

Husband

Top 10%:

Self Emp.

Other

Wife

Husband

Figure 6.10 Shares of Market Income in the Mid-1980s for All Households:

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Ireland 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Wages, Self-Employment Income and Other Private Income

Italy

1986

liiiiiiiiiiiiiiiiiiiil~~~ii France 1984

Luxembourg 1985

0% 20% 40% 60% 80% 100%

Bottom 50%:

- Other Private Inc.

c:::=::J Self Emp.

-Wages

Average for the first ten countries: Wages: 79.7% Self-Employed: 7.7% Other Private Income: 10.3%.

United States 1986

Sweden1987

Norway 1979

Netherlands 1987

Ireland 1987

Germany 1984

Canada 1987

Switzerland 1982

Australia 1985

Italy

1986

~~~~~i~~~~=~~i~~~i~~~~~~~ France 1984

Luxembourg 1985

0% 20% 40% 60% 80% 100%

Top 10%:

- Other Private Inc.

c::=J Self Emp.

-Wages

Average for the first ten countries: Wages: 75.8% Self-Employed: 16.5% Other Private Income: 7.7%.

101

Chapter 7

TAXES AND TRANSFERS

7.1 The distribution of equivalent income levels

This chapter considers the distribution of taxes and transfers, and the relative importance of each as components of equivalent disposable income at high, middle and low income levels. The results are necessarily approximate since information of taxes and transfers in the LIS datasets are not exactly comparable from country to country. For example, in France unemployment payments are included in salaries; in some other countries all (Switzerland) or part of child allowance is included in salaries. In some countries occupational pensions are included with public pensions - in any case the dividing line between public and "private" pensions is often ambiguous. What follows is the best description possible with existing data.

In the OECD countries both tax and transfer schedules embody assumptions that relate family size and composition to considerations of need and equity. Therefore it is appropriate to consider the contribution of these income components not to unadjusted income but to income adjusted for family size. As in Chapter 4, a simple equivalence scale is used which adjusts disposable income by dividing it by the square root of family size (a similar scale is used in other recent OECD studies) Adjusted disposable income is referred to as equivalent income. Taxes and transfers have been adjusted for family size in the same manner. The following analysis examines the role of equivalised income sources as components of equivalent income.

The role of tax and transfers by quintiles of equivalent income is examined first, followed by their impact on different relative income levels measured in terms of ratios to median equivalent income in order to focus on issues of poverty and low income. Appendix 7 presents more detail on the distribution of income sources at different equivalent income levels.

Because transfer payments for retirement are such a large part of total transfers, the latter part of the chapter examines separately the role of transfers to households with heads under 60 years of age and those with heads 60 years and over. Tables A 7.3 and A 7.4 in Appendix 7 provide additional information concerning demographic characteristics of the total sample, and of those with low and modest equivalent incomes.

Table 7.1 presents the distribution of all persons, with those under and over 60 grouped in low and modest, middle and high income categories as defined in the last paragraph of Chapter 2. Persons at the low and modest income level have equivalent incomes below 70 per cent of median equivalent income; high income persons have incomes 150 per cent or more of the median. To a considerable extent countries with large high income groups also have large low and modest income groups. Only the Netherlands diverges sharply from this pattern- it has a small modest income group and a high income group of average size.

The 60-years-old or more are more likely to have modest incomes than younger persons in every country except the Netherlands and Ireland. The discrepancy is particularly great in Australia, Switzerland, Norway, Austria and the United Kingdom in 1979. The modest income rates of the two groups are close in the United States in 1986 -an increase from 1979 in low income among younger persons and a decrease among older persons brought the difference down to only 5.8 points.

Table 7.2 focuses on individuals with low incomes, defined in Chapter 2 as those with equivalent incomes less than half of the median. Low income rates are around 5 per cent in such countries as Austria, Finland, the Netherlands (1987), Sweden (1981) and Norway (1979). A middle range of 8 or 9 per cent appears in such countries as France, Germany and Switzerland. Canada and Australia have rates of 12 to 14 per cent, and the United States stands out for high rates of 16.6 in 1979 increasing to 18.4 in 1986.

Low income rates are higher among older persons in all countries except Ireland, the Netherlands, and Sweden. Canada, France and the United Kingdom reverse this usual pattern from the earlier period to the later period.

103

Table 7.1 Percentage Distribution of Low and Modest, Middle and High Income Groups by Age

Total Persons under 60 Persons 60 and over

L&M M H L&M M H L&M M H

Australia 1981 26.2 52.2 21.7 21.9 55.2 22.9 48.0 36.9 15.1 1985 27.5 50.9 21.5 22.1 54.6 23.3 51.5 34.8 13.7

Belgium 1985 17.8 65.4 16.9 17.5 65.7 16.8 20.3 62.5 17.2 1988 19.8 66.2 14.1 16.3 69.2 14.6 32.1 55.5 12.4

Switzerland 1982 21.7 59.9 18.4 16.8 64.0 19.3 39.7 45.1 15.2 Canada 1981 25.5 54.6 19.9 22.8 56.6 20.6 39.1 44.4 16.5

1987 26.2 54.3 19.5 24.0 56.1 20.0 36.4 46.3 17.3 France 1979 22.7 57.6 19.7 20.2 59.6 20.3 32.8 49.8 17.4

1984 22.8 56.1 21.0 21.3 56.9 21.7 28.8 53.0 18.2 Germany 1984 21.5 62.1 16.5 19.4 63.3 17.3 28.1 58.0 13.8 Italy 1986 27.2 49.4 23.4 25.0 50.0 25.0 34.3 47.4 18.3 Ireland 1987 27.4 47.2 25.3 27.4 47.6 24.9 27.4 45.8 26.8 Luxembourg 1985 21.0 61.6 17.3 19.3 62.4 18.3 28.3 58.4 13.3 Netherlands 1983 16.8 61.8 21.4 17.3 60.7 22.0 14.7 66.5 18.8

1987 16.4 63.2 20.4 17.6 61.6 20.8 11.2 70.3 18.6 Norway 1979 19.0 68.3 12.7 13.9 73.0 13.0 36.2 52.3 11.5

1986 20.1 65.8 14.1 14.4 70.6 15.1 38.2 50.8 11.0 Sweden 1981 16.1 73.5 10.4 13.5 74.6 11.9 23.5 70.6 6.0

1987 20.0 69.5 10.5 15.7 72.2 12.1 32.7 61.7 5.6 UK 1979 24.8 55.8 19.4 17.7 60.3 22.0 51.1 39.0 9.9

1986 27.2 50.0 22.8 24.3 49.8 26.0 37.1 50.9 12.1 us 1979 28.6 50.1 21.4 26.2 51.9 21.9 38.9 42.0 19.1

1986 30.3 44.8 24.9 29.1 45.9 25.0 34.9 40.5 24.6

Finland 1987 18.7 70.3 11.0 14.0 74.0 12.0 42.2 52.1 5.7 Austria 1987 20.4 65.6 14.0 14.0 70.2 15.8 37.0 53.4 9.6

L&M: Low and modest income M: Middle income H: High income

Table 7.2 Percent Low Income by Age Group

Total Under 60 60 and over

Australia 1981 12.5 11.5 17.8 1985 12.3 10.8 18.7

Belgium 1985 2.9 2.4 6.5 1988 4.7 3.5 8.9

Switzerland 1982 8.0 5.9 15.7 Canada 1981 12.6 11.4 18.6

1987 12.2 12.3 ll.5 France 1979 8.2 7.6 10.7

1984 7.5 7.6 7.2 Germany 1984 6.5 5.8 8.5 Ireland 1987 10.7 11.2 9.0 Italy 1986 10.5 9.0 15.5 Luxembourg 1985 5.4 3.9 11.7 Netherlands 1983 6.6 7.0 4.8

1987 4.9 5.5 2.5 Norway 1979 5.0 4.9 5.4

1986 7.3 4.7 15.6 Sweden 1981 5.4 6.2 3.1

1987 7.6 8.0 6.5 UK 1979 9.2 7.0 17.3

1986 9.1 9.8 6.7 us 1979 16.6 15.2 22.7

1986 18.4 17.8 20.7

Finland 1987 5.0 4.0 9.8 Austria 1987 6.7 3.9 14.2

104

Slight increases appear in the size of the low income group among those under sixty in most countries for which there are two surveys- except France and the Netherlands. Among those sixty and over there are decreases in Canada, France, the Netherlands, the United Kingdom and the United States; Australia, Belgium and Sweden show increases.

7.2 The distribution of taxes on income and social contributions

Taxes as measured in the LIS dataset include income tax and employee social insurance contributions. Since income tax data for Italy and Luxembourg are unavailable, no results are presented for these countries. Results are not presented for France because data on employee social security payments (which are large) are unavailable.

Overall taxes amounted to around one-fifth of aggregate gross income in low tax countries. High tax countries paid thirty per cent or more of gross income in direct taxes.

Table 7.3 shows the percentage distribution of taxes by quintiles of equivalent income. For the countries with complete tax data, the share of total taxes paid by the highest quintile ranged from 36 per cent in Sweden (1981) to 53 per cent in the United States (1986). The bottom half of the equivalent income distribution pays from around one-sixth up to around 30 per cent of total taxes. It is difficult to make precise comparisons of the distribution of taxes because of differences in the way countries record taxes in the surveys used. In some countries taxes are imputed; in others the amount recorded is based on measured taxes (e.g. Sweden)1• This poses a particular problem when one focuses on the lowest disposable income groups which can contain a few individuals who paid very high taxes relative to their gross income in a particular survey year. Were we to look at taxes paid and transfers received by tax units and by gross income brackets, the tax and transfer rates would look quite different. The countries where the anomaly of high taxes and low disposable income occurs are readily apparent in Appendix Table A7.1 and A7.2- for example, the Netherlands, Sweden, and the United States.

Table 7.3 Distribution of Taxes by Quintile and Average Tax as a Percent of Median Equivalent Income

Average tax as a Bottom 2 3 4 Top Total per cent of mediane

equivalent income

Australia 1981 1.1 8.1 16.2 24.8 49.8 100.0 29.9 1985 0.7 7.6 16.3 24.2 51.2 100.0 32.2

Switzerland 1982 5.8 10.2 14.7 21.1 48.2 100.0 26.0 Canada 1981 1.8 9.3 16.7 25.8 46.5 100.0 19.0

1987 3.6 8.8 16.2 24.8 46.5 100.0 24.8 Germany 1984 5.5 10.4 17.0 23.4 43.7 100.0 36.0 Ireland 1987 7.0 12.2 17.6 23.8 39.3 100.0 29.5 Netherlands 1983 5.5 11.8 17.0 22.9 42.7 100.0 57.0

1987 10.3 10.0 16.2 22.3 41.2 100.0 67.2 Norway 1979 3.5 11.4 18.2 25.8 41.1 100.0 35.0

1986 3.7 13.2 19.2 25.7 38.1 100.0 32.5 Sweden 1981 10.1 13.1 17.7 23.3 35.8 100.0 42.7

1987 6.3 12.5 17.7 23.3 40.1 100.0 45.0 UK 1979 4.0 11.5 18.0 25.1 41.4 100.0 25.8

1986 4.5 8.1 15.9 25.0 46.4 100.0 31.0 us 1979 2.5 7.6 14.6 24.7 50.6 100.0 28.1

1986 3.8 6.9 13.9 22.6 52.7 100.0 30.6

Finland 1987 4.9 11.2 17.1 23.9 42.9 100.0 36.9

1. Income tax routines vary substantially by countries. Some countries (e.g. Sweden, Finland) record taxes actually paid in a given year. Some surveys (e.g. the United Kingdom) ask taxes paid in year t, when the taxes may in some cases be paid on the previous year's income (year t-1). And some countries (e.g. the United States and Australia) use microsimulation techniques to estimate tax liabilities based on reported income amounts. Taxes may further vary by disposable versus gross income bracket because families may have very high income and payroll deductions on one portion of income (e.g. wages and salaries) while also having large negative incomes due to losses in other forms of income (e.g. self-employment). Many households include two or more tax units as well and therefore reflect a mix of high income (and tax) and negative income.

105

Comparing the average level of direct taxes paid in different countries requires calculating equivalised taxes as a per cent of median equivalent income (taxes are equivalised by dividing each household's tax by the square root of its family size). The last column of Table 7.3 shows that low direct tax countries have a range from an average tax of around 20 to around 30 per cent of median disposable income. Two countries, the Netherlands and Sweden, have much higher average taxes.

The ratio of the share of taxes paid by the top two quintiles to those paid by the bottom three provides a rough index of the combined effect of the distribution of pre-tax income and the progressivity of the tax schedule. There is a slight tendency for countries with higher average taxes to have lower ratios - the Netherlands and Sweden in particular. The United States, Australia (1981), Canada and the United Kingdom (1986) all have higher ratios and lower average tax rates. Another perspective on distribution and progressivity is given in Table 7.4 which reports average equivalised tax in each quintile as a percent of median equivalent income.

Table 7.4 Average Taxes as a Percent of Median Equivalent Income by Quintile

Bottom 2 3 4 Top Overall average

Australia 1981 1.7 12.1 24.2 37.1 74.5 29.9 1985 1.1 12.2 26.2 38.9 82.5 32.2

Switzerland 1982 7.5 13.2 19.1 27.5 62.7 26.0 Canada 1981 1.7 8.8 15.8 24.4 44.0 19.0

1987 4.5 11.0 20.1 30.9 57.8 24.8 Germany 1984 9.9 18.8 30.5 42.1 78.8 36.0 Ireland 1987 2.7 8.5 20.4 37.4 78.7 29.5 Netherlands 1983 15.8 33.7 48.5 65.2 121.8 57.0

1987 34.6 33.7 54.4 75.0 138.6 67.2 Norway 1979 6.1 20.0 31.8 45.0 71.8 35.0

1986 6.1 21.5 31.1 41.9 62.0 32.5 Sweden 1981 21.4 28.0 37.8 49.7 76.4 42.7

1987 14.3 28.1 39.9 52.5 90.1 45.0 U.K. 1979 5.2 14.9 23.2 32.3 53.4 25.8

1986 7.0 12.5 24.7 38.9 72.1 31.0 u.s. 1979 3.5 10.7 20.5 34.8 71.2 28.1

1986 5.9 10.5 21.3 34.5 80.6 30.6

Finland 1987 9.0 20.7 31.5 44.2 79.2 36.9

7.3 The distribution of transfers

Social transfers in aggregate range from around six per cent of gross income (i.e. total income before tax) to one-quarter. In Australia, Switzerland, Canada, the United States and Norway in 1979, transfers amount to 10 per cent of gross income or less. Belgium, France, Italy, Luxembourg and Sweden have the highest ratio of social transfers, followed closely by the Netherlands and the United Kingdom in 1986. But since total income is net of some taxes in Belgium, France, Italy and Luxembourg it is difficult to make precise comparisons between these countries and the others. Tables A 7.1 and A 7.2 in Appendix 7 provide more information on transfers and income level.

The distribution of transfers over quintiles suggests two patterns. In some countries the proportion of transfers received is highest for the bottom quintile and declines when moving to higher quintiles (see Table 7.5). In this sense, transfers may be considered "targeted". It is essential to bear in mind that persons are classified here according to income including transfers (and minus taxes). Countries in this group, in decreasing order of targeting on lower income groups, are Australia, Switzerland, Canada, Norway, Ireland, the United Kingdom and the United States. Another group of countries shows little variation in share, though the higher rather than lower quintiles have the greater part. These countries, from the flattest pattern to the more elevated, are the Netherlands, Finland, Sweden, France, Luxembourg and Italy.

Except in Luxembourg and Italy, more than half of transfers go to persons with incomes below the median, but there is great variation among the other countries in the proportion of transfers to this group - ranging from almost three-quarters in Australia to slightly more than one-half in Sweden and the Netherlands.

The degree to which transfers are concentrated in the lowest quintile varies greatly, from around 40 per cent of total transfers in Australia and Switzerland, to less than 20 per cent in France, Italy, Luxembourg, and Sweden.

106

Table 7.5 Distribution of Transfers by Quintile and Average Transfers as a Percent of Median Equivalent Income

Average transfers as a Bottom 2 3 4 Top Total per cent of median

equivalent income

Australia 1981 42.8 22.2 13.3 12.5 9.2 100.0 10.8 1985 40.1 24.6 14.4 12.9 8.0 100.0 11.3

Belgium 1985 22.9 22.5 21.9 16.6 16.1 100.0 33.3 1988 21.5 23.6 20.1 16.1 18.7 100.0 34.9

Switzerland 1982 38.5 19.2 15.6 13.3 13.3 100.0 7.3 Canada 1981 33.0 22.9 17.9 14.1 12.1 100.0 10.1

1987 29.5 24.2 19.2 15.0 12.1 100.0 12.4 France 1979 19.7 21.2 18.8 17.7 22.6 100.0 22.2

1984 17.5 21.8 18.4 17.7 24.7 100.0 25.0 Germany 1984 21.8 22.2 16.7 21.0 18.3 100.0 19.8 Ireland 1987 32.0 21.9 21.3 15.2 9.6 100.0 20.5 Italy 1986 15.6 16.4 19.7 20.7 27.6 100.0 21.4 Luxembourg 1985 17.3 18.3 19.5 22.5 22.4 100.0 23.7 Netherlands 1983 21.8 21.8 18.4 20.4 17.6 100.0 28.5

1987 24.9 21.3 16.9 17.7 19.2 100.0 28.3 Norway 1979 34.0 20.9 16.4 13.6 15.1 100.0 13.5

1986 21.5 16.6 14.2 12.2 11.0 100.0 15.1 Sweden 1981 18.0 23.9 19.8 19.5 18.7 100.0 35.0

1987 15.2 25.8 21.7 19.9 17.4 100.0 35.5 UK 1979 30.6 20.0 17.4 17.0 15.0 100.0 18.5

1986 26.7 25.9 19.4 16.1 11.9 100.0 24.3 us 1979 29.7 21.1 17.4 14.7 17.1 100.0 8.9

1986 29.2 21.2 17.1 17.5 15.1 100.0 9.4

Finland 1987 25.9 22.6 18.2 15.8 17.6 100.0 27.7

Table 7.6 Average Transfers as a Percent of Median Equivalent Income by Quintile

Bottom 2 3 4 Top Overal average

Australia 1981 23.1 12.0 7.2 6.7 5.0 10.8 1985 22.5 13.8 8.1 7.3 4.5 11.3

Belgium 1985 33.3 32.2 30.5 28.6 22.3 21.3 1988 34.9 32.3 33.6 27.8 22.3 27.0

Switzerland 1982 14.1 7.0 5.7 4.9 4.9 7.3 Canada 1981 16.6 11.5 9.0 7.1 6.1 10.1

1987 18.3 15.0 11.9 9.2 7.5 12.4 France 1979 21.9 23.6 21.0 19.7 25.1 22.2

1984 21.9 27.2 22.9 22.1 30.8 25.0 Germany 1984 21.6 21.9 16.5 20.8 18.2 19.8 Ireland 1987 32.1 23.8 20.8 15.4 10.1 20.5 Italy 1986 16.6 17.6 21.1 22.1 29.5 21.4 Luxembourg 1985 20.5 21.7 23.1 26.6 26.5 23.7 Netherlands 1983 31.0 31.0 26.3 29.1 25.1 28.5

1987 35.2 30.2 23.9 25.1 27.2 28.3 Norway 1979 23.0 14.1 11.1 9.2 10.2 13.5

1986 21.5 16.6 14.2 12.2 11.0 15.1 Sweden 1981 31.6 41.8 34.7 34.2 32.8 35.0

1987 27.0 45.7 38.5 35.4 30.9 35.5 UK 1979 28.3 18.5 16.1 15.8 13.8 18.5

1986 32.4 31.4 23.5 19.6 14.5 24.3 us 1979 13.2 9.4 7.8 6.5 7.6 8.9

1986 13.7 9.9 8.0 8.2 7.1 9.4

Finland 1987 35.9 31.3 25.2 21.9 24.4 27.7

107

20

15

Q)

E 0 (.)

. !: 3: .Q 10 E Q)

e Q)

0...

SZ82

• 5

0 0

Figure 7.1 Low-income Rate of the Non-elderly by Transfers as a Percentage of Median Equivalent Income

US86

•

US79

•

CN87

• CN81 • • AS81

• AS85

IT86

• FR84

FR79 • • •

UK79 GE84 SW81 •

NW86 • •

NW79 LX85 •

BE85 •

I I I 5 10 15 20

Mean transfers below median group

108

IR87

• UK86

• SW87

• • NL83

• • NL87

• BEBB

I 25 30

In the first three of these latter countries, the top quintile actually receives more than its proportionate share of transfers. These are countries in which earnings-related pensions play an important role in the national social protection regime (compare the figures for those over and under 60 in Table 7.7).

Table 7.7 Average Transfers as a Percent of Median Equivalent Income Persons with below median income

Total under 60 60 and over

Australia 1981 15.3 11.3 30.6 1985 16.2 12.9 30.2

Belgium 1985 31.2 23.9 60.9 1988 31.9 24.6 57.5

Switzerland 1982 9.4 1.9 29.9 Canada 1981 13.2 10.2 24.9

1987 15.9 12.1 30.3 France 1979 22.8 18.0 38.3

1984 25.3 20.3 41.9 Germany 1984 20.9 12.6 43.3 Ireland 1987 31.3 26.4 49.1 Italy 1986 17.5 9.0 41.1 Luxembourg 1985 22.1 14.3 50.6 Netherlands 1983 30.4 25.4 51.2

1987 30.9 25.8 51.2 Norway 1979 17.6 11.0 34.1

1986 18.3 10.1 43.6 Sweden 1981 37.2 25.0 59.4

1987 39.0 24.9 65.4 UK 1979 22.4 18.3 31.8

1986 31.5 28.5 38.4 us 1979 10.5 6.7 24.2

1986 10.8 6.9 26.0

Finland 1987 32.2 21.4 65.3

The combined effect of aggregate levels of transfers and their concentration results in average social transfers to the bottom quintile, amounting to around 30 per cent of median equivalent income in Belgium, Ireland, the Netherlands, Sweden, and the United Kingdom (see Table 7.6). Transfers to the second quintile in these countries are also high. Switzerland and the United States are at the other extreme with average transfers to the bottom quintile, amounting to 13 and 14 per cent of the median. The other countries are in a range from 17 to 23 per cent.

7.4 The role of transfers

This section explores the role of transfers as a factor in variations among countries in the proportion of persons with low incomes- defined as less than half of median equivalent income. Table 7.2 shows a range from 4.9 per cent (the Netherlands) in the low income category to 18.4 per cent (United States, 1986). The range for the low and modest income groups combined (those with less than 70 per cent of median equivalent income) is from 16.1 per cent (Sweden, 1987) to 30.3 per cent (United States, 1986). To what extent are differences in these rates related to the size of transfers?

This issue is addressed here by examining the transfers received by the population with below median equivalent incomes (see Table 7.7). It is useful to examine separately working and retired persons. Transfers are the principal source of income for the latter (40.3 per cent of median equivalent income) and much smaller (16.8 per cent) for those of working age.

For those of working age there is no clear relation between rates of low or modest income and the level of transfers to families with working age heads. Figure 7.1 charts the relation of transfers to low income rates. Three countries with high low income rates- Australia, Canada and the United States- transfer 15 per cent or less of

109

25

US79

r- • USS6

• 20

CNS1

• r-

Q)

E 15 0 ()

.5 ;;: .2 r-c Q)

~ Q)

0. 10

'-

5

-

0 20

Figure 7.2 Low-income Rate of the Elderly by Transfers as a Percentage of Median Equivalent Income

ASS5

• ASS1 • UK79 •

SZS2 ITS6 NWS6 • • •

CNS7 LXS5

• • FR79

• IRS7 BESS

GES4 • • • FRS4

UKS6 • BES5 • • NW79

• NLS3 •

SWS1 NLS7 • •

30 40 50 60

Mean transfers below median group

110

SWS7

•

70

median equivalent income to persons with below median incomes; the average for the other countries is much higher (see Table 7.7). But, among the other countries there is a wide range in transfer ratio that bears very little relation to low income rates. For example, both Ireland and the Netherlands have transfer ratios around 30 per cent, yet Ireland's low income rate is only slightly lower than Australia's and Canada's while the Netherlands low income rate is quite low. The bivariate correlation between mean transfers to the non-elderly with below median incomes and the low income rate is 0.37. Without the outliers (Switzerland, Ireland and the United Kingdom in 1986) it rises to 0.65. For those of working age, most of the explanation of differences in low income rates must come from differences in the inequality of work income, or perhaps more exactly from the different ways of combining transfers and work income in different countries.

Four countries have very low working age low income rates and transfer ratios - Germany, Luxembourg, Norway, and Switzerland. These rates are necessarily the result of low work income inequality. Belgium, Finland, France, the Netherlands and Sweden also have low rates, but much higher transfer ratios. The combined effect of earnings and transfers produces those low rates.

Since transfers are the principal source of income for older persons it is to be expected that the level of transfers to the below median group is highly related to the low income rate - see Figure 7.2. The bivariate correlation between mean transfers to the elderly with below median incomes and their low income rate is 0.76. Countries with low average transfers to older persons have high low income rates- the United States, Australia, Canada, Switzerland, the United Kingdom in 1979. The other extreme consists of countries with very low rates and high levels of transfers- Sweden, the Netherlands, Germany, Finland, Belgium, France in 1984, and Ireland. Norway is somewhat of an outlier, perhaps because it had a much higher proportion of persons over 60 in the labour force and still remunerated. Italy and Luxembourg have more elderly low income than expected given the high levels of transfers to this group.

111

ChapterS

FUTURE RESEARCH PRIORITIES AND NEXT STEPS

This study has examined in detail various facets of the distribution of income and its components using the microdata available in the Luxembourg Income Study (LIS) and, where not available, special tabulations prepared from national survey data. That not all OECD countries are covered reflects either a lack of adequate data or the inability of national authorities to create the tabulations required.

Chapter 3 reviewed issues of comparability in LIS datasets and considered in particular the definitions of income used in different surveys, and of the component sources of total income. Where studies comparing aggregate income statistics with survey results for the national surveys used in the LIS are available, the report discussed the extent to which surveys capture total aggregate income in the country. The reader's understanding of the strengths and weakness of the national datasets is important for a proper assessment of the significance of the results reported in the substantive chapters.

The term "inequality" has been used as is customary in the field, but the existence of income differences does not necessarily have policy implications. The importance of the results for policy depend on social judgements, including the trade-off between the level of income and its distribution.

Chapter 4 considers various facets of the distribution of disposable income across LIS countries. First it consider the distribution of equivalised income using an intermediate equivalence scale which divides disposable income by the square root of family size. The distribution is characterised in terms of the relative situation of income deciles, and in terms of income classes defined as various proportions of median equivalised income. Two summary measures of income inequality are presented: the Gini and the Atkinson coefficients. Distributions of Lorenz curves are compared to highlight issues of dominance and the more complex patterns that arise which Lorenz curves cross.

After comparing equivalised distributions for the most recent time periods available, Chapter 4 considers evidence on changes in the various distribution measures during the 1980s for the smaller number of countries for which two datasets are available. No overall pattern appears; countries seem diverse in the extent to which there has been movement towards inequality.

The chapter finally considers the sensitivity of the results reported to the definition of equivalent income used, and to whether the household or person is the unit of observation (or weight). Two extreme equivalence scales are compared to the middle scale used in the first part of the chapter - no adjustment (that is, unadjusted disposable income) and per capita disposable income. It also compares household distributions to person-weighted distributions, noting that these variations in defining income distribution sometimes have considerable effect on the levels of inequality and comparison of distributions in different countries. Inequality is generally higher for both no adjustment and the per capita adjustment than the LIS middle equivalence adjustment. There is broad agreement among the three measures in the ranking of countries, but important changes in ranking still remain for some countries as the equivalence factor changes.

Chapter 5 compares these findings to those available in national studies of income distribution, as well as extending the range of countries to include Austria, Japan, Portugal and Spain. The findings in general appear consistent with those in Chapter 4, when taking account of differences in definition, although certain aspects need further attention. The national studies also provide fuller information about the trends over time.

The role of primary and market incomes is described in Chapter 6. Here, and in Chapter 7, care in interpreting results is important because of various incompatibilities in LIS data for different countries and periods. These range from minor to major in their impact on findings.

Three distributions are analysed: primary income of households with heads under 55 years of age, primary income of all households, and market income of all households. Their incomes of deciles are compared, and overall indices of inequality are presented. Most countries seem to have experienced a trend toward increasing inequality of primary and market incomes in the 1980s.

113

Chapter 6 also describes the role of the wage and salary income of different family members, and of self­employment in total primary income for the lower half of the distribution, and for the upper decile. The primary income sources of the latter are generally much more heterogeneous, with self-employment income and wives' earnings playing a more important role.

Chapter 7 considers the role of taxes on income and employee social insurance contributions. The pattern of taxes and transfers in different income quintiles is compared. The distribution of taxes and transfers by quintile is examined first, then the relative size of taxes and transfers to each quintile as a per cent of the overall median equivalent disposable income in each country/year. The results show marked divergences across countries not only in the level of transfers but in the degree to which transfers are targeted on the lower quintiles.

Finally, this chapter considers the impact of transfers on economic well-being separately for the working age population (head under 60 years of age) and persons living in households headed by an older person. To sharpen the focus on the impact on low income groups, only those with below median income are considered. As expected, transfers are the major source of income for older persons in many, but not all countries. They play less of a role for households with younger heads, but here too the range is wide with transfers making up a quarter of income in some countries and well under ten per cent in others.

The pattern of transfers and the low income share of the population is not particularly sharp for those with heads under 60. Here the role of earnings seems dominant. But there is a strong relation between average transfers to below median older persons and the low income rate.

Over the past decade substantial progress has been made in improving household income survey information and in making it more standardized and hence, more comparable across countries. Because of the willingness of countries to share their data and help standardize concepts and measures, vast improvements in the quantity and quality of cross-national data analysis are in progress. We are pleased that the Luxembourg Income Study (LIS) project has been able to contribute to this effort.

However, there is still much room for improvement, both in data quality and quantity. Concepts and measures need to become even more standardized across countries. The quality of income data needs more careful monitoring and reporting. Countries reluctant to join cross-national research projects such as LIS must be persuaded that they have more to gain than lose from such an exercise. Progress along these lines will greatly improve this, and like reports, in another decade.

The LIS is currently adding a third wave of data for the 1991-93 period. 1 This will certainly help identify trends that could only be hinted at with currently available information and will allow the examination of changes in the economic climate and in public policy since the mid-1980s which are not considered in this report (other than in Chapter 5). Moreover, new projects in the United States and Luxembourg are beginning to make longitudinal household "panel" income data sets comparable across the handful of countries with such surveys. New income data collection efforts are underway in many European Union countries. Thus, the quantity, quality, and flexibility of income comparisons across countries will see improvement in coming years.

Another important research thrust need not await new data. The major purpose of this book has been to lay out carefully and systematically the facts of the cross-national level and trend in income inequality as good as possible given the available data. The report does not however go very far in explaining the root economic, social policy and demographic factors which underlie these trends. Several analyses of specific sub-topics await further research, including at least the following:

- The long-term impact of increased market work by women. Clearly both labour force participation and long-term commitment to market work among women has been growing in all OECD countries, albeit at different speeds and trajectories. A set of gender-based studies which account for the impact of these changes on family income inequality is necessary; The distributive effect of population ageing on labour force participation, social budgets, trends in work, occupational pensions, and related issues should be addressed. This would complement the aggregate data and some sub-aggregate data which are already available in this area;

- The behaviour of earned income inequality both secularly and cyclically is of great interest. How different are returns to additional years of education across countries and how will this affect unemployment, worker dislocation, job loss and related issues?

- Changes in the structure of personal income tax are also important. Shortly after the most recent (1987) data year included in this study, several European countries severely reduced their top marginal income tax brackets. This no doubt had an effect on both the level of income and overall income inequality;

- Changes in the relative effectiveness of social policy, family income packaging and anti-poverty effectiveness are all in need of re-examination; and

1. Data from the 1970s are available in five nations: Canada, Sweden, the United Kingdom, the United States and Germany.

114

- Differences in the proportion of married couple and single parent households, and the trend in these variables.

A first step toward exploring these issues might come from a systematic application of standardization approaches to LIS microdata, so that country differences and trends over time would directly reflect changes in demographic structure, in patterns of labour force participation by gender, family role and age, and in levels of unemployment in these OECD countries. Because LIS data are freely available to social scientists throughout the world, there is ample opportunity for interested researchers to go beyond the findings in this report to explore further important policy and scientific issues.

115

REFERENCES

Assemat, J. (1989), "La qualite de la mesure des revenus au milieu des annees quatre-vingt", Melanges en l'honneur de Jacques Desabie, INSEE, Paris.

Assemat, J. and Glaude, M. (1989), "Source fiscale et/ou enquete par interview: !'experience fran9aise en matiere de mesure des bas revenus", paper presented at Eurostat Conference, Noordwijk.

Atkinson, A. B. (1970), "On the measurement of inequality", Journal of Economic Theory, Vol. 2, pp. 244-263.

Atkinson, A. B. (1983), The Economics of Inequality, 2nd Edition, Oxford University Press, Oxford.

Atkinson, A. B. (1993), "What is happening to the distribution of income in the UK?", Proceedings of the British Academy, Vol. 82, pp. 317-351.

Atkinson, A. B. and Micklewright, J. (1983), "On the reliability of income data in the family expenditure survey 1970-1977", Journal of the Royal Statistical Society, Series A, Vol. 146, pp. 33-61.

Atkinson, A. B. and Micklewright, J. (1992), Economic Transformation in Eastern Europe and the Distribution of Income, Cambridge University Press, Cambridge.

Atkinson, A. B. and Rein, M. (1993) (eds), Age, Work and Social Security, Macmillan, London.

Bauer, J. and Mason, A. (1992), "The distribution of income and wealth in Japan", Review of Income and Wealth, Vol. 38, pp. 403-428.

Begue, J. (1976), "Remarques sur une etude de l'OCDE concernant la repartition des revenus dans divers pays", Economie et Statistique, No. 84, pp. 97-104.

Berntsen, R. (1989), "Einkommensanalysen mit den Daten des Sozio- okonomischen Panels unter Verwendung von generierten Einkommensdaten", SFB 3, Discussion Paper 291, J-W Goethe University, Frankfurt.

Blau, F. and Kahn, L. (1992), "'The gender-earnings gap' some international evidence", National Bureau of Economic Research Working Paper #4224, National Bureau of Economic Research, Cambridge, MA, March.

Blundell, R. W. and Walker, I. (1988), "The changing structure of the labor force: married women and lone parents", paper presented to the Symposium on Population Change and European Society, Florence, Italy, December 8.

Brandolini, A. (1993), "A description and an assessment of the sample surveys of the personal distribution of incomes in Italy", Microsimulation Unit Discussion Paper 9303, University of Cambridge.

Brandolini, A. and Cannari, L. (1992), "Changes in income distribution in Italy, 1977-89", mimeo, Bank of Italy.

Brandolini, A. and Cannari, L. (1993), "Methodological appendix: the Bank ofltaly's survey of household income and wealth", in A. Ando, L. Guiso and I. Visco (eds), Savings and the Accumulation of Wealth, Cambridge University Press, Cambridge.

Brandolini, A. and Sestito, P. (1994), "Mutamenti nella distribuzione del reddito in Italia: 1977-1991", inN. Rossi (ed.), La transizione: 1992-1993, secondo rapporto CNEL sulla distribuzione e redistribuzione del reddito in Italia, mimeo.

Buhmann, B. I. (1988), Wohlstand und Armut in der Schweiz, Riiegger, Griisch.

Buhmann, B., Rainwater, L., Schmaus, G. and Smeeding, T. (1988), "Equivalence scales, well-being, inequality and poverty: sensitivity estimates across ten countries using the Luxembourg Income Study (LIS) database", Review of Income and Wealth, Vol. 34, pp. 115-142.

Callan, T. and Nolan, B. (1993), "Income inequality and poverty in Ireland in the 1970s and 1980s", ESRI Working Paper 43, ESRI, Dublin.

Callan, T., Nolan, B., Wehlan, B.J., Hannan, D.F., and Creighton, S. (1989), Poverty, Income and Welfare in Ireland, ESRI, Dublin.

117

Canceill, G. and Villeneuve, A. (1990), "Les inegalites de revenus: quasi statu quo entre 1979 et 1984 pour les salaries et les inactifs", Economie et Statistique, No. 230, pp. 65-74.

Cantillon, B., Marx, 1., Proost, D. and Van Dam, R. (1994), Indicateurs Sociaux: I985-I992, Centre for Social Policy, University of Antwerp.

Carbonaro, G. (1985), "Nota sulle scale di equivalenza", La poverta in Italia, lstituto Poligrafico dello Stato, Rome.

Central Statistical Office (1992), Family Spending: A Report on the I99I Family Expenditure Survey, HMSO, London.

Committee on Ways and Means, U.S. House of Representatives (1992), I992 Green Book, U.S. Government Printing Office, Washington, D.C.

Coulter, F.A.E., Cowell, F.A. and Jenkins, S.P. (1992), "Equivalence scale relativities and the extent of inequality and poverty", Economic journal, Vol. 102, pp. 1067-1082.

Cowell, FA. (1977), Measuring Inequality, Philip Allan, Oxford.

Cowell, F. A. (1992), INEQ, micro-computer programme, STICERD, London School of Economics.

Danziger, S., and Gottschalk, P. (eds) (1993), Uneven Tides, Rising Inequality in America, Russell Sage Foundation, New York.

Danziger, S., Vander Gaag, J., Taussig, M. K. and Smolensky, E. (1984), "The direct measurement of welfare levels", Review of Economics and Statistics, Vol. 66, pp. 500-505.

Davis, S. (1992), "Cross country patterns of change in relative wages", National Bureau of Economic Research Working Paper #4085, National Bureau of Economic Research, Cambridge, MA, June.

Department of Social Security (1992), Households Below Average Income: A Statistical Analysis I979-1988/89, HMSO, London.

Department of Social Security (1993), Households Below Average Income: A Statistical Analysis I979-I990/91, HMSO, London.

DiBiase, R., DiMarco, M. and DiNicola, F. (1993), "La predisposizione delle basi di dati e l'aggiomamento dei modelli di microsimulazione dell'ISPE", ISPE, Rome.

Dobroschke-Kohn, W. (1987), unpublished data from the German social security system, J-W Goethe University, Frankfurt.

Duncan, G., Smeeding, T. and Rodgers, W. (1993), "W(h)ither the Middle Class? A Dynamic view", in D. Papadimitriou and E. Wolff (eds), Economic Inequality at the Close of the 20th Century, Macmillan, New York.

Economic Report of the President (1994), United States Government Printing Office, Washington, D.C.

Epland, J. (1992), "lnntekstfordelingen i 80-arene", Okonomiske analyser, nr. 2, Statistik sentralbynl.

Ermisch, J. (1987), "Demographic aspects of the growing number of lone parent families", in Lone-Parent Families, The Economic Challenge, OECD Social Policy Studies No.8, pp. 27-42, OECD, Paris.

Fichtenbaum, R. and Shahidi, H. (1988), "Transition bias and the measurement of income inequality," Journal of Business and Economic Statistics, Vol. 6, No.3, July, pp. 335-337.

Gardiner, K. (1993), "A survey of income inequality over the last twenty years- how does the UK compare?", Welfare State Programme Discussion paper 100, LSE.

Garfinkel, I. and Haveman, R. (1977), Earnings Capacity, Poverty and Inequality, Academic Press, New York.

Gottschalk, P., and Joyce, M. (1991), "Changes in earnings inequality- an international perspective", LIS Working Paper 66, Luxembourg.

Gottschalk, P., and Joyce, M. (1992), "Is earnings inequality also increasing in other industrialised countries?" LIS/CEPS Working Paper No. 66, October.

Gouveia, M. and Tavares, J. (1993), "The distribution of household income and expenditure in Portugal: 1980 and 1990", mimeo.

Green, G., Coder, J., and Ryscavage, P. (1992), "International comparisons of earnings inequality for men in the 1980's", Review of Income and Wealth, Vol. 38, pp. 1-16.

Green, F., Henley, A., and Tsakalotos, E. (1992), "Income inequality in corporatist and liberal economies: a comparison of trends within OECD countries", University of Kent, Discussion Paper 92/13, forthcoming in International Review of Applied Economics.

118

Guger, A. (1989), "The distribution of household income in Germany", WIFO Working Paper, 35, Austrian Institute of Economic Research, Vienna.

Guger, A. (1989a), "Einkommensverteilung und Verteilungspolitik in bsterreich", in H. Abele, E. Nowotny, S. Schleicher and G. Winckler (eds.), Handbuch der Osterreichischen Wirtschaftspolitik, Manzsche Verlag, Vienna.

Gustafsson, B. and Palmer, E. E. (1993), "Changes in Swedish inequality: a study of equivalent income 1975-1991 ", University of Gothenburg.

Gustafsson, B. and Uusitalo, H. (1990), "Income distribution and redistribution during two decades: experiences from Finland and Sweden", in I. Persson (ed.), Generating Equality in the Welfare State, Norwegian University Press, Stavanger.

Hanefeld, U. (1987), Das Sozio-okonomische Panel- Grundlagen und Konzeption, Frankfurt.

Hauser, R. (1992), "Die personelle Einkommensverteilung in den alten und neuen BundesHindern vor der Vereinigung", in F. Buttler (ed.), Sozialpolitik im vereinten Deutschland II, Duncker & Humblot, Berlin.

Hauser, R. and Becker, I. (1993), "The development of the income distribution in the Federal Republic of Germany during the seventies and eighties", University of Frankfurt.

Hauser, R. and Nouvertne, U. (1980), "Poverty in rich countries: the case of Germany in comparison to Australia, Belgium, Great Britain, and Norway", SFB 3, Discussion Paper 39, J-W Goethe University, Frankfurt.

Hauser, R., Muller, K., Wagner, G. and Frick, J. (1991), "Incomes in East and West Germany on the Eve of Union", Discussion Paper 34, DIW Berlin.

Haveman, R. and Buron, L. (1993), "Escaping poverty through work: the problem oflow earnings capacity in the United States", Review of Income and Wealth, Vol. 39, pp. 141-158.

Henderson, R. (1975), Poverty in Australia, Australian Government Publishing Service, Canberra.

Jansson, K. (1990), Inkomst-ochformogenhetsfordelningen 1967-1987, Bilaga 19 till LU90, Finansdepartementet, Stockholm.

Jenkins, S. P. (1991), "Poverty measurement and the within-household distribution: agenda for action", Journal of Social Policy, Vol. 20, pp. 457-483.

Jenkins, S. P., and O'Higgins, M. (1987), "Measuring poverty in the European Community", mimeo, University of Bath.

Kapteyn, A., Kooreman, P. and Willense, R. (1987), "Some methodological issues in the implementation of subjective poverty definitions", Research Memorandum FEW 245, Department of Econometrics, Tilburg University.

Karoly, L. A. (1993), "The trend in inequality among families, individuals and workers in the United States: a twenty-five year perspective", in S. Danziger and P. Gottschalk (eds.), Uneven Tides, Rising Inequality in America, Russell Sage Foundation, New York.

Kassella, T., and Hochmuth, H. (1989), "Ein synthetisches Mikrodatenfile des Haushaltssektors", SFB 3, Discussion Paper 299, J-W Goethe University, Frankfurt.

Kemsley, W. F. F., Redpath, R. U. and Holmes, M. (1980), Family Expenditure Survey Handbook, HMSO, London.

Lazear, E. and Michael, R. (1980), "Family size and the distribution of real per capita income", American Economic Review, Vol. 70, pp. 91-107.

Levy, F. and Murnane, R. J. (1992), "U.S. earnings levels and earnings inequality: a review of recent trends and proposed explanations", Journal of Economic Literature, Vol. 30, pp. 1333-1381.

Lorenz, W. and Wagner, J. (1990), "A note on returns to human capital in the eighties: evidence from twelve countries", LIS/CEPS Working Paper No. 54, July.

Luethi, A. (1981), Messung Wirtschaftlicher Ungleichheit, Freiburg, Switzerland.

Lydall, H. F. (1979), "Some problems in making international comparisons of inequality", in Moroney, J. R. (ed.), Income Inequality: Trends and International Comparisons, D.C. Heath, Lexington.

Maurice, R. (ed). (1968), National Accounts Statistics: Sources and Methods, HMSO, London.

Mayer, S. E. and Jencks, C. (1993), "Recent trends in economic inequality in the United States", in D.B. Papadimitriou and E.N. Wolff (eds.), Poverty and Prosperity in the USA in the Late Twentieth Century, Macmillan, Basingstoke.

McClements, L. D. (1978), The Economics of Social Security, Heinemann, London.

Ministero del Lavoro (1983), Famiglia e reddito, lstituto Poligrafico dello Stato, Rome.

119

Mitchell, D. (1991), Income Transfers in Ten Welfare States, Avebury, Aldershot.

Nolan, B. (1978), "The personal distribution of income in the Republic of Ireland", Journal of the Statistical and Social Inquiry Society of Ireland.

Nygard, F. (1984), "Inkomstbildningen och inkomstfordelningen i Finland och Sverige under 1970-talet", in Statistika analyser av inkomstfordelningar, Abo Akademi, Statistka instiutionen, SerA: 197, pp. 1-11.

OECD (1988), Reforming Public Pensions, OECD, Paris.

OECD (1990), Tax and Benefit Position of Production Workers, OECD, Paris.

OECD (1991), Lone Parents, OECD, Paris.

OECD (1993), "Earnings inequality: changes in the 1980s", Chapter 5, Employment Outlook, OECD, Paris.

O'Higgins, M. and Ruggles, N. (1981), "The distribution of public expenditure and taxes among households in the United Kingdom", Review of Income and Wealth, Vol. 27 (3), pp. 198-326.

Palmer, J., Smeeding, T. and Jencks, C. (1988), "The uses and limits of income comparisons" in Palmer, J., Smeeding, T. and Torrey, B.B. (eds.), The Vulnerable, Urban Institute Press, Washington D.C., pp. 9-27.

Palmer, J., Smeeding, T. and Torrey, B.B. (1988), The Vulnerable, Urban Institute Press, Washington, D.C.

Pestieau, P. (1991), "How fair is the distribution of private pension benefits?", LIS Working Paper No. 71, LIS at CEPS, Luxembourg.

Radner, D. (1983), "Adjusted estimates of the size distribution of family money income", Journal of Business and Economic Statistics, Vol. 1, pp. 135-146.

Rainwater, L. (1974), What Money Buys, Inequality and the Social Meanings of Income, Basic Books, New York.

Rainwater, L. (1987), unpublished calculations performed on Dubnoff dataset, University of Massachusetts, Survey Research Center.

Ramprakash, D. (1975), "Distribution of income statistics for the United Kingdom, 1972/73: sources and methods", Economic Trends, No. 262, pp. 78-96.

Ramprakash, D. (1986), unpublished data, Central Statistical Office, London.

Reich, U. (1991), "Concept and definition of income in the national accounts", Review of Income and Wealth, Vol. 32, No.3, September, pp. 235-247.

Ringen, S. (1986), "Difference and similarity: two studies in comparative income distribution", Discussion Paper 2 (1986), Institutet fOr Social Forskning, Stockholm.

Ringen, S. (1991), "Households, standard of living, and inequality", Review of Income and Wealth, Vol. 37, pp. 1-13.

Rodrigues, C. (1993), "The measurement and decomposition of inequality in Portugal, 1980/81-1989/90", Microsimulation Unit Discussion Paper 9302, University of Cambridge.

Rossi, N. (1993) (ed.), La crescita ineguale: rapporto CNEL sulla distribuzione e Ia redistribuzione, 11 Mulino, Bologna.

Rossi, N. (1994), La transizione: 1992-1993, secondo rapporto CNEL sulla distribuzione e redistribuzione del reddito in Italia, mimeo.

Royal Commission on the Distribution of Income and Wealth (1979), Fifth Report on the Standing Reference, Report No. 8, HMSO, London.

Ruggles, P. (1990), Drawing the Line, Urban Institute Press, Washington, D.C.

Sandoval, V. (1989), "Income distribution in France", in G. Wolleb et al. (eds.), Trends and Distribution of Incomes, European Federation for Economic Research, Luxembourg.

Sandstrom, A. ( 1980), "On the effect of changing definitions on income components. A comparison of Finnish and Swedish data", Statistik Tidskrift, Vol. 18, pp. 487-500.

Sarpellon, G. (1982), "Rapporto sulla poverta in ltalia", in G. Sarpellon (ed.), La poverta in Italia, F. Angeli, Milan.

Saunders, P. (1993), "Economic adjustment and distributional change: income inequality in Australia in the Eighties", SPRC Discussion Paper No. 47, University of New South Wales.

Sawyer, M. (1976), "Income distribution in OECD countries", OECD Economic Outlook, OECD, Paris.

Schiiler, K. (1986), "Einkommensverwendung nach Haushaltsgruppen", Allgemeines Statistisches Archiv, Vol. 70.

Sen, A. K. (1973), On Economic Inequality, Clarendon Press, Oxford.

120

Smeeding, T. (1982), Alternative Methods for Valuing Selected In-Kind Transfers and Measuring their Impact on Poverty, U.S. Bureau of the Census, Technical Report 50.

Smeeding, T. (1993), "Cross-nation patterns of retirement and poverty among men and women in the 1980s: full stop or gradual withdrawal?", in A.B. Atkinson and M. Rein (eds.), Age, Work and Social Security, St. Martin's Press, New York, pp. 91-114.

Smeeding, T. and Coder, J. (1993), "Income inequality in rich countries during the 1980s", LIS Working Paper #88, LIS at CEPS, Luxembourg.

Smeeding, T., Rainwater, L. and O'Higgins, M. (1990), Poverty, Inequality, and the Distribution of Income in an International Context: Initial Research from the Luxembourg Income Study (LIS), Wheatsheaf Books, London.

Smeeding, T., Saunders, P., Coder, J., Jenkins, S., Fritzen, J., Hagenaars, A., Hauser, R. and Wolfson, M. (1993), "Poverty inequality and family living standard impacts across seven nations: the effect of non-cash subsidies", Review of Income and Wealth.

Stark, T. (1977), The Distribution of Income in Eight Countries, Royal Commission on the Distribution of Income and Wealth, Background Paper No. 4, HMSO, London.

Statistics Canada (1987), Low Income Cut-Offs, Statistics Canada, Ottawa.

Taylor, C. (1992), "Growth, inequality and the politics of discontent in the industrial countries", mimeo, Group of Thirty, Washington, D.C.

United Nations (1977), Provisional Guidelines on Statistics of the Distribution of Income, Consumption and Accumulation of Households, Study M 61.

U.S. Department of Commerce (1981), Money Income of Households and Persons in the United States: 1980, Current Population Reports, Series P-60, No. 132, U.S. Government Printing Office, Washington, D.C.

U.S. Department of Commerce (1987), Money Income and Poverty Status of Families and Persons in the United States: /986, Current Population Reports, Series P-60, No. 157, U.S. Government Printing Office, Washington, D.C.

U.S. Department of Commerce (1991), Poverty in the United States: 1990, Current Population Reports, Series P-60, No. 175, U.S. Government Printing Office, Washington, D.C.

U.S. Department of Commerce (1992), Money Income of Households, Families, and Persons in the United States: 1991, Current Population Reports, Series P-60, No. 132, U.S. Government Printing Office, Washington, D.C.

U.S. Department of Commerce (1992a), Poverty in the United States: 1991, Current Population Reports, Series P 60, No. 181, U.S. Government Printing Office, Washington, D.C.

U.S. Department of Commerce (1992b ), Measuring the Effects of Benefits and Taxes on Income and Poverty: 1979 to 1991, Current Population Reports, Series P-60, No. 182RD, U.S. Government Printing Office, Washington, D.C.

U.S. Departement of Comerce (1993), Poverty in the United States: 1992, Current Population Report, Series P-60, No. 185, U.S. Government Printing Office, Washington D.C.

Uusitalo, H. (1989), Income Distribution in Finland, Central Statistical Office of Finland, Helsinki.

Valenduc, Ch. (1989), "Estimation de la distribution des revenus primaires et de l'impact redistributif des cotisations et transferts de securite sociale", Ministere des Finances, Bulletin de Documentation, Brussels.

Valenduc, Ch. (1991), "La distribution des revenus disponibles de 1983 a 1986", Ministere des Finances, Bulletin de Documentation, Brussels.

Valenduc, Ch. (1994), "Synthese des etudes sur la Belgique", Ministere des Finances, Brussels.

Van der Gaag, J. and Smolensky, E. (1982), "True household equivalence scales and characteristics of the poor in the United States", Review of Income and Wealth, Vol. 28, pp. 17-28.

van Ginneken, W. and Park, J. (1984), Generating Internationally Comparable Income Distribution Estimates, ILO, Geneva.

van Praag, B. M.S., Hagenaars, A. J. and van Weerden, H. (1982), "Poverty in Europe", Review of Income and Wealth, Vol. 28, pp. 345-359.

Wahlstrom, S. (1987), "Minimum subsistence budget", Statistics Sweden, Stockholm.

Watts, H. W. (1967), "The isoprop index: an approach to the determination of differential poverty income thresholds", Journal of Human Resources, Vol. 2, pp. 3-18.

Wolfson, M. (1986), "Stasis amid change - income inequality in Canada 1965-1983", Review of Income and Wealth, Vol. 32, pp. 337-369.

121

Appendix I

LIS STAFF

Luxembourg Office

John Coder, Technical Director

Heinz Stapf, Operations Manager Caroline de Tombeur, Administrative Assistant Richard Randell, Koen Vleminckx (since Jan. 1994),

Database Specialists/OECD Project Assistants Michael Forster, LES Project Assistant Mark Cigrang, Cheri Minton, Technical Consultants

US Offices

Syracuse University Tim Smeeding, Head Gina Husak, Secretary Deborah Bailey, Assistant Inge O'Connor, Administrative Assistant

Harvard University Lee Rainwater, Head Cheri Minton, Assistant

Address:

Phone: Faxes: Internet:

LIS at CEPSIINSTEAD B.P. 65 L-7201 Walferdange LUXEMBOURG

(352) 33 32 33 519 (352) 30 27 05/33 25 19 CAROLINE @ POST.CEPS.LU

Address : Maxwell School Syracuse University 426 Eggers Hall Syracuse, NY 13244 USA

Phone: (315) 443-4526 Fax: (315) 443-1081 Internet: LISAA @ MAXWELL.SYR. EDU

Address: Harvard University 1380 William James Hall 33 Kirkland Street Cambridge, MA 02138 USA

Phone: (617) 495-0445 Fax: (617) 496-5794 Internet: MINTON@ ISR.HARVARD.EDU

The "Luxembourg Income Study" is a division of CEPSIINSTEAD (President : Gaston Schaber) in Luxembourg.

123

Appendix 2

UNIT AND HEAD DEFINITIONAL DETAILS

This appendix deals with population coverage issues, household definitions (Table A2.1), head and member definitions and related issues.

AS81 AS85 CN81 CN87 FR79 FR84 GE81 GE84 IT86 LX85 NL83 NL87 SW81

SW87

SZ82

UK79

UK86

US79

US86

Table A2.1 Household Definition

Persons in the same dwelling sharing eating facilities Persons in the same dwelling sharing eating facilities Any person or group of persons living in a dwelling Any person or group of persons living in a dwelling The household was formed by combining the tax returns of all household members. The household was formed by combining the tax returns of all household members. Related or unrelated persons living together in one housing unit. All persons living single or together form a household. Related persons living together in one housing unit. Persons sharing a housing unit and a common living space. Persons living and eating together form a household. Persons living and eating together form a household. Households were defined as either one adult or two adults (more than 18-years-old) with or without children (equal or less than 17-years-old). Households were defined as either one adult or two adults (more than 18-years-old) with or without children (equal or less than 17-years-old). No specific definition of households, because only persons who were liable for taxes were included in the survey. A person living alone, or a group of people living at the same address and having meals prepared together, and with common housekeeping A person living alone, or a group of people living at the same address and having meals prepared together, and with common housekeeping. Housing unit is the usual place of residence or, if for the moment they have no usual place of residence at the time of the interview. Housing unit is the usual place of residence or, if for the moment they have no usual place of residence at the time of the interview.

124

Population Coverage (Institutional Populations, Minority Populations, etc.)

Australia

The 1981 and 1985 Australian Income Surveys were conducted throughout Australia and covered private dwellings (about 15,000 houses, flats, etc.) and non-private dwellings (hotels, motels, etc.).

Information was obtained from each person aged 14 years and over in the selected dwelling. The following persons were excluded from the scope of the survey:

a) Members of the Australian defence forces living in military establishments (AS81 only); b) Certain diplomatic personnel of overseas governments customarily excluded from the census and

estimated populations; c) Overseas visitors vacationing in Australia; d) Members of non-Australian defence forces (and their dependents) stationed in Australia; and e) Students in boarding schools, patients in hospitals and sanatoria, and inmates of gaols, reformatories, etc.

Income questions were not posed to full-time school students aged 14 to 20 years, and persons overseas for the whole of 1981-82.

Austria

The Austrian 1987 Microcensus sampling frame was comprised of private dwellings. People living in institutions, including the military, were excluded. However, military personnel living with their families are included, and conscripts are counted at their previous residence.

Canada

The final stage sampling frames for the 1981 and 1987 surveys of consumer finances was comprised of a list of all private dwellings in the ten Canadian provinces. This sampling frame includes the total population of private household heads, with the exception of households on Native American reserves, and households in the Yukon and Northwest Territories.

The population living in institutions such as nursing homes for the aged and ill, prisons, etc., were not covered; however, military personnel living in military housing or with their families were covered. Foreign residents (defined as neither Canadian born nor landed immigrants), persons with a usual residence elsewhere, and those individuals living in seasonal dwellings, were excluded from the survey at the time of the interview.

Finland

The sample frame which was employed in the Finnish 1987 Income Distribution Statistics survey was the 1986 taxation register. This covers the entire resident population over 16 years of age, and also children under 16 who had a taxable income. All military personnel are included in the sampling frame, although people living permanently in institutions were excluded (about 1.5 per cent of the population).

France

Although the Revenus Fiscaux data are extracted from income tax files, the sampling frame for the survey consists of housing units recorded in the 1975 (or 1982) population census, updated to 1979 (or 1984) from a list of newly constructed housing units. A three-stage stratified sampling design was used to draw a sample of addresses, which were then given to the income tax administration, which matched the taxation records with these addresses.

Household heads who did not complete tax returns were not included in the survey. In addition, people living in institutions or mobile homes were not included in the survey. Some military personnel were included in the sampling frame, however this depended on where they registered their tax documents.

The geographic regions of Reunion, Martinique, Guadeloupe and Guyane were excluded from the Revenus Fiscaux survey.

125

Germany

The final stage sampling frame used for the SOEP 1984 survey was the list of registered voters in the German Federal Government Election of 1980. In addition, a subsample which included the main groups of guest workers (Spanish, Italian, Greek, Turkish, Yugoslav) was selected.

People living in institutions or other group situations are included in the survey in those cases where there are no extraordinary difficulties in obtaining an interview (e.g. persons living in homes for the elderly, dormitories, etc.).

Military personnel living in military housing are excluded from the sample. However, military personnel living with their families are included, but it is not possible to identify them separately.

All geographic areas in the country were included in the sample design.

Ireland

The Electoral register was employed as the sampling frame in the 1987 ESRI Survey of Income Distribution, Poverty and Usage of State Services. All people living in institutions (1-2 per cent of the population) were excluded. However, military personnel living with their families were included.

Italy

The final stage sampling frame employed in the Bank of Italy's 1986 Family Income Survey was a General Population Register. It did not include that segment of the institutionalised population residing in nursing homes for the aged or ill, prisons, or military installations. Military personnel living in military installations were not listed in the sampling frame, however, military personnel living with their families were included.

All geographic areas in the country were included in the sample design.

Luxembourg

The sampling frame for the PSELL 1985 survey consists of registers of the social security system held by L'inspection Generate de la Securite Sociale (IGSS), supplemented by other sources (Centre Commun d'Affiliation). This combined sample frame covers 97 per cent of the resident population of Luxembourg. Household heads who are of foreign nationality, and some people with very low and very high incomes who were not attached to Luxembourg social security system were not included in the sampling frame. These represent 3 per cent of the resident population.

People living in institutions such as prisons, homes for the elderly, or other group situations were not included in the survey. Military personnel were not included in the sampling frame.

All geographic areas in the country were included in the sample design.

The Netherlands

The sampling frames for the 1983 and 1987 AVO surveys consists of post office addresses of private housing units. It includes the total population of household heads.

People living in institutions or other group situations such as prisons, military installations were included in the sampling frame (although none of these institutions was actually selected for the sample). Other institutions, such as nursing homes for the aged and psychiatric wards were excluded.

All geographic areas in the country were included in the sample design.

Sweden

The sampling frame for the HINK 1981 and 1987 surveys consists of the General Population Register of all individuals 18 years of age or older.

People living in institutions such as prisons or hospitals are not included in the survey. Military personnel were included in the sampling frame. All geographic areas of the country were included in the sample design.

126

Switzerland

The sampling frame for the 1982 Swiss Income and Wealth Survey consists of a list of registered voters which is kept by the Swiss Statistical Office.

Foreign household heads without permanent residence permits were not included in the sampling frame. Military personnel, and people living in institutions or other group situations were included in the sampling frame.

All geographic areas in the country were included in the sample design.

United Kingdom

The final stage sampling frame for the 1979 FES consists of addresses drawn from electoral registers maintained for Parliamentary and Local Government elections. Although the sample design for the Northern Ireland survey has not been changed, it should be noted that in 1986 the Great Britain FES sample design was changed for the first time since 1967. Instead of using the Parliamentary Electoral Registers as the final stage sampling frame, the 1986 FES uses the "small users file" of the Postcode Address File (PAF). Consequently, care must be taken when comparing 1986 and later FESs with earlier ones.

It should be noted that since the final stage sampling frame for the 1979 FES is comprised of Electoral Registers maintained for Parliamentary and Local Government elections in the United Kingdom, all people not on this register were excluded from the survey. This group can initially be broken down into two categories:

a) People who were not legally eligible to vote (and consequently register) for Parliamentary and/or Local Government elections; and

b) People who were eligible to vote, but had not placed themselves on the electoral register.

The only exception to this occurs in those cases where the individual whose name was selected from the electoral register no longer lives at the household address given on the Electoral Register. In those cases where the household residents were not eligible to place their names on the Electoral Register the interviewer was nevertheless instructed to obtain an interview if possible.

No attempt was made to obtain interviews from households where it was discovered after the first interviewer call that the household contained:

a) Members of the Diplomatic Service of any country (except the United Kingdom); b) Members of the USA Forces; and c) Roman Catholic Priests if they are living in accommodation provided by the parish church.

In addition, because the FES is concerned with collecting information about domestic, not business expenditure, if a commercial establishment [such as public houses, hotels of all types, guest houses/commercial boarding houses, private households containing 4 or more boarders at the first interviewer call, and institutions (e.g. hostels, schools, prisons, hospitals, religious establishments)] fell into the sample, interviews were not obtained, except in those cases when at the selected address completely separate accommodation where the household is responsible for all its domestic expenses exists.

Despite its name, the Family Expenditure Survey is actually a survey of households, not families. Consequently, the survey unit is comprised of households. These are defined as:

a) One person living alone; or b) A group of people living at the same address having meals prepared together and with common

housekeeping.

Only private households are included. Resident domestic servants are included. The members of a household need not necessarily be related by blood or marriage.

Unoccupied units were not excluded from the sample frame as there was no indication in advance that any unit might be unoccupied.

Households were not excluded if some or all members were not British subjects.

The geographic regions of the Isles of Scilly, and three Scottish districts (Orkney, Shetland and Western Isles) and the island parts of Cunninghame, Argyll and Bute, Lochabar and Skye and Lochalsh districts were excluded from the sample frame. These areas were excluded because of travel costs and difficulty of accessibility.

United States

The sampling frame for the 1981 and 1982 CPS surveys consists of a list of all housing units compiled from the most recent decennial census of population and housing which is supplemented by lists of newly constructed housing units that did not exist when the census was conducted.

127

People resident in institutions (such as hospitals, nursing homes and long-term care facilities, homes for the aged, penitentiaries, housing for members of the military not living with their families, dormitories at schools, and schools) were not included in the survey.

All housing units containing one or more permanent residents are eligible for interview.

A group of persons, or a person living alone, in a housing unit is defined as a household without consideration of their family relationships.

The CPS universe is households and persons living in households.

The entire geographic area of the United States, including Hawaii and Alaska, is included in the CPS survey design. Puerto Rico and other territories are not included.

Definition of household and head of unit

General: For each country survey in the LIS database the head of household, if the designated person is male, is also designated as the household head by LIS, and no checks are made to determine who that person is. In those cases when a female is the designated household head in the original survey, checks are performed to see if there is a male spouse present. If so, the male person is designated by LIS as the household head.

Australia

The following describes the household definition, identification of household members, and how the unit head was designated in each of the original country surveys.

To determine family relationships, respondents were asked to nominate a person as head of the household, with relationships recorded relative to that person. Within households, respondents were classified into families and "income units".

Income units can be one of the following types: single adults, sole parents with dependent children, married couples or married couples with dependent children. Couples in "de facto" (cohabiting) relationships are coded identically to those legally married.

Families comprise persons related by blood, marriage (legal or de-facto) or adoption with the proviso that there can only be one couple or sole parent per family. Thus a household comprising a married couple and their sole parent daughter is coded as two families, as is a household comprising a husband and wife together with the wife's mother and her husband.

The family head on the unit record file, however, is not necessarily the person nominated as household head by the initial respondent. The family head is defined as the head of the primary income unit of the family. If this income unit is a couple, the husband is the head. The primary income unit is defined according to a pre-defined hierarchy, with married couples, for example, coming before individuals.

Austria

Children away from home and attending school or other educational institutions, and away from home for work or military duty are included as household members. The head of household is designated by the household itself.

Canada

A household member is a person who, during the survey week:

a) Regards the dwelling to be their usual place of residence; or b) Is staying in the dwelling and has no usual place of residence elsewhere.

A household is defined as any person or group of persons living in a dwelling. It may consist of one person living alone, a group of people who are not related but who share the same dwelling, or it may be a family.

128

A dwelling is defined as any set of living quarters which:

a) Is structurally separate from the living quarters of other dwellings; and b) Has a private entrance outside the building or a private entrance from a common hall or stairway inside

the building. The entrance must be one that can be used without passing through the living quarters of another dwelling.

In the case of families, the following individuals and their relationship to the head of the family can be identified in the Survey of Consumer Finances:

a) Head of family; b) Spouse; c) Son or daughter (natural, adopted or step); d) Grandchild; e) Son-in-law or daughter-in-law; f) Foster child (less than 18 years of age); g) Parent; h) Parent-in-law; i) Brother or sister; and j) Other relatives.

The head of household is always the head of the primary economic family.

The basic units of aggregation which are used in this survey are: Household, Economic Family, Census Family, and Individuals.

Within households, Economic Family units and individual relationships within that family, and Census Family Units and individual relationships within that family, can be identified.

Finland

The basic unit of aggregation is the household. A household is defined as consisting of persons who live and have meals together, or who share their income in other ways (i.e. children who usually live away from home and finance their studies by loan are classified as own households). Military conscripts and children who live away from home but whose educational expenses are covered by their parents are included in their parents' household. Men on military duty are included in the household depending on their previous circumstances.

France

All individuals who can be identified as living in a particular housing unit are considered members of the same household.

The basic unit of aggregation which was used in this survey was the household unit, which is an aggregation of fiscal units (foyer fiscal).

The sampling unit can be divided into subgroups in those situations where several families share the same housing unit.

If there is one family, the head is the male adult. If there is more than one family in a household, the head is the oldest working person, or the oldest person when no household member is working.

Other than the sampling unit head the following individuals can be identified in the Revenus Fiscaux:

a) Head of household; b) Spouse (legal or not); c) Children; d) Grandchildren; e) Elderly persons of the family (father or uncle of the head); f) Siblings; g) Friends, pensioners, sub-tenants, foster children; h) Servants (domestic), employed persons; and i) Undefined

It is possible to identify the relationship of these individuals to the sampling unit head.

129

Germany

Households are housing and economic units, comprised of one or more individuals. Household members need not be related to each other.

The basic unit of aggregation which was used in the SOEP was the individual. The overall sampling unit can be divided into subgroups in situations where several families share the same housing unit.

The survey unit head was defined as the person with the best knowledge of household living conditions (not in general the oldest person).

Individuals other than the sampling unit head can be identified, and information about the relationship of individuals to the head of household is available.

Italy

In order to be considered a member of the survey unit, household members must live together. However, people who live together solely for economic reasons are not considered household members. Consequently, households are defined in terms of family relationships.

Other than the sampling unit head, spouses, children and other relatives can be identified.

The basic unit of aggregation which was used in this survey was families.

The overall sampling unit cannot be divided into subgroups in situations where several families share the same housing unit.

The survey unit head is usually the husband or father, but if he is abroad or in another region of Italy the household head is the person who is economically responsible for the family.

Luxembourg

The basic unit of aggregation which was used in this survey was the household. Information from individual and group files have been aggregated at the household level.

Households were defined as a housing unit comprised of persons living together, and sharing a common living space such as a kitchen.

In the original file, the overall sampling unit is divisible into subgroups in situations where several families share the same housing unit. The survey unit head was the male in the case of couples, or the owner of the housing unit.

Individuals other than the sampling unit head which can be identified are in the original data file of the PSELL. The following relationships of individuals to families within the household can be identified:

a) The relationship of each member to the head of household is encoded; and b) The relationship of each household to others in the household (such as partner, father, mother, child, etc.).

The Netherlands

A household is comprised of persons living and eating together. At the same address there can be more than one household. A person sleeping regularly at another address (e.g. drafted soldiers, students) is not considered to be part of his or her family. There are no rules governing the number of nights one has to be at home to be considered a household member - this is left up to the discretion of the interviewer.

Two basic units of aggregation were used in the AVO:

a) Parents with all children living at home; and b) Parents with children from 0 to 17 years of age.

The overall sampling unit can be divided into subgroups in situations where several households share the same housing unit. 1 The household head who was identified by the respondents should be over 18 years. If the respondent could not decide, the interviewer suggested the rent payer or home ow~r, the person with the highest income, or the oldest person successively.

1 Only five cases.

130

Sweden

The basic units of aggregation used in this survey were individuals and households. These were further aggregated into groups based on sex, age, profession and income.

Households were defined as either one adult or two adults (more than 18 years old) with or without children (equal or less than 17 years old). Persons older than 17, if in the sample of households, constitute a household of their own. Consequently, if the parent is sampled, children over 17 are not considered members of the household.

The overall sampling unit is not divisible into subgroups in situations where several families share the same housing unit.

All household members can be identified in this survey.

Switzerland

All individuals who can be identified as living in a particular housing unit are considered members of the same household.

The basic unit of aggregation which was used in this survey was the tax unit. In order to determine whether or not a person is considered a member of the survey unit (for example, children, students away from home and attending a school or other educational institution, husbands away from home at work, or elderly family members living in a nursing home) they had to be supported members of the tax unit. There is no specific definition of households or families, because only persons who were liable for taxes were included in the survey.

The overall sampling unit cannot be divided into subgroups in situations where several families share the same housing unit.

United Kingdom

A person (in a household) who usually has one meal a day in that household is normally regarded as a household member. Persons who spend only part of their time in the household are considered members provided they usually spend at least four nights a week in the household. However, when a married person is a member, the husband (or wife) is usually counted as a member provided that either he or she usually goes home at least one night a week or will be staying with the household for all or most of the record-keeping period. A child under 16 at boarding school is also regarded as a member provided holidays are spent with the household. A person staying temporarily with the household, or who has been living with the household for only a short time, is considered a member provided he or she stays with the household for at least one month from the start of records.

The survey unit head is the head of household. The head of household must be a member of that household. He or she is the person, or the husband of the person, who:

a) owns the household accommodation; b) is legally responsible for the rent of the accommodation; c) has the household accommodation as an emolument or prerequisite; or d) has the household accommodation by virtue of some relationship to the owner who is not a member of the

household.

When two members of different sex have equal claim, the male is taken as head of household. When two members of the same sex have equal claim, the elder is taken as head of household. The following relationships to the household head can be identified: wife or husband; son or daughter [including stepson/daughter; son-in-law or daughter-in-law; father or mother; father-in-law or mother-in-law; brother or sister; grandson or grand-daughter; other relative (e.g. niece, nephew, brother-in-law)].

United States

A person is considered a member of a household (survey unit) if that housing unit is his/her usual place of residence, or if at the time of the interview he/she has no usual place of residence. The usual place of residence is the place at which a person usually eats and sleeps, and to which that person is free to return to at any time.

This definition includes persons who usually live in the sample unit but who are temporarily absent, such as unmarried children away at school who live there when not at school, persons travelling on business, seamen, railroad workers, persons away on vacation, or temporarily in hospital.

Also included as household members are lodgers, servants, farm workers, and other employees who live in the unit and consider it their usual residence.

131

Citizens of foreign countries living in the United States are defined as household members using these same definitions.

Members of the military who are away on assignment and who do not eat or sleep in the housing unit are not included as household members, even though they may be married to one of the current household members.

Individuals constitute the basic unit of aggregation which was used in this survey.

Families are defined as groups of people within the household who are related by blood, marriage, or adoption.

In addition, subgroups within the household are established based solely on relationship.

Persons are defined as related if they are bound by blood, marriage, or adoption.

All related persons living within the same housing unit are grouped together to a family even though the family may include several generations, or several separate married couples, such as a husband and wife and the wife's parent.

If a housing unit contains two unrelated family groups, each family group is defined separately according to the relationships of the members of each group to the family head.

The head of a household that does not contain two or more related persons (family) must be an unrelated individual.

Children (up to age 18), students

Australia

Dependent children were defined as all unmarried persons living with their parent(s) and either under 15 years of age, or full-time students aged 15-20 years. Dependent children who were at school (but not in higher education) were not asked any income questions. Since all persons aged 14 and over were included in the unit record file, alternative definitions of children are possible.

Couples in de facto (cohabiting) relationships were coded as if married. De facto relationships were defined as existing where a couple live together in a married situation, but are not legally married. For example, responses such as "fiancee" and "living with my girlfriend/boyfriend" are coded as de-facto.

Austria

Children are defined according to age and economic dependence (operationalised in terms of the right to receive a family allowance). Children are preschool children, students, children at school, apprentices, or other supported persons not above the age of 27 years.

Spouses are either legally married persons, or people living in consensual unions.

Canada

In this survey natural children, adopted children, guardianship children and foster children can be identified, but it is not possible to distinguish between the children of the head versus children of the spouse.

Spouses are defined as legally married, and common-law couples.

Finland

Children are defined as all persons under the age of 18 years. However, if a person under 18 years is the head of a household, or the spouse of the head, he or she is not considered a child.

France

In this survey, children are defined as all supported persons in a family who have never been married and were under the age of 18 years before January 1, 1979 (or 1984), or were born after 1979 (or 1984). If they are disabled, the age is not taken into consideration. Individuals who were supported by their parents and are between the ages of 18 and 21 years, or who were in military service, or were studying and are between the ages of 18 and 25 years,

132

were also defined as children. In addition to this, children who are married but are supported by their family and under 21 years of age (or under 25 years if studying) are considered children.

Spouses are defined as the legally married partner of the household head. Cohabiting individuals are also considered spouses if this information was available from the census form. However, in this case they are considered to be two separate fiscal units.

Variables exist which allow one to link the household head and the fiscal unit head.

Germany

In the SOEP no specific rule is used to define children. However, information on individual relationships to the head of household is available (e.g. own or adopted children). Hence it is possible to identify children - including adult children - of the head.

The survey also provides information which allows one to distinguish between children of the unit head (or spouse), and children of some other household member, or of someone living outside the household.

Spouses are not explicitly defined. However, as with children, spouses can be identified through their relationship to the head of household.

Italy

In this survey, information such as age and relationship to the head is available, however, no specific rule is used to define children.

The survey does not provide information to distinguish between children of the unit head (or spouse), and children of some other household member, or of someone living outside the household.

In this survey spouses are defined as the legal marriage partner of the household head.

Luxembourg

Children are defined as all persons under 16 years and not married. Persons over 16 years, not married, and in training (school or college) are also considered children. However, their education must not have been interrupted for more than 1 year. The PSELL survey provides information to distinguish between children of the unit head (or spouse), and children of some other household member, or of someone living outside the household.

In this survey the definition of spouses includes persons who are legally married, and cohabiting partners of the head.

The Netherlands

In the AVO, children are defined as all persons under the age of 18 years.

The survey provides information to distinguish between children of the unit head or partner, son/daughter in law, and grandchildren.

In this survey partners or spouses are initially identified by the interviewer.

Other than children and spouses, it is also possible to identify father/mother in law, other family of the head of household, and those who are not related to the head of household.

Sweden

Children are defined as all persons under the age of 18 who have never been married.

The survey does not provide information to distinguish between children of the unit head (or spouse), and children of some other household member, or of someone living outside the household.

Spouses are defined as people who are cohabiting, as well as people who are married.

Switzerland

Children were registered in the Swiss Income and Health Survey, but no further information is available about them. This includes characteristics such as their sex and age. Neither does the survey provide information to distinguish between children of the unit head (or spouse), and children of some other household member, or of someone living outside the household.

133

Spouses are defined in terms of their relationship to the person who is liable for the household taxes. Other types of family or household arrangements were classified as comprising of two or more tax units. For example: non-married couples; married couples with a child of 18 years or more; divorced; single parent families; and female-headed households.

Other than children and spouses, it is also possible to identify grandparents and other relatives.

United Kingdom

Adults are defined as persons who have reached the age of 18, or who are married. Children are defined as persons who are under 18 years of age and unmarried.

United States

In the CPS, children are defined by their relationship to household and family heads. They are included in the household in which they reside even if their parents are separated or divorced and the children are supported mainly by the absent spouse's income (dependents for tax purposes).

A person of any age can be considered an "own" child if his/her parent is the household or family head.

Other persons of any age who are related to the household or family head but not an own child are defined as other relatives.

Foster children are not considered children of the head but unrelated individuals.

In the CPS, spouses must be legally married to be so defined. "Married" therefore means legally married. Cohabiting men and women are not defined as related in the survey even if they have children.

Persons who are not related to any other person in the household are not members of any family and are considered to be unrelated individuals.

134

Appendix 3

MEASURES OF INEQUALITY

This appendix describes briefly the measures of inequality used in the report. For fuller discussion, the reader is referred to Atkinson (1983), Cowell (1977) or Sawyer (1976).

The description is given for the case where the data are grouped at intervals; the application to micro-data follows directly by treating each observation as a group of one. There are assumed to ben observations of incomes y1, Yr"' y" where these are ranked in increasing (non-decreasing) order. There are J: people with incomes yi where

n LJ; = N (A3.1) i =I

and the mean income is

(A3.2)

(where we have micro-data,.t; = 1 and n = N).

The Lorenz curve is formed by calculating for all k the proportion of the population with income less than or equal to yk:

(A3.3)

and their share of total income:

(A3.4)

(The generalised Lorenz curve is calculated without dividing by y.) Where the distribution is located at discrete points, as assumed above, this generates a piece-wise Lorenz curve. Where the data are grouped in intervals, so that we have .t; people with incomes between y(l and yi' then an interpolation procedure has to be applied [see, for example, Atkinson and Micklewright, 1992, pp. 279-83, or the manual for the INEQ programme (Cowell, 1992)].

The Gini coefficient is the area between the Lorenz curve and the diagonal, relative to the triangle formed by the diagonal and the axes. It may be calculated from the formula for the relative mean difference:

(A3.5)

135

The Atkinson index, I, is calculated from the formula:

[ ]

11(1-e)

I -I=: ~ J; (y; /yJf·efN (A3.6)

where e is a parameter (e i= 1). The value taken in the text is e = 0.5.

Where e = 1.0, the index is calculated as

1-I:exp[~ J;log.(Y,Iy)!N] (A3.7)

136

Appendix4

DATA TYPE, QUALITY AND CONSISTENCY

This appendix identifies the types of surveys the LIS employs (Table A4.1), the sampling frame for each (Table A4.2), the response rates and LIS sample sizes (Table A4.3), and survey coding procedures (Table A4.4). It also discusses the technical terms "hot decking", "cold decking", "top and bottom coding" and other technical terms.

Several technical terms are used in Chapter 3. They are presented here in order of appearance.

Item non-response. This occurs when a respondent is unable or unwilling to answer the amount of a particular type of income received, or to give a particular demographic fact (e.g. the age).

Imputations. This is a technique used to estimate the amount of income of a particular type which a unit should receive, or a tax which they should pay. These are usually types of income which are universal (e.g. child allowances) or which follow a simple rule (e.g. payroll taxes as a percentage of earnings). The rules are used to impute or assign incomes to households. Hot-deck and cold-deck imputations are two particular imputation techniques used to correct for item non-response. Hot-decking refers to using an algorithm to assign a value to the non-respondent based on the value reported by the nearest survey respondent whose demographic characteristics match those of the non-respondent. Cold-decking refers to assigning an item value to a non-respondent based on the average value reported by all respondents with the same demographic characteristics reported by the item non­respondent.

Differential non-response. Refers to the fact that both non-respondents (those who refuse to participate in the survey) and item non-respondents (those who participate but refuse to answer certain questions) are a non-random sample of the population. For instance, in the United States, the high-income elderly are particularly likely to under-report or not report their incomes (Radner, 1983). Non-response may vary systematically by age, region, income level or other characteristics depending on the particular question or the particular survey examined (see, for instance, Atkinson and Micklewright, 1983).

Bottom-coding and top-coding. These refer to survey rules to externally impose minimum and maximum values for total income or for specific income components (e.g. earnings, self-employment). For instance, some countries have no bottom codes for self-employment income, thereby allowing for losses (negative income)- see Table A4.5. Other countries impose zero as the minimum value for an income variable.

Some countries impose top-codes as well. For instance, individual earnings cannot be more than 299,000 in the U.S. survey. Other countries impose maximums on total income, ages (e.g. age 80 is the maximum age of a household head in several countries), or other variables to protect the privacy of respondents.

137

ASS! ASS5 BESS BESS CNSI CNS7 FR79 FRS4 GES4

ITS6 LXS5 NLS3 NLS7 NW79

NW86

OSS7 SWSI SWS7 SZS2 UK69 UK79 UKS6 US79 USS6

Table A4.1 Primary Objective of the Survey*

Measurement of personal and household income distributions and housing cost Measurement of personal and household income distributions and housing cost To analyse the income distribution, poverty and the effectiveness of the Belgian social security system To analyse the income distribution, poverty and the effectiveness of the Belgian social security system To measure the composition, earnings levels, and distribution of income in Canada To measure the composition, earnings levels, and distribution of income in Canada Measurement of household income distribution Measurement of household income distribution Provide information on the dynamics of individual welfare, and to evaluate the social impact of government social policy Provide information on the economic behavior of families To measure unemployment, poverty, labor force participation, income, and family expenditure To measure income household composition and the use of public services To measure income, household composition, and the use of public services Provide information on the structure and the distribution of income and property for households and individuals Provide information on the structure and the distribution of income and property for households and individuals Microcensus, multi-purposes To measure income distribution To measure income distribution To measure income distribution Provide information on spending patterns for the United Kingdom Retail Prices Index Provide information on spending patterns for the United Kingdom Retail Prices Index Provide information on spending patterns for the United Kingdom Retail Prices Index To provide estimates of employment, unemployment, and other characteristics of the labor forces To provide estimates of employment, unemployment, and other characteristics of the labor force

Primary objective of the survey refers to the main purpose for which the data were collected.

ASS! ASS5 BESS BESS CNSI CNS7 FR79 FRS4 GES4 IRS7 ITS6 LXS5 NLS3 NLS7 NW79 NWS6 OSS7

SWSI SWS7 SZS2 UK69 UK79 UKS6 US79 USS6

Table A4.2 Sampling Frame*

Block sampling within Census collector's districts Block sampling within Census collector's districts Population registers of selected municipalities Population registers of selected municipalities Address list of all private dwellings Address list of all private dwellings Census list of housing units Census list of housing units Electoral registers Electoral registers General Population Register Social Security System Register Post office address list of private households Post office address list of private households Central Population Register Central Population Register Stock of Dwellings (Census of Population and Housing 19SI +newly built dwellings since 19Sl), supplemented by population living in institutions > 50 persons Taxation register Taxation register Electoral registers Electoral registers Electoral registers Post office address lists Address lists of Census Address lists of Census

For those surveys which utilized a multi-stage sample design, only the final stage sampling frame is listed in this table.

138

Table A4.3 Response Rates and LIS Sample Sizes

Response Rate LIS Sample Size (unweighted)

FR79 Not applicable (based on tax file II 044 documents)

FRS4 Not applicable (based on tax file 12 693 documents)

GES4 SOEP Wave 1: 61.7%; 5 Ill SOEP Wave 2: S8.9%

ITS6 61.4% 8 022 LXSS 61.2% 2 054 NL83 56 9% 4 833 NL87 60.7% 4190 NW79 75.0% 10414 SZ82 Not applicable (based on tax file 7 036

documents) SW81 86.9% 9 625 SW87 86.1% 9 530 US79 95% to 96.5% 15 174 USS6 95% to 96.5% 12 158 AS81 93.6% 15 985 AS85 94.1% 8 014 UK79 67% 6 888 UK86 69% 7 178 CN81 73.0% 15 136 CN87 73.0% II 518

Table A4.4 Data Transformations

Values replaced Other from Simulation,

Hot or Cold Adjustments for Alternative Estimation or Decking Non-Response Data Sources Imputation of Values

ASS! Yes No No Yes ASSS Yes No No Yes BESS Yes No No Yes BES8 Yes No No Yes CNS1 Yes No No Yes CNS7 Yes No No Yes FR79 No No No Yes FRS4 No No No Yes GES4 Yes No No Yes IR87 No Yes No Yes IT86 Yes No No No LXSS Yes Yes No No NLS3 No No No Yes NLS7 No No No Yes NW79 No No No No NWS6 No No No No 0887 Yes No No No SWS1 No No No No SWS7 No No No No SZ82 No No No No UK69 No No No No UK79 No No No No UKS6 No No No No US79 Yes No No Yes US86 Yes No No Yes

139

Table A4.5 Descriptives of LIS Second Wave Datasets with Negative Disposable Income (DPI)

Dataset Mean DPI Minimum Maximum Mean Percentage DPI DPI Negative Cases with

DPI Negative DPI

Canada 1987 30 028.73 -69 810 165 630 -15 507 0.087 France 1994 110 286.40 -I 082 060 7 850 600 -68 486 0.430 Ireland 1987 11 497.24 -54 714 173 930 -5 698 1.002 Netherlands 1983 31 377.47 -2 980 171 680 -1 083 0.014

1987 30 875.76 -18 760 137 620 -1 330 0.021 Norway 1986 16 588.20 -118 633 I 158 133 -42 667 0.047 Sweden 1987 98 681.87 -650 500 3 551 600 -134 814 0.290 United Kingdom 1986 8 722.55 -41 846 103 559 -2 736 0.012 United States 1986 23 425.38 -33 920 269 110 -6 335 0.276

Note: Amounts are in national currency units per year that were used in the LIS database.

140

...... -!:>-

Appendix 5 CORRESPONDENCE OF MICRO-INCOME DATA MEASURES WITH SNA CATEGORIES

UN Distribution Statistics Australia 81-82

Australia 85-86

Austria 1987

1. Primary income gross of consumption of fixed capital

a) compensation of employees i) wages and salaries ii) employers' contribu­

tions to social security and similar schemes

b) income of members from producers' co­operation

c) gross entrepreneurial income of unincorporated enterprises, including withdrawals from entrepreneurial income of quasi-corporate enterprises

2. Property income received

a) imputed rent of owner­occupied dwellings

b) interest

c) dividends

d) rent, royalties, patents, copyrights, etc.

Differs Differs Differs

Not Not applicable applicable N.A.

Included Included N.A. under c) under c)

Differs Differs N.A.

N.A. N.A. N.A.

Available Available N.A.

Included Corresp. N.A. under d)

Differs Differs N.A.

3. Current transfers and other benefits received

Belgium 1985

Differs

N.A.

N.A.

Belgium 1988

Differs

N.A.

N.A.

Finland 1987

Differs

N.A.

N.A.

France 1984

Corresp.

N.A.

Differs

Germany 1981

Germany Ireland 1984 1987

Italy 1986

Corresp. Corresp. Available Differs

Available Available N.A.

Not Not N.A.

N.A.

Not

Netherlands Norway 1987 1979

Sweden 1987

Differs Corresp. Differs

Imputed N.A.

Included N.A.

N.A.

N.A. separable separable separable under c)

N.A. N.A. Differs Differs Available Corresp. Available Differs Available Available Differs

Available Available Corresp. N.A. Corresp. Available N.A. Available Available Differs N.A .

Available Available Available Differs Available Available Available Differs

Included Included Corresp. Differs Included Included Available Differs under b) under b) under b) under b)

Available Available Available Available Corresp. Available Available Differs

Available Available

Included under b)

Available

Differs

Included under b)

Included under b)

Available Differs

U.K. 1986

u.s. 1986

Corresp. Differs

Differs Available

Available Not clear

Available Differs

Available N.A.

Available Available

Available Available

Differs Differs

a) social security benefits Available Available Available Available Available Available N.A. Available Available Available Available Differs Available Differs Available Available

b) pensions and life insurance - annuity benefits

c) other current transfers

4. Direct taxes paid

Differs

Differs

N.A.

Corresp. Differs

Differs N.A.

Corresp. N.A.

5. Social security and pension fund contributions

a) social security Not Not N.A. contributions applicable applicable

b) pension fund contributions Not Not N.A. applicable applicable

N.A. =not available.

Available Available Available Available Available Differs Differs Available Available Available Available Available Available

Available Available Available Available Available Available Available N.A. Available Available Available Available Available

N.A. N.A. Available Available Available Available Available N.A. Available Available Available Available Available

N.A. N.A. Available N.A. Differs Available Available N.A. Available Available Available Available Available

N.A. N.A. Available N.A. See Sa See5a Available N.A. Available N.A. N.A. Available N.A.

Appendix 6

QUALITY OF INCOME DATA

This appendix describes the comparisons of the LIS survey data with information from national accounts and administrative sources (such as those on benefit receipt or on the distribution of taxable income) where such comparisons are available. This information contained here provides the basis for the summary table (Table 3.7) in the main text which compares survey data with national accounts aggregates, but the appendix also includes some comparisons with other sources.

These comparisons aim to help assess the quality of income data. The following points should nevertheless be kept in mind:

a) the definition and/or coverage of these external sources may differ from that of the LIS dataset -discrepancies between the two sets of results do not imply inadequacies in the dataset; and

b) comparisons are not generally directly based on the LIS dataset but on the original survey, so that there may be (minor) discrepancies, particularly where different grossing-up or other procedures were followed.

The appendix is organised by country in alphabetical order. Individual items are grouped in line with the categories used in the text discussion:

- wages and salaries; - self employment income; - property income; and - transfers.

Australia

The main strength of the Australian surveys is the very high response rate resulting from the compulsory nature of the survey. Fieldwork procedures were also very thorough, respondents were requested to refer to records expected to result in income levels close to those recorded for taxation purposes. Where respondents to the survey could not supply precise details of annual income from some sources, income from the sources concerned was derived from other known sources (e.g. pension rates) or was imputed from a similar source from other respondents with similar characteristics (hot-decking). Negative incomes are recorded as zero.

In addition, incomes which may be considered unusually high (and possibly identifying individuals) have been adjusted. This has generally been done by selecting a number of persons with very high incomes (in a given demographic category) and recoding their incomes to their mean income (i.e. the mean income of the persons with high incomes). This should leave the totals unaffected.

Nonetheless, it appears that the survey suffers from under-recording of unearned income, as is typical for most other income surveys. It should be noted that one reason for discrepancies between the National Accounts and survey estimates is that the National Accounts explicitly adjust for the under-reporting of income in taxation and similar data. A comparison of the survey income aggregates with those available from the Australian National Accounts is shown for 1981-82 in Table A6.1 and for 1985-86 in Table A6.2. 1 Although total household disposable income recorded by the survey comprises only 82 per cent in 1981-82 and 79 per cent in 1985-86 of that recorded by the National Accounts, most of this difference is explained by the conceptual differences in the two data sources. Though such conceptual differences apply to all the items in this table, some key issues of data quality are identifiable.

1. These calculations are based on the unit record file rather than the LIS database. In particular, the imputation of income tax has been done using a model developed by the Social Policy Research Centre, rather than the imputation model used by LIS. (These calculations were not done on the LIS file because some cases with fluctuating incomes have been removed by LIS.)

142

Table A6.1 Comparison of Australian Income Survey with National Accounts Aggregates, 1981-82

Wages and Salaries Supplements to W&S TOTAL wages and salaries

Gross Operating Surplus of Unincorporated Enterprises Dwelling Rent (NA includes imputed rent of owner/occupiers) Imputed Interest on Life and Superannuation Funds Superannuation Receipts Other Interest Dividends TOTAL interest and dividends

Personal Benefit Payments to Residents Current Grants to Non-Profit Institutions Third Party Insurance Transfers Unreq. Transfers from OS Other Income Sources TOTAL HOUSEHOLD INCOME

Personal Income Tax Other Transfers to Govt and OS TOTAL HOUSEHOLD DISPOSABLE INCOME

National Accounts

(NA) $m

80401 5 143

85 544

11482

2495

3 002 na

8 081 1045 9 126

13 250 1 516

655 798

na 127 868

21 205 3 158

103 505

Income Survey

$m

74 135 156

4 291

14 338

971

na 1 189 2 537 2 086 4 623

9 985 na na na

717 106 114

21 112 na

85 002

92.2 3.0

86.8

124.9

38.9

31.4 199.6 50.7

75.4

83.0

99.6

82.1

Source: l98I-82 Income and Housing Survey, Unit Record File, Australian National Accounts, National Income and Expenditure 1987-88, and SPRC calculations.

a) In 1981-82 (but not 1985-86), wages received by persons from their own limited liability company have been grouped with self-employment income, whereas the convention followed by the national accounts is to classify this income as wages and salaries. 2 This may account for the shortfall in wages and salaries in 1981-82. In 1985-86 the total is essentially the same in the two sources.

b) Recorded income from unincorporated enterprises is naturally expected to be under-recorded, on account of the explicit adjustment made in the national accounts for potential under-recording of self-employment income in taxation statistics. This is the case with the survey recording 84 per cent in 1985-86. On the other hand, the inclusion in 1981-82 of wages received by persons from their own limited liability company would cause the income survey to be too high - this apparently more than offsets the under­recording.

c) Income from interest is very much under-recorded, although this is offset to a small extent by an over­estimate of dividend income.

d) Only three-quarters (in 1981-82) or two-thirds (1985-86) of personal benefits payments to residents are recorded in the survey. Much of this difference stems from the inclusion of benefits such as medical and pharmaceutical benefits in the National Accounts measures. However, there is also some apparent under­recording of income support payments (see below).

Further details comparing the survey estimates of selected aggregate income support benefits received during the year with Department of Social Security (DSS) expenditure statistics are shown in Table A6.3. While the exclusion of the institutionalised population from the scope of the survey makes these two sources of data not entirely compatible, some problems of under-recording are apparent. Income from Unemployment Benefits, for example, is apparently under-recorded by $939m, or 30 per cent. Similar rates of under-recording apply to supporting parents benefit and widows pension. (The exclusion of the institutionalised will mainly affect the comparison with age and invalid pension expenditure.)

2. In the 1985-86 survey (AS85) the income survey coding has been changed to correspond to the National Accounts basis.

143

Table A6.2 Comparison of Australian Income Survey with National Accounts Aggregates, 1985-86

Income component National Income Ratio Accounts (NA) Survey Survey iNA

% $m $m

Wages and Salaries 112 931 113 662 100.6 Supplements to W&S 9 264 813 8.8 TOTAL wages and salaries 122 195 114 475 93.7

Gross Operating Surplus of Unincorporated Enterprises 13 958 11 683 83.7 Dwelling Rent (NA includes imputed rent of owner/occupiers) 3 912 1 320 33.7 Imputed Interest on Life and Superannuation Funds 6 326 na Superannuation Receipts na 2 230 Other Interest 15 045 7 788 51.8 Dividends 1450 3 215 221.7 TOTAL interest and dividends 16495 11 003 66.7

Personal Benefit Payments to Residents 22 932 15 223 66.4 Current Grants to Non-Profit Institutions 2 883 na Third Party Insurance Transfers 1443 na Unreq. Transfers from OS 1 820 250 13.7 Other Income Sources na 593 TOTAL HOUSEHOLD INCOME 191 964 156 777 81.7

Personal Income Tax 32 714 34 965 106.9 Other Transfers to Govt and OS 5 551 na TOTAL HOUSEHOLD DISPOSABLE INCOME 153 699 121 812 79.3

Source: 1986 Income Distribution Survey, Unit Record File, Australian National Accounts, National Income and Expenditure 1987-88, and SPRC calculations.

Table A6.3 Comparison with Department of Social Security Expenditure Aggregates

Pension/Benefit

Sickness Benefit Unemployment Benefit Supporting Parent Benefit Special Benefit Family Income Supplement Family Allowance Wifes/Carers Pension* Invalid Pension* Age Pension* Handicapped Child's Allowance Widows Pension TOTAL

DSS Expenditure

$m

392 3 122 1 238

108 49

1 537

I 674 5 897

31 925

14 973

Survey Income

$m

344 2 183

888 120 60

1 453 326

1 176 5 262

32 585

12429

Difference

$m %

-48 -12 -939 -30 -350 -28

12 11 11 21

-84 -5 326

-498 -30 -635 -11

1 4 -340 -37

-2,544 -17

Note: * In the DSS data, expenditure on Wifes/Carers Pension is included with expenditure on Invalid and Age Pensions (depending upon which pension the spouse is receiving).

Sources: DSS Annual Report, 1985-86, DSS Ten Year Statistical Summary, 1976 to 1986, 1986 Income Distribution Survey, Unit Record File.

Similar rates of under-recording apply to current data on numbers of persons in receipt of the different pensions or benefits. However, it is not known whether the discrepancy is mainly a result of non-recording of income receipts by respondents, or under-sampling of the pensioner/beneficiary population.

144

It may also be noted that the synthetic modelling of personal income tax seems to be reasonably successful in replicating national aggregates. The slight over-estimation in 1985-86 may reflect the fact that the survey recorded income tax liability for the year, whereas the National Accounts estimates are for payments. (This is a different, though similar, algorithm from that used in the LIS database).

Belgium3

The 1985 and 1988 Belgian surveys in the LIS dataset are respectively the first and second waves of a panel study of all private households in Belgium. The survey population includes resident foreigners, and excludes people in institutions, as well as persons without permanent address. It is estimated that the survey population covers more than 98 per cent of the total Belgian population.

About 13 per cent (822) of households had missing data on at least one income component. In about 300 of these cases, there is an estimate by the respondent of his or her total household income, and the missing income component was imputed from this, if the result appeared plausible. In other cases, missing data were imputed using estimates of average income within classes. Control variables were: age, sex, (former) profession, region and position in household for labour income and pensions; sex, position in household and having worked or not for other replacement incomes. For two categories (earned income of self-employed and white collar workers) a hot­deck method was used, in order not to reduce the variance. If family allowances were missing, these were estimated by a programme incorporating the administrative rules for granting family allowances.

No government survey exists in which household and/or individual earned incomes are measured. Comparisons with administrative data are very difficult, because they mostly measure gross income (before taxes), while the survey asks for net income.

Table A6.4 Comparison of Distribution of Net Taxable Income in Belgium: Tax Statistics Versus Estimates on the Basis of CSP-Survey Data

Net taxable yearly income (x 1.000 Bfr.)

0-100 100-250 250-350 350-500 500-600 600-800 800-1000 1000-1250 1250-1500 1500-2000 2000+ Total

Tax statistics for 1984

4.5% 8.9% 12.3% 23.2% 11.8% 17.3% 10.5% 6.0% 2.6% 1.8% 1.1%

100.0%

Source: I. Nicais, Methoden van Studiefinanciering, Dee! III, HIVA, Leuven, 1987, p. 7.

Estimates on basis of CSP-survey 1985

3.0% 12.6% 14.4% 20.0% 8.7% 13.8% 9.9% 7.6% 4.9% 3.0% 2.1%

100.0%

A research group at the University of Louvain (HIVA) has estimated the distribution of net-taxable incomes on the basis of the survey-data, and has compared this to official statistics (Table A6.4). It appears that both lower and higher incomes are over-represented. These differences may be due to:

a) Under-reporting of incomes to the tax authorities, and other forms of tax-evasion; b) Non-inclusion of certain people with low incomes in the official statistics; and c) Different definitions of the tax-unit.

3. The information provided here has been extracted from a document provided to LIS by Karel van den Bosch.

145

Canada

LIS data come from the Survey of Consumer Finances (SCF). Households in the monthly Labour Force Survey for April 1988 were sent income questionnaires by mail two weeks prior to the April Labour Force Survey. Respondents then completed the forms and gave their responses to interviewers by telephone. Participation in the survey was required by law. The privacy of respondents is assured by the Statistics Act Chapter 15, Statutes of Canada 1970-71-72. This act prohibits the disclosure by Statistics Canada of any particulars which could reveal the identity of any individual.

One income questionnaire was provided to each person 15 years of age and over. Each individual 15 years and over should complete his or her questionnaire. However, another individual (the household contact for the Labour Force Survey) may relate the completed information to the interviewer during the telephone interview. Respondents were asked, when applicable, to consult their income tax forms. It is not known what proportion of the respondents did this.The population living in institutions such as nursing homes for the aged and ill, prisons, etc., were not covered by the sampling frame; however, military personnel living in military housing or with their families were covered. Foreign residents (defined as neither Canadian born nor landed immigrants), persons with usual residence elsewhere, and those individuals living in seasonal dwellings, were excluded from the survey at the time of the interview.

Adjustments were made to income data in the Survey of Consumer Finances. In addition, hot deck imputation was performed by categorising responding and non-responding individuals by demographic and labour force characteristics, and using "donor" values to complete "recipient" records. No other adjustments for non-response were performed, nor were any values replaced by consulting alternative data sources.

Weights were assigned to each case in the Survey of Consumer Finances in order to insure that the sum of the survey weights are representative of independent population estimates by Province, age and sex categories. The assigned weights do not help adjust for missing income data or non-sampling errors related to income. (Missing income data were imputed through a hot-deck procedure.)

The Canadian Survey of Consumer Finances was compared with National Accounts. These estimates are reported in Table A6.5, and the findings may be summarised as follows:

Item

I. 2. 3.

5.

6. 7.

8.

9. 10. 11. 12.

Table A6.5 Comparison of SCF Estimates in Canada to Adjusted Personal Income in National Accounts, 1987

(Millions of dollars)

SCF SCP %NA

Adjusted

Wages and Salaries 100.0 259 852 Military Pay and Allowances 50.2 I 576 Farm Income 92.1 4 320 Non-farm Income from Self-employment 89.9 17 602 including Net Income from Roomers and Boarders TOTAL self-employment 90.4 21 922 Interest, Dividends and Miscellaneous 47.7 18 591 Investment Income Government Transfers (Sum of 7 to II) 75.5 39 841 A. Family Allowances 94.9 2 618 B. Child Tax Creditsb 89.0 I 361 C. Sales Tax Creditsb 110.0 329 Old Age Security and Guaranteed Income 95.8 12 582 Supplement Canada/Quebec Pension Benefits 87.6 7 667 Unemployment Insurance 74.9 7744 Other Government Transfers'· 47.1 7 540 Total (Sum of 1+2+3+4+5+6) 90.1 341 782

NA Adjusted

259 950 3 138 4690

19 572

24262 38 989

52 800 2 758 1 529

299 13 128

8 757 10 335 15 994

379 139

a. SCF = Social Assistance and Provincial Income Supplements plus Other Income from Government Sources plus Provincial Tax Credits. NA =Item 6- (Item 7+8+9+10).

b. The Child Tax Credits figures and Sales Tax Credits were obtained from Revenue Canada, not from National Accounts Statistics Canada.

c. SCF amounts include receipts for family units whose major source of income is military pay. These family units are not included in published tables.

Source: <<SCF/National Accounts Reconciliation.>> No date. Statistics Canada, Households Surveys Division.

146

a) Wage and salary income is fully covered; b) The shortfall for self-employment income was relatively small in 1987 by international standards; c) Investment income exhibited a considerable shortfall, which may reflect the fact that the top end of the

income distribution curve was under-represented in the sample; and d) There was a significant shortfall in government income sources that are not universal programmes,

including unemployment insurance.

Finland

The sampling frame employed in the Finnish Income Distribution Survey (IDS) is a taxation register. It was however 1 to 2 years out of date when the sample was drawn. Immigrants who entered the country after the register was established, and young people between 15 and 16 years are under-represented. The non-response rate was higher amongst one-person households than families with children, and higher in urban than in rural areas.

The survey estimates were compared with national accounts and administrative data. However, the household population in the IDS includes only private households while other statistics cover the whole population, including people living in institutions. In addition to this, the household composition is defined in the end of the survey year, so that the income of those who have died during the survey year is not included. These exclusions affect above all the aggregate estimates of pensions.The survey figures are compared with the external totals in Table A6.6. The main conclusions are:

a) Cash wages and salaries are identical in the IDS and tax records. The total sum of wages and salaries in the IDS exceeds that of national accounts because the IDS aims at estimating earnings in-kind at market values, while the national accounts report the taxation value. In a broad sense, other substitutes for wages and salaries such as benefit gained through the below-market interest lending by the employer, as well as possible over-compensation of employees' business travel or other costs should also be included, but these are covered neither by the IDS nor by the national accounts;

b) The coverage of entrepreneurial income is smaller than that of the national accounts. This is true for income from forestry because the IDS relies on interviews to collect this information. Income information from agriculture and other businesses is collected mainly from tax records. Aggregate income is less than that recorded as entrepreneurial income in the national accounts;

c) The survey estimate of property incomes suffers from both under- and over-coverage. The information on dividends, rents, royalties and patents relies on tax records and their coverage is comparable to the administrative source and to the National Accounts. The coverage of interest income in IDS is 32 per cent of the level in the national accounts. Interest income is mostly tax-free and data cannot be collected from administrative files, and is highly under-reported in the interviews. The imputed income from owner­occupied dwellings exceeds that of national accounts mainly because a different basis is used for estimating the gross rent of owner-occupied dwellings. IDS uses the government recommendations for moderate rents, which exceed the average rent level in the regulated Finnish market; and

d) There is a modest shortfall in government transfers, mainly attributed to the exclusion of people who live in institutions or who have died during the survey year.

Table A6.6 Comparison of Survey Estimates to National Accounts in Finland 1987

Item

1. Wages and Salaries 2. Self-employment Income 3. Property Income 4. Government Transfers All Income

Survey/ Admin.

Adjusted %

101.5 73.4 82.5 90.6 93.5

Source: Information supplied by A. Salomaki, Government Institute for Economic Research, Helsinki.

147

Survey (FIM

million)

177 807 25 018 12 876 54438

270 139

France

Table A6.7 Comparison of French Survey Estimates to National Accounts 1984 (Millions de francs)

National accounts of households

All households Ordinary households

ENQUETE/REVENUS FISCAUX (RF) Amount Coverage

(Income tax revenue survey) RF

Wages and Salaries 1 511 862 1 490 502 (Net of employees' social contribution) Unemployment

benefit 59 425 59425

Total 1 478 330 1 571 287 1 549 927 95.4%

Self Employment Income Agriculture 44 387 86 295 86 295 Industry/commerce !14 897 Non commercial 75 289 268 276 268 276 Managers & assoc. 2 279 Secondary Income 2564

Total 239 416 354 571 354 571 67.5%

Property Income Income from property (rent: urban + landed) 60 613 63 131 63 131 Income from capital 53 941 228 466 228 466

Total 114 254 291 597 291 597 39.2%

Government Transfer Pensions (incl. 516 435 old age minimum) 480 617 148 079 Maternity-Family !10 041 Daily sickness allowance Health (total) (excl. maternity and 279 379 work accident) 9 599 Employment (excl. unemployment benefits) 0 7 666 Other 0 3 191

954 750 919 219

Total 600 257 of which paid by public

administration 857 010 70.0% 890 163

Total 2 432 257 3 107 591 3 053 105 79.7%

Note : For all ordinary households (field of survey), estimates are very rough. This is why the decimal of the coverage rate is not significant.

Germany

The 1984 survey is the second wave of the Socio-Economic Panel (SOEP) study. Adjustments to the original data of the SOEP were made in order to improve data quality. New variables were generated in order to supplement the income data and correct non-plausible values among the different income components. In addition, hot-decking was employed in order to correct item non-response and to complete information. (See Berntsen, 1989, Section 2.3.2.)

An evaluation of the quality of income data was done by comparing SOEP data with income data from other sources [National Income Accounts, Income and Consumer Survey, DIW Calculations- see Berntsen (1989) and Kassella and Hochmuth (1989)]. The latter authors stress the difficulty in constructing appropriate national

148

Table A6.8 Comparison of German Survey Estimates to National Accounts Control Totals 1983

(Millions of DM)

Item

I. Wages and Salaries 2. Self-employment and property income 3. Transfers 4. Total Gross Income

Source: Kassella and Hochmuth, 1989, Table 14

SOEPas% National Accounts

108.8 36.3 50.6 76.9

SOEP

799 560 126 040 179 713

1 105 313

National Accounts

734 902 347 086 355 504

I 437 491

accounts aggregates in order to make a comparison which arises from the different concepts employed. The national accounts figures, for example, must be adjusted to exclude private non-profit organisations, and use is made of the work of Schuller (1986).4 Similarly, there are problems regarding the institutional population, which in principle is covered by the Panel data, but the response rate is low (Hanefeld, 1987, p. 166).

The results of the comparison are shown in Table A6.8, and the main conclusions are:

Ireland

Self-employment and property income are combined in the national accounts estimates used by Kassella and Hochmuth (1989). According to the separate figures given by Berntsen, the coverage of self­employment income was 35.5 per cent and property income 28.8 per cent. Kassella and Hochmuth note that the category "self employment and property income" is obtained in the national accounts as a residual, and is subject to considerable revision. But a sizeable part of the shortfall on self-employment income is attributed to the under-representation of high earners, and of property income to under-recording. There is substantial under-representation of transfer income. The detailed analysis of Berntsen (1989, Table 2) shows that the estimates for pensions (56.6 per cent) and unemployment insurance (37.4 per cent) were somewhat higher, while unemployment assistance ( 44.0 per cent) and social assistance (38.4 percent) were lower.

The survey sample, randomly selected from the electoral register, was based on persons residing in private households, thus excluding those living in institutions [see Callan, Nolan et al. (1989)].

Several checks were done to test the general reliability of income variables gathered in the survey. Since national accounts data are not available for the household sector, income checks centred mainly on comparisons with administrative statistics on income tax and social security transfers. The results are presented below.

The level and distribution of wage income is close to that reported by the Revenue Commissioners for the nearest fiscal year. This is shown in Table A6.9. (Note that the income cut-off of £5,000 is used because many incomes below this are not included in Revenue Commissioner statistics).

The main problems concern self-employment and investment income. Self-employment incomes are considered less reliable than wage income data. In particular, farm incomes had to be estimated from a special farm questionnaire. It should be noted that the reference year for farm income was 1986 - the low point for farm incomes in the 1980s. Investment income, and the tax yield on investment income, are underestimated.

Table A6.10 shows that the aggregate level of transfers was about 90 to 95 per cent of actual social security expenditures and, despite some misclassification of widow's pensions as old age pensions, its distribution across social welfare schemes is also close to official statistics.

4. The difference between the figures cited here in Table A6.6 and those in Berntsen (1989, Table 2) are that the latter uses a different set of national accounts estimates. The Berntsen table provides greater detail.

149

Table A6.9 Distribution of Tax Units over Ranges of «Total Income» in Ireland

Range of <<total income>> RC ESRI RC ESRI

More than Less than Number of cases Income Tax

£ £ ('OOOs) ('OOOs) £m £m

5 000 6000 60.2 63.9 36.8 31.9 6000 7 000 61.2 64.5 58.9 62.0 7 000 8 000 67.0 79.8 90.1 103.6 8 000 9 000 60.8 63.1 104.0 100.6 9 000 10000 53.4 58.9 111.7 117.9 1000 12500 100.7 107.6 278.1 277.1

12 500 15 000 70.8 62.1 260.6 217.1 15 000 17 500 47.0 46.8 216.4 209.9 17 500 20000 29.1 24.1 163.8 137.3 20000 25 000 30.7 37.5 222.8 274.2 25 000 30000 13.5 17.1 134.3 172.2 30 000 35 000 6.0 7.4 76.6 97.6 35 000 40000 2.4 4.4 37.7 72.2 40000 50000 1.6 2.1 32.5 43.0 Over 50000 1.2 2.1 48.4 60.5

Totals 605.5 641.4 1 872.6 1 977.1

Table A6.10 Expenditure on Social Welfare Schemes in Ireland

Social Welfare Scheme

Old Age

Illness

Old Age Contributory Pension & Retirement Pension Old Age Non-Contributory Pension & Blind Pension Widow's Contributory Pension Widow's & Orphan's Non-Contributory Pension Single Woman's Allowance

Disability Benefit Invalidity pension Injury Benefit Disablement Benefit Disabled Persons Maintenance Allowance Domiciliary Care Allowance

Unemployment Unemployment Benefit Unemployment Assistance

Family Income Support

Total

Orphan's Allowance (Contributory) Maternity Benefit Deserted Wife's Benefit Deserted Wife's Allowance Unmarried Mother's Allowance Prisoner's Wife's Allowance Supplementary Welfare Allowance Family Income Supplement

Actual expenditure on scheme

1986

£m

339.8° 258.3° 210.8

39.5 4.5

223.8 86.2

7.7 15.9 55.6

5.3

237.1 391.5

1.3 17.2 23.5 14.8 36.7

1.2 43.4

3.0 2 017.1

Estimated expenditure based on sample

Annual Current incomes incomes

£m £m

310.1 321.7 332.4 341.3 139.0 141.4 32.3 33.0

2.4 2.4

152.1 165.6 111.1 112.7

4.2 1.0 26.9 25.1 41.8 40.7

3.7 3.4

265.2 281.6 390.2 405.8

2.5 0.5 9.8 9.0

10.7 11.6 12.9 16.2 18.0 20.7

nil nil 12.7 21.9

1.4 1.7 I 879.4 1 955.2

Actual expenditure on scheme

1987

£m

363.5" 265.1° 223.6

42.7 4.5

218.1 95.9

7.8 20.2 57.7

5.5

236.3 417.2

1.3 20.0 28.0 17.0 44.0

nil 32.5 4.4

2 106.7

a Adjusted to exclude age group resident in institutions: if a greater (lesser) proportion of pensioners than of non-pensioners is resident in institutions, these numbers are overestimates (underestimates).

150

Italy

LIS data are taken from the Bank of Italy regular survey of family incomes. The quality of the data was evaluated by comparing the estimates of disposable income with the national accounts (Brandolini, 1993, and Brandolini and Cannari, 1993). For this purpose, the national accounts for the household sector are used,5 and two different methods to calculate disposable income.

The results for 1989 are shown in Table A6.11 and suggest that:

a) Total earnings appear to be well represented in aggregate, and if anything to be over-represented; b) Under-reporting is substantial for self-employment income. This may in part be explained by the use of

net operating surplus in the national accounts, and of cash withdrawals in the survey question; and c) The coverage of property income has changed over time, and with this there has been a change in the

apparent extent of shortfall. An improvement in the ratio of survey to national accounts in 1987 and 1989 is attributed in part to the use of estimated interest receipts (based on reported asset holdings) in place of the information supplied by the respondents.

The position is summarised by Brandolini as follows:

"There are significant differences between survey data and national accounts aggregates, though in the last years the quality of the survey has greatly improved" (1993, p. 32).

Brandolini and Cannari (1993) investigate how far the difference may be attributed to under-reporting by those declaring income from a source, as distinct from non-reporting of the source. Dividing the total income by the number of equivalent workers, or the number of pensions (for transfer income), they find that the result is to:

"tum the overestimation of earnings into an underestimation by about 10 per cent, and to reduce the shortfall of self-employment income to 29 per cent in 1987 and 44 per cent in 1989; the average pension, on the other hand, appears to be estimated correctly" (Brandolini and Cannari, 1993, p. 10).

Table A6.11 Comparison of Bank of Italy Survey Estimates to National Accounts Control Totals 1989

Item

1. Wages and Salaries after tax 2. Self-employment income after tax 3. Property income after tax 4. Transfers after tax 5. Total Disposable Income

(Billions of lire)

Bank of Italy as % of National Accounts

106.9 53.1 78.4 74.3 80.6

Bank of Italy

328 799 127 347 138 858 124 469 719 472

Note: Bank of Italy data using family-based grossing-up; national accounts using reconstructed taxes. Source: Brandolini (1993, Tables 12 and 13).

United Kingdom

National Accounts

307 582 239 894 177 107 167 548 892 131

The LIS dataset is derived from the Family Expenditure Survey (FES). According to the official report on the survey:

"It is thought that averages of household income recorded in the FES are too low, principally because certain forms of income, including investments, occupational pensions or self-employment, may be under­estimated" (Central Statistical Office, 1992, p. 74).

5. Brandolini discusses the choice between the total household sector and the "consumer households" sector, which excludes sole proprietorships with fewer than 20 employees.

151

Reference is made to the now dated study by Atkinson and Micklewright (1983), which is the source for Table A6.12. This study makes the following adjustments to the national accounts figures in order to reach more closely comparable totals:

- The national accounts estimate of self-employment income is lagged by the average delay in the reporting of self-employment income in the FES, and adjusted for items not included in taxable profits and for the alternative basis in calculating farm income (following Ramprakash, 1975).

- The national accounts estimate for occupational pensions is adjusted to exclude lump-sum payments, refunds of contributions and other payments by life insurance companies unrelated to pension business.

- The national accounts estimate for investment income is adjusted to exclude accrued interest on national savings certificates and lagged to allow for the difference in timing.

- The results of the comparison may be summarised as follows: - There is a moderate under-statement of earnings in the FES, which occurs despite the fact that pay receipts

are consulted in about 75 per cent of all cases (Kemsley et al., 1980). According to the official report, this is attributable to:

"understatement of earnings by women in part-time employment and an under-representation of the highest 1 per cent of earners"(Central Statistical Office, 1992, p. 74).

- There is a considerable shortfall for self-employment income. This is attributed tentatively in proportions one third/two thirds to a lower response rate by the self-employed and under-reporting by respondents (Atkinson and Micklewright, 1983, p. 41).

- The FES total for investment income is only about half that recorded in the national accounts estimate. It reflects a combination of factors, including under-reporting by respondents and under-representation of top incomes.

- There is understatement of occupational (private) pensions and, to a smaller extent, of state transfers. In the latter case, the shortfall was marked for sickness and injury benefit and for "other benefits."

Table A6.12 Comparison ofFES Estimates to National Accounts Control Totals in the United Kingdom 1977

(£million)

Item

1. Wages and Salaries 2. Self-employment income 3. Property income 4. Occupational pensions 5. Transfers 6. Total of lines 1-5

FES as% of National Accounts

93.7 75.7 50.6 74.5 90.9 89.0

Note: Figures refer to basic FES estimates with no age-weighting. Source: Atkinson and Micklewright (1983), Tables 2, 3, 4, 5 and 7.

United States

FES

69 260 5 664 2 897 2483

10 693 90997

National Accounts

73 934 7 479 5 722 3 334

11 759 102 228

The income data are derived from the March Current Population Survey (CPS). Some attempt is made to adjust the CPS for two of the three main sources of non-sampling error: household non-response and item non­response. No attempt is made to adjust for errors present in the responses. All missing survey responses were imputed using hot deck methodology. In the case of household non-response, the weights of interviewed households are adjusted upward by characteristics that were measurable for both non-interviewed and interviewed units. Since very little information is available for the non-interviewed households (only location and race of head), this adjustment has limited effect.

152

A comparison has been made of CPS estimates of aggregate income and independent estimates derived from the National Income and Product Accounts, data from the Social Security Administration, and other sources. The independent estimates have been adjusted to attempt to remove income received by the institutional population, by the deceased, and by persons not resident in the United States at the time of interview; any "in-kind" payments; and any lump-sum payments.

The results are summarised below for the different categories:

a) Wage and salary income has a total close to 100 per cent; b) Self-employment income is below the independent estimate, particularly for farm income, although the

Census Bureau comments that "the apparent extreme under-reporting problem for farm income helps to illustrate the difficulty in developing meaningful independent estimates" (U.S. Department of Commerce, 1991, p. 215). The estimate shown in Table A6.13 is based on data from the U.S. Department of Agriculture, whereas the farm income reported on tax returns was only $10.5 billion, on which basis the CPS reported figure would be an over-statement;

c) Property income is seriously under-reported; and d) Social security is relatively well-reported, but there are significant problems with the means-tested

programmes and with unemployment compensation, veterans' payments and with workers' compensation. Private pensions appear to be seriously under-reported. On the other hand, the Census Bureau suggests that this may partly be due to the inclusion of lump-sum payments in the independent estimate.

The relative accuracy of the different categories is generally in line with the extent of item non-response, as shown in Table A6.14. This indicates the proportion of respondents with income from the source who failed to supply information on the amount. At the same time, the item non-response is not negligible for wages and salaries. This raises the question of the accuracy of the imputation. For item non-responses, imputation procedures are used to replace the non-response to the question by an answer that was characteristic of other households with similar characteristics. Information provided to the LIS by John Coder of the U.S. Bureau of the Census shows that the current imputation procedures correct only slightly for this bias, due to the fact that non-respondents have, on average, higher levels of income than respondents.

Table A6.13 Comparison of March Current Population Survey (CPS) Estimates of Aggregate Income with Independent Estimates 1987

(Billions of dollars)

Item CPS/ CPS Independent Independent estimate

%

1. Wages and Salaries 99.4 2 202.4 2 215.9 2. Non-farm self-employment 78.5 172.5 219.8 3. Farm self-employment 33.5 15.7 46.9 4. TOTAL self-employment (2+3) 70.6 188.2 266.7 5. Interest income 55.2 134.9 244.4 6. Dividend income 52.7 38.8 73.7 7. Net rents and royalties 72.5 29.2 40.3 8. TOTAL property income (5+6+7) 56.6 202.9 358.4 9. Social security 92.3 178.7 193.6 10. Supplemental Security Income 82.8 9.5 11.5 11. Aid to Families with Dependent Children 72.8 11.9 16.4 12. Veterans' payments 68.5 9.7 14.2 13. Unemployment compensation 74.6 10.4 14.0 14. Workers' compensation 64.8 9.2 14.2 15. TOTAL government transfers 86.9 229.4 263.9

(9+10+1 1+12+13+14) 16. Private pensions 46.0 57.8 125.8 17. Federal and military pensions 99.1 40.3 40.7 18. State and local government pensions 78.8 20.3 25.7 19. TOTAL occupational pensions 61.6 ll8.4 192.2 20. TOTAL (1+4+8+15+19) 89.2 2 941.4 3 297.1

Source: U.S. Department of Commerce (1991), p. 216.

153

Evidence about the accuracy of reporting of wages and salaries in the CPS is provided by the Census Bureau study which has compared the reported wage and salary income for an exact match of CPS and income tax returns (for married couples filing joint returns in 1985). This showed that the mean income reported in the CPS was 98.0 per cent of that in the tax returns, but that there was the result of errors in both directions. Although 51 per cent reported within 5 per cent of the tax return, 19 per cent had differences of more than 20 per cent.

Table A6.14 Non-Response Rates by Category oflncome in March Current Population Survey (CPS)

Item

Wages and Salary Non-farm self-employment Farm self-employment Interest income Dividend income Rental income Social Security Aid to Families With Dependent Children Unemployment compensation Veteran's payments Private pensions Public pensions Child support

Non-response Rate for Amount

17.6 24.2 25.3 28.7 33.4 28.3 19.2 15.0 15.0 20.2 20.8 20.4 14.5

Source: Information supplied to LIS by John Coder of the US Bureau of the Census.

154

Appendix 7

BACKGROUND TABLES ON INCOME SOURCES

AND DEMOGRAPHIC CHARACTERISTICS

The first two tables in this appendix (Tables A7.1 and A7.2) present information on the composition of income by deciles of equivalent and by income levels phrased as a percentage of median equivalent income.

The following tables describe characteristics of the total samples in the LIS database. In Table A7.3 information is presented on the percentage of household heads under 60 years of age and 60 years and older, and the percentage of households with some earnings (wage, salary and self-employment income combined) and with no earnings.

Table A 7.4 presents the percentage distribution of the six family types in the database -unmarried male heads without children, married couples without children, unmarried female heads without children, unmarried male heads with children, married couples with children, and unmarried female heads with children.

Both tables present this information for the total sample, and for low income households (less than 50 per cent of median equivalent income) and those with modest incomes (between 50 per cent and 70 per cent of median income).

155

Table A7.1 Percentage of Income Sources of Gross Income by Decile

Country Income Primary Other Private Social Taxes Decile Income Income Transfers

ASS! I JS.7 14.3 67.0 1.7 ASS! 2 37.7 15.0 47.3 5.3 ASS! 3 63.S 16.6 19.6 11.9 ASS! 4 77.6 12.3 10.1 16.2 ASS! 5 S2.3 11.4 6.4 IS.9 ASS! 6 S4.4 9.9 5.7 20.6 ASS! 7 S3.4 11.9 4.6 21.7 ASS1 s S4.7 11.6 3.7 22.S ASS1 9 S2.S 14.5 2.7 24.5 ASS1 10 76.S 21.9 1.4 29.S ASS5 1 23.9 s.o 67.5 0.6 ASS5 2 37.0 12.0 47.1 3.9 ASS5 3 60.3 10.1 20.5 9.1 ASS5 4 70.5 6.6 s.s 14.1 ASS5 5 72.6 5.6 5.S 16.0 ASS5 6 73.2 4.9 4.6 17.2 ASS5 7 73.1 4.S 4.2 1S.O ASS5 s 72.1 6.4 2.S 1S.7 ASS5 9 72.4 6.1 1.9 19.6 ASS5 10 62.0 12.6 O.S 24.6 CNS1 1 3S.6 7.5 53.9 1.4 CNS1 2 60.3 7.S 31.9 5.4 CNS1 3 75.4 7.2 17.4 9.1 CNS1 4 Sl.3 7.3 11.4 12.2 CNS1 5 S5.0 6.0 S.9 13.4 CNS1 6 S7.4 5.4 7.2 14.7 CNS1 7 SS.5 6.3 5.2 15.9 CNS1 s S9.0 6.6 4.3 16.7 CNS1 9 SS.5 S.3 3.3 17.4 CNSJ 10 S3.9 14.0 2.1 19.7 CNS7 I 44.1 5.9 50.0 11.9 CNS7 2 60.0 6.5 33.5 7.S CNS7 3 70.2 7.4 22.4 11.6 CNS7 4 S0.6 5.1 14.3 14.4 CNS7 5 S3.0 5.7 11.3 16.S CNS7 6 S4.2 6.7 9.1 17.5 CNS7 7 S7.4 6.0 6.5 19.1 CNS7 s S7.S 6.S 5.4 20.3 CNS7 9 SS.2 7.S 4.1 21.2 CNS7 10 S7.1 10.7 2.2 24.0 FR79 1 50.S -0.6 49.S 1.9 FR79 2 56.3 1.3 42.3 0.7 FR79 3 64.9 J.S 33.2 1.1 FR79 4 71.7 1.0 27.3 1.7 FR79 5 75.4 1.2 23.4 2.7 FR79 6 so.s 1.5 17.7 3.9 FR79 7 83.0 1.3 15.7 5.0 FR79 s 85.0 1.6 13.4 6.4 FR79 9 85.7 2.5 l1.8 8.6 FR79 10 84.7 6.9 S.5 1S.8 FRS4 1 45.5 -1.2 55.7 4.0 FR84 2 52.7 1.4 45.9 0.5 FR84 3 60.7 1.7 37.6 0.8 FR84 4 65.6 2.4 32.0 1.5 FR84 5 73.3 1.9 24.8 2.6 FR84 6 7S.1 2.0 19.9 3.4 FR84 7 81.3 2.5 16.2 4.7 FR84 8 80.7 3.5 15.8 6.4 FRS4 9 82.9 3.3 13.8 8.8 FR84 10 77.1 12.1 10.7 18.5 GE84 1 43.4 5.2 51.4 10.5 GES4 2 73.9 1.9 24.1 19.2 GE84 3 71.8 2.8 25.4 17.4 GE84 4 76.3 2.9 20.8 21.0 GES4 5 84.7 1.8 13.5 22.9 GE84 6 S5.1 2.4 12.5 24.3

156

Table A7.1 (contd) Percentage of Income Sources of Gross Income by Decile

Country Income Primary Other Private Social Taxes Decile Income Income Transfers

GE84 7 83.1 2.0 14.9 24.8 GE84 8 85.9 3.4 10.7 26.1 GE84 9 88.3 3.5 8.2 28.4 GE84 10 83.6 10.5 5.9 30.3 IR87 1 11.9 6.2 72.8 9.0 IR87 2 29.4 2.5 63.8 4.3 IR87 3 50.2 2.9 38.2 8.7 IR87 4 57.3 2.4 29.3 10.9 IR87 5 63.6 2.9 19.8 13.7 IR87 6 65.1 2.8 16.9 15.2 IR87 7 69.6 3.3 11.0 16.2 IR87 8 71.1 2.5 7.6 18.8 IR87 9 72.4 2.8 5.0 19.8 IR87 10 72.9 5.2 2.2 19.7 IT86 I 55.2 1.2 43.6 0 IT86 2 68.6 1.3 30.1 0 IT86 3 70.3 1.5 28.2 0 IT86 4 79.2 1.6 19.2 0 IT86 5 77 2.1 20.9 0 IT86 6 76.8 2.5 20.8 0 IT86 7 80 2.6 17.4 0 IT86 8 81.7 3.2 15.1 0 IT86 9 81.4 5.2 13.4 0 IT86 10 78.6 8.6 12.8 0 LX85 I 52.1 2.7 45.2 4.8 LX85 2 71.9 0.8 27.3 7.7 LX85 3 71.2 1.5 27.2 7.7 LX85 4 73.7 2.3 24 7.3 LX85 5 71.3 3.5 25.2 6.8 LX85 6 80.2 1.7 18.1 7.8 LX85 7 76 3.8 20.2 7.5 LX85 8 76.5 3.5 20 6.6 LX85 9 78.4 4.9 16.7 6.2 LX85 10 81.7 7 11.3 6.3 NL83 1 34.9 4.6 60.5 21.3 NL83 2 54.1 1.5 44.5 26.8 NL83 3 63.5 1.3 35.2 28.4 NL83 4 76.4 2.8 20.8 32.1 NL83 5 79.2 2.8 18.1 33.1 NL83 6 78.5 4 17.5 33.7 NL83 7 78.5 5.6 15.9 33.4 NL83 8 78.9 7 14 34.3 NL83 9 84.3 5.2 10.5 35.7 NL83 10 85.6 8.6 5.8 38.9 NL87 I 52.8 3.8 43.3 45.6 NL87 2 61.8 1.2 37 36.3 NL87 3 60.5 2.1 37.3 27.2 NL87 4 79.5 2.9 17.6 33.3 NL87 5 81.6 2.8 15.6 34.4 NL87 6 80.1 4.5 15.4 37.5 NL87 7 83 4.1 13 37.1 NL87 8 83.3 5 11.7 37.9 NL87 9 85.4 4.3 10.3 38.5 NL87 10 84.8 8.5 6.7 43.9 NW79 1 30.2 11.5 58.3 11.3 NW79 2 59.2 4.7 36.1 13 NW79 3 79.4 3.3 17.3 18.6 NW79 4 84.6 3.3 12.1 21.9 NW79 5 87.1 3.3 9.5 24 NW79 6 89.1 2.9 8 25.7 NW79 7 91.1 2.8 6.1 27.8 NW79 8 91.4 3.1 5.5 28.2 NW79 9 90.9 3.7 5.4 29.2 NW79 10 91.6 4.9 3.6 31 SW81 I 64.1 -1.3 37.2 34.8 SW81 2 48.9 4.5 46.6 23.2

157

Table A 7.1 ( contd) Percentage of Income Sources of Gross Income by Decile

Country Income Primary Other Private Social Taxes Decile Income Income Transfers

SW81 3 52.2 3.4 44.5 24.5 SW81 4 63.3 2.1 34.6 26.8 SW81 5 71.7 1.7 26.6 27.3 SW81 6 72.6 1.7 25.7 28.6 SW81 7 76.3 1.5 22.2 30 SW81 8 78.9 1.3 19.8 30.1 SW81 9 83 0.8 16.3 31.1 SW81 10 85.1 1.8 13 34.8 SW87 1 53.4 11.1 35.5 35.6 SW87 2 42.1 8.2 49.7 22.2 SW87 3 46.8 5.5 47.7 25.1 SW87 4 55.3 4.4 40.4 27.5 SW87 5 66.2 3.4 30.4 28.8 SW87 6 70.1 2.8 27.1 29.5 SW87 7 74.7 2.6 22.7 30.4 SW87 8 76.8 2.8 20.4 31.9 SW87 9 81.8 2.5 15.7 33.7 SW87 10 83.5 5.8 10.7 38.2 SZ82 1 46.6 12.8 40.6 26.1 SZ82 2 73.9 7.8 18.4 12.2 SZ82 3 83.5 6.8 9.7 13.8 SZ82 4 88 6 6 14.8 SZ82 5 87.4 7 5.6 15.7 SZ82 6 88.9 6.9 4.2 16.4 SZ82 7 90.4 6.2 3.4 17.8 SZ82 8 90.1 6.8 3 17.7 SZ82 9 88.8 8.9 2.3 19.2 SZ82 10 84.5 14.2 1.3 20.8 UK79 1 25.9 6.6 67.5 11.5 UK79 2 45.6 6.2 48.2 9.4 UK79 3 68.8 5.5 25.7 15.0 UK79 4 78.3 3.8 17.9 18.2 UK79 5 80.3 4.8 14.9 18.3 UK79 6 83.5 3.8 12.7 20.0 UK79 7 84.4 4.3 11.3 19.9 UK79 8 86.0 4.6 9.3 20.9 UK79 9 87.9 4.3 7.8 22.4 UK79 10 89.6 5.9 4.5 21.7 UK86 1 22.2 8.4 69.4 25.0 UK86 2 28.8 5.3 65.8 7.8 UK86 3 47.0 6.6 46.4 12.6 UK86 4 62.4 6.9 30.8 16.9 UK86 5 71.2 6.7 22.1 19.9 UK86 6 75.6 7.3 17.2 20.6 UK86 7 78.3 8.0 13.7 22.4 UK86- 8 82.9 7.1 10.0 23.5 UK86 9 86.4 7.1 6.5 25.2 UK86 10 85.4 10.3 4.3 25.8 US79 1 41.8 10.3 47.9 10.3 US79 2 66.2 5.6 28.3 8.8 US79 3 79.6 6.2 14.2 11.6 US79 4 84.8 6.0 9.2 14.7 US79 5 87.1 6.0 6.9 16.2 US79 6 88.3 5.6 6.1 18.5 US79 7 90.0 5.6 4.4 20.0 US79 8 90.6 5.8 3.6 22.4 US79 9 90.2 6.9 2.8 23.8 US79 10 84.2 13.0 2.8 28.4 US86 1 49.9 5.3 44.8 22.1 US86 2 63.8 5.6 30.6 10.7 US86 3 77.9 6.5 15.6 12.3 US86 4 83.9 6.5 9.6 14.2 US86 5 86.3 6.8 6.8 17.1 US86 6 86.0 7.8 6.2 18.4 US86 7 87.4 7.6 4.9 19.8 US86 8 87.4 8.2 4.5 20.8 US86 9 88.2 9.1 2.7 23.4 US86 10 82.7 15.2 2.0 28.8

158

Table A7.2 Percentage of Income Sources of Gross Income by Income Level (Percentage of Median Equivalent Income)

Percent of Gross Income

Income Primary Other Social Country Level Income Priv. Income Transfers Taxes

Australia 1981 <50 22.8 13.5 63.7 2.2 >=50 <70 49.5 16.4 34.2 8.2 >=70 <100 78.8 12.4 8.8 17.1 >=100 <150 84.3 11.0 4.7 21.7 >= 150 79.5 18.6 2.0 27.4

1985 <50 24.5 8.0 66.4 1.0 >=50 <70 50.1 11.2 32.0 6.6 >=70 <100 71.2 6.4 7.6 14.8 >=100 <150 72.9 5.3 3.8 17.9 >= 150 66.1 10.0 1.3 22.5

Belgium 1985 <50 32.9 -9.9 77.0 n.a >=50 <70 43.0 1.1 55.9 n.a >=70 <100 62.5 0.9 36.6 n.a >=100 <150 79.3 1.3 19.4 n.a >= 150 85.0 4.1 10.9 n.a

1988 <50 22.3 2.4 75.2 n.a >=50 <70 42.4 1.6 55.9 n.a >=70 <100 61.6 1.2 37.2 n.a >=100 <150 78.0 1.7 20.3 n.a >= 150 78.5 6.7 14.8 n.a

Switzerland 1982 <50 48.3 13.5 38.2 33.3 >=50 <70 71.5 8.0 20.4 11.8 >=70 <100 86.8 6.6 6.6 14.9 >=100 <150 89.8 6.8 3.5 17.4 >= 150 85.8 12.7 1.5 20.4

Canada 1981 <50 43.8 7.4 48.8 2.1 >=50 <70 67.2 7.7 25.1 7.0 >=70 <100 82.5 6.7 10.8 12.4 >=100 <150 88.4 6.2 5.4 15.9 >= 150 85.8 11.6 2.6 18.7

1987 <50 46.0 6.5 47.4 10.2 >=50 <70 64.2 7.1 28.8 9.3 >=70 <100 81.1 5.5 13.4 15.4 >=100 <150 86.8 6.5 6.7 19.1 >= 150 87.4 9.6 3.0 22.9

France 1979 <50 48.2 -1.2 53.0 2.5 >=50 <70 58.3 1.3 40.4 0.7 >=70 <100 71.6 1.3 27.0 2.0 >=100 <150 83.1 1.5 15.4 5.2 >= 150 85.0 5.4 9.6 15.3

1984 <50 40.9 -2.6 61.7 6.2 >=50 <70 53.8 1.4 44.8 0.5 >=70 <100 67.9 2.1 30.0 1.8 >=100 <150 80.2 2.7 17.1 4.9 >= 150 79.2 8.8 12.0 14.8

Germany 1984 <50 38.8 6.6 54.6 9.9 >=50 <70 69.4 2.2 28.4 17.1 >=70 <100 78.5 2.5 19.0 21.0 >=100 <150 85.2 2.7 12.1 25.5 >= 150 84.9 8.4 6.7 29.9

Ireland 1987 <50 11.4 6.1 74.1 8.4 >=50 <70 40.3 2.5 50.9 6.3 >=70 <100 60.0 2.8 24.9 12.3 >=100 <150 68.5 2.9 12.3 16.3 >= 150 72.4 4.0 3.9 19.6

159

Table A 7.2 ( contd) Percentage of Income Sources of Gross Income by Income Level (Percentage of Median Equivalent Income)

Percent of Gross Income

Income Primary Other Social Country Level Income Priv. Income Transfers Taxes

Italy 1986 <50 56.1 1.1 42.8 0.0 >=50 <70 71.3 1.3 27.4 0.0 >=70 <100 76.0 1.9 22.1 0.0 >=100 <150 79.3 2.7 18.0 0.0 >= 150 80.1 6.9 13.0 0.0

Luxembourg 1985 <50 43.5 3.4 53.1 3.9 >=50 <70 69.9 1.0 29.0 7.4 >=70 <100 71.6 2.6 25.8 7.2 >=100 <150 77.0 3.4 19.5 7.2 >= 150 81.1 6.0 12.9 6.3

Netherlands 1983 <50 39.4 5.4 55.2 20.8 >=50 <70 44.8 2.5 52.6 25.4 >=70 <100 73.1 2.2 24.6 31.1 >=100 <150 78.8 5.6 15.6 33.8 >= 150 84.7 7.2 8.1 37.5

1987 <50 71.3 2.1 26.6 62.3 >=50 <70 53.2 2.3 44.5 37.1 >=70 <100 74.4 2.7 22.9 32.1 >=100 <150 82.1 4.6 13.2 37.5 >= 150 85.1 6.7 8.2 41.7

Norway 1979 <50 20.7 21.6 57.7 16.2 >=50 <70 52.2 5.0 42.8 11.8 >=70 <100 83.7 3.4 12.9 21.6 >=100 <150 90.6 3.2 6.3 27.7 >= 150 91.7 4.6 3.8 30.9

Sweden 1981 <50 79.1 -8.1 29.1 44.6 >=50 <70 50.1 4.6 45.4 25.2 >=70 <100 62.3 2.4 35.2 26.0 >=100 <150 78.1 1.3 20.6 30.1 >= 150 85.0 1.8 13.2 34.5

1987 <50 57.3 11.3 31.4 38.8 >=50 <70 43.0 8.6 48.4 23.5 >=70 <100 57.1 4.3 38.6 27.3 >=100 <150 76.2 2.7 21.1 31.5 >= 150 83.6 5.7 10.7 38.1

United Kingdom 1979 <50 24.1 6.7 69.3 11.6 >=50 <70 53.0 6.0 41.0 11.1 >=70 <100 77.9 4.5 17.6 17.9 >=100 <150 84.9 4.2 10.9 20.3 >= 150 88.8 5.3 5.8 22.0

1986 <50 23.2 8.9 68.0 27.6 >=50 <70 33.8 5.9 60.2 9.3 >=70 <100 66.4 6.7 26.9 18.2 >=100 <150 78.6 7.6 13.8 22.2 >= 150 85.7 8.8 5.6 25.4

United States 1979 <50 53.8 6.9 39.3 7.8 >=50 <70 76.8 6.3 16.9 11.9 >=70 <100 85.9 6.0 8.0 15.3 >=100 <150 89.7 5.6 4.7 20.3 >= 150 86.9 10.3 2.8 26.4

1986 <50 57.2 5.3 37.4 15.6 >=50 <70 76.9 6.5 16.6 11.9 >=70 <100 85.4 6.7 7.9 15.9 >=100 <150 87.0 7.8 5.2 19.5 >= 150 85.2 12.1 2.6 25.8

Note: n.a = not available.

160

Table A7.3 Age and Earning Status Distribution for Total Sample and Persons with Low and Modest Incomes

Age Earning Status

Total Sample: Under 60 60 and Over None Some

Australia 1981 83.7 16.3 21.2 78.8 1985 81.6 18.4 18.8 81.2

Belgium 1985 80.0 20.0 22.0 78.0 1988 78.0 22.0 22.4 77.6

Canada 1981 83.3 16.7 10.9 89.1 1987 82.3 17.7 12.4 87.6

France 1979 80.1 19.9 13.1 86.9 1984 80.0 20.0 16.1 83.9

Germany 1984 76.3 23.7 18.9 81.2

Ireland 1987 79.3 20.7 20.4 79.6

Italy 1986 76.5 23.6 16.7 83.3

Luxembourg 1985 81.1 18.9 17.8 82.3

Netherlands 1983 81.9 18.1 26.9 73.1 1987 81.4 18.6 25.8 74.2

Norway 1979 77.1 22.9 12.5 87.5

Sweden 1981 74.4 25.6 17.0 83.0 1987 74.8 25.3 17.9 82.1

Switzerland 1982 78.6 21.4 14.9 85.1

United Kingdom 1979 78.8 21.2 18.0 82.0 1986 77.3 22.7 27.9 72.1

United States 1979 81.5 18.5 13.0 87.0 1986 80.6 19.4 14.6 85.4

Low Income: Australia 1981 76.9 23.1 70.1 29.9

1985 72.1 27.9 67.8 32.2

Belgium 1985 64.9 35.1 69.0 31.0 !988 58.0 42.0 77.5 22.5

Canada 1981 75.4 24.6 40.7 59.3 1987 83.3 16.7 37.7 62.3

France 1979 74.1 25.9 28.2 71.8 1984 80.9 19.1 37.7 62.3

Germany 1984 68.6 31.4 58.5 41.5

Ireland 1987 82.7 17.3 62.3 37.7

Italy 1986 65.3 34.7 39.2 60.8

Luxembourg 1985 59.1 40.9 56.1 43.9

Netherlands !983 86.8 13.2 78.8 21.2 1987 90.7 9.3 47.6 52.4

Norway 1979 75.3 24.7 35.7 64.3

Sweden 1981 85.4 14.6 23.0 77.0 1987 78.4 21.6 24.1 75.9

Switzerland !982 58.1 41.9 54.3 45.7

United Kingdom 1979 60.1 39.9 73.0 27.0 1986 83.2 16.8 66.6 33.4

United States 1979 74.6 25.4 40.5 59.5 1986 78.1 21.9 39.6 60.4

Modest Income: Australia 1981 63.8 36.2 45.6 54.4

1985 60.3 39.7 44.3 55.7

Belgium 1985 70.6 29.4 50.3 49;7 1988 66.3 33.7 51.2 48.8

Canada 1981 73.5 26.5 22.8 77.2 1987 68.5 31.5 25.7 74.3

161

Table A7.3 (contd) Age and Earning Status Distribution for Total Sample and Persons with Low and Modest Incomes

Age Earning Status

Modest Income (contd): Under 60 60 and Over None Some

France 1979 69.6 30.4 25.2 74.8 1984 71.8 28.2 25.5 74.5

Germany 1984 69.1 31.0 30.2 69.8

Ireland 1987 77.2 22.8 46.8 53.2

Italy 1986 73.4 26.6 26.9 73.1

Luxembourg 1985 79.9 20.1 25.6 74.4

Netherlands 1983 82.5 17.5 56.1 43.9 1987 86.0 14.0 53.5 46.5

Norwaye 1979 49.7 50.3 47.0 53.0

Sweden 1981 51.1 48.9 42.6 57.4 1987 46.8 53.2 48.3 51.7

Switzerland 1982 62.5 37.5 32.2 67.9

United Kingdom 1979 54.1 45.9 46.1 53.9 1986 61.9 38.1 66.2 33.8

United States 1979 75.1 24.9 19.0 81.0 1986 76.7 23.3 19.1 80.9

162

Table A7.4 Family Type Distribution for Total Sample and Persons with Low and Modest Incomes

Unmarried Married Unmarried Unmarried Married Unmarried male head couple female head male head couple female without without without with with head with

Total Sample: children children children children children children

Australia 1981 5.8 24.5 8.0 0.7 55.7 5.3 1985 7.2 25.9 8.4 1.9 52.3 4.3

Belgium 1985 n.a. n.a. n.a. n.a. n.a. n.a. 1988 4.0 44.1 5.7 0.8 44.3 1.1

Canada 1981 5.6 24.7 8.5 0.9 54.5 5.8 1987 6.8 28.1 9.2 0.7 50.3 4.9

France 1979 4.1 27.7 7.9 1.0 56.7 2.6 1984 4.7 29.2 8.4 1.0 53.6 3.1

Germany 1984 5.8 36.2 11.6 0.3 44.1 2.0

Ireland 1987 4.9 16.3 7.0 1.4 66.8 3.6

Italy 1986 3.5 34.3 6.6 0.4 53.0 2.1

Luxembourg 1985 4.2 31.2 8.4 0.5 52.1 3.4

Netherlands 1983 4.2 30.1 7.3 0.5 54.8 3.1 1987 6.1 30.1 9.7 0.5 50.0 3.6

Norway 1979 5.5 16.1 8.2 4.0 60.3 5.9

Sweden 1981 13.5 26.4 13.5 0.6 41.0 5.1 1987 14.3 27.7 14.0 0.8 38.6 4.5

Switzerland 1982 10.1 23.1 14.9 2.1 46.7 3.0

United Kingdom 1979 4.4 26.0 8.7 0.9 55.1 4.8 1986 6.3 29.6 9.3 1.1 47.8 5.8

United States 1979 6.2 24.0 9.0 1.0 49.7 10.0 1986 6.3 26.5 10.0 1.3 46.8 9.1

Low Income:

Australia 1981 9.9 9.0 18.9 0.9 38.9 22.3 1985 10.9 10.5 20.8 1.4 35.6 20.9

Belgium 1985 n.a. n.a. n.a. n.a. n.a. n.a. 1988 8.3 37.6 21.0 1.2 30.6 1.3

Canada 1981 9.3 11.3 19.2 1.2 38.5 20.5 1987 11.9 11.5 16.8 1.7 38.8 19.2

France 1979 8.1 21.0 17.3 2.0 44.0 7.4 1984 7.7 23.4 13.4 0.9 46.0 8.5

Germany 1984 11.5 24.4 26.4 0.0 27.8 9.9

Ireland 1987 8.2 8.7 5.2 0.8 68.4 8.7

Italy 1986 4.8 24.4 12.6 0.2 54.5 3.6

Luxembourg 1985 4.0 28.5 17.2 1.3 38.5 10.5

Netherlands 1983 9.1 25.3 11.3 0.4 50.2 3.7 1987 15.9 12.2 19.1 0.9 45.9 6.0

Norway 1979 12.4 10.5 17.3 9.2 31.7 18.9

Sweden 1981 35.7 9.0 18.4 0.2 28.9 7.7 1987 37.7 7.9 34.4 0.6 15.9 3.4

Switzerland 1982 26.0 11.9 37.4 3.0 12.8 8.9

United Kingdom 1979 6.9 21.0 20.6 1.4 37.9 12.2 1986 5.6 15.6 10.1 1.8 54.3 12.6

United States 1979 8.0 12.3 16.9 1.3 29.5 32.1 1986 6.7 10.9 16.6 1.9 35.5 28.8

Modest Income:

Australia 1981 4.6 31.4 9.2 0.7 48.5 5.6 1985 4.8 35.9 8.6 1.9 44.5 4.3

Belgium 1985 n.a. n.a. n.a. n.a. n.a. n.a. 1988 4.4 41.5 10.2 1.3 39.7 3.0

Canada 1981 6.2 22.6 10.5 0.9 52.1 7.7 1987 6.7 22.3 16.4 0.7 47.4 6.4

163

Table A7.4 (contd) Family Type Distribution for Total Sample and Persons with Low and Modest Incomes

Unmarried Married Unmarried Unmarried Married Unmarried male head couple female head male head couple female

without without without with with head with Modes Income (contd): children children children children children children

France 1979 4.4 25.0 11.0 0.6 55.5 3.4 1984 4.1 23.7 11.6 1.8 53.7 5.2

Germany 1984 4.7 21.5 19.4 0.8 49.4 4.2

Ireland 1987 5.3 9.1 10.8 1.1 69.0 4.7

Italy 1986 2.0 31.4 7.8 0.5 56.8 1.5

Luxembourg 1985 2.3 18.4 10.9 0.5 63.7 4.3

Netherlands 1983 3.4 21.0 7.2 0.3 56.1 12.1 1987 3.5 14.0 8.8 1.2 57.7 14.7

Norway 1979 9.7 18.2 25.7 5.5 31.6 9.3

Sweden 1981 18.1 17.1 29.8 0.3 26.4 8.4 1987 19.5 18.3 36.8 0.6 18.2 6.5

Switzerland 1982 6.8 19.0 22.3 2.4 46.0 3.5

United Kingdom 1979 5.0 27.9 17.2 0.9 42.4 6.7 1986 6.0 24.7 16.0 1.1 38.7 13.4

United States 1979 6.6 16.2 11.4 1.6 50.3 13.9 1986 6.3 19.6 12.4 1.5 50.0 10.1

164