TO P I N C O M E S OV E R T H E T W E N T I E T H C E N T U RY
Top Incomes over the Twentieth Century:
A Contrast Between European and English-Speaking Countries
Edited by A . B. AT K I N S O N , NuYeld College, Oxford,
and T. P I K E T T Y,
PSE, Paris
1
3
Great Clarendon Street, Oxfordox26dp
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto With oYces in
Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York
!Oxford University Press, 2007 The moral rights of the author have been asserted
Database right Oxford University Press (maker) First published 2007
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above You must not circulate this book in any other binding or cover
and you must impose the same condition on any acquirer British Library Cataloging in Publication Data
Data available
Library of Congress Cataloguing in Publication Data Data available
Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain
on acid-free paper by Biddles Ltd, King’s Lynn, Norfolk ISBN 0–19–928688–4 978–0–19–928688–1
1 3 5 7 9 10 8 6 4 2
Contents
List of Figures, Tables, and Boxes vi
Preface xv
1 Top Incomes over the Twentieth Century:
A Summary of Main Findings 1
T. Piketty
2 Measuring Top Incomes: Methodological Issues 18
A. B. Atkinson
3 Income, Wage, and Wealth Inequality in
France, 1901–98 43
T. Piketty
4 The Distribution of Top Incomes in the
United Kingdom 1908–2000 82
A. B. Atkinson
5 Income and Wage Inequality in the
United States, 1913–2002 141
T. Piketty and E. Saez
6 The Evolution of High Incomes in
Canada, 1920–2000 226
E. Saez and M. Veall
7 The Distribution of Top Incomes in Australia 309
A. B. Atkinson and A. Leigh
8 The Distribution of Top Incomes in New Zealand 333 A. B. Atkinson and A. Leigh
9 Top Incomes in Germany Throughout the
Twentieth Century: 1891–98 365
F. Dell
10 Top Incomes in the Netherlands over the
Twentieth Century 426
W. Salverda and A. B. Atkinson
11 Income and Wealth Concentration in Switzerland
over the Twentieth Century 472
F. Dell, T. Piketty, and E. Saez
12 Long Term Trends in Top Income Shares in Ireland 501 B. Nolan
13 Towards a UniWed Data Set on Top Incomes 531
A. B. Atkinson and T. Piketty
List of Figures, Tables, and Boxes
F I G U R E S
1.1 The fall of top capital incomes in France, 1913–98 10 1.2 The top 1% income shares in France, the UK, and the US, 1913–98 12 1.3 Wealth concentration in Paris and France overall, 1807–1994 16 2.1 Share of top 1% and overall Gini coeYcient in US, 1947–2002 20 2.2 ‘Taxable capacity’ of top 1% in the UK, 1937–2000 21
2.3 Globally rich as % world population, 1910–92 25
2.4 Personal income control totals for the UK, 1908–99 31 2.5 Interpolation into open upper interval, UK 2000 data 33
3.1 The top decile income share in France, 1900–98 48
3.2 The income shares of fractiles P90–95, P95–99, and P99–100 in France,
1900–98 49
3.3 The top decile and the top percentile wage shares in France, 1913–98 53
3.4 Factor shares in France, 1913–98 58
3.5 The average estate left by fractiles P90–95 and P99.99–100 in France, 1902–94 60 3.6 EVective average income tax rates in France, 1915–98 62 4.1 Share of total gross income of the top 0.05%, 0.1%, and 0.5% in the UK,
1908–2000 92
4.2 Share of total gross income of the top 1%, 5%, and 10% in the UK,
1908–2000 95
4.3 EVect on share of top 1% of adjustment for retained earnings,
UK 1937–65 101
4.4 Shares within shares, UK 1918–2000 102
4.5 Pareto-Lorenz coeYcients, UK 1908–2000 103
4.6 Share of total personal after tax income of the top 0.05%, 0.1%, and 0.5%,
UK 1937–2000 105
4.7 Share of total personal after tax income of the top 1%, 5%, and 10%,
UK 1937–2000 106
4.8 Percentage reduction in after tax shares compared with before tax shares,
UK 1937–2000 106
4.9 Composition of adjusted total income, UK 1949–2000 108 4.10 Composition of income for diVerent groups, UK 1937–98 109
4.11 Composition of income of top 1%, UK 1937–2000 110
4.12 Contribution to share of top 1%, UK 1949–2000 111
4.13 Shares of top earners and top wealth holders in UK, 1923–2000 112
5.1 The top decile income share, US 1917–2002 147 5.2 The income shares of P90–95, P95–99 and P99–100 in US, 1913–2002 147
5.3 The top 0.01% income share, US 1913–2002 149
5.4 Income composition of top groups within the top decile in US, 1929
and 1999 151
5.5 The capital income share in the top 0.5% in US, 1916–99 153 5.6 Capital income in the corporate and personal sector, US 1929–2003 154
5.7 The top 0.1% wealth share in US, 1916–2000 156
5.8 The top decile wage income share, US 1927–2002 159 5.9 Wage income shares for P90–95, P95–99, and P99–100 in US, 1927–2002 159 5.10 Shares of oYcers compensation and wage shares, P99.5–10 and
P99–99.9 in US, 1917–60 160
5.11 CEO pay vs. average wage income, US 1970–2003 163
5.12 Top 0.1% income shares in the US, France, and the UK, 1913–98 166 5A.0 Average real income and consumer price index, US 1913–2002 170 5A.1 Average real income of bottom 90% and top 1% in US, 1917–2002 174 5A.2 Top 1% income shares in US: The role of capital gains 1913–2002 175 6.1 Average real income and consumer price index in Canada, 1920–2000 230
6.2 Top income shares in Canada, 1920–2000 233
6.3 The income shares of the top income groups in Canada and US 1920–2000 234 6.4 Capital income in the corporate and the personal sector in Canada,
1926–2000 237
6.5 Salary vs. wage earners in manufacturing sector in Canada, 1915–48 239 6.6 Income composition of top groups within the top decile in Canada,
1946 and 2000 240
6.7 The share of wage income in upper income groups in Canada, 1946–2000 241 6.8 The top wage income shares in Canada, 1972–2000 243 6.9 The top 1% wage income share of Quebec Francophones vs. allWlers from
the rest of Canada, 1982–2000 246
6.10 Top 1% wage income share for individuals and families in Canada,
1982–2000 247
6.11 The role of stock options in the surge in top wage income shares in
Canada, 1995–2000 248
6.12 Mobility of high incomes in Canada, 1982–2000 251
6.13 Marginal income tax rates in Canada for various percentiles, 1920–2000 253 6.14 Marginal tax rates and income share for the top 0.1% in Canada and US,
1960–2000 254
6A.1 Income shares with and without capital gains of top income groups
in Canada, 1972–2000 262
6A.2 Average income tax rates in Canada within top decile, 1920–2000 264
6A.3 Average income tax rates in Canada within top percentile, 1920–2000 264 7.1 Shares of top 1%, 0.5%, and 0.1%, Australia 1921–2002 317 7.2 Comparing Victoria, 1912–23, with Australia, 1921–31 318 7.3 Share of next 4% and second vintile in Australia, 1921–2002 319
7.4 Shares within shares in Australia, 1921–2002 320
7.5 Pareto-Lorenz coeYcients, Australia 1921–2002 321
7.6 Fraction of income from salary and wages, Australia 1954–2002 322 7.7 Contributions to share of top 1%, Australia 1954–2002 322 8.1 Shares of top 1%, 0.5%, and 0.1% in New Zealand, 1921–2002 342 8.2 Shares of next 4% and second vintile in New Zealand, 1921–2002 342 8.3 Comparison with other top income groups in New Zealand, 1921–2002 344 8.4 Comparison with other studies of New Zealand: shares of top 10%
and 20%, 1921–2002 348
8.5 Shares within shares in New Zealand, 1921–2002 349 8.6 Pareto-Lorenz coeYcients, New Zealand 1921–2002 349 9.1 Series of Mu¨ller and Geisenberger (1972) for Prussia 367
9.2 Share of the top decile, Germany 1891–1998 371
9.3 Share of P90–95 and P95–99, Germany 1891–1998 376
9.4 Share of the top percentile, Germany 1891–1998 376 9.5 Share of P99–99.5, P99.5–99.9, and P99.9–99.9, Germany 1891–1998 378
9.6 Share of the top 0.01%, Germany 1891–1998 378
9.7 Share of the top percentile within the top decile, France, US, and
Germany 1891–1998 379
9.8 Share of P99.99–100 in top percentile, Germany 1891–1998 379 9.9 Share of the bottom part of the top decile (P90–99), France, US,
and Germany 1891–1998 380
9.10 Share of the top part of the top decile (P99–100), France, US,
and Germany 1891–1998 380
9.11 Sources of income in top income groups in Germany, 1928 381 9.12 Sources of income in top income groups in Germany, 1932 382 9.13 Sources of income in top income groups in Germany, 1936 382 9.14 Sources of income in top income groups in Germany, 1992 383 9.15 Sources of income in top income groups in Germany, 1998 383
9F.1 German DAX index, 1988–2000 395
9F.2 German DAX index, 1950–2002 395
9F.3 Implicit capital gains in the last bracket, German tax data, 1961–98 396 9G.1 Evolution of the overall Prussian population; evolution of the share
of tax units actuallyWling tax returns, 1891–1918 403 9G.2 Overall population, tax units, Weimar Republic, and Third Reich, 1925–38 403
9G.3 Overall population, households, and tax units, Federal Republic of
Germany, 1946–2002 404
9H.1 Net personal income of private households and total taxable income
Federal Republic of Germany, 1950–98 406
9H.2 Aggregates of the German national accounts after the Second World War and adjusted net personal income of private households, 1950–2004 406
9H.3 Unemployment in Germany, 1925–38 409
9H.4 Net personal income of private households and total taxable income,
Weimar Republic and Third Reich 1925–38 410
9H.5 Average tax unit income over the twentieth century in Germany 414 10.1 Years for which data in the Netherlands, 1914–99 433 10.2 Real gross average tax unit income and consumer prices Netherlands,
1914–2000 441
10.3A Gross income shares of top 10%, 5%, and 1%, Netherlands 1914–99 442 10.3B Gross income shares of top 0.5% and 0.1%, Netherlands 1914–99 442 10.3C Gross income shares of next 4% and second vintile group, Netherlands
1914–99 443
10.4A Gross income shares within shares, Netherlands 1914–99 444 10.4B Gross income Pareto-Lorenz coeYcients of gross incomes, Netherlands
1914–99 445
10.5 Disposable income shares within shares, Netherlands 1959–99 447 10.6 Ratio of disposable income to gross income top shares, Netherlands
1959–99 447
10.7 Capital income shares within gross income of top 10%, 1%, and 0.1%,
Netherlands 1952–99 450
10.8 Composition of top shares by source of income, Netherlands 1952, 1977,
and 1999 450
10.9A Wage income contributions to gross income of top 10%, 1%, and 0.1%,
Netherlands 1952–99 452
10.9B Wage income contributions to gross income of top 1% and 0.1%,
Netherlands 1952–99 453
10.9C Wage income contributions to gross income of top 0.1%, Netherlands
1952–99 453
10.10 EVective tax rates on gross income of top 10%, 1%, and 0.1%, Netherlands
1914–99 457
10.11 Relative eVective tax rates on gross income of top 10%, 1%, and 0.1%
(average¼1), Netherlands 1914–99 458
10B.1 Tax units (x 1000), Netherlands 1914–99 466
10B.2 Control totals of gross income and known gross income as % of national accounts personal income total, Netherlands 1914–99 467 11.1 Average real income and consumer price index in Switzerland, 1901–2000 488 11.2 Top 10% and top 5% income shares in Switzerland, 1933–96 488
11.3 Top 1%, top 5–1%, and top 10–5% income shares in Switzerland, 1933–96 489 11.4 Top 0.1%, top 0.5–0.1%, and top 1–0.5% income shares in Switzerland,
1933–96 490
11.5 Shares within shares in Switzerland, 1933–63 490
11.6 The top 0.1% income share in France, the US, and Switzerland, 1933–97 491 11.7 Top 10–5%, top 5–1%, and top 1% wealth shares in Switzerland, 1913–97 492 11.8 Top 1–0.5%, top 0.5–0.1%, and top 0.1% wealth shares in Switzerland,
1913–97 493
11.9 The top 1% wealth share in the US and Switzerland, 1915–2000 493 11.10 The fraction of foreign income earners and non-residents in top
income groups Switzerland, 1957–91 496
12.1 Share of top 0.1% in total income, Ireland 1922–90 511 12.2 Shares of top 1% and top 0.5% in total income, Ireland 1938–2000 512 13.1A Share of top 10% in English speaking countries 540 13.1B Share of top 10% in continental European countries 540 13.2A Share of top 1% in English speaking countries 541 13.2B Share of top 1% in continental European countries 541 13.3A Share of top 0.1% in English speaking countries 542 13.3B Share of top 0.1% in continental European countries 542 13.4A Share of top 1% in income of top 10% in English speaking countries 543 13.4B Share of top 1% in income of top 10% in continental European countries 544
TA B L E S
1.1 Raw top income tabulations, France 1919 (originally published in
Bulletin de statistique et de legislation compare,March 1923, tome 93) 4 1.2 Raw income composition tabulations, France 1919 (originally published
inBulletin de statistique et de legislation compare,March 1923, tome 93). 6 1.3 The age proWle of wealth at death in Paris, 1817–1994 17 2.1 Example of income tax data: UK super-tax 1911–12 26 3.1 Income growth and income shares in France, 1900–10 and 1990–98 50 3.2 The impact of progressive taxation on capital accumulation 64
3A.1 Top income shares in France, 1900–98 (I) 71
3A.2 Top income shares in France, 1900–98 (II) 73
3A.3 Sources for French income tax data, 1915–98 75
3A.4 Income and population totals for France, 1900–98 77 4.1 Shares in total before tax income, UK 1908–2000 93 4.2 Shares in total after tax income, UK 1937–2000 104
4A.1 Sources for UK super-tax and surtax data, 1908–72 115
4A.2 Sources of UK SPI data, 1918–2000 117
4B.1 UK control totals for tax units (individuals) and income, 1908–2000 126 4C.1 Derivation of control totals (£ million) for income in the
UK 1945/46–2000/01 132
4C.2 Derivation of control totals (£ million) for income in the UK,
1908/09–44/45 135
5.1 Thresholds and average incomes in top income groups, US 2000 144 5.2 Shares of each occupation within the top 1% in US, 1916 152 5A.0 Reference totals for tax units and income, US 1913–2002 171 5A.1 Top fractiles income shares (excluding capital gains), US 1913–2002 176 5A.2 Top fractiles (deWned excluding capital gains) income shares
(including capital gains), US 1913–2002 179
5A.3 Top fractiles (deWned including capital gains) income shares
(including capital gains), US 1913–2002 182
5A.4 Top fractiles income levels (excluding capital gains), US 1913–2002 185 5A.5 Top fractiles (deWned excluding capital gains) income levels
(including capital gains), US 1913–2002 188
5A.6 Top fractiles (deWned including capital gains) income levels
(including capital gains), US 1913–2002 191
5A.7 Income composition by fractiles of total income, US 1916–99 199 5A.8 Capital gains by fractiles of total income, US 1916–2002 207
5B.1 Aggregate series on wage income, US 1917–2002 211
5B.2 Top wage income shares, US 1927–2002 215
5B.3 Average salary and threshold for each fractile (in 2000 dollars),
US 1927–2002 217
5B.4 CEO pay vs. average wage, US 1970–2003 220
6.1 Thresholds and average incomes in top groups within the top decile
in Canada in 2000 229
6.2 Marginal tax and US eVects on Canadian top income shares, 1920–2000 256 6A.1 Reference totals for population, income, and inXation in Canada, 1920–2000 259
6B.1 Top income shares in Canada, 1920–2000 266
6B.2 Top income shares including capital gains in Canada, 1972–2000 269 6B.3 Top fractile income levels (excluding capital gains) in Canada, 1920–2000 271 6C.1 Shares of total tax returns in each occupation in Canada, 1920–41 278 6C.2 Shares of each occupation within the top 10% in Canada, 1942 279 6C.3 Income composition by fractiles of total income (excluding capital gains)
in Canada, 1946–2000 280
6C.4 Share of capital gains in total income for upper groups in Canada,
1972–2000 285
6D.1 Aggregate series on wages in Canada, 1972–2000 288
6D.2 Shares of wage income for upper groups in Canada, 1972–2000 289 6D.3 Average wage income and threshold for each fractile (in 2000 Canadian
dollars) in Canada, 1972–2000 292
6D.4 Top wage income shares, Francophones in Quebec vs. allWlers from rest
of Canada, 1982–2000 294
6D.5 The role of stock options in top wage income shares in Canada, 1995–2000 296
6E.1 High income mobility in Canada, 1982–2000 298
6F.1 Marginal income tax rates in Canada, 1920–2000 301 6F.2 Average tax rates in upper groups in Canada, 1920–2000 304
7.1 Top income shares, Australia 1921–2002 315
7.2 Top income shares, Victoria, Australia, 1912–23 318
7A.1 Population totals for Australia, 1912–2002 324
7B.1 Personal income totals for Australia, 1912–2002 327 7C.1 Sources of income tax data for Australia, 1921–2002 329 7C.2 Sources of income tax data for Victoria, Australia, 1912–23 330
8.1 Top income shares, New Zealand 1921–2002 340
8.2 Top income percentiles (% mean), New Zealand 1921–2002 345 8A.1 Sources of income tax data for New Zealand, 1921–2002 352 8B.1 New Zealand population totals (thousands), 1921–2002 356 8C.1 New Zealand personal income totals and coverage, 1921–2002 359 8D.1 New Zealand comparison groups for top income shares, 1921–2002 361
9A.1 Income tax publications used, Germany 384
9C.1 Tax units (Tu) in the micro-data set for Germany in the 1990s 385 9C.2 The accuracy of quantile estimation for Germany in the 1990s 386 9F.1 Capital gains and the various aggregates, Germany 1992 392 9F.2 Capital gains and the various aggregates, Germany 1995 393 9F.3 Capital gains and the various aggregates, Germany 1998 394 9G.1 Tax units (Tu) control total for Prussia, 1891–1918 399 9G.2 Tax units (Tu) control total, Germany 1891–1998 401
9H.1 Income control total for Prussia, 1891–1918 411
9H.2 Income control total, 1891–1998 412
9I.1 Nominal thresholds and nominal average income of top income groups,
Prussia 1891–1918 415
9I.2 Nominal thresholds and nominal average income of top income groups,
Germany 1925–38 416
9I.3 Nominal thresholds and nominal average income of top income groups,
Federal Republic of Germany 1950–98 (1) 417
9I.4 Nominal thresholds and nominal average income of top income groups,
Federal Republic of Germany 1950–98 (2) 418
9I.5 Nominal thresholds and nominal average income of top income groups,
Federal Republic of Germany 1950–98 (3) 419
9I.6 Top income shares, Germany 1891–1998 (1) 420
9I.7 Top income shares, Germany 1950–98 (2) 422
9I.8 Top income shares, Germany 1950–98 (3) 423
10.1 Overview of income tax data sources for the Netherlands 432 10.2 Top shares in gross income, Netherlands 1914–99 434 10.3 Top shares in disposable income by range of disposable income,
Netherlands 1959–99 446
10.4 Composition of gross income top shares by source of income,
Netherlands 1952–99 449
10.5 Composition of aggregate gross income by socio-economic category
of receiving tax unit, Netherlands 1952, 1977, and 1999 451 10.6 EVective top share tax rates, Netherlands 1914–99 455 10A.1 Sources for data on total gross income and summary statistics,
Netherlands 1915–99 460
10A.2 Sources for data on disposable income and summary statistics,
Netherlands 1959–99 463
10B.1 Population totals (thousands), Netherlands 1914–99 464 10B.2 Reference income totals (million guilders) and prices, Netherlands 1914–99 468
11.1 Reference totals for population, income, and inXation in Switzerland,
1901–2002 480
11.2 Top income shares in Switzerland, 1933–95/96 484
11.3 Top wealth shares in Switzerland, 1913–97 486
11.4 Fraction of non-residents and residents with income abroad in top
income groups in Switzerland, 1949/50–1991/92 495
11.5 Capital income earned through Swiss accounts and tax evasion, 1950–2002 497 12.1 Sur-tax payers classiWed by income ranges, Ireland 1936–37 503 12.2 Personal income classiWed by income ranges, Ireland 1938 and 1943 504 12.3 Income tax payers classiWed by income ranges, Ireland 2000 505 12.4A Control totals for number of tax units, Ireland 1922–2000 507
12.4B Control totals for income, Ireland 1922–2000 509
12.5 Shares of top income groups, Ireland 1922–2000 513 12.6 Top income shares estimated from ‘gross incomes’, Ireland 1989/90–2000 519 12.7 Composition of top incomes, Ireland 1989/90 and 2000 520 12.8 Share of top income groups in top incomes, Ireland 1938–2000 521 12A.1 Source of income data used in deriving ‘total’ income shares,
Ireland 1922–2000 523
12A.2 Source of income data used in deriving ‘gross’ income shares,
Ireland 1989–2000 526
12A.3 Estimated indices for national income in money terms, ‘general prices’,
and ‘real income’ 526
13.0 Key features of estimates for ten countries 533
13.1 Shares in total before tax income, France 545
13.2 Shares in total before tax income, UK 547
13.3 Shares in total before tax income, US 549
13.4 Shares in total before tax income, Canada 551
13.5 Shares in total before tax income, Australia 553
13.6 Shares in total before tax income, New Zealand 555
13.7 Shares in total before tax income, Germany 557
13.8 Shares in total before tax income, Netherlands 559 13.9 Shares in total before tax income, Switzerland 561
13.10 Shares in total before tax income, Ireland 563
B OX E S
2.1 Pareto distribution 24
10.1 Summary of approach adopted in Netherlands estimates 440
Preface
The origins of this volume, and the companion volume to follow, lie in the study of top incomes in France over the twentieth century published by one of us (TP) in 2001. The study used data from income tax and other sources to show the evolution of income inequality over a much longer continuous period than had previously been investigated (see Piketty 2001). This study, summarized in Chapter 3, inspired the other editor (ABA) to examine the same topic for the United Kingdom, and the results are presented in Chapter 4. Piketty and Emmanuel Saez extended the comparison further by making estimates for the United States (summarized in Chapter 5). Since then, the fruitfulness of income tax data in providing long run evidence about the top of the distribution has led to estimates being constructed for a sizeable number of countries (covered here in Chapters 6 to 12 and in a forthcoming second volume).
The aim of the project is to assemble in one place the studies of top incomes for a wide range of countries (ten in this volume). A number of the chapters are based on research that has already been published in journal articles (see the Bibliography, Chapters 1 and 2 in this volume), but the present versions contain more extensive accounts of the sources and methods as well as further and, in some cases, more recent results. Present journal editorial practice does not typically allow space for full documentation of methods, but we believe that it is important that these be recorded and discussed. The preparation of new economic data such as those presented here involves a large number of operations and recourse to a diversity of sources. Along the way, the data constructor has inevitably had to make assumptions and corrections; it is not simply a matter of copying tables. If this process is not documented in full, then the reader is unable to assess the validity of theWnal series. We have therefore encouraged authors to explain their methods in detail.
The volume is not intended to be a comparative study. Although a number of the chapters refer to evidence for other countries, it will be clear that each country studied has its own speciWcities with regard to systems of income taxation, to the ways in which data are collected, and to the wider processes of income deter- mination. We cannot assume that the series are fully homogeneous across countries, and the literature on cross-country growth regressions warns us of the pitfalls in merging data without regard to the speciWcities of both data and reality. The emphasis is therefore on the historical experience of each of the ten countries. At the same time, as discussed in Chapter 1, the studies presented here represent a necessary Wrst stage in any comparative analysis. The series were constructed by using the same raw data sources for all countries and applying the same methodology to derive the Wnal series. Although fully homogenous, cross-country data sets do not exist, we have done our best to make our database
as homogenous as possible, and to provide users with adequate guidance and technical information. We have therefore, in the Wnal chapter (Chapter 13), assembled the key series for the ten countries. In the second volume, we hope to cover the Nordic countries, countries from Southern Europe, India, China, Brazil, and Indonesia, which will extend considerably the range of experience.
The bibliographic references for theWrst two chapters are grouped together, but we have kept separate bibliographies for the individual country chapters (even though this means some duplication,) on the grounds that some readers may only be interested in one country, and wish to see the sources for that country collected together.
A number of the chapters were presented at a conference organized as part of the CHANGEQUAL network meeting at NuYeld College, Oxford, in September 2003. Atkinson worked on theWnal preparation of the manuscript while holding a Chaire Blaise Pascal at ENS-PSE. The editors would like to thank Lin Sorrell and Cathy Douglas for their help at NuYeld, and the authors for their contributions and patience.
A.B. Atkinson and T. Piketty
R E F E R E N C E
Piketty, T. (2001).Les hauts revenus en France au XXe sie`le: ine´galite´s et redistributions, 1901–1998. Paris: Grasset.
1
Top Incomes Over the Twentieth Century:
A Summary of Main Findings 1
T. Piketty
1 . 1 I N T RO D U C T I O N
This introductory essay presents some of the key Wndings and perspectives emerging from the detailed country chapters published in this volume. All chapters are part of a collective research project on the long-run dynamics of income and wealth distribution. The general objective of this project was to construct a high quality, long-run, international database on income and wealth distribution using historical tax statistics. The resulting database now includes annual series covering most of the twentieth century for over 20 (mostly Western) countries. The present volume focuses upon the contrast between continental European countries and English-speaking countries and includes ten case studies:
France, UK, US, Canada, Australia, New Zealand, Germany, the Netherlands, Switzerland, and Ireland. A forthcoming volume will complete the study by covering Scandinavian and Northern Europe (including Sweden, Finland, and Norway), Southern Europe (including Italy, Spain, Portugal), as well as a number of Latin American (including Argentina, Brazil) and Asiatic countries (including India, China, and Indonesia).
The primary motivation for this project was a general dissatisfaction with existing income distribution databases. The international databases on inequality that existed were not high quality (they display little homogeneity over time or across countries),2they are not long-run (typically they cover only a couple of isolated years per country, generally restricted to the post-1970 or post-1980 period), and they almost never oVer any decomposition of income inequality into a labour income and a capital income component. This latter feature of existing data sets is unfortunate, because the economic mechanisms at work can be very
1 The references to this chapter are given at the end of Chapter 2.
2 See, e.g., the Atkinson-Brandolini (2001) criticism of the World Bank (Deininger-Squire) sec- ondary database. The database is ‘secondary’ in the sense that it is based on the collection of inequality measures computed by others using various income data sets and methodologies for diVerent countries and time periods. In contrast, our inequality measures were computed by ourselves using the same primary data sources and methodology for all countries and time periods.
diVerent for the distribution of labour income (demand and supply of skills, labour market institutions, etc.) and the distribution of capital income (capital accumulation, credit constraints, estate taxation, etc.), so that it is fairly heroic to test for any of these mechanisms using such data. The fact that existing database are not long run is also most unfortunate, because structural changes in income and wealth distributions are relatively slow and very often span over several decades. In order to properly understand such changes, one needs to be able to put them into broader historical perspective.3
Our database also suVers from strong limitations (in particular, our long-run series are generally conWned to top income and wealth shares and contain little information about bottom segments of the distribution), and fully homogenous, cross-country data sets do not exist. However, our database has the following advantages:
. we use the same raw data sources for all countries and apply the same methodology to derive theWnal series;
. the series are typically annual and cover a long-run of years;
. the data are mostly broken down by income source.
This means that they oVer a unique opportunity to understand better the dynamics of income and wealth distribution and the two-way interaction be- tween inequality and growth.
We should stress that the main objective of the chapters collected in this volume is to describe how the series were constructed, and to oVer Wrst cut analysis of the long-run dynamics of inequality in each individual country. Such analytical narratives and detailed case studies are useful, but in our view they should be seen as complements (rather than substitutes) to a more systematic statistical exploitation of the complete database, which we do not oVer in this volume. We very much hope that future researchers will use our database to explore causal mechanisms in a more systematic way, and in particular that our data will contribute to renew the literature on cross-country inequality/growth regressions.4
The rest of this introductory essay is organized as follows. In section 1.2, we brieXy present the basic data and methodology used to construct the database.
Section 1.3 presents some of the main descriptiveWndings and conclusions, with particular emphasis to the Kuznets’ curve debate. Section 1.4 attempts to illus- trate how our database could potentially be used to renew the cross-country structural analysis of the interplay between inequality and growth, with better hopes of success than the previous literature. We then discuss some of the prospects for extending the database using additional published historical tax tabulations and collecting historical individual tax data (Section 1.5).
3 This wasWrst stressed by Kuznets (1955).
4 One of the key reasons why the literature on cross-country inequality/growth regressions failed to deliver robust conclusions (see, e.g., Banerjee and DuXo (2003) for a critical appraisal) is the poor quality of existing databases.
1 . 2 . C O N S T RU C T I N G A N EW DATA B A S E : P R I M A RY DATA A N D M E T H O D O LO G Y
Household income surveys are a relatively recent venture: they virtually did not exist on a national basis prior to 1950, and in most countries they are not available in a homogenous, machine-readable format until the 1970s–80s. The only data source that is consistently available on a long-run basis is tax data. Progressive income tax systems were set up in most Western countries at the beginning of the twentieth century (1913 in the US, 1914 in France, etc.), and in all countries with an income tax system the tax administration started compiling and publishing tabulations based on the exhaustive set of income tax returns.5These tabulations generally report for a large number of income brackets the corresponding number of taxpayers, as well as their total income and tax liability. They are usually broken down by income source: capital income, wage income, business income, etc.
In order to give a sense of what our primary data sources look like, we reproduce on Table 1.1 the raw top income tabulations for France in 1919, as they were originally published by the Finance Ministry. One can see for instance on this table that 181 French taxpayers reported tax income above one million francs in 1919 (a pretty large income at that time). We also reproduce on Table 1.2 the raw income composition tabulations for France in 1920. One can see that out of the 722 million French francs reported by French taxpayers with individual income above 1 million francs in 1920, 322 million francs took the form of ‘revenus des valeurs et capitaux mobiliers’ (interest and dividend income), 356 million francs took the form of ‘be´ne´Wces industriels et commerciaux’ (business income), and only 2.2 million francs took the form of ‘traitements publics et prive´s, salaires, etc.’ (wage income).
One can then use standard Pareto extrapolation techniques to compute top fractiles thresholds and average incomes using such data. This methodology is described in a detailed manner in Chapter 2. Here it is suYcient to recall that the Pareto law for top incomes is given by the following distribution function:
1!F(y)¼(k=y)a (k>0,a>1) (1:1) The corresponding density function is given by f(y)¼aka=y(1þa). The key property of Pareto distributions is that the ratio between the average incomey* (y) of individuals (or households or tax units) with income aboveyandydoes not depend on the income thresholdy:
y*(y)¼h Z
z>y
zf(z)dz i
=h Z
z>y
f(z)dz i
¼h Z
z>y
dz=zai
=h Z
z>y
dz=z(1þa)i
¼ay=(a!1) (1:2) i:e:y$(y)=y ¼b,with b¼a=(a!1)
5 Full details about the administrative publications where the raw tabulations were originally published are given in the country chapters.
Table1.1Rawtopincometabulations,France1919(IMPOˆTGE´ NE
´ RAL
SURLEREVENU) MONTANTDESDE´ DUCT
IONS CATE´ G
ORIES DEREVENUS.
NOMBRE decontribuables inscrits dansles roˆles.
MONTANT des revenus impose
´s.
pour situation de famille.
pour chargesdefamilleMONTANT brut del’impoˆt.
MONTANT despØnalitØs et droits ausus. 1,500fr.5,000fr. 12345678 fr.fr.fr.fr.fr.fr. 6,100a`10,000francs130,7871,170,324,800123,915,0007,110,00034,406,0003,805,400170,500 10,100a`20,000193,6792,851,910,400417,507,00025,410,000194,082,00021,056,000759,100 20,100a`30,00058,8941,477,045,800137,517,0008,983,50097,740,00018,687,300755,300 30,100a`50,00039,9741,529,512,70093,711,0006,235,50079,134,00040,061,4001,025,200 50,100a`100,00023,8821,592,572,50062,733,0003,354,00046,894,00094,486,6001,907,700 100,100a`200,0009,4871,517,031,00021,768,0001,513,50050,530,000142,413,8002,820,500 200,100a`300,0002,289556,396,9006,651,000315,0005,456,00099,524,900965,900 300,100a`500,0001,388527,734,8003,204,000138,0003,080,000126,024,7001,228,500 500,100a`1million576387,082,9001,380,00046,5001,318,000130,956,9001,680,800 Au-dessusde1million181451,968,100420,00013,500336,000206,785,300883,400 Totaux467,13711,867,588,900868,911,00055,119,500492,776,000883,801,20012,177,000 (contd.)
Table1.1(Contd.) MAJORATION dueparlescontribuablescØlØbataires. (25p.100.)
MAJORATION dueparlesmØnagessansinfants. (10p.100.) Nombre de contribuables supportant la majoration.
Montant desrevenus des inte
´re´sse
´s Produit dela majoration.
Nombre de contribuables supportant la majoration.
Montant des revenus des inte
´resse´s.
Produits dela majoration.
MONTANT des dØductions pourchargesde famille.
PRODUITNET total del’impo ˆt.
910111213141516 fr.fr.fr.fr.fr.fr. 45,190340,334,700430,70011,900111,048,30070,000105,6004,570,800 21,602301,518,900875,60029,401413,678,300354,800727,80022,518,600 5,162130,728,5001,026,2006,712168,608,500470,500096,20020,225,100 3,398132,038,6001,073,9004,225168,148,800431,0001,801,20040,777,300 2,049143,370,6002,067,8002,407168,390,6001,034,8005,028,80095,868,100 74699,947,6003,173,800904125,934,3001,557,00010,161,500139,805,600 16739,950,6001,886,30019646,776,500893,0004,704,10098,566,000 11433,245,2002,153,30012345,315,2001,137,0003,080,000127,403,800 3524,508,5002,087,5004529,941,9001,086,1001,518,000134,493,500 2349,247,6005,993,8001733,763,3001,506,700336,000214,833,200 Totaux78,4921,294,870,80020,770,90055,9301,511,005,7008,590,90028,620,200896,719,800 Note:Tableaupre´sentant,
a`ladatedu30avril1922,lade
´composition,
parcate
´gor
iesderevenus,desre´sultatsdesroˆlese´tablisautitredel’anne
´e 1920(revenusde1919) Source:OriginallypublishedinBulletindestatistiqueetdelegislationcompare,March1923:vol.93.
Table1.2Rawincomecompositiontabulations,France1919(IMPOˆTGE
´ NE
´ RAL SURLEREVENU.) CATE´ GORIES
DE
´ COM
POSITIONDESREVENUSGLOBAUXSUIVANTLESDIVERSESSOURCESD’OU`ILSPROVIENNENT(a). derevenus.
MONTANT total des revenue globaus (A)
REVENUS des proprie
´te´s baˆties REVENUS des propriete
´s non baˆties REVENUS des valeurs et capilaux mobiliers.
BE´ NE
´ FCLES de l’exploitation agricols. Montant.Proportion.Montant.Proportion.Montant.Proportion.Montant.Proportion. 12345678910 millions.millions.%millions.%millions.%millions.% 6,100a`10,000fr...1,100655.9313.114813.5181.6 10,100a`20,0003,8322055.31002.649713.0822.1 20,100a`30,0002,0441270.3633.130117.7472.3 30,100a`50,0002,1321426.7622.940221.7401.9 50,100a`100,0002,2811430.3592.658625.0361.6 100,100a`200,0001,803975.4301.751428.5181.0 200,100a`300,000751344.5101.323331.050.7 300,100a`500,000699203.781.122732.560.9 500,100a`1million......530173.240.718635.140.8 Au-dessusde1million...722121.750.732244.630.4 Totauxetmoyennes16,8978685.53752.43,53622.22591.6 (contd.
Table1.2(Contd.) BE´ NE
´ FICES Industriels et commerciaux.
BE´ NE
´ FICES del’exploitation minie
`res.
TRAITEMENTS publicset prive´s, salaires, etc.
PENSIONS de rentosvirgu
`res BE´ NE
´ FICE
S des professions non commerciales.
REVENUS des charges et oYces. Montant.Proportion.Montant.Proportion.Montant.Proportion.Montant.Proportion.Montant.Proportion.Montant.Proportion. 111213141516171819202122 millions.%millions.%millions.%millions.%millions.%millions.% 1079.710.166560.4353.2242.230.3 82021.630.11,87348.9721.91483.8260.7 65131.830.163431.0291.41025.0271.3 70535.020.150223.5221.01014.7341.6 80039.020.142213.5170.7924.0371.6 80044.430.225714.280.4503.1201.1 35347.010.18711.030.4182.471.0 31449.210.1689.730.3142.030.5 20050.210.2438.120.281.5.’’’’ 35649.330.4162.2’’’’50.7’’’’ Totauxetmoyennes5,35833.7209.14,56728.71891.25683.61671.0 Notes:IMPOˆTE´ TABLEAUTITREDEL’ANNE
´ E
1921,—BE
´ NE
´ FICES
BYREVENUSRE
´ ALISE
´ S
AUCOURSDEL’ANNE
´ E
1920.Tableaupre
´sentant, pourlescontribuablesinscritsdantslesroˆlese´mis 1stjanvier1921au30avril1922,lade´composition
durevenuglobal(Revenusde
´clare
´s sculement)lesdiVe´rentessourcesderevenus. (a)Avanttoutede´duction
autitredeschargesgrovantlerevenuglobal.(Contributionsdirectesassimile
´es,
pertesre´sultantd’unde
´ficit
d’exploitation,inte
´reˆts dedellos,etc.). (b)Auconeconcordancenepeutexisterentrelemontantderevenusindique
´s
aupre´senttableauetlemontantdesrevenusquelservidebaseauximpo
ˆts ce´dulaires
pourl’ane
´e 1921.Tous contribuablesassujettisauximpo
ˆts ce´dulaires
nesontpas,eneVet,possiblesdel’impo
ˆt ge´ne
´ral
et,investement,certainsrevenusentrantdanslacompositiondurevenuglobalsoumia`l’impo
ˆt ge´ne
nesantpasfrappe
´sparl’impo ˆtc
e´dulaireparcequeleurmontantnede
´ponse
paslasommeaffranchiedel’impo
ˆtdanslace
´dule correspondante. Source:OriginallypublishedinBulletindestatistiqueetdelegislationcompare,March1923:vol.93.
That is, ifb¼2, the average income of individuals with income above1100,000 is 1200,000, and the average income of individuals with income above11 million is 12 million. Although this law is only an asymptotic approximation (in practice, estimatedb coeYcients vary slightly with y), it works remarkably well for top incomes, as wasWrst noted by Vilfredo Pareto (1896, 1896–97) in the 1890s using tax tabulations from Swiss cantons. In this volume, we do not address the interesting issue as to why this law holds, and we solely use is as an interpolation technique allowing us to compute top fractile thresholds and average incomes from grouped income data. It is important to note that although thebcoeYcient is (almost) invariant with y for a given country and a given year, it does vary substantially over time and across countries.6A higherbcoeYcient means a fatter upper tail of the income distribution, which generally implies higher inequality (for a constant mean). For instance, thebcoeYcient declined from about 2.3–2.4 to about 1.7–1.8 in France during the twentieth century, as top income shares dropped. ThebcoeYcient went through a similar decline in all countries where inequality dropped, and it started rising again in countries where inequality rose since the 1970s, e.g. in the United States (where thebcoeYcient is now back to about 2.3–2.4).
Pareto extrapolation techniques are fairly powerful, but they do not allow extrapolation on income ranges for which we have no data. In that respect, one major limitation of tax data is that the income of individuals not subject to the tax is excluded from the data. Prior to the Second World War, the proportion of individuals subject to progressive income taxation hardly exceeded 10–15% in most countries, so that one can only compute top decile income series (and above) over the entire period. In order to construct top fractile income shares series from top fractile income data, one needs a total income denominator, which can be computed using aggregate income sources (national accounts and their ancestors). Constructing homogenous numerator and denominator series requires special care and raises a number of issues, many of which are addressed in Chapter 2.
1 . 3 B A S I C D E S C R I P T I V E F I N D I N G S : T H E K U Z N E T S ’ C U RV E , 5 0 Y E A R S L AT E R
TheWrst economist to use these data sources and methodology in a systematic way was Kuznets (1953).7He exploited US income tax tabulations covering the
6 Most authors refer toa¼b/(b!1) (rather thanb) as the ‘Pareto coeYcient’. Note, however, that thebcoeYcient has a more intuitive economic meaning. One could for instance refer tob!1 as the
‘income advantage of the rich’ (IAR) coeYcient. During the twentieth century the IAR coeYcient declined from 130–140% to 70–80% in France, i.e., the income advantage of the rich nearly halved.
7 Earlier authors (e.g., Bowley 1914 and Stamp 1916) used income tax data in a sophisticated way (see Chapter 4), but Kuznets was apparently theWrst scholar to use control totals to construct top income shares series.
1913–48 period and computed corresponding top decile and top percentile income shares series. These were the Wrst long-run income distribution series ever produced (income distribution had been at the centre of speculative eco- nomic thought at least since the time of Ricardo and Marx, but few data were available). Unsurprisingly, these series had a major impact on economic thinking, especially after Kuznets (1955) proposed his famous ‘Kuznets curve’ theory in order to account for the 1913–48 decline in income inequality that he witnessed for the United States. According to this theory (which Kuznets himself viewed as highly speculative),8income inequality should follow an inverse-U shape along the development process,Wrst rising with industrialization and then declining, as more and more workers join the high productivity sectors of the economy.
In a sense, all what we are doing in this project is to extend and generalize what Kuznets did in the early 1950s—except that we now have 50 more years of data, and over 20 countries instead of one. In addition, note that Kuznets had access to a fairly limited data processing technology, which probably explains why he did not use all available data as systematically as possible. In particular, Kuznets did not fully use the tabulations broken down by income source, and his top income shares series are only deWned for total income (for instance, he did not compute separate series for wage income or capital income).
The fact that we have 50 more years of data, over 20 countries and series broken down by income source led us to adopt a fairly diVerent perspective than Kuznets as to why income inequality dropped in Western countries during theWrst half of the twentieth century. First, as one can see on Figure 1.1, where we plot the basic series for the French case, the decline in top income shares witnessed by Kuznets for the US also took place in France, but it came to an end right after the Second World War. The secular decline in income inequality took place during a very particular and politically chaotic period, namely during the 1914–45 period (and especially during both World Wars and the early 1930s). This raises serious doubts about a gradual, Kuznets type explanation. If the decline in income inequality was due to a continuous reallocation process between from a low productivity to a high productivity sector (say, from rural to urban sector, as in Kuznets’ original model), then it is hard to understand why the timing of the fall should be so particular.
Next, and most importantly, one can see from Figure 1.1 that the 1914–45 drop in top income shares is entirely due to the fall of top capital incomes: top wage shares actually did not decline at all. One gets the same picture by using other inequality measures, e.g., by looking at the top decile share rather than the top percentile share. In particular, the striking fact that the wage distribution in a country like France has been extremely stable in the long run during the twentieth century appears to be very robust, irrespective of how one measures wage inequality (for instance, the 90–10 ratio—and not only top wage shares—has also remained stable in the long run); see Piketty (2003) and Chapter 3. Labour
8 ‘This is perhaps 5% empirical information and 95% speculation, some of it possibly tainted by wishful thinking’ (Kuznets 1955: 26).
