Piketty 2003

Income Inequality in France, 1901–1998
Author(s): Thomas Piketty

Source: Journal of Political Economy, Vol. 111, No. 5 (October 2003), pp. 1004-1042
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/376955 .
Accessed: 24/01/2015 12:42
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journal
of Political Economy.
http://www.jstor.org
This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

All use subject to JSTOR Terms and Conditions
Income Inequality in France, 1901–1998
Thomas Piketty
Ecole des Hautes Etudes en Sciences Sociales, Paris-Jourdan, and Centre for Economic Policy
Research
This paper uses data from income tax returns (1915–98), wage tax
returns (1919–98), and inheritance tax returns (1902–94) in order to
compute homogeneous, yearly estimates of income, wage, and wealth
inequality for twentieth-century France. The main conclusion is that
the decline in income inequality that took place during the first half
of the century was mostly accidental. In France, and possibly in a
number of other countries as well, wage inequality has been extremely
stable in the long run, and the secular decline in income inequality
is for the most part a capital income phenomenon. Holders of large
fortunes were badly hurt by major shocks during the 1914–45 period,
and they were never able to fully recover from these shocks, probably
because of the dynamic effects of progressive taxation on capital ac-
cumulation and pretax income inequality.
I. Introduction
The primary objective of this research is to document trends in income
inequality in France during the twentieth century. Did income distri-
bution become more unequal or more equal in France over the course
of the 1901–98 period? What are the specific periods in which income
inequality increased or declined, and what income deciles were most
affected by these trends?
I am grateful to seminar participants at Columbia, Harvard, Massachusetts Institute of

Technology, Chicago, London School of Economics, and Paris for lively discussions. I also
thank an editor and two anonymous referees of this Journal for their helpful comments.
I gratefully acknowledge financial support from the MacArthur Foundation. This paper
presents some of the results that are exposed in a more detailed manner in a book in
French (Piketty 2001a). All series used in this book and in this paper can be downloaded
at http://www.cepremap.ens.fr/piketty.
[Journal of Political Economy, 2003, vol. 111, no. 5]

䉷 2003 by The University of Chicago. All rights reserved. 0022-3808/2003/11105-0006$10.00
1004

income equality 1005
The second objective of this work is obviously to understand these
facts. What are the economic mechanisms and processes that allow us
to understand the way income inequality evolved in France over the
course of the twentieth century? According to Kuznets’s (1955) influ-
ential hypothesis, one should expect income inequality to decline spon-
taneously in advanced capitalist countries, as more and more workers
join the high-paying sectors of the economy. Can this model account
for what happened in France during the 1901–98 period, or at least
during the first half of the twentieth century?
One advantage of looking at France is that French data sources allow
for a detailed analysis of inequality trends. In particular, I was able to
construct fully homogeneous yearly series running from World War I
until the late 1990s for both income inequality and wage inequality; to
my knowledge, this has not been done for any other country. I can
therefore distinguish precisely between the trends that are due to
changes in the wage structure and those that are due to changes in the
concentration of capital income. This allows me not only to better un-
derstand the French experience but also to reinterpret the experience
of other countries. The main conclusion is that the decline in income
inequality that took place during the first half of the twentieth century
was mostly accidental. In France, and possibly in a number of other
countries as well, wage inequality has actually been extremely stable in
the long run, and the secular decline in income inequality is for the
most part a capital income phenomenon. Holders of large fortunes were
badly hurt by major shocks during the 1914–45 period, and they were
never able to fully recover from these shocks, probably because of the
dynamic effects of progressive taxation on capital accumulation and
pretax income inequality.
The rest of this paper is organized as follows. Section II describes my
data sources and outlines my methodology. Section III presents the basic
facts that characterize my income inequality series and that need to be
explained. Section IV attempts to account for these facts. In Section V,
I briefly discuss whether my French conclusions can be applied to other
developed countries. Section VI presents concluding comments.
II. Data Sources
This work relies on three major types of data sources: data from income
tax returns (1915–98), data from wage tax returns (1919–98), and data
from inheritance tax returns (1902–94).

1006 journal of political economy
A. Income Tax Returns (1915–98)
The most important data source is the income tax. A general income
tax was enacted in France in 1914. It took effect for the first time in
1915 (i.e., taxpayers reported their 1915 incomes at the beginning of
1916), and it has applied every year ever since. Most important, the
French tax administration has been compiling every year since 1915
(including the World War II era) summary statistics based on the tab-
ulation of all individual income tax returns. The raw materials produced
by the tax administration have had the same general form since 1915:
the tabulations indicate the number of taxpayers and the amount of
their taxable income as a function of a number of income brackets (the
number of brackets is usually very large, especially at the top of the
distribution). This basic table is available for each single year of the
1915–98 period.1
One important limitation of these annual tables is that they include
only those households whose income is high enough to be taxable under
the general income tax system.2 In France, less than 5 percent of the
total number of households had to pay the income tax during the first
few years of the income tax system, and the percentage of taxable house-
holds fluctuated around 10–15 percent during the interwar period. This
percentage then rose steadily from 10–15 percent in 1945 up to 50–60
percent in 1975 and finally stabilized around 50–60 percent since the
1970s. It is therefore impossible to use these data in order to produce
estimates of the entire income distribution, and one needs to concen-
trate on top fractiles.
The methodology that I applied to the raw data can be described as
follows.3
1
The complete technical characteristics of these raw statistical materials, as well as the
exact references of the official statistical bulletins and administrative archives in which
these data were originally published by the French Ministry of Finance, are given in the
book from which this paper is extracted (see Piketty 2001a, app. A, pp. 555–91).
2
For simplicity, I shall always refer to tax units as “households” in the context of this
paper. In actual fact, these are two different concepts (just as in the United States): one
nonmarried couple makes two tax units but one household, etc. All estimates reported
here were computed in terms of tax units (i.e., the “top decile income share” denotes
the income share going to the top decile of the tax unit distribution of income per tax
unit, etc., with no adjustment for the varying size of these tax units). The key point,
however, is that the average number of tax units per household has been fairly stable
since 1915 (around 1.3) and that the income profile of this ratio has been fairly stable
since 1915 (as a first approximation). Tax data on the number of dependents and married
couples per tax bracket also show that the income profile of average household size appears
to have been relatively stable in the long run (in spite of a sharp fall in average household
size).
3
The methodology is fully described in Piketty (2001a, app. B, pp. 592–646). In par-
ticular, the book provides a detailed account of the many technical adjustments that were
made to the tax data in order to take into account changes in tax law and to ensure
homogeneity of the series. It includes all necessary information and intermediate com-
putations to reproduce my estimates, from the raw data to my final series.

i) I used the basic tables produced by the tax administration in order
to compute the Pareto coefficients associated with the top of the French
income distribution for each year of the 1915–98 period. These struc-
tural parameters then allowed me to estimate for each single year of
the 1915–98 period the average incomes of the top 10 percent of the
income distribution (i.e., the top decile, which I denote P90–100), the
top 5 percent of the income distribution (P95–100), the top 1 percent
(P99–100), the top 0.5 percent (P99.5–100), the top 0.1 percent
(P99.9–100), and the top 0.01 percent (P99.99–100), as well as the
average incomes of the intermediate fractiles (P90–95, i.e., the bottom
half of the top decile; P95–99, i.e. the next 4 percent; etc.) and the
income thresholds corresponding to the ninetieth percentile, the ninety-
fifth percentile, and so forth (P90, P95, etc.). For the years 1915–18,
because of the small number of taxable households, I estimated the
incomes of only fractiles P99–100 and above. The Pareto interpolation
technique has been used by other researchers working with historical
tax data,4 and the estimates that I obtain for the French case appear to
be as precise as those obtained in other countries (thanks to the large
number of income brackets used by the tax administration).5
ii) I then used French national income accounts in order to estimate
total and average household income for the entire population (taxable
and nontaxable), and I used these estimates to compute series for the
share of fractile P90–100 in total income, the share of fractile P95–100
in total income, and so forth and the share of fractile P99.99–100 in
total income. This methodology (i.e., using tax returns to compute the
level of top incomes and using national accounts to compute the average
income denominator) is also standard in historical studies on income
inequality (see, e.g., Kuznets 1953). The income concept that I have
used for both the numerator and the denominator is pretax, prede-
ductions taxable income.6 Finally, note that I obtained average estimates
of top income shares for the 1900–1910 period by using the rough
estimates of the income distribution that were made by the French tax
4
See, e.g., Kuznets (1953) and Feenberg and Poterba (1993), who applied Pareto in-
terpolation techniques to U.S. income tax returns data over the 1913–48 and 1950–89
periods.
5
I used large micro files of individual tax returns (including all taxpayers above a certain
income threshold) available for the 1980s–90s in order to make sure that my interpolation
technique was indeed very reliable (see Piketty 2001a, app. B, pp. 599–601).
6
The adjustments that I made to national accounts series to ensure that I use the same
income concept both at the numerator and at the denominator are described in Piketty
(2001a, app. G, pp. 693–720), where I also offer a detailed comparison of existing national
accounts series. Official accounts series from the French national statistical institute
(INSEE) start in 1949, and for earlier periods I have relied for the most part on the
retrospective national accounts published by Villa (1994) and on the very well documented
income accounts published by Dugé de Bernonville (1933–39).

administration prior to World War I for revenue projection purposes
(these estimates probably understate inequality a little bit).7
B. Wage Tax Returns (1919–98)

One important feature of the income tax system that was enacted in
France in 1914–17 is that, in addition to the general income tax set up
in 1914, it also included a number of taxes levied separately on each
income source. In particular, there was a “wage tax,” that is, a progressive
tax levied on individual wages, which was first applied in 1917. Individual
wages were declared by employers, who had to file wage tax returns
indicating the annual amount of wages paid to each individual em-
ployee. In 1919, the French tax administration started compiling sum-
mary statistics based on these wage returns. The basic statistical infor-
mation is similar to that contained in the income tax tables: the wage
tables indicate for a large number of earnings brackets the number of
workers and the total amount of their wages (all sectors and occupations,
including government employees, are included). The French tax ad-
ministration stopped compiling these wage tables in 1939, so that these
series cover only the 1919–38 period. In 1947, INSEE decided to use
these wage tax returns to compile new series of annual statistical tables.8
The INSEE tables look like the tax administration tables of the interwar
period (they indicate for each wage bracket the number of wage earners
and the total amount of wages), with the important difference that they
cover the entire wage distribution, not only top wages.9
I have used these raw data in the same way as the income tax data.
Pareto interpolation techniques allowed me to compute the average
wage of the top 10 percent of the wage distribution, the top 5 percent,
the top 1 percent, and so forth (fractiles were defined according to the
total number of wage earners, taxable and nontaxable), and I have used
independent estimates of the total wage bill (coming mostly from the
national accounts) in order to compute series of the top wage shares.10
7
The adjustments that I made to these 1900–1910 estimates on the basis of the data
generated by the first few years of the income tax are described in Piketty (2001a, app.
I, pp. 738–41).
8
The tax on wages was actually repealed in 1948, but the tax administration has kept
using these returns to make sure that income tax payers report the right wage.
9
The 1919–38 tables cover only those wage earners whose wage is high enough to be
taxable under the wage tax system (about 15–20 percent of all workers during the interwar
period).
10
All technical details are given in Piketty (2001a, app. D, pp. 657–76). Unlike the
annual income tables published by the tax administration (which had never been used
to compute long-run inequality series until the present study), wage tables had already
been used to produce series on interdecile ratios for the post-1950 period (see Baudelot
and Lebeaupin 1979; Bayet and Julhès 1996). These authors did not compute top wage
share series, however. Most important, pre–World War II wage tables had never been used

C. Inheritance Tax Returns (1902–94)
A progressive inheritance tax was enacted in France in 1901, and it has
been in force every year ever since. Before 1901, the inheritance tax
was purely proportional, so that the tax administration did not need
information on total estates and did not bother ranking individual es-
tates and compiling statistical tables. In 1901, the tax administration
started using inheritance tax returns to compile tables indicating the
number of estates and the amount of these estates as a function of a
number of estate brackets. These tables were compiled almost every year
between 1902 and 1964 (with an interruption during World War I and
the early 1920s). Since 1964, similar tables have been compiled only in
1984 and in 1994. I have used these raw data in order to compute series
for the average estate of the top 10 percent of the estate distribution,
the top 5 percent, the top 1 percent, and so forth (fractiles were defined
according to the total number of adult decedents, taxable and
nontaxable).11
III. The Basic Facts

Consider first the evolution of the top decile income share (see fig. 1).
The basic fact is that income inequality in France declined significantly
over the course of the twentieth century. According to my estimates,
the share of total household income received by the top decile dropped
from about 45 percent at the beginning of the twentieth century to
about 32–33 percent in the 1990s. That is, the average income of the
top 10 percent was about 4.5 times larger than the average income of
the entire population at the beginning of the twentieth century, and it
was about 3.2–3.3 times larger than the average income of the entire
population in the 1990s.
Next, one can see immediately from figure 1 that this secular decline
has been far from steady. The top decile income share dropped during
World War I and subsequently recovered during the 1920s and the first
half of the 1930s. In 1935, that is, at the height of the Great Depression
in France, the top decile income share was slightly below 47 percent.12
The income share received by the top decile then started to fall sharply
in 1936, and even more so during World War II. The top decile income
until the present study (the very existence of these tables had probably been forgotten,
just as the income tables).
11
All technical details are given in Piketty (2001a, app. J, pp. 744–71). These inheritance
tables had never been used to construct long-run wealth inequality series until the present
study.
12
According to my estimates, the top decile income share during the entire century has
never been as high as in 1935. Note, however, that my average estimates for the 1900–1910
decade probably understate inequality a little bit.

Fig. 1.—The top decile income share in France, 1900–1998. Source: Author’s computations based on income tax returns (see App. table A1, col.
P90–100, and Piketty [2001a, app. B, table B14, pp. 620–21]).
share fell to a nadir in 1944–45 (about 29–30 percent). As far as the
postwar period is concerned, three subperiods need to be distinguished.
The top decile income share increased from 1945 (29–30 percent) to
1967–68 (36–37 percent). Then it declined until 1982–83, when it
reached 30–31 percent. It has then increased somewhat since the early
1980s (32–33 percent in the 1990s). Note, however, that most of the
action took place before 1945. Since World War II, income inequality
in France (as measured by the top decile income share) appears to have
been fluctuating around a constant mean value of about 32–33 percent,
with no trend. In other words, most of the secular decline occurred
during a specific time period (1914–45). These were times of crisis for
the French economy, with two world wars and the Great Depression of
the 1930s. This definitely does not look like a gradual, Kuznets-type
process.
Moreover, and most important, my series show that the secular decline
of the top decile income share is almost entirely due to very high in-
comes. The income share of fractile P90–95 has been extremely stable
in the long run: between 1900 and 1998, that share has always been
fluctuating around a mean value of about 11–11.5 percent of total house-
hold income (which means that these households always get about 2.2–
2.3 times the average income) (see fig. 2). The income share of fractile
P95–99 has experienced a modest secular decline, from about 15 per-
cent of total household income at the beginning of the twentieth century
to about 13–13.5 percent during the 1990s, that is, a drop of about 10
percent (see fig. 2).
In contrast, the top percentile income share has dropped by more
than 50 percent. The share of total income received by the top 1 percent
was about 20 percent at the beginning of the twentieth century, and it
was only about 7–8 percent during the 1990s (see fig. 2). That is, the
average income of the top 1 percent was about 20 times larger than the
average income of the entire population at the beginning of the century,
and it was about seven to eight times larger at the end of the century.
Moreover, my series clearly show that the higher one goes within the
top percentile of the income distribution, the larger the secular decline
(see table 1). The most extreme case is that of the top 0.01 percent:
this income share has dropped from about 3 percent at the beginning
of the century to about 0.5–0.6 percent since 1945. In fact, the average
real income of the top 0.01 percent has not increased at all during the
entire twentieth century: expressed in 1998 French francs, it is about
15 percent lower in 1990–98 than it was in 1900–1910. During the same
time period, the average real income of the entire population, as well
as the average real income of fractile P90–95, has been multiplied by
about 4.5 (see table 1). According to my series, almost 90 percent of
the secular decline of the top decile income share is due to the top

Fig. 2.—The income share of fractiles P90–95, P95–99, and P99–100 in France, 1900–1998. Source: Author’s computations based on income tax
returns (see App. tables A1, A2; Piketty [2001a, app. B, tables B14, B15, pp. 620–22]).
TABLE 1
Income Growth and Income Shares in France, 1900–1910 and 1990–98
Share of Total
Decline of Top
Decile Share
Income Share
Corresponding to
(%) Difference
Income Each Fractile
Fractiles Growth 1900–1910 1990–98 Points Percent (%)
A. Top Fractiles
P0–100 4.48 100.0 100.0 .0 .0
P90–100 3.23 45.0 32.4 ⫺12.6 ⫺28.0 100.0
P95–100 2.77 34.0 21.0 ⫺13.0 ⫺38.3 103.2
P99–100 1.84 19.0 7.8 ⫺11.2 ⫺59.1 88.9
P99.5–100 1.54 15.0 5.2 ⫺9.8 ⫺65.6 78.1
P99.9–100 1.12 8.0 2.0 ⫺6.0 ⫺75.0 47.6
P99.99–100 .83 3.0 .6 ⫺2.4 ⫺81.6 19.4
B. Intermediate Fractiles
P0–90 5.51 55.0 67.6 12.6 22.9
P90–95 4.65 11.0 11.4 .4 3.6 ⫺3.2
P95–99 3.95 15.0 13.2 ⫺1.8 ⫺12.0 14.2
P99–99.5 2.94 4.0 2.6 ⫺1.4 ⫺34.4 10.9
P99.5–99.9 2.02 7.0 3.2 ⫺3.8 ⫺54.9 30.5
P99.9–99.99 1.30 5.0 1.4 ⫺3.6 ⫺71.1 28.2
P99.99–100 .83 3.0 .6 ⫺2.4 ⫺81.6 19.4
Source.—Author’s computations based on income tax returns (see Piketty 2001a, tables 2-1, 2-2, pp.128–29).
Note.—“Income growth” refers to the ratio between the average household incomes of 1990–98 and 1900–1910
(both expressed in 1998 French francs).
percentile, and more than half of the top percentile drop is due to the
top 0.1 percent (see table 1).
The timing of the fall of very top incomes is also striking. Between
1945 and 1998, the income share of the top 1 percent has been fairly
stable (see fig. 2). The secular fall took place exclusively during the
1914–45 period, and especially during the 1930s and World War II. It
is interesting to note that the deflationary years of the Great Depression
had a very different impact on moderately high incomes and on very
top incomes. While the income shares of fractiles P90–95 and P95–99
(the “upper middle class”) increased sharply during the early 1930s, the
income shares of fractiles P99–100 and above (the “rich”) fell (see fig.
2 and App. tables A1 and A2). I shall come back to this below.
IV. Accounting for the Facts

The key facts that need to be explained are the following: the secular
decline in the top decile income share took place during a specific time
period (i.e., between 1914 and 1945, and mostly during the 1930s and
World War II), and it is due for the most part to the sharp drop in the
top percentile income share (and, to a significant extent, to the sharp

drop in the top 0.1 percent income share). How can one account for
these facts?
A. Income Composition Patterns

One first needs to be aware of the large differences in income com-
position that have always characterized the various subfractiles of the
top decile. Every single year of the 1915–98 period, tax return tabula-
tions show that the share of wage income declines continuously from
fractile P90–95 to fractile P99.99–100, whereas the share of capital in-
come (dividends, interest, and rents) rises continuously from fractile
P90–95 to fractile P99.99–100. The shape of the self-employment income
share is intermediate between the wage share and the capital share: it
rises until fractile P99.5–99.9 (approximately) and declines afterward.
These variations in income composition within the top decile are truly
enormous. Whereas the households of fractile P90–95 have very little
capital or self-employment income (about 80–90 percent of their in-
come consists of wages), the households of fractile P99.99–100 rely for
the most part on their capital and self-employment income (typically,
more than 60 percent of their income consists of capital income, and
an extra 20 percent consists of self-employment income). Tax return
tabulations also distinguish between rents, dividend, and interest in-
come, and my detailed series show that top capital incomes consist
mostly of dividends (the share of interest and rents in total income is
basically flat within the top decile, and the share of interest and rents
in total capital income is steeply downward sloping).13 Large capital
owners are predominantly shareholders, not bondholders or landlords.14
These composition patterns suggest that the secular decline in income
inequality is primarily a capital income phenomenon. That is, the frac-
tiles relying mostly on wage income did not experience any significant
decline in the long run (or experienced a limited decline), whereas the
fractiles relying mostly on their capital income experienced major
shocks between 1914 and 1945 (wars, inflation, and depression), from
which they never fully recovered. This interpretation is consistent with
the fact that the capital share at the level of fractile P99.99–100 was as
small as 15 percent in 1945–46 and that the incomes of the top 0.01
percent were mostly made of self-employment income (more than 70
percent of total income) during those years. This is the only instance
during the entire century in which capital income is not the dominant
source of income for very top incomes (capital income returned to its
13
For the detailed composition series, see Piketty (2001a, tables B16–B18, pp. 625–34).
14
It is interesting to note that large capital owners were already predominantly share-
holders (and to some extent bondholders, but very rarely landlords) at the beginning of
the twentieth century.

dominant position during the late 1940s and early 1950s, albeit at a
somewhat lower level than during the interwar period). This clearly
shows that the large drop in top income shares observed between 1914
and 1945 was due, to a large extent, to the fall of top capital incomes.
The fact that the capital share is particularly low at the end of World
War II is also consistent with macroeconomic data. Available series on
factor shares do indeed show that the capital share in French corporate
value-added has never been as low as in 1944–45 (see fig. 4 below).
French gross domestic product has never been as low during the twen-
tieth century as in 1944–45 (fights between the Germans and the Allies
took place over significant portions of the French territory after D-Day,
and firms were completely disorganized), and the big wage increase
implemented by the provisional government implied that there was
almost nothing left for profits.
The composition patterns derived from tax returns also allow me to
account for the sharp divergence between moderately high incomes
and very top incomes observed during the deflationary Great Depression
of the early 1930s. Given that fractiles P90–95 and P95–99 mostly rely
on wages, one should indeed expect these fractiles to benefit from the
fall in prices: real wages did increase during the 1929–35 period (thanks
to the nominal rigidity of wages and the fall in prices), at a time when
real output was falling. Moreover, the high-wage employees (and es-
pecially the government employees) of fractiles P90–95 and P95–99 were
shielded from unemployment, which hurt mostly low-wage workers
(such as low-skill manufacturing or rural workers). Conversely, given
that fractiles P99–100 and above mostly relied on capital income and
business profits, one should indeed expect these fractiles to lose out in
the recession (the capital share fell sharply during the early 1930s). This
process reversed in 1936, when the Front Populaire decided to devalue
the French franc and to put an end to the deflationary strategy. The
high-wage employees of fractiles P90–95 and P95–99 started to lose
ground (inflation pushed their real wages down), and the fall of the
profit holders of fractiles P99–100 and above was temporarily halted.
This again shows that one needs to distinguish between the different
subfractiles of the top decile in order to account properly for the in-
equality facts (this is true both for long-run trends and for short-run
fluctuations).
B. The Long-Run Stability of Wage Inequality

Before I further explore the nature of the shocks suffered by capital
owners during the 1914–45 period and the reasons why they never
managed to fully recover from these shocks, it is important to make
sure that the capital income view of the inequality facts is the right one.

That is, I need to show that wage inequality did not play any significant
role in the secular decline of the top decile income share.
My wage series demonstrate that wage inequality in twentieth-century
France has been extremely stable in the long run. The share of the total
wage bill received by the top decile of the wage distribution has always
fluctuated around a mean value of about 25–26 percent, and the share
of the total wage bill received by the top 1 percent of the wage distri-
bution has hovered near 6–7 percent (see fig. 3).
Note that the wage shares of the top decile and top percentile were
substantially below their secular mean in 1919 (when my annual series
start) and during the early 1920s. But there is ample occupational and
sector-specific evidence showing that this was not a “normal” situation.
The wage structure did narrow substantially during World War I in
France (low-wage workers enjoyed nominal pay increases that were sig-
nificantly higher than those obtained by high-wage workers), and one
can show that the top decile and top percentile wage shares at the eve
of World War I were very close to their secular mean.15
More generally, the fact that wage inequality has been extremely stable
in the long run does not mean that the French history of wage inequality
was smooth and steady during each single decade of the twentieth cen-
tury. Both world wars led to significant compressions of the wage struc-
ture. But the point is that, after each world war, the wage share received
by high-wage workers quickly recovered its prewar level. My wage series
also confirm that the deflationary depression of the early 1930s led to
a widening of wage inequality: high-wage workers benefit from the nom-
inal rigidity of their wages and from the fact that they are less exposed
to unemployment than low-wage workers. In the same way as with the
income series, this process ends in 1936, when the Front Populaire
decided to put an end to the deflationary strategy. The 1967–68 and
1982–83 turning points are also visible in my wage series. Wage disper-
sion significantly widened between 1950 and 1967–68, and the sharp
increases in the minimum wage implemented in the summer of 1968
and during the 1970s led to a significant decline in wage inequality until
1982–83, when the newly elected socialist government decided to freeze
the minimum wage (wage dispersion has increased somewhat since
then). In other words, wage inequality in France during the twentieth
century has been going up and down for all sorts of reasons in the short
and medium run, but it has always reverted to its secular mean. No
long-run trend can be detected in the series.
The contrast between the long-run evolution of the share of total
15
See Piketty (2001a, pp. 188–91, 199–200). The estimates for 1913 reported in fig. 6
below (26 percent for the top decile share and 6.5 percent for the top percentile share)
were computed on the basis of these occupational and sector-specific data (and in par-
ticular on the basis of public-sector data).

Fig. 3.—The top decile and top percentile wage shares in France, 1913–98. Source: Author’s computations based on wage tax returns (see Piketty
2001a, app. D, tables D7, D16, cols. P90–100, P99–100, pp. 664, 675).
income received by the top percentile of the income distribution (fig.
2) and the long-run evolution of the share of the total wage bill received
by the top percentile of the wage distribution (fig. 3) is particularly
striking. While the top percentile income share has declined sharply
from about 20 percent at the beginning of the century to about 7–8
percent in the 1990s, the top percentile wage share has always been
near 6–7 percent.
My wage inequality series therefore confirm that the capital income
interpretation of the inequality facts is the right one. The secular decline
in the top percentile income share is due for the most part to the sharp
drop in the level of the top capital incomes received by the affluent.
Had this level remained constant (relative to the average income), there
would have been no secular decline in the top percentile income share.16
Another advantage of looking at wages is that data are available on
the entire distribution, and not only on the average and on the top
decile. For the 1950–98 period, one can compute annual series for all
percentile ranks of the wage distribution. By looking at the evolution
of ratios such as P10 to the average wage, P50 to the average wage, and
P90 to the average wage during this period, one can see that the entire
distribution of wages has been extremely stable in the long run, and
not only the top decile and top percentile shares.17 Again, one does
observe important fluctuations in the short run and medium run: the
P90/P10 ratio rose sharply between 1950 and 1968, then declined
sharply between 1968 and 1982–83, and finally rose somewhat since
1982–83.18 But these short- and medium-run fluctuations cancel out in
the longer run, in the same way as for top decile and percentile wage
shares.
The same phenomena seem to have occurred during the 1900–1950
period. Available wage returns data do not allow me to estimate annual
series for lower deciles prior to 1950, but occupational and sector-spe-
cific wage data can to some extent serve as a proxy. During the first half
of the twentieth century, agricultural workers were very numerous
16
Strictly speaking, this is more than the data can actually say: depending on the trends
in family structure and correlations between the various types of incomes, a given trend
in wage inequality can translate into various trends in income inequality. But the gap
between fig. 2 and fig. 3 is simply too big to be undone by that kind of bias. Moreover,
note that the correlation of wages between spouses has probably been trending upward
during the twentieth century (as a consequence of the upward trend in female partici-
pation), so that a stable level of wage concentration should actually give rise to an in-
creasing level of income concentration (everything else equal).
17
During the 1950–98 period, P10 has always been fluctuating around 45–50 percent
of average wage, P50 around 80–85 percent of average wage, and P90 around 160–70
percent of average wage (see Piketty 2001a, app. D, table D12, p. 671).
18
The fact that the turning points of postwar trends in wage inequality coincide with
the breaks in French minimum-wage policy was already apparent in the series compiled
by Baudelot and Lebeaupin (1979) and Bayet and Julhès (1996).

(around 30 percent of all wage earners in 1900, down to 20 percent in
1930, 10 percent in 1950, and 1 percent in 1998), and very low wages
were concentrated in this sector. By using the lowest wages observed in
the agricultural sector as a proxy for P10, one finds that the P10/average
wage ratio was already around 45–50 percent in 1900 and 1930, that is,
around the same mean level as during the 1950–98 period.19 That is,
migration from the low-wage rural sector to the high-wage urban sector
did not lead to a structural compression of wage inequality. Low-wage
rural workers disappeared, but they were replaced by low-wage urban
workers, so that the hierarchy did not change very much in the long
run. This evidence stands in contrast to the theoretical predictions of
Kuznets’s two-sector development model, according to which one
should expect inequality to decline as more and more workers join the
high-paying urban sector of the economy.
C. The Robustness of Wealth Leveling
As was already noted above, the fact that capital owners experienced
major shocks during the 1914–45 period (and especially during the
1930s and World War II) is fully consistent with the general economic
history of France during that period. In a sense, what happened between
1914 and 1945 is just the normal consequence of an extraordinary re-
cession. Capital income generally tends to be procyclical, and it is natural
to expect capital owners to suffer a lot from the Great Depression and
the war and to be at their secular low in 1944–45, at a time when the
French gross national product was also at a century low.
In fact, what really needs to be explained is why capital owners never
managed to fully recover from the shocks of the 1914–45 period.
One explanation would simply be that capital owners were confronted
during the 1914–45 period with major shocks to their capital holdings
(and not only to their capital income) and that it takes a long time to
reconstitute the level of fortunes and capital income that capitalists
enjoyed before these shocks. The shocks to capital holdings took three
main forms: inflation, bankruptcies, and destructions.
First, one must bear in mind that inflation did act as a powerful capital
tax. The French consumer price index was multiplied by a factor of
more than 100 between 1914 and 1950, which means that bondholders
19
See Piketty (2001a, pp. 214–15; app. H, tables H2–H4, pp. 726–28). These P10 esti-
mates for 1900 and 1930 were computed by using as proxies wages for low-skill agricultural
workers and rural female domestic workers. I used only money wage estimates and did
not try to take into account in-kind payments (which were quite important for agricultural
and domestic workers). The resulting estimates should therefore be considered as a lower
bound for the true P10 in 1900 and 1930: the true P10/average wage ratio might have
declined somewhat between 1900 and 1950, but it certainly did not rise.

were fully expropriated by inflation. The same process applied, in a less
extreme way, to real estate owners and landlords. Rent control was severe
during both world wars, and the real value of rents was divided by 10
between 1913 and 1950.20 Further, the 1914–50 inflationary process was
something entirely new for the economic agents of the time. There had
been virtually no inflation since the Revolutionary and Napoleonic wars
(the average annual inflation rate between 1815 and 1914 was 0.3 per-
cent), and the government suddenly started to print vast quantities of
money after 1914 to pay for the huge budget deficits brought on by
World War I.
Next, the “recession” induced by the Great Depression of the 1930s
and by World War II was not a “normal” recession. Real GDP declined
by 20 percent between 1929 and 1935 and by 50 percent between 1929
and 1944–45 (see Piketty 2001a, app. G, table G1, p. 695). Many firms
faded and disappeared during that time (much more than during a
“normal” recession). Bankruptcies were particularly numerous in man-
ufacturing and in finance. Large fortunes have always comprised far
more equity shares than bonds or real estate during the twentieth cen-
tury. The impact of the bankruptcies of the 1930s and of World War II
on top fortunes was therefore probably even larger than the impact of
inflation.21
Finally, and most importantly, the physical destructions induced by
both world wars were truly enormous in France. According to the best
available estimates, about one-third of the capital stock was destroyed
during World War I and about two-thirds during World War II. This
reflects the fact that the bombing technology was far more destructive
during World War II than during World War I. According to these
estimates, the capital stock/national income ratio was around 5 at the
eve of World War I, and it then fell to 3.5 in 1934 and 1.2 in 1949.22
It is also important to recall that the French government enacted a
broad nationalization program in 1945. The nationalization process of-
ten was straight expropriation: prices for shares were often set at an
arbitrary, low level so as to punish the “capitalists,” who were often
accused of “collaboration” with the Vichy government. A leading ex-
20
See Piketty (2001a, app. F, table F1, pp. 690–91). On the history of rent control
legislation in France since 1914, see Hirsch (1972) and Taffin (1993).
21
It is unfortunately very difficult to quantify the impact of bankruptcies on the distri-
bution of wealth. It is known that the annual number of bankruptcies more than doubled
between 1929 and 1935 (see INSEE 1966, pp. 170–71), but there is no systematic infor-
mation about the individuals who own these firms and their rank in the wealth distribution.
22
See Piketty (2001a, p. 137). These estimates are due to Sauvy (1965–75, 2:442; 1984,
2:323), who uses estimates of the capital stock computed by Cornut (1963, p. 399). These
estimates are not fully homogeneous (the 1949 capital stock is probably underestimated
somewhat; see INSEE [1958, pp. 34–35]), but they are broadly consistent with the inde-
pendent computations by Divisia, Dupin, and Roy (1956, 3:62), who also find that World
War II destruction was about twice as large as World War I destruction.

ample of this kind of punitive nationalization-expropriation process was
the car company Renault.23 At the same time, in 1945, the provisional
government decided to implement a one-shot tax on capital holdings,
with rates up to 20 percent on top fortunes (and 100 percent on those
fortunes that experienced substantial nominal increases during the
war!) (see Piketty 2001a, p. 138).
In other words, there are good reasons to believe that the accumu-
lation process for large capital holdings was to a large extent set back
to zero (or close to zero) in 1945. This interpretation is consistent with
the composition patterns described in subsection A above: in 1945, very
top incomes were mostly made up of new entrepreneurs, simply because
the old capitalists had disappeared.
But such an explanation cannot be the full story. More than 50 years
have elapsed since 1945, and it would seem that this is a sufficiently
long time period for capitalists to recover from the 1914–45 shocks (at
least partially). The point is that the top percentile income share did
not rise at all during the 1945–98 period (see fig. 2). Apparently, some-
thing important has changed over the course of the twentieth century:
it just seems impossible to accumulate individual fortunes as large as
those that were accumulated in the past.
It is also important to emphasize that the decline in top capital in-
comes is the consequence of a decreased concentration of capital in-
come and not of a decline in the share of capital income in the economy
as a whole. According to national accounts, the share of capital income
(dividends, interest, and rent) in aggregate household income is ap-
proximately the same at the end of the twentieth century as at the
beginning of the twentieth century, that is, about 20 percent (see fig.
4). This is not too surprising given the well-known long-run stability of
the capital share in corporate value-added. Note, however, that while it
took only a few years for the capital share in corporate value-added to
recover from the 1944–45 secular low, it was only in the 1980s–90s that
the capital share in aggregate household income reached the levels
observed in the interwar period and at the eve of World War I (see fig.
4). This important time lag is due to a mixture of two factors. First of
all, retained earnings were unusually high during the reconstruction
23
Unfortunately, there does not seem to exist any systematic, quantitative study of the
1945 nationalization process. Divisia et al. (1956, 3:73–76) describe a number of interesting
examples of nationalization/expropriation, but they do not attempt to quantify the process
at the national level. Similarly, Andrieu, Le Van, and Prost (1987) offer a detailed analysis
of the political context of the nationalization policies, but they do not try to quantify their
importance. I shall come back below to the complicated issue of the long-run impact of
the 1945 nationalizations.

Fig. 4.—Factor shares in France, 1913–98. Source: Author’s computations based on national accounts (see Piketty 2001a, app. G, tables G3–G6, G9,
pp. 703–5, 710–13).
24
period in France (1950s–60s), and the profit share was unusually low
during the 1970s.25 This explains why distributed dividends and interest
income did not return to their pre–World War I and interwar levels (as
a percentage of household income) until the 1980s–90s. Next, several
decades were needed for the real value of rents to recover from the
1914–50 inflation. Here again, one needs to wait until the 1980s–90s to
see the rent index/CPI ratio and the share of rents in household income
returning to their pre–World War I levels.26 These time lags demonstrate
the importance of the 1914–45 shocks. But the key point is that aggre-
gate capital income has now fully recovered from these shocks, whereas
top capital incomes did not recover.
One could also wonder whether the decline of top capital incomes
could simply be the consequence of fiscal manipulation and tax evasion.
I have performed two kinds of checks in order to make sure that fiscal
manipulation and tax evasion can be only a small part of the story (at
most) and that the observed trends do indeed describe a real economic
phenomenon.
First, I adjusted the capital income figures reported in tax returns so
as to match the capital income totals coming from national accounts.
The general conclusion is that the observed trends are simply too large
to be explained by this kind of factor. Whatever way one makes the
adjustment, the trends are still very large (for detailed computations,
see Piketty [2001a, pp. 408–48]). In fact, all available information sug-
gests that tax evasion in France has never been as high as in the interwar
period, that is, at the time in which reported incomes at the very top
of the distribution were much higher than they were in the 1990s. If
one looks at the (tax return capital income ⫹ legally tax-exempt capital
income)/national accounts capital income ratios, which can be viewed
as a measure of tax evasion, then one finds ratios over 90 percent for
the 1980–90s versus 60–70 percent for the interwar period. This is con-
sistent with the fact that the tax administration had much less investi-
24
High retained earnings during the 1950s–60s were due primarily to the high invest-
ment needs of companies. This situation was exacerbated by the fact that retained earnings
were close to zero during the 1930s (i.e., companies did not cut dividends as much as
they should have during the Great Depression) (see Malissen 1953; Piketty 2001a, pp.
62–63).
25
The fall in the profit share was due primarily to the big wage push of the 1970s (the
minimum wage was increased by 130 percent in real terms between 1968 and 1982–83,
whereas GNP increased by only 40 percent!). The profit share started recovering when
wages were frozen in 1982–83.
26
One key reason why it took so long is that French landlords can (partially) adjust
their rent to market conditions only when they have a new tenant. Note that high inflation
(wage-driven) during the 1970s temporarily halted this recovery process (the same as for
dividends).

gative power before World War II than it has today. Tax evasion therefore
seems to amplify the trends rather than to reduce them.27
Next, I have used inheritance tax return data in order to test whether
the leveling of fortunes is a real economic phenomenon. The results
are spectacular (see fig. 5). Whereas the average estate left by the fractile
P90–95 of the estate distribution has been multiplied by about 3.2 in
real terms between 1900–1910 and the 1990s, the average estate left by
the fractile P99.99–100 of the estate distribution during the 1990s is
only one-fourth of what it was in 1900–1910. The decline in capital
concentration seems truly astonishing. Inheritance tax returns are ob-
viously subject to fiscal manipulation and tax evasion, but the trends
are so enormous that these explanations can be only a small part of the
story. One would need to assume that the reporting rate was 100 percent
at the beginning of the twentieth century and less than 10 percent at
the end of the twentieth century! This does not seem plausible. More-
over, in the same way as for income tax returns, it is likely that tax
evasion was actually larger at the beginning of the twentieth century
and during the interwar period than later in the century. It is also
important to note that the inheritance tax and the gift tax were unified
in France in 1942. One important consequence is that my pre-1942 top
estates estimates exclude inter vivos gifts, whereas my post-1942 estimates
do include inter vivos gifts. This again tends to amplify the trend rather
than to reduce it (inter vivos gifts were already quite important at the
beginning of the twentieth century).
Inheritance series show that the decline in top fortunes is the con-
sequence of a decreased concentration of wealth and not of a decline
in aggregate wealth in the economy as a whole. Top estates never re-
covered from the shocks, but lower estates did recover perfectly well
and were able to compensate the fall in top estates. This is consistent
with macroeconomic estimates showing that the capital stock/national
income ratio was about five in the late 1990s, that is, at about the same
level as at the eve of World War I.28 In other words, both capital income
and the capital stock have returned to their pre–World War I levels. The
distribution has changed, not the aggregates.
Although the French tax administration did not compile inheritance
27
I have also checked that legally tax-exempt capital income (which has become more
and more important over time) and capital gains (which were excluded from my basic
series altogether) can be only a small part of the story. For instance, tax return data show
that capital gains represent an average income supplement of about 25 percent for fractile
P99.99–100 (see Piketty 2001a, pp. 420–31; app. A, pp. 586–88). This is a nonnegligible
amount in absolute terms, but this is not going to explain why the income share of fractile
P99.99–100 has been divided by five during the twentieth century.
28
For the 1999 figures, see INSEE (2001, pp. 34, 38): 36,583/6,951p5.2. The capital
stock estimate for 1999 is not fully homogeneous with the estimates given above for 1913,
1934, and 1949, but the orders of magnitude seem right.

Fig. 5.—The average estate left by the fractiles P90–95 and P99.99–100 in France, 1902–94 (1998 French francs). Source: Author’s computations
based on inheritance tax returns (see Piketty 2001a, app. J, table J-9, p. 763).
tax tables until 1901, a number of inheritance series (based on samples
of tax returns collected by historians) are available for the nineteenth
century. Those series show that wealth concentration increased sharply
in France between 1815 and 1914 (top estates rose more than lower
estates) and that wealth inequality did not start declining until World
War I. This seems to confirm my “accidental” interpretation of the in-
equality decline: no “spontaneous” downward trend was taking place
before the shocks.29
Finally, there is plenty of anecdotal evidence suggesting that the de-
cline of top capital incomes is indeed a real economic and social phe-
nomenon. Individuals living off large capital incomes were plentiful in
the literature of the nineteenth century and the early twentieth century
(see, e.g., the novels by Stendhal, Balzac, and Proust), whereas they have
virtually disappeared from the literary scene since World War II. It is
also interesting to note that “rentiers” disappeared from French census
questionnaires in 1946: since the 1946 census, one can no longer de-
scribe oneself as a rentier (this category was used in all censuses through
1936). Another interesting piece of evidence is the evolution of the
number of household workers and domestic servants. At the eve of
World War I, household workers and domestic servants were very nu-
merous in France: about 900,000 to 1 million according to the censuses,
that is, around 5 percent of the labor force. This number fell suddenly
in the aftermath of World War I and during the 1930s (down to about
700,000, 3.5 percent of the labor force), and even more so in the af-
termath of World War II. The number of household workers and do-
mestic servants has stabilized around 200,000 since the 1950s–60s, that
is, about 1 percent of the labor force, one-fifth the number at the eve
of World War I.30 The parallelism between this evolution and the evo-
lution of top income shares is striking. It is particularly important to
note that the number of household workers and domestic servants was
relatively stable at the eve of World War I. The obvious interpretation
is that this number suddenly started falling together with the number
of wealthy households that could afford having domestic servants.31
29
Inheritance series for the nineteenth century can be found in Daumard (1973) and
Bourdieu, Postel-Vinay, and Suwa-Eisenmann (2001). Morrisson (2000) reports top income
share estimates according to which income inequality declined somewhat in France be-
tween 1860 and 1900. But these estimates are based on macroeconomic data alone and
do not take into account the rise in wealth concentration that took place during this
period. On these issues, see Piketty (2001a, pp. 535–42).
30
For detailed series on the number of household workers and domestic servants since
the 1901 census, see Piketty (2001a, app. H, pp. 726–28).
31
The labor cost of domestic servants has increased at a slightly higher rate than per
capita income in the long run (see Piketty 2001a, pp. 86–87), but the gap seems far too
small to explain why the number of domestic servants was divided by five across the century.
In any case, labor costs cannot explain why the number of servants dropped so suddenly
after World War I (there was no sudden variation in labor costs).

D. The Role of Progressive Taxation
How can one account for the fact that large fortunes never recovered
from the 1914–45 shocks, whereas smaller fortunes did recover perfectly
well? The most natural and plausible candidate for an explanation seems
to be the creation and the development of the progressive income tax
(and the progressive inheritance tax). The large fortunes that generate
the top capital incomes observed at the beginning of the twentieth
century were accumulated during the nineteenth century, at a time in
which progressive taxation did not exist and capitalists could use almost
100 percent of their pretax income to consume and to accumulate.32
The conditions faced by twentieth-century capitalists to recover from
the shocks incurred during the 1914–45 period were quite different.
The top marginal rate of the income tax in France was set to only 2
percent in 1915, but it quickly reached very high levels (over 60 percent)
during the interwar period, and it stabilized around 60–70 percent after
1945. These high marginal rates applied to only a small fraction of
incomes, but the point is that they were to a large extent designed to
hit the incomes of the top 1 percent (and even more so the top 0.1
percent and 0.01 percent) of the income distribution, that is, the in-
comes that depend primarily on capital income and capital accumula-
tion. Effective average tax rates have always been fairly moderate at the
level of fractile P90–95: less than 1 percent during the interwar period
and between 5 percent and 10 percent since World War II. In contrast,
effective average tax rates borne by fractile P99.99–100 reached 30 per-
cent during the interwar period and stabilized around 40–50 percent
since World War II (see fig. 6).33 It is therefore not surprising that
progressive taxation had a substantial impact on capital accumulation
at the very top and a negligible impact for smaller fortunes.
Needless to say, these numbers are not sufficient to prove in a rigorous
way that the dynamic effects of progressive taxation on capital accu-
mulation and pretax income inequality have the “right” quantitative
magnitude to account for the observed facts. One would need to know
more about the savings rates of capitalists, how their accumulation strat-
32
Before the creation of a progressive income tax in 1914, personal taxation relied on
individual characteristics such as housing rents, the number of doors and windows, etc.
Effective tax rates were roughly proportional and never exceeded 3–4 percent of income
(see Caillaux 1910, pp. 208–9; Piketty 2001a, pp. 236–39). Note also that there did exist
an inheritance tax during the nineteenth century, but it was purely proportional and the
rate was only 1 percent (see below).
33
The large year-to-year variations in fig. 6 (especially for top incomes) show how chaotic
the history of the income tax has been in France. For instance, the 1968 and 1981 spikes
correspond to the large tax increases on the rich that were voted in the aftermath of the
1968 general strike and of the 1981 socialist electoral victory. I offer a detailed historical
account of these politico-economic developments over the 1914–98 period in Piketty
(2001a, chap. 4, pp. 233–334).

Fig. 6.—Effective average income tax rates in France, 1915–98. Source: Author’s computations based on income tax returns and income tax laws
(see Piketty 2001a, app. B, table B-20, pp. 636–37).
egies have changed since 1945, and so forth. Note, however, that the
orders of magnitude do not seem unrealistic, especially if one assumes
that the owners of large fortunes, whose pretax incomes and lifestyles
were already severely hit by the 1914–45 shocks, were not willing to
reduce their consumption down to very low levels and to increase their
savings so as to counteract the rise in tax rates.34
In fact, in the most standard economic models of capital accumula-
tion, the behavioral response tends to amplify (and not to counteract)
the rise in tax rates. That is, a rise in tax rates imposed on very top
incomes leads wealthy taxpayers to increase their consumption and to
reduce their savings. In the Barro-Becker dynastic model of capital ac-
cumulation, this behavioral effect is so large that large fortunes com-
pletely disappear in the long run. Progressive taxation leads to a trun-
cated wealth distribution in the long run, in the sense that there is
nobody above the top marginal rate threshold (for a formal proof of
this result, see Piketty [2001b, pp. 30–32]). In less extreme and more
realistic models of capital accumulation, the impact of progressive tax-
ation is smaller (large fortunes do not completely disappear). But the
impact is still substantial. For instance, simple computations show that
a capitalist will deplete his wealth at a very high rate if he keeps the
same consumption after progressive taxation is introduced. In the ab-
sence of taxation (say, before World War I), the capital stock of a cap-
italist consuming each year the full return (say, 5 percent) to his capital
stock is stationary. But if an effective tax rate of 30 percent is suddenly
introduced (say, in the interwar period) and if this capitalist keeps con-
suming the full before-tax return to his capital stock, then he will need
to consume some of his capital stock each year: 18 percent of the initial
capital stock is destroyed after 10 years, 42 percent after 20 years, and
so forth, and there is no capital left after 35 years.35
Consider now the more interesting case of a capitalist (or a would-
be capitalist) in 1945, and assume that this capitalist is ready to devote
a large fraction of his income to capital accumulation. How much can
he accumulate in 50 years? The point is that progressive taxation dras-
tically reduces the assets that one can accumulate, including for capi-
talists adopting relatively low living standards (see table 2). For instance,
34
Existing evidence shows that the negative shocks incurred between 1914 and 1945
and the rise in progressive taxation induced wealthy French families to reduce drastically
their savings rate between 1873–1913 and 1946–53 (see Perrot 1961). Note, however, that
this research by Perrot relies on a few hundred private account books from wealthy French
families, and it would need to be supplemented by extensive new research based on larger
samples.
35
This cumulative process would take place at an even faster pace in case of higher
returns or higher tax rates (see Piketty 2001b, table 3). This mechanism is trivial, but I
believe that it did contribute to amplifying the shocks incurred by capital owners during
the 1914–45 period.

TABLE 2
Impact of Progressive Taxation on Capital Accumulation
rp5%, rp5%, rp5%, rp10%, rp10%, rp10%,

tp0% tp30% tp50% tp0% tp30% tp50%
cp100% 1.0 .0 .0 1.0 .0 .0
cp80% 3.1 .3 .0 24.3 .0 .0
cp60% 5.2 1.7 .5 47.6 5.1 .0
cp40% 7.3 3.0 1.5 70.8 13.2 3.1
cp20% 9.4 4.3 2.5 94.1 21.3 7.3
cp0% 11.5 5.6 3.4 117.4 29.5 11.5
Note.—This table reads as follows: Assume that a capitalist’s consumption level is equal to a fixed fraction c (say,
cp20 percent) of the full return r (say, rp5 percent) to his capital stock. In the absence of taxation (tp0 percent),
his capital stock will be multiplied by 9.4 after 50 years; with an effective tax rate of tp50 percent, his capital stock will
be multiplied by 2.5 after 50 years (I assume that the capitalist keeps the same absolute consumption level during 50
years). The corresponding formula is given by
xn p
c
1⫺t (
⫹ [1 ⫹ (1 ⫺ t)r]n # 1 ⫺
c
1⫺t ).
with a 5 percent before-tax return and a consumption level equal to 40

percent of the before-tax return to the initial capital stock, one can
accumulate in 50 years a fortune that is about five times as large with
a 0 percent tax rate as with a 50 percent tax rate. That is, the initial
capital stock is multiplied by 7.3 after 50 years in the absence of taxation,
and the initial capital stock is multiplied by only 1.5 with a tax rate of
50 percent. This tax rate of 50 percent corresponds approximately to
the average effective tax rates faced by fractile P99.99–100 in France
since World War II, and the factor of five corresponds approximately
to the secular decline in the income share of fractile P99.99–100.
Note also that these simple simulations do not take into account the
impact of the progressive inheritance tax. During the nineteenth cen-
tury, the French inheritance tax was strictly proportional, with a fixed
1 percent tax rate. A progressive inheritance tax was introduced in 1901,
but tax rates remained low until World War I: at the eve of the war, top
tax rates did not exceed 5 percent. In the same way as with the pro-
gressive income tax, the top rates of the progressive inheritance tax
suddenly reached nontrivial levels in the aftermath of World War I. One
can compute that the effective tax rate faced by fractile P99.99–100 of
the estate distribution was about 20–25 percent during the interwar
period (or even 30–35 percent during the early 1920s), 30–35 percent
during the 1950s, 15–20 percent during the 1960s–70s, and again 30–35
percent during the 1980s–90s (see Piketty 2001a, app. J, pp. 767–71).
Note, however, that the long-run impact of the progressive inheritance
tax on capital accumulation, though important, has probably been less
drastic than the impact of the progressive income tax. Because the
income tax applies every year and has cumulative effects, an effective
income tax rate of 50 percent can reduce by a factor of five the size of

fortunes that one can accumulate in 50 years. In contrast, under the
assumption that the inheritance tax is paid once every 50 years (on
average), an effective inheritance tax rate of 50 percent reduces by a
factor of two the size of fortunes that one can accumulate in 50 years.
Finally, it is worth emphasizing that it is not that easy to find con-
vincing explanations (other than the introduction of progressive taxa-
tion) that can account for the nonrecovery of large fortunes. For in-
stance, explanations based on hypothetical changes in before-tax returns
to capital do not seem to work. All capital holders should have been
hit by a reduction in before-tax asset returns. The point is that large
fortunes were unable to recover from the 1914–45 shocks, whereas for-
tunes that were slightly smaller did recover perfectly well. One needs
an explanation that applies only to the top of the distribution and
nowhere else, and progressive taxation looks like an obvious candidate.
Another possible explanation would be the existence of a large public
sector in France after the nationalizations of 1945. But the negative
impact on private capital accumulation would seem to apply to all capital
holders, or at least to broader segments of the wealth distribution than
simply the very top. Moreover, one should not exaggerate the impor-
tance of the public sector in postwar France. For instance, the output
share of nationalized firms was around 15–20 percent in the manufac-
turing sector during the postwar period.36 This is a substantial share in
absolute terms, but this does not seem sufficient to explain the mag-
nitude of the observed trends. Although there was a public sector in
postwar France, the point is that private capital accumulation could
freely take place in at least 80–85 percent of the manufacturing sector.
It is also interesting to note that Carré, Dubois, and Malinvaud (1972,
pp. 614–15), in their standard account of postwar growth in France,
have pointed out that the bulk of the growth performance came from
manufacturing subsectors in which there was almost no nationalized
firm. This suggests that there were plenty of economic opportunities to
accumulate large fortunes with little interference with the public sector.
If it is assumed that the rise of progressive taxation is indeed the right
explanation for the observed facts (or at least for a significant fraction
of the observed facts), what was the economic impact of the nonrecovery
of large fortunes? More generally, what were the consequences for the
performance of the French economy of the shocks incurred by capital
owners during the 1914–45 period and the structural decline in the
concentration of wealth? It is obviously very difficult to give a satisfactory
36
According to estimates given by Delion and Durupty (1982, p. 191), this output share
was around 15–20 percent between 1945 and 1982, and it shortly reached 30 percent
between 1982 and 1986 (following the nationalizations of 1982), before being drastically
reduced following the privatizations of 1986–87. Nationalized firms have been privatized
one by one since 1986–87, and the public-sector share is now converging toward 0 percent.

answer to such a complex question. One could try to construct a his-
torical micro data base on French firms so as to compare the growth
performance of firms with different levels of capital dispersion and
different levels of exposure to shocks during the 1914–45 period. In
the meantime, one can make a number of simple remarks on the basis
of available macroeconomic data.
First of all, the decline in wealth concentration does not seem to have
been an obstacle to growth. Growth rates were extremely high from the
late 1940s to the 1970s, and this period is now referred to as the Trente
Glorieuses (the “30 Glorious Years”) in France.37 Needless to say, these
very high growth rates are to a large extent the consequence of the
abysmal economic performance of the 1914–45 period (which was itself
the consequence of the two world wars and the Great Depression).
During the Trente Glorieuses, France was simply catching up with the most
advanced capitalist countries, and in particular with the United States.
According to Maddison’s (1995, pp. 194–97) estimates, the ratio be-
tween U.S. GDP per capita and French GDP per capita (both expressed
in purchasing power parity terms) was about 1.4–1.5 at the eve of World
War I, up to 1.8 in 1950, and down to 1.2–1.3 in the late 1970s (this
ratio stabilized around 1.2–1.3 during the 1980s–90s). Of course, one
cannot rule out the possibility that French growth rates would have been
even higher during the Trente Glorieuses if capital concentration had
remained at the same level as in 1914. Note, however, that several mac-
roeconomic historians have suggested that the decline in wealth con-
centration might have had a positive growth impact. For instance, Carré
et al. (1972, pp. 457–59, 620) have pointed out that wealth redistribution
during the 1914–45 period (in particular the inflation-induced redis-
tribution from creditors to debtors) might have favored the develop-
ment of new firms and new generations of entrepreneurs. In the pres-
ence of credit constraints, high capital concentration can indeed entail
negative consequences for productive efficiency, and wealth redistri-
bution under certain conditions can have positive efficiency effects. This
is all very hypothetical, however, and extensive research based on new
micro data sets would be necessary to test these hypotheses.
It is also important to emphasize that the rise of progressive taxation
had apparently no negative impact on aggregate capital accumulation.
As was already noted above, the capital stock/national income ratio
seems to have fully recovered from the 1914–45 shocks, with a ratio
around five both at the eve of World War I and in the late 1990s (see
subsection C above). That is, the fall of large fortunes was compensated
37
The term Trente Glorieuses was coined by Fourastié (1979). Average real household
income in France grew at about 5 percent per year between 1948 and 1978 (see Piketty
2001a, p. 72).

by rapid accumulation at intermediate and moderately high wealth lev-
els, so that the structural decline in capital concentration seems to have
had little impact on the average capital stock. It is interesting to note
that this is exactly what the Barro-Becker dynastic model of capital ac-
cumulation would predict. In the presence of progressive taxation, dy-
nastic preferences with a fixed rate of time preference imply that capital
deaccumulation by the wealthy will be fully compensated by increased
accumulation from individuals with lower wealth (for a formal proof,
see Piketty [2001a, pp. 30–32]). This does mean, however, that there is
no efficiency cost: aggregate capital stock will recover in the long run,
but it might well be inefficiently low during the transition. The analysis
of the efficiency properties of progressive taxation in less extreme and
more realistic models of capital accumulation is an issue that would
deserve further research.
Finally, it is important to note that although progressive taxation
seems to have had a substantial dynamic impact on capital concentra-
tion, its static impact on income inequality has been more moderate.
During the 1990s, the after-tax top decile income share was quite close
to the before-tax top decile share (30 percent vs. 33 percent). This
reflects the fact that effective income tax rates have always been fairly
moderate for the vast majority of top decile taxpayers (e.g., effective tax
rates have never exceeded 5–10 percent at the level of fractile P90–95).
It is not surprising that the impact is larger for higher incomes: during
the 1990s, the after-tax top percentile income share is about 25 percent
smaller than the before-tax top percentile income share (6 percent vs.
8 percent). At the level of fractile P99.99–100, after-tax income shares
are more than 40 percent smaller than before-tax income shares during
the 1990s (0.35 percent vs. 0.6 percent).38 It looks as though progressive
taxation was designed to hit top capital incomes rather than to reduce
drastically the top decile income share as a whole.39
V. How Specific Is the French Experience?

Available historical series on income inequality in other European coun-
tries and in the United States are too scarce and incomplete to draw
definite conclusions about the differences and similarities between the
French experience and other developed countries’ experiences.
Existing European series are particularly fragile. For most European
countries (and in particular for Germany and for the United Kingdom),
38
Series on after-tax income shares were computed by applying effective tax rate series
to pretax income share series (see Piketty 2001a, table B22, pp. 640–41).
39
This conclusion would not be dramatically altered by the inclusion of nontaxable
income transfers (most income transfers [pensions, unemployment benefits, etc.] are
taxable and are therefore already taken into account in the before-tax series).

there are only a couple of heterogeneous estimates of income inequality
covering only a small number of years over the course of the twentieth
century.40 Note, however, that there are two key points on which all
existing estimates seem to be consistent with my French findings. First,
the secular decline in the top decile income share seems to have oc-
curred in all European countries during a specific time period, that is,
between 1914 and 1945 (and especially during the 1930s and World
War II). Next, the substantial 1914–45 decline in the top decile share
seems to be due for the most part to the top percentile share.41 Existing
estimates also suggest that countries with greater war destruction ex-
perienced a larger decline of their top centile income share (e.g., total
decline was apparently larger in Germany than in the United Kingdom),
which again is consistent with my explanation. This would seem to imply
that the 1914–45 decline in inequality in all European countries was an
accidental, capital income phenomenon (for the most part).
The U.S. case is particularly interesting, especially since available se-
ries are less scarce than for European countries. Kuznets’s series appear
to be very much in line with my French findings. Kuznets (1953) used
U.S. tax return statistics to construct annual 1913–48 series on top in-
come shares, and these series constitute the most valuable source of
information on U.S. inequality dynamics during the first half of the
twentieth century. Kuznets’s series show that the significant decline in
the top decile income share that took place between 1913 and 1948 is
almost entirely due to the sharp decline in the top percentile income
share. The total decline of the top percentile income share, though
very significant, seems smaller than what I found in France. This is
consistent with the capital income explanation: world wars induced a
much more severe shock on capital holders in France than in the United
States (unlike the Great Depression of the 1930s, which was more severe
in the United States). Kuznets’s series also confirm that the decline in
40
For a recent survey on historical research on inequality in European countries, see
Morrisson (2000). The only European countries for which annual, long-term income
inequality series are available seem to be Denmark and the Netherlands (unfortunately,
these series do not offer a complete decomposition of the top decile; nor do they offer
composition estimates or separate estimates for wage inequality). In particular, Morrisson
offers only three (very heterogeneous) estimates of income inequality for twentieth-century
France: one for 1900 (this is the 1900–1910 estimate referred to above); one for 1929
(this estimate comes from Sauvy [1965–75], who gives no details about his sources and
methodology; this estimate is vastly inconsistent with the income tax return data for 1929:
for instance, Sauvy underestimates the number of incomes above 600,000 francs by a factor
of four; see Piketty [2001a, app. I, pp. 741–42]); and one for 1975 (this estimate comes
from a 1975 income survey). The estimates reported for Germany and the United Kingdom
suffer from the same limitations.
41
Existing series do not usually offer a complete decomposition of the top decile income
share (see the estimates reported by Morrisson [2000]). But whenever such decomposi-
tions are available, top fractile income shares account for a disproportionate share of the
total decline in the top decile share.

inequality was not a linear, continuous process: the top percentile in-
come share dropped during World War I, recovered during the 1920s,
and dropped again during the Great Depression and World War II.
Unfortunately, Kuznets did not attempt to go beyond the top per-
centile (nor did he give systematic estimates of income composition for
the various fractiles). Most important, Kuznets did not construct sepa-
rate series for wage inequality (there was no separate wage tax in the
United States, so the data are less rich than in France). It is therefore
impossible to undertake the same kind of test that I did for France. In
particular, it is impossible to know whether U.S. wage inequality declined
significantly during the 1900–1950 period (which would mean that what
happened was not just an accidental capital income phenomenon).
Since the time of Kuznets, several economists have collected long-term,
occupational wage data in order to shed light on this issue.42 These data
do show that there was significant wage compression during both World
War I and World War II (as in France). However, these data do not
allow any strong conclusion regarding the existence of a more general
equalizing trend during the 1900–1950 period.43 At this stage, the only
well-established part of the story appears to be the capital income side.
It is interesting to note that Kuznets himself, in his 1955 article, started
by proposing an interpretation of his 1953 series that was very much in
line with the capital-income interpretation that I have advocated in this
paper. Kuznets emphasized the shocks incurred by capital owners during
the 1914–45 period, and he mentioned explicitly the dynamic impact
of progressive taxation on capital accumulation and income inequality.
But, by the end of his article (which was also his presidential address
to the American Economic Association), Kuznets formulated a com-
pletely different theory. He argued that there could well exist an en-
dogenous mechanism forcing inequality to decline in advanced capitalist
countries: in a two-sector model of economic development, one should
indeed observe inequality to rise when only a small fraction of the
population benefits from the incomes generated by the high-produc-
tivity sector and to decline when most workers join the high-productivity
42
See, e.g., Williamson and Lindert (1980), Goldin and Margo (1992), and Goldin and
Katz (1999). For a recent survey, see Lindert (2000).
43
Given the large changes in workforce composition, it is problematic to use occupa-
tional wage ratios to analyze long-run trends in wage inequality. In France, the ratio
between the average wage of managers and the average wage of production workers has
declined enormously in the long run (during both the 1900–1950 and the 1950–98 pe-
riods), although the top decile and top percentile wage shares have been roughly constant
(the explanation for this paradox is simply that the number of managerial jobs has in-
creased a lot; see Piketty [2001a, pp. 203–10]). To my knowledge, there does not exist
any U.S. wage inequality series expressed in terms of fractiles prior to 1940 (starting in
1940, censuses asked a question on wages).

44
sector. Kuznets had basically no empirical evidence to support this
theory: this “is perhaps 5 per cent empirical information and 95 per
cent speculation, some of it possibly tainted by wishful thinking” (1955,
p. 26). Although this optimistic theory quickly became popular, it is
important to recall that the theory of the Kuznets curve is not supported
by Kuznets’s series. Kuznets himself believed more strongly in the effect
of shocks and progressive taxation than in the Kuznets curve, and the
first part of his theory seems to have been overly neglected by
economists.
Regarding the more recent period, there exists one important diver-
gence between U.S. and French inequality dynamics. Top income shares
have been increasing sharply in the United States since the 1970s (see
Feenberg and Poterba 1993, 2000), whereas my series show that they
have been flat in France. The very steep rise in top incomes observed
in the United States since the 1970s seems to be due to large increases
in high-skill wages and executive compensation. The large decline in
top tax rates observed in the United States since the 1970s also provides
a test for the theory of progressive taxation and capital accumulation.
One should expect the decline in top tax rates to facilitate the accu-
mulation of large fortunes and the resurgence of top capital incomes
during the next few decades. This issue deserves more attention in future
research.45
VI. Concluding Comments

In this paper, I have presented new inequality series on France during
the twentieth century. The main conclusion is that the decline in income
inequality that took place during the first half of the twentieth century
was mostly accidental. In France, and possibly in a number of other
developed countries as well, wage inequality has actually been extremely
stable in the long run, and the secular decline in income inequality is
for the most part a capital income phenomenon: holders of large for-
tunes were badly hurt by major shocks during the 1914–45 period, and
they were never able to fully recover from these shocks, probably because
of the dynamic effects of progressive taxation on capital accumulation
and pretax income inequality.
More research is needed in order to better understand the deter-
minants of long-run inequality dynamics. First, it would be useful to
44
Kuznets also mentioned that with a higher variance of earnings in the urban sector,
it might take a long time before inequality starts declining (and it might not decline at
all).
45
After the present study on France was completed, Emmanuel Saez and I constructed
similar long-run series for the United States (see Piketty and Saez, in press). These series
broadly confirm the interpretation advocated in the present paper.

construct similar inequality series for other countries. The raw statistical
materials that I have used to construct my French series are to some
extent available in other countries, and these materials have been un-
derused by economists so far. Next, the dynamic interplay between pro-
gressive taxation, capital accumulation, and income inequality would
need to be analyzed more carefully, from both an empirical and a the-
oretical standpoint. I hope that the empirical findings presented in this
paper will contribute to the stimulation of future research in this area.
Appendix
TABLE A1
Top Income Shares in France, 1900–1998: Top Fractiles
P90–100 P95–100 P99–100 P99.5–100 P99.9–100 P99.99–100

1900–1910 45.00 34.00 19.00 15.00 8.00 3.00
1915 18.31 14.49 7.90 3.03
1916 20.65 16.52 9.39 3.79
1917 20.09 16.05 8.89 3.44
1918 17.95 14.28 7.67 2.87
1919 42.25 33.84 19.50 15.36 8.26 2.81
1920 39.59 31.41 17.95 14.12 7.63 2.86
1921 39.70 31.04 17.32 13.49 7.23 2.65
1922 41.54 32.50 17.87 13.84 7.26 2.51
1923 43.54 34.15 18.91 14.68 7.61 2.61
1924 42.14 32.27 17.96 13.91 7.05 2.39
1925 44.07 33.63 18.16 14.00 7.07 2.38
1926 42.06 32.34 17.82 13.73 6.98 2.41
1927 42.95 32.47 17.45 13.43 6.87 2.35
1928 42.75 32.19 17.27 13.24 6.77 2.33
1929 41.59 30.90 16.15 12.39 6.25 2.16
1930 41.08 30.14 15.31 11.59 5.79 1.93
1931 41.12 29.67 14.63 10.95 5.37 1.77
1932 43.44 31.06 14.80 10.89 5.22 1.67
1933 44.87 31.95 14.95 10.92 5.20 1.69
1934 46.01 32.68 15.28 11.17 5.31 1.71
1935 46.61 33.10 15.40 11.21 5.31 1.74
1936 44.10 31.58 14.74 10.77 5.17 1.74
1937 42.90 30.21 14.46 10.67 5.24 1.83
1938 42.52 29.79 14.27 10.49 5.05 1.75
1939 38.24 27.21 13.30 9.98 4.99 1.73
1940 39.11 27.85 13.35 9.89 4.90 1.65
1941 38.70 27.37 12.88 9.33 4.27 1.30
1942 35.04 24.90 11.53 8.26 3.64 1.06
1943 32.26 22.68 10.13 7.13 3.01 .84
1944 29.42 20.18 8.37 5.75 2.32 .61
1945 29.70 19.58 7.54 5.04 1.96 .51
1946 32.87 22.34 9.22 6.35 2.61 .72
1947 33.20 23.05 9.22 6.31 2.59 .68
1948 32.35 21.46 8.75 6.00 2.43 .63
1949 32.20 21.70 9.01 6.25 2.61 .70
1950 31.97 21.62 8.98 6.23 2.60 .70
1951 32.93 22.06 9.00 6.19 2.55 .68

TABLE A1
(Continued )
P90–100 P95–100 P99–100 P99.5–100 P99.9–100 P99.99–100

1952 33.19 22.35 9.16 6.27 2.53 .65
1953 32.89 22.10 9.00 6.13 2.48 .65
1954 33.53 22.55 9.14 6.20 2.45 .64
1955 34.42 23.16 9.33 6.30 2.48 .65
1956 34.36 23.11 9.37 6.29 2.46 .65
1957 34.74 23.38 9.37 6.28 2.44 .64
1958 34.05 22.76 9.01 6.02 2.34 .60
1959 35.88 24.14 9.46 6.27 2.37 .60
1960 36.11 24.40 9.71 6.48 2.45 .62
1961 36.82 24.92 9.88 6.57 2.48 .64
1962 35.88 24.16 9.46 6.25 2.34 .58
1963 36.41 24.43 9.43 6.19 2.29 .56
1964 36.84 24.75 9.56 6.28 2.30 .56
1965 37.15 24.94 9.58 6.27 2.30 .56
1966 36.46 24.41 9.36 6.14 2.26 .57
1967 36.21 24.27 9.36 6.16 2.29 .59
1968 34.80 23.08 8.77 5.76 2.15 .56
1969 33.96 22.48 8.55 5.61 2.09 .55
1970 33.14 21.95 8.33 5.45 2.02 .53
1971 33.35 22.10 8.47 5.57 2.07 .53
1972 33.03 21.97 8.52 5.63 2.11 .55
1973 33.90 22.61 8.87 5.90 2.26 .62
1974 33.33 22.09 8.50 5.60 2.09 .53
1975 33.41 22.06 8.48 5.56 2.08 .54
1976 33.19 21.91 8.44 5.53 2.08 .54
1977 31.68 20.71 7.79 5.11 1.94 .51
1978 31.38 20.56 7.80 5.11 1.93 .50
1979 31.03 20.42 7.82 5.15 1.97 .52
1980 30.69 20.11 7.63 5.01 1.91 .50
1981 30.73 20.04 7.55 4.95 1.89 .50
1982 29.93 19.37 7.07 4.61 1.72 .44
1983 30.43 19.53 6.99 4.51 1.63 .40
1984 30.52 19.57 7.03 4.51 1.65 .41
1985 31.05 19.96 7.20 4.66 1.70 .43
1986 31.39 20.30 7.44 4.85 1.81 .46
1987 31.73 20.66 7.75 5.13 1.98 .53
1988 32.09 20.90 7.92 5.28 2.06 .57
1989 32.42 21.31 8.21 5.51 2.20 .62
1990 32.64 21.45 8.23 5.52 2.20 .62
1991 32.44 21.18 7.97 5.30 2.07 .57
1992 32.23 20.90 7.75 5.12 1.97 .54
1993 32.22 20.81 7.65 5.05 1.94 .53
1994 32.37 20.90 7.71 5.10 1.98 .55
1995 32.41 20.93 7.70 5.08 1.96 .54
1996 32.25 20.79 7.59 5.01 1.92 .53
1997 32.42 20.93 7.70 5.10 1.98 .55
1998 32.50 20.98 7.72 5.10 1.97 .55
Source.—Author’s computations based on income tax returns (see Piketty 2001a, app. B, table B14, pp. 620–21).
1038

TABLE A2
Top Income Shares in France, 1900–1998: Intermediate Fractiles
P90–95 P95–99 P99–99.5 P99.5–99.9 P99.9–99 P99.99–100

1900–1910 11.00 15.00 4.00 7.00 5.00 3.00
1915 3.82 6.59 4.87 3.03
1916 4.14 7.13 5.60 3.79
1917 4.04 7.16 5.45 3.44
1918 3.68 6.60 4.80 2.87
1919 8.41 14.33 4.15 7.10 5.45 2.81
1920 8.18 13.46 3.83 6.49 4.77 2.86
1921 8.66 13.72 3.83 6.26 4.58 2.65
1922 9.04 14.63 4.03 6.58 4.74 2.51
1923 9.38 15.25 4.22 7.08 4.99 2.61
1924 9.86 14.31 4.05 6.86 4.66 2.39
1925 10.44 15.47 4.16 6.93 4.69 2.38
1926 9.72 14.52 4.09 6.75 4.58 2.41
1927 10.48 15.02 4.02 6.56 4.52 2.35
1928 10.56 14.92 4.03 6.47 4.44 2.33
1929 10.69 14.75 3.77 6.13 4.09 2.16
1930 10.94 14.83 3.72 5.80 3.86 1.93
1931 11.45 15.04 3.69 5.57 3.61 1.77
1932 12.38 16.26 3.90 5.68 3.54 1.67
1933 12.92 17.00 4.02 5.72 3.51 1.69
1934 13.33 17.39 4.12 5.86 3.60 1.71
1935 13.50 17.71 4.19 5.90 3.57 1.74
1936 12.51 16.85 3.97 5.60 3.43 1.74
1937 12.69 15.75 3.79 5.44 3.41 1.83
1938 12.73 15.52 3.78 5.44 3.30 1.75
1939 11.03 13.91 3.32 4.99 3.26 1.73
1940 11.25 14.51 3.45 5.00 3.25 1.65
1941 11.32 14.49 3.55 5.06 2.97 1.30
1942 10.14 13.37 3.27 4.62 2.58 1.06
1943 9.58 12.55 3.00 4.12 2.18 .84
1944 9.24 11.81 2.62 3.43 1.71 .61
1945 10.12 12.04 2.50 3.08 1.45 .51
1946 10.52 13.12 2.88 3.73 1.90 .72
1947 10.16 13.83 2.91 3.72 1.91 .68
1948 10.88 12.71 2.76 3.57 1.80 .63
1949 10.50 12.69 2.76 3.64 1.91 .70
1950 10.35 12.64 2.76 3.62 1.90 .70
1951 10.87 13.05 2.82 3.63 1.88 .68
1952 10.84 13.19 2.89 3.74 1.88 .65
1953 10.80 13.10 2.86 3.65 1.83 .65
1954 10.99 13.41 2.94 3.75 1.81 .64
1955 11.26 13.83 3.02 3.82 1.83 .65
1956 11.25 13.74 3.08 3.83 1.81 .65
1957 11.36 14.01 3.09 3.84 1.80 .64
1958 11.29 13.75 2.99 3.68 1.74 .60
1959 11.74 14.68 3.19 3.90 1.77 .60
1960 11.71 14.69 3.23 4.03 1.83 .62
1961 11.90 15.05 3.31 4.09 1.84 .64
1962 11.71 14.70 3.21 3.92 1.76 .58
1963 11.98 15.00 3.24 3.90 1.73 .56
1964 12.09 15.19 3.28 3.97 1.74 .56
1965 12.21 15.36 3.31 3.97 1.74 .56
1039

TABLE A2
(Continued )
P90–95 P95–99 P99–99.5 P99.5–99.9 P99.9–99 P99.99–100

1966 12.04 15.05 3.22 3.88 1.70 .57
1967 11.93 14.92 3.20 3.86 1.70 .59
1968 11.72 14.31 3.02 3.60 1.60 .56
1969 11.48 13.94 2.94 3.52 1.54 .55
1970 11.19 13.63 2.87 3.44 1.49 .53
1971 11.25 13.63 2.90 3.50 1.54 .53
1972 11.06 13.45 2.89 3.51 1.56 .55
1973 11.29 13.74 2.98 3.64 1.63 .62
1974 11.23 13.59 2.90 3.51 1.55 .53
1975 11.35 13.59 2.92 3.48 1.54 .54
1976 11.28 13.47 2.91 3.45 1.54 .54
1977 10.97 12.92 2.68 3.17 1.43 .51
1978 10.82 12.77 2.69 3.18 1.43 .50
1979 10.62 12.59 2.67 3.18 1.45 .52
1980 10.59 12.47 2.62 3.11 1.41 .50
1981 10.69 12.49 2.61 3.06 1.39 .50
1982 10.56 12.30 2.46 2.89 1.28 .44
1983 10.91 12.53 2.49 2.88 1.23 .40
1984 10.95 12.54 2.51 2.87 1.24 .41
1985 11.09 12.76 2.54 2.95 1.28 .43
1986 11.10 12.86 2.59 3.04 1.34 .46
1987 11.07 12.91 2.62 3.15 1.44 .53
1988 11.19 12.98 2.64 3.21 1.49 .57
1989 11.11 13.10 2.70 3.31 1.57 .62
1990 11.19 13.22 2.71 3.32 1.57 .62
1991 11.26 13.20 2.67 3.23 1.50 .57
1992 11.33 13.15 2.63 3.15 1.43 .54
1993 11.40 13.16 2.60 3.11 1.41 .53
1994 11.47 13.19 2.60 3.13 1.43 .55
1995 11.48 13.23 2.61 3.13 1.42 .54
1996 11.45 13.20 2.58 3.08 1.40 .53
1997 11.49 13.23 2.60 3.12 1.43 .55
1998 11.52 13.27 2.62 3.13 1.42 .55
Source.—Author’s computations based on income tax returns (see Piketty 2001a, app. B, table B15, pp. 621–22).
References
Andrieu, Claire; Le Van, Lucette; and Prost, Antoine. Les nationalisations de la

Libération: De l’utopie au compromis. Paris: Presses Fondation Nationale Sciences
Politiques, 1987.
Baudelot, Christian, and Lebeaupin, Anne. “Les salaires de 1950 à 1975.” Econ-
omie et Statistiques, no. 113 (July/August 1979), pp. 15–22.
Bayet, Alain, and Julhès, Martine. “Séries longues sur les salaires.” INSEE-Résultats
457. Série Emploi-revenus, no. 105. Paris: INSEE, April 1996.
Bourdieu, Jérôme; Postel-Vinay, Gilles; and Suwa-Eisenman, Akiko. “Wealth Ac-
cumulation and Inequality in France, 1800–1940.” Manuscript. Paris: Fédér-
ation Paris-Jourdan, 2001.
Caillaux, Joseph. L’impôt sur le revenu. Paris: Berger-Levrault, 1910.
Carré, Jean-Jacques; Dubois, Paul; and Malinvaud, Edmond. La croissance fran-

çaise: Un essai d’analyse économique causale de l’après-guerre. Paris: Éditions du
Seuil, 1972.
Cornut, Paul. Répartition de la fortune privée en France par département et nature de
biens au cours de la première moitié du XXe siècle. Paris: Armand Colin, 1963.
Daumard, Adeline. Les fortunes françaises au XIXe siècle: Enquête sur le répartition et
la composition des capitaux privés à Paris, Lyon, Lille, Bordeaux et Toulouse d’après
l’enregistrement des déclarations de succession. Paris: Mouton, 1973.
Delion, André G., and Durupty, Michel. Les nationalisations 1982. Paris: Eco-
nomica, 1982.
Divisia, François; Dupin, Jean; and Roy, René. A la recherche du franc perdu. 2 vols.
Paris: Blanchard, 1956.
Dugé de Bernonville, Léo. “Les revenus privés.” In La France économique en 193[x].
Special annual issues of Revue d’Economie Politique 47–53 (May/June 1933–
May/June 1939).
Feenberg, Daniel R., and Poterba, James M. “Income Inequality and the Incomes
of Very High Income Taxpayers: Evidence from Tax Returns.” In Tax Policy
and the Economy, vol. 7, edited by James M. Poterba. Cambridge, Mass.: MIT
Press, 1993.
———. “The Income and Tax Share of Very-High Income Households, 1960–
1995.” A.E.R. Papers and Proc. 90 (May 2000): 264–70.
Fourastié, Jean. Les Trente Glorieuses: Ou, la révolution invisible de 1946 à 1975.
Paris: Fayard, 1979.
Goldin, Claudia, and Katz, Lawrence. “The Returns to Skill across the Twentieth
Century United States.” Manuscript. Cambridge, Mass.: Harvard Univ., Dept.
Econ., 1999.
Goldin, Claudia, and Margo, Robert A. “The Great Compression: The Wage
Structure in the United States at Mid-century.” Q.J.E. 107 (February 1992):
1–34.
Hirsch, Anita. “Le logement.” In Histoire économique de la France entre les deux
guerres, vol. 3, edited by Alfred Sauvy. Paris: Fayard, 1972.
INSEE. “Quelques données statistiques sur l’imposition en France des fortunes
privées.” Etudes Statistiques (quarterly suppl., Bull. Mensuel Statistique) 1 (Jan-
uary–March 1958): 33–37
———. Annuaire statistique de la France 1966—Résumé rétrospectif. Paris: INSEE,
1966.
———. “Compte nationaux: Rapport sur les comptes de la Nation 2000.” INSEE-
Résultats 743. Série Economie générale, no. 189. Paris: INSEE, June 2001.
Kuznets, Simon. Shares of Upper Income Groups in Income and Savings. New York:
NBER, 1953.
———. “Economic Growth and Economic Inequality.” A.E.R. 45 (March 1955):
1–28.
Lindert, Peter H. “Three Centuries of Inequality in Britain and America.” In
Handbook of Income Distribution, vol. 1, edited by Anthony B. Atkinson and
François Bourguignon. Amsterdam: Elsevier Sci., 2000.
Maddison, Angus. Monitoring the World Economy. Paris: OECD, 1995.
Malissen, Marcel. L’autofinancement des sociétés en France et aux États-Unis. Paris:
Libraire Dalloz, 1953.
Morrisson, Christian. “Historical Perspectives on Income Distribution: The Case
of Europe.” In Handbook of Income Distribution, vol. 1, edited by Anthony B.
Atkinson and François Bourguignon. Amsterdam: Elsevier Sci., 2000.
Perrot, Marguerite. Le mode de vie des familles bourgeoises, 1873–1953. Paris: Armand
Colin, 1961.

Piketty, Thomas. Les hauts revenus en France au XXe siècle: Inégalités et redistributions,
1901–1998. Paris: Grasset, 2001. (a)
———. “Income Inequality in France, 1901–1998.” Discussion Paper no. 2876.
London: Centre Econ. Policy Res., 2001. (b)
Piketty, Thomas, and Saez, Emmanuel. “Income Inequality in the United States,
1913–1998.” Q.J.E. (in press).
Sauvy, Alfred, ed. Histoire économique de la France entre les deux guerres. 4 vols. Paris:
Fayard, 1965–75. 2d ed. 3 vols. Paris: Economica, 1984.
Taffin, Claude. “Un siècle de politique du logement.” In La Société française:
Données sociales 1993. Paris: INSEE, 1993.
Villa, Pierre. “Un siècle de données macro-économiques.” INSEE-Résultats, nos.
303–4. Série Economie générale, nos. 86–87. Paris: INSEE, April 1994.
Williamson, Jeffrey G., and Lindert, Peter H. American Inequality: A Macroeconomic
History. New York: Academic Press, 1980.


Piketty 2003

Uploaded by

Copyright:

Available Formats

Piketty 2003

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Piketty 2003

Uploaded by

Copyright:

Available Formats

Income Inequality in France, 1901–1998

Author(s): Thomas Piketty

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

I am grateful to seminar participants at Columbia, Harvard, Massachusetts Institute of

[Journal of Political Economy, 2003, vol. 111, no. 5]

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

II. Data Sources

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

B. Wage Tax Returns (1919–98)

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

III. The Basic Facts

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

IV. Accounting for the Facts

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

A. Income Composition Patterns

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

B. The Long-Run Stability of Wage Inequality

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

C. The Robustness of Wealth Leveling

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

rp5%, rp5%, rp5%, rp10%, rp10%, rp10%,

with a 5 percent before-tax return and a consumption level equal to 40

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

V. How Specific Is the French Experience?

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

VI. Concluding Comments

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

P90–100 P95–100 P99–100 P99.5–100 P99.9–100 P99.99–100

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

P90–100 P95–100 P99–100 P99.5–100 P99.9–100 P99.99–100

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

P90–95 P95–99 P99–99.5 P99.5–99.9 P99.9–99 P99.99–100

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

P90–95 P95–99 P99–99.5 P99.5–99.9 P99.9–99 P99.99–100

Andrieu, Claire; Le Van, Lucette; and Prost, Antoine. Les nationalisations de la

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

This content downloaded from 137.149.200.5 on Sat, 24 Jan 2015 12:42:58 PM

You might also like