Capital in the Twenty-First Century (108 page)

10
. See in particular David Card and Alan Krueger,
Myth and Measurement: The New Economics of the Minimum Wage
(Princeton: Princeton University Press, 1995). Card and Krueger exploited numerous
cases in which neighboring states had different minimum wages. The pure “monopsony”
case is one in which a single employer can purchase labor in a given geographical
area. (In pure monopoly, there is a single seller rather than a single buyer.) The
employer then sets the wage as low as possible, and an increase in the minimum wage
does not reduce the level of employment, because the employer’s profit margin is so
large as to make it possible to continue to hire all who seek employment. Employment
may even increase, because more people will seek work, perhaps because at the higher
wage they prefer work to illegal activities, which is a good thing, or because they
prefer work to school, which may not be such a good thing. This is precisely what
Card and Krueger observed.

11
. See in particular
Figures 8.6
–
8
.

12
. This fact is crucial but often neglected in US academic debate. In addition to the
work of Goldin and Katz,
Race between Education and Technology
, see also the recent work of Rebecca Blank,
Changing Inequality
(Berkeley: University of California Press, 2011), which is almost entirely focused
on the evolution of the wage difference associated with a college diploma (and on
the evolution of family structures). Raghuram Rajan,
Fault Lines
(Princeton: Princeton University Press, 2010), also seems convinced that the evolution
of inequality related to college is more significant than the explosion of the 1 percent
(which is incorrect). The reason for this is probably that the data normally used
by labor and education economists do not give the full measure of the overperformance
of the top centile (one needs tax data to see what is happening). The survey data
have the advantage of including more sociodemographic data (including data on education)
than tax records do. But they are based on relatively small samples and also raise
many problems having to do with respondents’ self-characterization. Ideally, both
types of sources should be used together. On these methodological issues, see the
online technical appendix.

13
. Note that the curves in
Figure 9.2
and subsequent figures do not take account of capital gains (which are not consistently
measured across countries). Since capital gains are particularly large in the United
States (making the top centile’s share of national income more than 20 percent in
the 2000s if we count capital gains), the gap is in fact wider than indicated in
Figure 9.2
. See, for example, Supplemental Figure S9.3, available online.

14
. New Zealand followed almost the same trajectory as Australia. See Supplemental Figure
S9.4, available online. In order to keep the figures simple, I have presented only
some of the countries and series available. Interested readers should consult the
online technical appendix or the WTID for the complete series.

15
. Indeed, if we include capital gains, which were strong in Sweden in the period 1990–2010,
the top centile’s share reached 9 percent. See the online technical appendix.

16
. All the other European countries in the WTID, namely, the Netherlands, Switzerland,
Norway, Finland, and Portugal, evolved in ways similar to those observed in other
continental European countries. Note that we have fairly complete data for southern
Europe. The series for Spain goes back to 1933, when an income tax was created, but
there are several breaks. In Italy, the income tax was created in 1923, but complete
data are not available until 1974. See the online technical appendix.

17
. The share of the top thousandth exceeded 8 percent in the United States in 2000–2010
if we omit capital gains and 12 percent if we include them. See the online technical
appendix.

18
. The “0.1 percent” in France and Japan therefore increased from 15 to 25 times the
national average income (that is, from 450,000 to 750,000 euros a year if the average
is 30,000), while the top “0.1 percent” in the United States rose from 20 to 100 times
the national average (that is, from
$
600,000 a year to
$
3 million). These orders of magnitude are approximate, but they give us a better sense
of the phenomenon and relate shares to the salaries often quoted in the media.

19
. The income of “the 1 percent” is distinctly lower: a share of 10 percent of national
income for the 1 percent means by definition that their average income is 10 times
higher than the national average (a share of 20 percent would indicate an average
20 times higher than the national average, and so on). The Pareto coefficient, about
which I will say more in
Chapter 10
, enables us to relate the shares of the top decile, top centile, and top thousandth:
in relatively egalitarian countries (such as Sweden in the 1970s), the top 0.1 percent
earned barely twice as much as the top 1 percent, so that the top thousandth’s share
of national income was barely one-fifth of the top centile’s. In highly inegalitarian
countries (such as the United States in the 2000s), the top thousandth earns 4 to
5 times what the top centile earns, and the top thousandth’s share is 40 to 50 percent
of the top centile’s share.

20
. Depending on whether capital gains are included or not. See the online technical
appendix for the complete series.

21
. See, in particular,
Table 5.1
.

22
. For Sweden and Denmark, in some years in the period 1900–1910, we find top centile
shares of 25 percent of national income, higher than the levels seen in Britain, France,
and Germany at that time (where the maximum was closer to 22 or 23 percent). Given
the limitations of the available sources, it is not certain that these differences
are truly significant, however. See the online technical appendix.

23
. For all the countries for which we have data on the composition of income at different
levels, comparable to the data presented for France and the United States in the previous
chapter (see
Figures 8.3
–
4
and
8.9
–
10
), we find the same reality.

24
. See Supplemental Figure S9.6, available online, for the same graph using annual
series. Series for other countries are similar and available online.

25
.
Figure 9.8
simply shows the arithmetic mean of the four European countries included in
Figure 9.7
. These four countries are quite representative of European diversity, and the curve
would not look very different if we included other northern and southern European
countries for which data are available, or if we weighted the average by the national
income of each country. See the online technical appendix.

26
. Interested readers may wish to consult the case studies of twenty-three countries
that Anthony Atkinson and I published in two volumes in 2007 and 2010:
Top Incomes over the Twentieth Century: A Contrast Between Continental European and
English-Speaking Countries
(Oxford: Oxford University Press, 2007), and
Top Incomes: A Global Perspective
(Oxford: Oxford University Press, 2010).

27
. In China, strictly speaking, there was no income tax before 1980, so there is no
way to study the evolution of income inequality for the entire twentieth century (the
series presented here began in 1986). For Colombia, the tax records I have collected
thus far go back only to 1993, but the income tax existed well before that, and it
is entirely possible that we will ultimately find the earlier data (the archives of
historical tax records are fairly poorly organized in a number of South American countries).

28
. The list of ongoing projects is available on the WTID site.

29
. When digital tax files are accessible, computerization naturally leads to improvement
in our sources of information. But when the files are closed or poorly indexed (which
often happens), then the absence of statistical data in paper form can impair our
“historical memory” of income tax data.

30
. The closer the income tax is to being purely proportional, the less the need for
detailed information about different income brackets. In
Part Four
I will discuss changes in taxation itself. The point for now is that such changes
have an influence on our observational instruments.

31
. The information for the year 2010 in
Figure 9.9
is based on very imperfect data concerning the remuneration of firm managers and
should be taken as a first approximation. See the online technical appendix.

32
. See Abhijit Banerjee and Thomas Piketty, “Top Indian Incomes, 1922–2000,”
World Bank Economic Review
19, no. 1 (May 2005): 1–20. See also A. Banerjee and T. Piketty, “Are the Rich Growing
Richer? Evidence from Indian Tax Data,” in Angus Deaton and Valerie Kozel, eds.,
Data and Dogma: The Great Indian Poverty Debate
(New Delhi: Macmillan India Ltd., 2005): 598–611. The “black hole” itself represents
nearly half of total growth in India between 1990 and 2000: per capita income increased
by nearly 4 percent a year according to national accounts data but by only 2 percent
according to household survey data. The issue is therefore important.

33
. See the online technical appendix.

34
. In fact, the principal—and on the whole rather obvious—result of economic models
of optimal experimentation in the presence of imperfect information is that it is
never in the interest of the agents (in this case the firm) to seek complete information
as long as experimentation is costly (and it is costly to try out a number of CFOs
before making a final choice), especially when information has a public value greater
than its private value to the agent. See the online technical appendix for bibliographic
references.

35
. See Marianne Bertrand and Sendhil Mullainathan, “Are CEOs Rewarded for Luck? The
Ones without Principals Are,”
Quarterly Journal of Economics
116, no. 3 (2001): 901–932. See also Lucian Bebchuk and Jesse Fried,
Pay without Performance
(Cambridge, MA: Harvard University Press, 2004).

1
. In particular, all the data on the composition of income by level of overall income
corroborate this finding. The same is true of series beginning in the late nineteenth
century (for Germany, Japan, and several Nordic countries). The available data for
the poor and emergent countries are more fragmentary but suggest a similar pattern.
See the online technical appendix.

2
. See esp.
Table 7.2
.

3
. The parallel series available for other countries give consistent results. For example,
the evolutions we observe in Denmark and Norway since the nineteenth century are very
close to the trajectory of Sweden. The data for Japan and Germany suggest a dynamic
similar to that of France. A recent study of Australia yields results consistent with
those obtained for the United States. See the online technical appendix.

4
. For a precise description of the various sources used, see Thomas Piketty, “On the
Long-Run Evolution of Inheritance: France 1820–2050,” Paris School of Economics, 2010
(a summary version appeared in the
Quarterly Journal of Economics,
126, no. 3 [August 2011]: 1071–131). The individual statements were collected with
Gilles Postel-Vinay and Jean-Laurent Rosenthal from Parisian archives. We also used
statements previously collected for all of France under the auspices of the Enquête
TRA project, thanks to the efforts of numerous other researchers, in particular Jérôme
Bourdieu, Lionel Kesztenbaum, and Akiko Suwa-Eisenmann. See the online technical appendix.

5
. For a detailed analysis of these results, see Thomas Piketty, Gilles Postel-Vinay,
and Jean-Laurent Rosenthal, “Wealth Concentration in a Developing Economy: Paris and
France, 1807–1994,”
American Economic Review
96, no. 1 (February 2006): 236–56. The version presented here is an updated version
of these series.
Figure 10.1
and subsequent figures focus on means by decade in order to focus attention on long-term
evolutions. All the annual series are available online.

6
. The shares of each decile and centile indicated in
Figures 10.1
and following were calculated as percentages of total private wealth. But since private
fortunes made up nearly all of national wealth, this makes little difference.

7
. This method, called the “mortality multiplier,” involves a reweighting of each observation
by the inverse of the mortality rate in each age cohort: a person who dies at age
forty represents more living individuals than a person who dies at eighty (one must
also take into account mortality differentials by level of wealth). The method was
developed by French and British economists and statisticians (especially B. Mallet,
M. J. Séaillès, H. C. Strutt, and J. C. Stamp) in 1900–1910 and used in all subsequent
historical research. When we have data from wealth surveys or annual wealth taxes
on the living (as in the Nordic countries, where such taxes have existed since the
beginning of the twentieth century, or in France, with data from the wealth tax of
1990–2010), we can check the validity of this method and refine our hypotheses concerning
mortality differentials. On these methodological issues, see the online technical
appendix.

8
. See the online technical appendix. This percentage probably exceeded 50 prior to
1789.

9
. On this question, see also Jérôme Bourdieu, Gilles Postel-Vinay, and Akiko Suwa-Eisenmann,
“Pourquoi la richesse ne s’est-elle pas diffusée avec la croissance? Le degré zéro
de l’inégalité et son évolution en France: 1800–1940,”
Histoire et mesure
18, 1/2 (2003): 147–98.