THE ECONOMICS OF INEQUALITY

Thomas Piketty
Translated by Arthur Goldhammer

The Belknap Press of Harvard University Press

CAMBRIDGE, MASSACHUSETTS
LONDON, ENGLAND
2015
 Copyright © 2015 by the President and Fellows of Harvard College
All rights reserved

First published as L’économie des inégalités

copyright © Éditions La Découverte, Paris, France, 1997, 2008, 2014

Jacket design by Graciela Galup

978-0-674-50480-6 (hardcover)
978-0-674-91558-9 (EPUB)
978-0-674-91557-2 (MOBI)
978-0-674-91556-5 (PDF)

The Library of Congress has cataloged the printed edition as follows:

Piketty, Thomas, 1971–
[L’économie des inégalités. English]
The economics of inequality / Thomas Piketty ; translated by Arthur Goldhammer.
pages cm
Includes bibliographical references and index.
1. Income distribution. 2. Equality—Economic aspects. I. Title.
HB523.P54713 2015
339.2'2—dc23 2015008813

Book design by Dean Bornstein

 Contents

Note to the Reader

Introduction
1. The Measurement of Inequality and Its Evolution
2. Capital-Labor Inequality
3. Inequality of Labor Income
4. Instruments of Redistribution

References
Contents in Detail
Index
 Note to the Reader

This book was written and first published in 1997. It was subsequently updated for
several new editions, most recently in 2014. It should be noted, however, that the
overall structure has not been changed since 1997 and that the work essentially reflects
the state of knowledge and data available at that time. As a consequence, this book does
not fully take into account the results of the past fifteen years of international research on
the historical dynamics of inequality. In particular, recent research has demonstrated
that there are important historical variations in the capital-income ratios and the capital
shares in national income, and not only in the concentration of capital ownership at the
individual level. That is, the macroeconomic or functional distribution of national
income and national wealth is substantially less stable than what I was taught in
graduate school and what I report in this book. The large historical variations in top
income shares also receive insufficient treatment in the present book, because the
corresponding research became fully available only recently. Readers interested in a
detailed account of that more recent research and the lessons that can be drawn from it
are advised to consult the World Top Incomes Database (available online) and my book
Capital in the Twenty-First Century (Belknap Press, 2014).
 Introduction

The question of inequality and redistribution is central to political conflict. Caricaturing

only slightly, two positions have traditionally been opposed.
The right-wing free-market position is that, in the long run, market forces, individual
initiative, and productivity growth are the sole determinants of the distribution of
income and standard of living, in particular of the least well-off members of society;
hence government efforts to redistribute wealth should be limited and should rely on
instruments that interfere as little as possible with the virtuous mechanisms of the market
—instruments such as Milton Friedman’s negative income tax (1962).
The traditional left-wing position, passed down from nineteenth-century socialist
theory and trade-union practice, holds that the only way to alleviate the misery of the
poorest members of capitalist society is through social and political struggle, and that
the redistributive efforts of government must penetrate to the very heart of the
productive process. Opponents of the system must challenge the market forces that
determine the profits of capitalists and the unequal remuneration of workers, for
instance, by nationalizing the means of production or setting strict wage schedules.
Merely collecting taxes to finance transfers to the poor is not enough.
This left-right conflict shows that disagreements about the concrete form and
desirability of redistributive policy are not necessarily due to contradictory principles
of social justice but rather to contradictory analyses of the economic and social
mechanisms that produce inequality. Indeed, there exists a certain consensus in regard to
the fundamental principles of social justice: if inequality is due, at least in part, to
factors beyond the control of individuals, such as inequality of initial endowments
owing to inheritance or luck (which cannot be attributed to individual effort), then it is
just for the state to seek in the most efficient way possible to improve the lot of the least
well-off (that is, of those who have had to contend with the most adverse factors).
Modern theories of social justice have expressed this idea in the form of a “maximin”
principle, according to which a just society ought to maximize the minimum
opportunities and conditions available within the social system. The maximin principle
was formally introduced by Serge-Christophe Kolm (1972) and John Rawls (1972), but
one finds it more or less explicitly formulated in much earlier works—for example, in
the traditional idea that everyone should be guaranteed the broadest possible range of
equal rights, a concept widely accepted at the theoretical level. Often, the real conflict
is about the most effective way to improve the actual standard of living of the least
well-off and about the extent of the rights that can be granted to all in the name of
abstract principles of social justice.
Hence only a detailed analysis of the socioeconomic mechanisms that generate
inequality can sort out the competing truth claims of these two extreme versions of
redistribution and perhaps contribute to the elaboration of a more just and effective set
of policies. The purpose of this book is to present the current state of knowledge as a
first step toward that end.
The contrast between the left- and right-wing views sketched above highlights the
importance of different systems of redistribution. Should the market and its price system
be allowed to operate freely, with redistribution effected solely by means of taxes and
transfers, or should one attempt to alter the structure of the market forces that generate
inequality? In the jargon of economics, this contrast corresponds to the distinction
between pure redistribution and efficient redistribution. Pure redistribution occurs
when the market equilibrium is “Pareto efficient,” meaning that it is impossible to alter
the allocation of resources and output in such a way that everyone gains, yet social
justice nevertheless calls for redistribution from the better-off to the worse-off. Efficient
redistribution occurs when the existence of market imperfections allows for direct
intervention in the production process to achieve Pareto-efficient improvements in the
allocation and equitable distribution of resources.
In contemporary political conflict, the distinction between pure and efficient
redistribution is often conflated with the distinction between redistribution on a modest
scale and redistribution on a large scale. The traditional right-left conflict has grown
more complicated over time, however. For instance, some on the left advocate a
“guaranteed basic income” for all citizens, to be financed by taxes without direct
intervention in the market. This guaranteed basic income differs from Friedman’s
negative income tax solely by virtue of size. Broadly speaking, therefore, the question of
how redistribution is to be achieved is separate from the question of the extent of
redistribution. In this book I will try to show that it is best to treat the two questions
separately, because they involve different analytical considerations and lead to different
answers.
To pursue these issues further, it is useful to begin by reminding the reader of the
history and extent of today’s inequality. Doing so will enable us to identify the principal
sets of facts that any theory of inequality and redistribution must take into account
(Chapter 1). The next two chapters (2 and 3) present the leading analyses of the
mechanisms that produce inequality, emphasizing both the political stakes involved in
the intellectual conflict between opposing theories and the observed or observable facts
that can help us decide which theories are correct. Chapter 2 looks first at inequality
between capital and labor, a fundamental inequality that has deeply influenced the
analysis of the social question since the nineteenth century. Chapter 3 deals with
inequality of income from labor itself, which has perhaps become (if it hasn’t always
been) the central question in regard to contemporary inequality. It will then be possible
to delve more deeply into the key issue, namely, the conditions under which
redistribution becomes possible and the tools for achieving it (Chapter 4). Special
attention will be paid to inequality and redistribution in France, although the relative
paucity of available data and analyses (in sharp contrast to the attention devoted to
unemployment, the “social fracture,” and other central issues of French political debate
in the 1990s) will force us at times to rely mainly on studies of other countries,
especially the United States, to illustrate, confirm, or refute the theories discussed.
 {ONE}

The Measurement of Inequality and Its Evolution

What orders of magnitude can we associate with contemporary inequality? Is the income
of the rich in a given country twice that of the poor? Ten times as great? Or a hundred
times? How does the income gap in one country or period compare with that in other
places at other times? Was the income gap in 1950 the same as in 1900 or 1800? Has
inequality with respect to unemployment become the major form of inequality in the
Western world in the 1990s?
 Different Types of Income
What are the various sources of household income? Table 1.1 breaks down the income
of 24 million French households in 2000 into various categories: wages, self-
employment income (earned, for example, by farmers, merchants, doctors, lawyers, and
so on), pensions, other transfer income (family allowances, unemployment insurance,
welfare), and capital income (dividends, interest, rent, etc.).
What do we learn from Table 1.1? First, 58.8 percent of total household income
comes in the form of wages. If we add to this the 5.8 percent of income consisting of
self-employment compensation, we find that nearly two-thirds of total household income
is compensation for labor. Social income accounts for another 30 percent of the total,
and for more than two-thirds of retiree income. Finally, income from household wealth
(capital income such as dividends, interest, and so on) accounts for roughly 5 percent of
the total. As is well known, however, capital income is not accurately reported in
household income surveys. National accounts based on dividend and interest data
provided by firms and banks yield a higher estimate of the share of capital income in
total household income, on the order of 10 percent (INSEE, 1996b, pp. 26–29). In any
case, all sources agree that labor income accounts for at least six or seven times as
large a share of total household income as capital income. This is a general feature of
the income distribution in all Western countries (Atkinson et al., 1995, p. 101). But the
5 to 10 percent share of household income derived from capital underestimates the
share of capital income in total national income, since a substantial portion of the
capital income of firms is not distributed to households (see Chapter 2).
 The importance of these various types of income is obviously not the same for rich
and poor. To analyze this further, it is useful to distinguish between different deciles of
the income distribution: the first decile, denoted D1 in Table 1.1, includes the bottom 10
percent of the household income distribution. The second decile, D2, includes the next
10 percent, and so on, up to the top decile, D10, which represents the 10 percent of
households with the highest income. To refine this description, we also use the notion of
centiles: the first centile includes the bottom 1 percent of households, and so on up to
the hundredth centile. Each decile includes a subgroup of the population: some 2.4
million households per decile and 240,000 per centile in the case of France in the year
2000. One can calculate various characteristics for each decile or centile: average
income, for example. This should not be confused with the notion of upper income limits
for each group. To capture this statistic, we use the letter P: for example, P10 represents
the level of income below which we find 10 percent of all households; P90 is the upper
limit below which we find 90 percent of all households; and so on. In Table 1.1, P90–
95 represents the subset of all households with incomes between the top of the 90th
centile and the top of the 95th centile, that is, the first half of the tenth decile, whereas
P95–100 represents the second half of the tenth decile, which includes the five top-
earning centiles.
Table 1.1 shows that the households in D1 consist largely of modest retirees and
unemployed workers: the wages they receive account on average for 18 percent of their
income, while nearly 80 percent consists of social income. The share of wages in total
income increases with income level, while the share of retirees and unemployed
decreases, until we reach the top 5 percent (P95–100), where capital income and
nonwage compensation account for a substantial share of the total (I make no distinction
between “wages” and “salary” throughout: both refer to income from labor). Nonwage
compensation is intermediate in nature between labor income and capital income, since
it remunerates both the labor of the farmer, doctor, or merchant and the capital invested
in his or her business. Still, labor income continues to account for a very large share of
the total income of households at the top of the distribution: the top 5 percent take more
of their income in wages than in income from capital, no matter how the latter is
estimated. One has to go even higher in the income hierarchy to reach a level where
labor income no longer accounts for the largest share (Piketty, 2001).
 Wage Inequality
How are wages, which represent the lion’s share of household income, distributed?
Table 1.2 describes wage inequality among full-time private-sector workers in France
in 2000 (a group of some 12.7 million individuals).
The bottom 10 percent of the wage distribution (D1) earned on average an income
roughly equal to the minimum wage, or about €890 per month (net of taxes) in 2000. The
median wage (denoted P50, by definition the wage level below which lies 50 percent of
the sample) was €1,400. This was higher than the average wage of the fifth decile
(€1,310), since the fifth decile consists of workers between P40 and P50. It was also
lower than the average wage overall, which was €1,700 in 2000, because the top half of
the wage distribution is always “longer-tailed” than the bottom half, so that very high
earners inevitably lift the average wage above the median. Furthermore, the best-paid
10 percent, who earn at least €2,720 per month, earn an average wage of €4,030, or
nearly twice as much as the next lower 10 percent (€2,340).
One practical indicator of total wage inequality is the P90/P10 ratio, that is, the ratio
of the lower limit of the tenth decile to the upper limit of the first decile. In the case of
France in 2000, the P90/P10 ratio was 2,720/900 or roughly 3.0: to belong to the top-
earning 10 percent, one had to make at least three times as much as the least well paid.
This indicator should not be confused with the D10/D1 ratio, that is, the ratio of the
average wage of the tenth decile to that of the first decile, which is by definition always
higher and which in France in 2000 was 4,030/890, or 4.5: the best-paid 10 percent in
France earned on average 4.5 times as much as the worst-paid 10 percent. Table 1.2
also allows us to calculate the total wages paid to the top 10 percent: since the average
wage of D10 was 2.37 times the average wage (4,030/1,700 = 2.37) and the number of
workers in D10 is by definition 10 percent of the total number of workers, it follows
that D10 received 23.7 percent of total wages.
TABLE 1.2
Wage inequality in France, 2000

Average monthly wage in euros

Average 1,700
D1 890 900 P10
D2 1,100
D3 1,110
D4 1,210
D5 1,310 1,400 P50
D6 1,450
 D7 1,620
D8 1,860
D9 2,340 2,720 P90
D10 4,030

Note: D1 represents the worst-paid 10 percent; D2 the next 10 percent, and so on. P10 is the limit dividing D1 and
D2, P50 the limit dividing D5 and D6, and P90 the limit dividing D9 and D10. In other words, the worst paid all
earned less than €900 a month, with an average income of €890, whereas the best-paid 10 percent all earned more
than €2,720, with an average income of €4,030. These figures represent monthly wages excluding bonuses net of
social charges (and CSG/CRDS) for full-time, private-sector workers.
Source: DADS (Annual declaration of social data), INSEE, 2002, p. 10.

Other indicators are also used in order to capture the overall inequality of the
distribution and not just the gap between the extreme deciles: for instance, the Gini
coefficient or the Theil and Atkinson indices (Morrisson, 1996, pp. 81–96).
Nevertheless, interdecile indicators (such as P90/P10, D10/D1, P80/P20, etc.) are by
far the simplest and most intuitive. The P90/P10 indicator has the merit of being
available in reliable numbers for many countries, hence it will be cited frequently in
this chapter.
For a more complete view of wage inequality, one would need to include figures for
public-sector wages in addition to private-sector wages. In France, the 4.1 million full-
time employees of the public sector earn slightly more on average than private-sector
workers, and public-sector wages are significantly less widely dispersed: for example,
the P90/P10 ratio for civil-service workers was 2.6 (INSEE, 1996d, p. 55).

International Comparisons
Is a P90/P10 ratio of 3:1 typical of wage inequality everywhere? Table 1.3 gives the
P90/P10 ratio for fourteen OECD countries in 1990.
The table shows that France, with a P90/P10 ratio of 3.1 in 1990, occupied a middle
position between Germany and the Nordic countries on the one hand and the English-
speaking countries on the other. In the former, the ratio was generally around 2.5,
dipping as low as 2 in Norway, 2.1 in Sweden, and 2.2 in Denmark, while in the latter it
was as high as 3.4 in the United Kingdom, 4.4 in Canada, and 4.5 in the United States.
For all the countries shown, the figures in Table 1.3 concern only full-time employees.
This is an important detail, because including part-time workers (of whom there were
some 3.1 million in France in 2000) systematically leads to larger P90/P10 ratios. For
example, the OECD figures including intermittent and part-time workers in the United
States in 1990 give a P90/P10 ratio of 5.5, but only 4.5 when these workers are left out
(Katz et al., 1995, fig. 1; Lefranc, 1997, table 1), as is the case with other countries
(OECD, 1993, p. 173). In short, P90/P10 ratios range from 2 to 4.5, which is
considerable variation for countries at very similar levels of economic development.

TABLE 1.3
Wage inequality in OECD countries in 1990, measured by the P90/P10 ratio

Country Ratio between best-paid and worst-paid 10 percent

Norway 2.0
Sweden 2.1
Denmark 2.2
Netherlands 2.3
Belgium 2.3
Italy 2.4
Germany 2.5
Portugal 2.7
Japan 2.8
France 3.1
United Kingdom 3.4
Austria 3.5
Canada 4.4
United States 4.5

Note: For example, in Germany the best-paid 10 percent earn at least 2.5 times as much as the worst-paid 10
percent.
Sources: OECD, 1993, pp. 170–173; US data, Katz et al., 1995, fig. 1.
 Income Inequality
How does this inequality between workers translate into inequality of household
income? The answer is not simple, because one has to add nonwage compensation of
the self-employed (some 3 million individuals in France in 2000), social transfers, and
capital income, and then individual wage earners, non–wage earners, and their children
have to be grouped together to form households. Table 1.4 presents the results for
France in 2000.
The average monthly household income in France was €2,280, but 10 percent of
households had less than €790 of monthly income, while 10 percent had more than
€4,090, for a P90/P10 ratio for total household income of 5.2 (compare with the
P90/P10 ratio for wages of 3.0). The top 5 percent of households, with monthly income
above €5,100, had an average monthly income of €7,270.
TABLE 1.4
Income inequality in France, 2000

Monthly income in euros

Average 2,280
D1 540 790 P10
D2 930 1,070 P20
D3 1,190 1,330 P30
D4 1,480 1,610 P40
D5 1,760 1,920 P50
D6 2,080 2,240 P60
D7 2,430 2,630 P70
D8 2,880 3,150 P80
D9 3,570 4,090 P90
P90–95 4,520 5,100 P95
P95–100 7,270

Notes: See Tables 1.1 and 1.2. The poorest 10 percent of households have income of less than €790 per month,
averaging €540. The wealthiest 5 percent have income above €5,100 per month, with an average of €7,300. Monthly
income calculated by dividing annual income by twelve, including wages, self-employment income, pensions,
transfers, and capital income. Income is net of social charges (including CSG/CRDS) but not of other direct taxes
(income tax, housing tax).
Source: “2000 family budget survey,” INSEE (author’s calculations).

The fact that inequality of household income is greater than wage inequality is quite
common, and it was made worse in France in 2000 by the fact that many households
included unemployed workers. In general, however, the explanation of this difference
depends on factors other than unemployment.
First, nonwage income and especially capital income are much more unequally
distributed than wages. Typically, the share of capital income claimed by the wealthiest
10 percent of households is on the order of 50 percent of total capital income, as is the
share of total wealth owned by the wealthiest 10 percent, whereas the share of total
wages going to the highest-paid 10 percent falls between 20 and 30 percent (23.7
percent in France in 2000). The share of capital income in total income is low,
however, so the share of income going to the wealthiest 10 percent of households was
just 26 percent of total household income in France in 2000. These very large
disparities of wealth, much greater than wage and income inequalities, are also much
less well known. Inequalities of wealth cannot be explained solely by inequalities of
present and past income. Behavioral differences with respect to savings and
accumulation also play a part (accounting for nearly half of wealth inequality in 1992,
according to Lollivier and Verger [1996]). These specific difficulties of accounting for
wealth inequality explain why measures of inequality are often limited to inequalities of
wages and income.
But the main reason why income inequality is always significantly greater than wage
inequality is totally different: it comes from the fact that a majority of low-income
households are households living on small pensions, often consisting of one person,
whereas high-income households are generally couples, often with two earners and
children living at home. If one were to calculate the P90/P10 ratio not for household
income but for household income adjusted for household size in order to measure
inequality in standards of living rather than of income as such, one would find a ratio of
4.3–4.4 rather than 5.2, depending on how the adjustment was made (INSEE, 1996b, p.
16). If one is interested in disposable income, then the effect of taxes on income also
needs to be taken into account, which the figures in Table 1.4 do not do. This would
reduce the P90/P10 ratio by about 10 percent, since the income tax paid by a household
with income at the P90 level, about €4,090 per month, would average about 10 percent,
while households at the P10 level pay no income tax (INSEE, 1995, p. 19). (On the
redistributive effect of taxes and transfers, see Chapter 4.) We would thus arrive at a
ratio for disposable household income adjusted for household size of 3.5–4, slightly
greater than wage inequality.

International Comparisons
How does a P90/P10 ratio of 3.5–4 compare with what we see in other countries?
International comparisons are unfortunately much more difficult to do for household
incomes than for wages: it is often difficult to include exactly the same income
categories in all countries. Culminating an ambitious effort to compile comparable data
for many countries, the Luxembourg Income Study (LIS) was published in 1995 at the
behest of the OECD (Atkinson et al., 1995).
The P90/P10 ratios in Table 1.5 are for disposable income, that is, income
accounting for taxes and transfers, and adjusted for household size. France is assigned a
ratio of 3.5 (and not 5.2 as in Table 1.4) for that reason, along with the fact that the
OECD study used 1984 fiscal data rather than the “Family Budget Survey” of 2000. The
same international contrast observed earlier in relation to wage inequalities occurs
again here: Germany, Belgium, Sweden, and Norway, which had wage ratios of 2–2.5,
have income ratios of 2.7–3, while the United Kingdom, United States, and Canada,
which had wage ratios of 3.5–4.5, have income ratios of 3.8–5.9, with the United States
topping the chart at 5.9. France again occupies an intermediate position.

TABLE 1.5
Income inequality in OECD countries

Country P90/P10 ratio

Sweden 2.7
Belgium 2.8
Norway 2.9
Germany 3.0
France 3.5
United Kingdom 3.8
Italy 4.0
Canada 4.0
United States 5.9

Notes: In Sweden, to belong to the top 10 percent in income, one must earn at least 2.7 times as much as anyone in
the bottom 10 percent. The table shows the P90/P10 ratio for disposable income adjusted for household size
(Atkinson et al., 1995). Years: 1984 (Germany, France), 1985 (Australia), 1986 (US, Italy, Norway, UK), 1987
(Canada, Sweden), 1988 (Belgium).
Source: LIS, Atkinson, Rainwater, and Smeeding, 1995, p. 40.

It is extremely difficult to compare these figures with the few available inequality
indicators from outside the developed world. Conditions vary widely: the South
American countries stand out for having even higher levels of inequality than the most
inegalitarian Western countries, whereas most Asian countries, as well as the less-
developed countries of Africa, generally have levels of income inequality equal to or
less than those found in the least inegalitarian Western countries (Morrisson, 1996, pp.
145–172). It is also difficult to compare levels of inequality found in the Communist
bloc, because income often includes compensation in kind in one form or another and is
difficult to quantify in monetary terms. The available indicators seem to show real
income gaps quite comparable to the average in capitalist countries and generally
greater than in the more egalitarian capitalist countries (Morrisson, 1996, p. 140).
 Inequalities in Time and Space
Are P90/P10 wage and income ratios of 3:1 or 4:1 between people living in one
country at one point in time negligible compared with the gap between a citizen of an
advanced country in 1990 and of the same country in 1900 or a citizen of India in 1990?
Table 1.6 indicates the average purchasing power of a French blue-collar worker and of
an executive from 1870 to 1994 measured in 1994 francs (that is, accounting for changes
in the cost of living).
These figures should obviously be viewed cautiously: the further back we go in time,
the more problematic the idea of a synthetic cost index becomes, because modes of
consumption change so much. Still, orders of magnitude may be taken as significant:
between 1870 and 1994, the purchasing power of a worker increased roughly eightfold.
This spectacular increase in standard of living during the last century of capitalism was
more or less comparable in all the Western countries. For example, the hourly wage of
an American worker increased by a factor of eleven between 1870 and 1990, for an
average rate of increase of about 2 percent per year (Duménil and Lévy, 1996, chap.
15), which is approximately the same as in France if we take the decrease in annual
hours of work into account.

This 10:1 ratio between 1990 and 1870 is approximately equivalent to, or slightly
less than, the ratio of the average income of a Western citizen in 1990 to that of a
Chinese or Indian citizen, using the best available estimates of purchasing power parity
(Drèze and Sen, 1995, p. 213). The gaps in GDP per capita, which are often four to five
times greater, don’t actually make much sense, because they are expressed in terms of
official exchange rates with the advanced economies, and these rates are a very poor
gauge of actual differences in purchasing power. A 10:1 gap between the average
standard of living in the wealthiest countries and that in the poorest countries probably
comes closer to reality.
To sum up, inequality between the top 10 percent and bottom 10 percent in a given
country, as measured by the P90/P10 ratio, is on the order of 3–4, and this is two to
three times smaller than the gap in standard of living between the end of the nineteenth
century and the end of the twentieth century and than the gap between the richest and
poorest countries. These two forms of inequality are therefore not incomparable, even if
one is undeniably larger than the other.
 The Historical Evolution of Inequality
Are these 3:1 ratios between top and bottom income groups in one country and 10:1
ratios between rich and poor countries destined to remain steady, increase, or decrease?
Although Marx and other nineteenth-century social theorists did not quantify
inequality this way, they were certain of the answer: the logic of the capitalist system
was to increase inequality between social classes—between capitalists and
proletarians—constantly. The gap between rich and poor countries would also grow.
These predictions were soon challenged, however, even within the socialist camp. In
the 1890s Eduard Bernstein insisted that Marx’s proletarianization thesis did not hold
because the social structure was clearly becoming more diverse and wealth was
spreading to ever broader segments of society.
It was not until after World War II, however, that it became possible to measure the
decrease in wage and income inequality in the Western countries. New predictions
were soon forthcoming. The most celebrated was that of Simon Kuznets (1955):
according to Kuznets, inequality would everywhere be described by an inverted U
curve. In the first phase of development, inequality would increase as traditional
agricultural societies industrialized and urbanized. This would be followed by a second
phase of stabilization, and then a third phase in which inequality would substantially
decrease. This pattern—of growing inequality in the nineteenth century followed by
declining inequality after that, has been well studied in the case of the United Kingdom
(Williamson, 1985) and the United States (Williamson and Lindert, 1980). In the latter,
for example, one finds that the share of total wealth owned by the wealthiest 10 percent
rose from 50 percent in 1770 to a maximum of 70–80 percent in the late nineteenth
century, before declining to about 50 percent, a level typical of contemporary wealth
inequality. The available sources suggest that the pattern was similar in all the Western
countries.
FIGURE 1.1. The fall of rentiers and the stability of the wage hierarchy in France, 1913–2005. Sources: Piketty, 2001;
Landais, 2007.

The most recent research on France and the United States (Piketty, 2001; Piketty and
Saez, 2003; Landais, 2007) shows, however, that the sharp decrease in inequality
observed over the course of the twentieth century is in no sense the consequence of a
“natural” economic process. Only inequality of wealth decreased (the wage hierarchy
showing no tendency toward compression over the long run), and this decrease was due
mainly to shocks incurred by wealth owners in the period 1914–1945 (wars, inflation,
the Great Depression). The concentration of wealth and capital income did not return to
the astronomical level achieved on the eve of World War I, however. The most likely
explanation involves the fiscal revolution of the twentieth century. The impact of the
progressive income tax (created in France in 1914) and the progressive estate tax
(created in 1901) on the accumulation and reconstitution of large fortunes seems to have
prevented a return to nineteenth-century rentier society. If contemporary societies have
become societies of managers—that is, societies in which the top of the income
distribution is dominated by the “working rich” (people who live mainly on their labor
income rather than on income derived from capital accumulated in the past), it is
primarily a consequence of particular historical circumstances and institutions. The
Kuznets curve is thus not the “end of history” but the product of a specific—and
reversible—historical process.

From Laws of History to Uncertainties

The fatal blow to the Kuznets curve—the idea of an inverted U-shaped curve linking
development inexorably to increasing and then decreasing inequality—came in the
1980s, when inequality began to increase in the advanced economies. This inversion of
the Kuznets curve spelled an end to the notion that there was a grand historical law
governing the evolution of inequality, at least for a time. It also encouraged extensive
and detailed research into the complex mechanisms that might explain why inequality
increases in some periods and decreases in others.
Table 1.7 describes the evolution of wage inequality in the Western economies since
1970. Inequality actually increased only in the United States and United Kingdom, but
wage inequality ceased to decrease everywhere in the 1980s. This set the Western
countries apart from less developed countries, where no such tendency has been
detected (Davis, 1992). In the United States, the P90/P10 ratio for wages increased by
about 20 percent between 1980 and 1990, for a total increase of nearly 50 percent over
the entire period, which is considerable in view of the habitually slow rate of change of
wage inequality. The result was that wage inequality in the United States returned to
levels last seen between the two world wars (Goldin and Margo, 1992). As a logical
consequence of this evolution, wealth inequality, which had been decreasing until 1970,
seems to have begun to rise again (Wolff, 1992).
The case of the United Kingdom was very different, because wage inequality there
was quite low in 1970, very close to Scandinavian levels. After increasing slightly in
the second half of the 1970s, the P90/P10 ratio jumped nearly 30 percent between 1980
and 1990, and in the 1990s the United Kingdom joined the United States at the head of
the pack in inequality. In the Nordic countries, inequality held steady, with P90/P10
ratios on the order of 2–2.5, despite a slight upward trend. France was a somewhat
special case, since French wage inequality was the highest in the Western world in
1970 before decreasing rapidly in the 1970s and then stabilizing between 1980 and
1990, with a very slight increase since 1983–1984: the P90/P10 ratio was 3.1 in 1984,
then rose to 3.2 in the period 1984–1995 (INSEE, 1996a, p. 48). Hence it was only in
the 1970s that the wage distribution in the United States became more unequal than in
France, while the United Kingdom did not surpass France until the late 1980s, and then
by just a hair (Table 1.7). Although wage inequality in Italy was significantly lower than
in France initially, its evolution in the period 1970–2000 resembled that of France: after
decreasing rapidly in the 1970s and early 1980s, the P90/P10 ratio began to climb again
in 1984 (Erickson and Ichino, 1995).
TABLE 1.7
The growth of wage inequality since 1970, as measured by the P90/P10 ratio

Country 1970 1980 1990

 Country 1970 1980 1990

Germany – 2.5 3.5

United States 3.2 3.8 4.5
France 3.7 3.2 3.2
Italy – 2.3 2.5
Japan – 2.5 2.8
United Kingdom 2.5 2.6 3.3
Sweden 2.1 2.0 2.1

Note: To belong to the best-paid 10 percent in the United States, one had to earn at least 3.2 times as much as
anyone in the worst-paid 10 percent in 1970, compared with 4.5 times as much in 1990.
Sources: Germany, Italy, Japan, Sweden: OECD, 1993, pp. 170–173. France: INSEE, 1996a, p. 48. United States,
United Kingdom: Katz et al., 1995, fig. 1.

From Wages to Incomes

The evolution of income inequality has not been measured as well as that of wage
inequality. Nevertheless, the Luxembourg Income Study data do allow us to describe in
broad outline the evolution of the P90/P10 ratio for disposable household income
adjusted for household size (Atkinson et al., 1995, p. 47). The countries in which wage
inequality increased also saw increases in income inequality: in the United States, for
example, the P90/P10 ratio for income rose from 4.9 to 5.9 between 1979 and 1986,
and in the United Kingdom from 3.5 to 3.8. Conversely, inequality increased moderately
in the Nordic countries, rising from 2.8 to 2.9 in Norway and 2.5 to 2.7 in Sweden,
following the similarly modest increase in wage inequality. In France, the P90/P10 ratio
has remained stable at 3.5 since the early 1980s after a sharp decrease in the 1970s.
Since the early 1990s, there has been a slight upward trend in the P90/P10 ratio in
France, although the increase was just barely statistically significant (INSEE, 1996b,
pp. 36–37). In all Western countries, it is clear that the downward trend of the previous
period has been reversed: income inequality, like wage inequality, ceased to decrease
everywhere in the 1980s and 1990s, and it increased significantly in those countries
where wage inequality resumed its upward trend. The Kuznets curve is definitely dead.
It would be a mistake, however, to interpret the evolution of income inequality as a
simple mechanical consequence of the evolution of wage inequality, even though the
latter is undeniably the main force at work (Gottschalk, 1993). For example, nearly half
of the increase in US household income inequality between 1970 and 1990 was in fact
due to increased correlation of the incomes of members of the same household: in other
words, high earners are increasingly likely to marry other high earners, whereas the
lowest earners are often single women with children (Meyer, 1995). Furthermore, taxes
and transfers have evolved in different ways in different Western countries since the
1970s: whereas the United States and United Kingdom adopted policies that tended to
increase income inequality, other countries adopted policies that sought to limit it. It is
striking to compare the United States and Canada: although the labor market and wage
inequality evolved in similar ways in both countries, the P90/P10 ratio for household
income held steady at around 4 in Canada but rose from 4.9 to 5.9 in the United States
(Atkinson et al., 1995, p. 47). The reasons for this are complex, but different fiscal and
social policies explain a great deal (Card and Freeman, 1993).

Inequality with Respect to Employment

More broadly, it would be misleading to suggest that the evolution of inequality in a
country like France since the 1970s can be summed up by saying that the P90/P10 wage
ratio has remained more or less constant. And while many countries, including France,
have kept the P90/P10 ratio for disposable household income relatively constant, this
has been possible only because social transfers have been able to compensate for the
loss of income owing to the growing number of the unemployed. Without transfers
(including unemployment insurance, basic income support, etc.), income inequality
would have increased as much as in the English-speaking countries, despite the stability
of the wage income distribution: labor income inequality among all individuals of
working age (and not just the employed) increased considerably in France after 1980, at
a rate comparable to that observed in the English-speaking countries (Bourguignon and
Martinez, 1996). Actual inequality of income from labor (whether due to employment
inequality or wage inequality) has thus increased in all Western countries since the
1970s.
Is it really possible to draw a clear distinction between the English-speaking
countries, where rising income inequality is supposedly a matter of increasing wage
inequality, and other countries, where it is supposedly a matter of unequal risk of
unemployment? Official figures might seem to support this view: the 1996
unemployment rate was 5.6 percent in the United States and 7.5 percent (and rapidly
declining) in the United Kingdom, compared with 10.3 percent in Germany, 12.1
percent in Italy, and 12.2 percent in France (where 3 million people were unemployed
in a working-age population of around 25 million [OECD, 1996, A24]). High growth in
the late 1990s significantly reduced unemployment everywhere but left the geographical
variation intact: in 2000, the unemployment rate was 4 percent in the United States and
10 percent in France (OECD, 2000).
The problem with this type of comparison, however, is that the notion of
unemployment is not an adequate measure of the phenomenon of underemployment. In
the United States, for example, there has been a substantial withdrawal of less-skilled
individuals from the labor market (and from the group of people classified in official
statistics as “actively seeking work”) since the 1970s. This is entirely due to the
collapse of low-wage employment opportunities (Juhn et al., 1991; Topel, 1993). Many
people of working age have thus found themselves excluded from the labor market, yet
they are not counted in unemployment statistics. One striking manifestation of this can be
seen in the impressive increase of the prison population. In 1995, 1.5 million
individuals were incarcerated in US prisons, compared with 500,000 in 1980; it is
estimated that 2.4 million will be incarcerated in 2000 (Freeman, 1996). This aspect of
underemployment, entirely neglected in official unemployment statistics, is not a minor
matter, since these 1.5 million prisoners represented 1.5 percent of the US working age
population in 1995. In France, by comparison, the prison population was just 60,000, or
0.3 percent of the working age population. It would of course be simplistic to suggest
that the growth of crime in the United States since 1970 can be explained entirely by the
evolution of wage inequality. Clearly, however, it was more difficult to be a model
proletarian in the United States in 1995 than it was in 1970, given that the wage of the
tenth centile fell by nearly 50 percent compared with that of the ninetieth centile.
It is therefore tempting to conclude that underemployment is in fact as high in the
United States as in the European countries where unemployment is high. This is
misleading, however, since the phenomenon of hidden underemployment is unfortunately
not limited to the United States. It takes other forms in Europe, less visible perhaps but
no less significant. Consider, for example, the fact that only 67 percent of the working
age population is classified as belonging to the active population in France in 1996,
compared with 77 percent in the United States, 75 percent in the United Kingdom, and
only 68 percent in Germany and 60 percent in Italy (OECD, 1996, A22). This indicator,
known as the labor market participation rate, is highly imperfect because it mixes
together a wide range of phenomena, such as the female participation rate and the
percentage of early retirees, but it nevertheless points to a real problem. For example, it
is well known that, in France, in order to decrease the number of unemployed by one,
more than one job must be created, indeed something close to two jobs, since some
fraction of newly created jobs will in fact be taken by people who were not counted as
part of the active population but who were prepared to enter the labor market if an
appropriate job presented itself. In addition, involuntary part-time work (by people
working part-time who claim to want to work more) has been increasing rapidly in
France (CSERC, 1996, p. 50). Uncertainties such as these reveal the limits of our
ability to correctly measure inequality with respect to employment, which is a
fundamental feature of contemporary inequality.
 {TWO}

Capital-Labor Inequality

Since the industrial revolution, and in particular since the work of Karl Marx (1818–
1883), the question of social inequality and redistribution has been posed primarily in
terms of the opposition between capital and labor, profits and wages, employers and
employees. Inequality is thus described as a contrast between those who own capital,
that is, the means of production, and those who do not and must therefore make do with
what they can earn from their labor. The fundamental source of inequality is thus said to
be the unequal ownership of capital. Initially, the two terms of this basic inequality,
capitalists and workers, are conceived as homogeneous groups, and inequality of
income from labor is regarded as a secondary matter. This view of inequality as purely
a question of capital versus labor has exerted, and will continue to exert, a profound
influence on the way redistribution is conceptualized, even in countries that have not
gone so far as to abolish private ownership of capital.
The special attention accorded to labor-capital inequality should come as no
surprise. Indeed, the mere fact that a share of productive income goes to capital might
seem to contradict basic principles of justice and immediately raise the question of
redistribution: Why should a person who inherits ownership of capital receive income
denied to those who inherit only their labor power? Absent any market inefficiency, this
would amply justify a pure redistribution of capital income to labor income (using the
distinction between pure and efficient redistribution discussed in the Introduction). How
large should this pure redistribution be, and with what tools might it be accomplished?
What does the history of this kind of redistribution and of the capital-labor split tell us?
Social justice is not the only reason to worry about the capital-labor split, however.
Is the unequal distribution of wealth among individuals and countries not only unjust but
also inefficient, because it reproduces itself by limiting the ability of the poor to invest
and thus close the gap between themselves and the rich? If so, how can capital be
efficiently redistributed?
 The Share of Capital in Total Income
The question seems simple. A nation produces what it produces using a certain quantity
of capital (machinery, infrastructure, etc.) and a certain quantity of labor (hours
worked). What determines the share of output going, respectively, to capital (in the form
of dividends and interest paid to the owners of capital) and labor (in the form of wages
paid to workers), and what can government do to redistribute these shares? This
question, and especially the role of the price system in determining the capital-labor
split, has given rise to unusually acrimonious intellectual and political controversy,
especially among economists.

The Question of Capital/Labor Substitution

Suppose, to begin with, that the technology that allows the nation to produce from
available quantities of capital and labor is characterized by what economists call fixed
coefficients: in order to produce 1 unit of output, exactly 1 unit of capital and n units of
labor are required. In other words, in order to operate a particular machine properly,
exactly n workers are required, neither more nor less.
When posed in these terms, the distribution of income between capital and labor is a
purely distributive question; efficiency does not come into it. The only decision to be
made is how to divide the 1 unit produced between the two factors of production,
capital and labor. Or, to put it another way, between the owner of the machine and the n
workers, independent of the production process itself. Market forces and the price
system play no part in determining the allocation of resources at the macroeconomic
level, because no matter what price firms pay for each unit of capital and labor, they
must employ n workers and 1 machine to produce 1 unit of output. In particular, the
level of employment is fixed: it is entirely determined by the capital stock, that is, by the
productive capacity of the economy. Without any governmental redistribution, the
capital-labor split will depend on the negotiating power of unions and the ability of the
employer to appropriate a portion of the product: in short, on the relative power of
capitalists and workers. The key fact here is that the prices paid for capital and labor
will have no effect on the level of production or the volume of employment. The capital-
labor split is the outcome of a pure distributional conflict.
Under these conditions, there is no reason to ask how capital/labor redistribution is
carried out: the question of redistributive instruments does not arise. Income might be
redistributed to labor by increasing the wage paid to each worker: for example, by
raising the minimum wage or supporting union demands for higher wages. Or the same
thing might be accomplished by raising taxes on capital to finance a fiscal transfer to
each worker (or to reduce the taxes paid by workers). Direct redistribution from
profits to wages is completely equivalent to fiscal redistribution (via taxes and
transfers); in either case there is no interference with the so-called primary
distribution, which is determined by firms, because the total quantities of capital and
labor employed in production are fixed.
Of course, it may be desirable to give the owners of capital incentives to invest and
accumulate more capital in order to increase the economy’s future productive capacity,
and this may limit the desirable degree of redistribution from capital to labor. But this
objection (whose practical implications we will examine later) applies to any attempt
to decrease capital’s share of total income, regardless of whether redistribution is
achieved directly or indirectly (by taxes and transfers). Since the capital-labor split is a
purely distributive matter, how it is achieved is of no importance; only the result
matters.
What Capital/Labor Substitution Means
Different conclusions would follow if it were possible to vary the proportions of
capital and labor in the production process. Suppose that 1 unit of capital no longer
requires exactly n units of labor and that we can increase output by adding labor,
because some tasks done by machines can also be done by workers. More generally,
even if a given firm with a given production technology cannot efficiently use more than
n workers per machine, it might still be possible for other firms in other sectors of the
economy to use less capital-intensive and more labor-intensive technologies. For
example, the service sector, which generally uses more labor and less capital than
industry, might grow larger, so that for a given stock of capital, the economy as a whole
would employ more workers. Thus the possibility of substituting labor for capital or
vice versa is not simply a matter of technological opportunity but also a consequence of
structural transformation of a society’s modes of production and consumption.
If capital can be substituted for labor and vice versa, then (in contrast to the case of
fixed-coefficient technology) the prices of capital and labor can play an important role
in determining the quantities of each of the two factors of production used at the
macroeconomic level. In a market economy, firms will hire more workers as long as
doing so brings in more money than it costs, hence as long as the marginal productivity
of labor, which is defined as the additional output obtained by using one additional unit
of labor with the same quantity of capital, is greater than the price of labor, which is
determined by the costs (in wages, social charges, bonuses, and so on) the firm incurs
for employing an additional worker.
The same is true for capital, with the price of capital measured by the costs (interest
and dividends, depreciation, maintenance, etc.) the firm incurs for using one additional
unit of capital. Labor-intensive firms will grow more rapidly than capital-intensive
firms if the price of labor is low compared with the price of capital, because consumer
demand for labor-intensive goods will increase if their price is low (and conversely).
In other words, the quantities of labor and capital used in a market economy, and
therefore the levels of output and employment, will depend on the prices of capital and
labor: prices play an allocative and not just a distributive role.
The concept of the capital-labor split and of the role played by the price system,
centered on the idea of permanent adjustment of the amount of capital and labor used by
firms as a function of the prices they face, was explicitly introduced by the so-called
marginalist economists in the 1870s. Their notion of the marginal productivity of the
factors of production broke with the thinking of classical economists such as David
Ricardo and Karl Marx, who implicitly reasoned in terms of fixed-coefficient
technology, so that the available capital stock entirely determined the economy’s
productive capacity and level of employment, thus making the capital-labor split a pure
distributive conflict. The opposition between the classical and marginalist theories of
the capital-labor split resurfaced in the 1950s and 1960s in the so-called “Cambridge
capital controversy,” which pitted economists in Cambridge, England, who insisted that
the split was in essence a purely distributive conflict and emphasized the role of
bargaining power, against other economists in Cambridge, Massachusetts, who argued
that the relative prices of capital and labor also played an allocative role, drawing on
Robert Solow’s idea of the aggregate production function, a mathematical
representation of the possibility of capital-labor substitution at the level of the economy
as a whole.
Redistribution: “Fiscal” or “Direct”?
What are the implications of capital-labor substitutability for redistribution? If one tries
to redistribute capital income to labor by increasing workers’ wages, thereby increasing
the price of labor, firms (and therefore the economy as a whole) will use less labor and
more capital, so that the level of employment will decrease, and labor’s share of total
income will increase less than the initial wage increase might have led one to believe.
This would not happen with fiscal redistribution (as described earlier): if one taxed the
profits of firms (or capital income paid by firms to capital-owning households), one
could finance a fiscal transfer or tax decrease to achieve the same redistribution of
income to workers without increasing the labor costs of firms and thus without
triggering a substitution of capital for labor deleterious to employment.
The essential difference between these two types of redistribution is that the
contribution of firms is not calculated in the same way: direct redistribution requires
firms to contribute to redistribution in proportion to the number of workers they employ,
whereas fiscal redistribution requires firms to contribute only in proportion to their
profits, no matter how much capital or labor they employ to produce those profits.
Fiscal redistribution thus makes it possible to separate the price the firm pays for labor
from the price that workers charge for their services and thus to preserve the allocative
role of the price system while still redistributing income. By contrast, with direct
redistribution, these two prices are necessarily equal, so that redistribution inevitably
has undesirable allocative consequences.
The foregoing argument shows why it is important to distinguish between the means
of redistribution and the amount. Regardless of the amount of redistribution desired,
fiscal redistribution is preferable to direct redistribution in a market economy where
capital-labor substitution is possible. The argument also shows that redistributions are
not all alike: some are more efficient than others, in the sense that they allow the same
degree of improvement in the standard of living of workers without decreasing the level
of employment. The key point is that one cannot judge the effects of any form of
redistribution simply by looking at who pays: it is essential to consider the effects of the
proposed redistribution on the economic system as a whole.
Furthermore, redistributions based on taxes and transfers are not all alike: it is not
enough to look at who pays any given tax to judge the actual consequences of a
particular form of redistribution. The “fiscal incidence” of the tax matters. For example,
increasing the social charges paid by firms for each employed worker is tantamount to
increasing the price of labor, unless firms lower their wages to compensate for the
increased charges, thereby negating the redistributive effect. By contrast, increasing
taxes on corporate profits (retained or distributed) does not increase the price of labor
to the firm and thus makes it possible to finance the same social transfer payments in a
more efficient way than by increasing social charges. Taxes paid by firms do not all
have the same effects in terms of effective redistribution: if the ultimate incidence of a
tax is truly to fall on capital, its amount must depend on the level of capital used or on
the income to capital.
This argument also illustrates a key result of modern economic theory: in pure (as
opposed to efficient) redistribution, where redistribution is justified by considerations
of pure social justice rather than by any supposed market failure, redistribution should
be achieved by means of taxes and transfers rather than by manipulation of the price
system. This is a very general idea: for example, it is more efficient to help the poor
cope with high prices by means of fiscal transfers than by establishing price controls,
because price controls lead to shortages and rationing. We will encounter this idea
again when we analyze inequality and redistribution of labor income in Chapter 3.
The Elasticity of Substitution between Capital and Labor
In the special case of redistribution between capital and labor, the implications of the
finding that fiscal redistribution is preferable to direct redistribution depend, however,
on the degree of capital/labor substitutability and therefore on the importance of the
allocative role played by the price system. No one would argue that capital-labor
substitution is completely impossible. In order to determine whether fiscal
redistribution is truly superior to direct redistribution, we must ask two questions: How
much variation is possible in the amounts of capital and labor used at the
macroeconomic level? And what influence do the prices of capital and labor have on
the amounts of each the economy uses? If capital/labor substitutability is low, direct
distribution has the advantage of being simple and transparent: why let the market
determine the income flowing to capital and labor and establish a complex system of
taxes and transfers to redistribute that income if a comparable result can be obtained by
directly requiring firms to divide revenues between capital and labor in a manner
deemed to be just?
To measure the degree of capital/labor substitutability and the allocative role of the
price system, economists rely on the notion of elasticity of substitution between capital
and labor. This is defined as the percentage reduction in the quantity of capital used by
firms when the price of capital increases by 1 percent relative to the price of labor. This
elasticity measures not only the choices of individual firms (to lay off workers, say, if
the price of labor increases and to hire additional workers if it decreases) but, more
importantly, the macroeconomic consequences of these individual firm-level decisions
(for example, a labor-intensive sector may hire new workers less rapidly and grow
more slowly if the price of labor rises and hire more rapidly and grow faster if the price
of labor falls).
A high elasticity indicates that it is easy, if need be, for the economy as a whole to
replace capital by labor, and vice versa: in this case, one says that there is a high
elasticity of substitution between capital and labor. If the elasticity of substitution is
greater than 1, a 1 percent increase in wages reduces the quantity of labor used by more
than 1 percent, so that labor’s share of income decreases. An elasticity equal to 1
percent corresponds to the case in which the two effects exactly balance each other, so
that labor’s share of income is constant, regardless of the price of labor and capital.
This corresponds to the case of a Cobb-Douglas production function, named for the two
economists who developed it in the 1920s after observing the division of income
between profits and wages in American and Australian industry, from which they
concluded that the function they found correctly accounted for what they observed (see
Douglas, 1976, for a posthumous overview). We will see later whether their analysis is
consistent with more recent data and studies. Conversely, an elasticity of substitution of
less than 1 brings us closer to the case of fixed-coefficient technology: in this case, the
marginal productivity of capital and labor falls off rapidly as one departs from the norm
of n workers per machine, hence capital’s share of income decreases and labor’s share
of income increases when the price of labor rises. The extreme case of completely fixed
coefficients corresponds to zero elasticity of substitution: no deviation from the norm of
n workers per machine is possible. The capital-labor split then becomes a purely
distributive problem, and we are faced with the distributive conflict described by
classical theory, as discussed earlier.
The debates stirred by high unemployment in Europe in the 1980s and 1990s
illustrate the political issues raised by the question of elasticity of substitution between
capital and labor. Many commentators suggested that the considerable increase of taxes
on labor (especially in the form of social contributions) and the decrease of taxes on
capital (lower taxes on profits and tax exemptions granted to the income from many
forms of capital owned by households) contributed to the increase in unemployment in
Europe after 1970 by increasing the cost of labor and thus encouraging firms to use
more capital and less labor, or at least by reducing their incentive to use more labor and
penalizing the development of labor-intensive sectors. This led to proposals to shift
some of the burden on labor to capital, for example, by requiring firms to pay social
contributions not on their total wage bill but on their profits, or by broadening the base
of social contributions from wages to capital income (in France, the Generalized Social
Contribution, or CSG, is an example of this). The practical value of such proposals
depends entirely on the quantitative magnitude of the elasticity of substitution between
capital and labor. If this elasticity is high, then such measures can indeed finance social
spending while creating additional jobs, thus leading to more efficient redistribution.
But if this elasticity is low, then such tax reform measures are illusory. And if one really
wants to make capital pay more, then why not increase wages, which would not
decrease the level of employment because that level is fixed, instead of inventing new
taxes to replace social contributions?
The Elasticity of Capital Supply
The elasticity of substitution between capital and labor is therefore the crucial
parameter for deciding what instruments of redistribution are most effective, but it does
not answer the question of the desirable extent of redistribution from the workers’ point
of view. In fact, when considering the optimal level of redistribution, whether by direct
or fiscal means, one needs to consider the effect of redistribution on the future stock of
capital in the economy. A decrease in capital’s share of income, whether due to higher
taxes on capital or increased wages, may decrease the ability of firms to finance new
investment and reduce the incentives of households to save and invest their savings in
firms.
How important in practice are these negative effects of redistribution on saving and
capital accumulation? The extreme traditional position is that they are so important that
it is in the interest of workers not to reduce the income of capital, because any
redistribution of income from capital to labor will always decrease the capital stock so
much that the productivity of labor, and therefore wages, will also decrease, even if
wages are supplemented by fiscal transfers (Judd, 1985; Lucas, 1990b). In that case, a
pragmatic conception of social justice, typified by the Rawlsian maximin principle
discussed in the Introduction, leads to the conclusion that the state should do nothing to
redistribute income from capital to labor whether by direct or fiscal means: any attempt
to reduce inequality would ultimately work against the least well off and would
therefore not be just. The redistributive efforts of government should therefore be
limited to reducing labor income inequality and forget about inequality between capital
and labor.
This argument may be logically correct, but it is not confirmed by empirical studies.
To measure the effects in question, economists rely on the notion of “elasticity of the
supply of capital.” If the rate of return on invested capital decreases by 1 percent, by
how much does the supply of capital—that is, the quantity of savings that households
decide to invest in firms—decrease? Empirical estimates of this elasticity find in
general that it is fairly close to zero: when the return on capital decreases, households
attempt to preserve future income by saving more, and in practice this seems to balance
or even outweigh the fact that a lower return on investment makes immediate
consumption more attractive than future consumption out of savings. In the jargon of
economics, one says that the income effect compensates for the effect of substitution
between present and future consumption (Atkinson and Stiglitz, 1980, chaps. 3–4). In
fact, the high interest rates and reduced taxes on capital income seen in the 1980s and
1990s did not lead to unusually high rates of saving—quite the contrary. As long as the
elasticity of capital supply is zero or close to it, that is, as long as the capital stock is
relatively independent of the extent of redistribution, then fiscal redistribution permits,
and social justice recommends, as extensive a redistribution as possible between
capital and labor. If the elasticity of substitution between capital and labor is
nonnegligible, such ambitious redistribution cannot be achieved efficiently through
direct means, which, as discussed earlier, tend to reduce the level of employment
unnecessarily.
It is true, however, that estimates of the elasticity of the supply of capital measure
only some of the negative potential effects of redistribution, because in practice only
part of investment comes directly from household savings. Another important, even
predominant, part comes directly from the profits that firms do not distribute to
shareholders and creditors. Relying on retained earnings is often more efficient than
soliciting outside savings. Therefore, we must also take into account the effects of
capital-labor redistribution on the financial structure of firms and their ability to invest
out of earnings in order to arrive at a global estimate of the elasticity of the supply of
capital and therefore at the optimal level of redistribution from the standpoint of social
justice.
 A more fundamental objection is that, even if the elasticity of the supply of capital is
really low, taxation of capital income causes significant problems in a world in which
savings and investment are internationally mobile and states choose their level of
redistribution independently while attempting to attract as much investment capital as
possible. Tax competition makes the supply of capital highly elastic for each state taken
in isolation, even if the true elasticity of the global supply of capital is low. In fact, lack
of coordination between governments largely accounts for the significant decrease in the
taxation of capital in Europe in the 1980s and 1990s. Without fiscal federalism—or
taxation of capital at the broadest possible geographical level—it is impossible to
achieve an optimal redistribution from capital to labor in the social justice sense.
Are Capitalists and the Price System Necessary?
If one could precisely measure the elasticity of substitution between capital and labor
and the elasticity of the supply of capital, it would be possible in principle to determine
the optimal amount of capital-labor redistribution and the best instruments to achieve it.
The intellectual and political conflict over redistribution is about more than just the
measurement of elasticities, however. Indeed, this whole conceptual framework
implicitly assumes that we accept the rules of the market economy and the allocative
role of the price system. This is obvious in the case of the elasticity of capital supply
(why should society give in to the threat of capitalist households to save less if they
deem the rate of return on capital to be too low?). It is just as important, though, when it
comes to the elasticity of capital/labor substitution: why should firms use more capital
and less labor if the relative price of labor rises? Wouldn’t it be enough to prohibit
firms from laying off workers, or simply to require firms to take into account the
collective goals of high employment and social justice, subject to oversight by works
councils and a vigilant public? To accept the allocative role of the price system and
insist that fiscal redistribution is preferable to direct redistribution is tantamount to the
claim that individual self-interest is the adequate guide to making correct resource
allocation decisions in a complex economy. Traditionally, the left has rejected such
fatalism and placed its faith in the possibility of organizing the economy in a more
solidaristic way. Hence the left remains skeptical of taxation as the primary means of
achieving social justice. In Chapter 3, we will see how this skepticism has influenced
thinking about the redistribution of income from labor.
For now, consider the fact that this refusal to embrace the logic of the price system
and of fiscal distribution (rather than belief in a low elasticity of capital-labor
substitution) explains why much of the European left and trade-union movement was
unenthusiastic toward, if not downright hostile to, the tax reform proposals of the 1980s
and 1990s aimed at decreasing the tax burden on labor (such as the French CSG, as
discussed earlier). These proposals were based on the idea that if labor is abundant,
then a low price of labor and high price of capital may not be the worst way of
encouraging firms to use less capital and more labor and encouraging consumers to
consume more labor-intensive goods and fewer capital-intensive goods. How could it
be otherwise in a world in which a wide range of different goods and services are
produced, and it is difficult to determine the precise capital and labor content of each
good? In other words, prices serve as a signal conveying information to various
economic actors that would be difficult to convey in any other way, as illustrated by the
universally acknowledged failure of central planning. But the issue is sufficiently
complex, and the fatalistic acceptance of individual self-interest sufficiently depressing,
that not everyone is willing to accept the logic of this argument.
The debate about the price system, individual self-interest, and other possible forms
of economic organization, which raises questions that can never be fully answered on
the basis of historical data alone, obviously stands on a different plane from the
question of empirical estimates of the elasticity of the capital supply and capital/labor
substitution, as well as on another plane from the controversy between classical and
marginalist theory over the capital-labor split. Nevertheless, the two debates have not
always been totally independent. As discussed above, a low elasticity of capital/labor
substitution makes the price system less useful. If the capitalist mode of production is
simply a mechanism for matching fixed quantities of capital and labor, n workers per
machine, why is it necessary for anyone to own the machine? If the owner of the
machine does nothing but claim a share of what it produces, then he could be eliminated
by collectivizing the means of production. Saving could be replaced by taking a
sufficient sum out of national income for the purpose of increasing the stock of machines
and matching them to an appropriate number of workers: there would be no need of
capitalists to accomplish this. This is obviously what Marx concluded from his
observations of the capitalist economy, whose operation seemed terribly simple.
Conversely, to insist on the possibility of substituting capital for labor, as marginalist
economists do, is to emphasize the complexity of the modern economy and introduce the
existence of choices, which someone must make, and this allows one to argue that the
price system and private property are legitimate in the absence of some other system for
solving complex problems of allocation. That is why the debate about capital/labor
substitution has often been seen as a more general debate having to do with the
legitimacy of capitalism and the price system, first in Marx’s controversies with
marginalist economists in the 1870s and later in the Cambridge capital controversy of
the 1950s and 1960s.
Although this blurring of the boundaries between different debates is
comprehensible, it is also unfortunate. The legitimacy of the price system is not simply a
question of capital/labor substitutability, if only because the price system can play a
useful role in deciding what goods and services to produce even if there is no
possibility of substituting capital for labor. Conversely, as noted earlier, an answer to
the question of capital/labor substitution cannot tell us what the optimal level of
redistribution from capital to labor is. If we take the market economy as given, the real
issue in the controversy between the classical theory and the marginalist theory of the
capital-labor split is the issue of direct versus fiscal redistribution.

A Compromise between Short-Term and Long-Term Theories?

Can historical data, particularly concerning the capital-labor split, help to resolve the
controversy between classical and marginalist theory?
It is not always easy to move back and forth between theoretical notions such as
national income, profits and wages, or capital and labor and the empirical notions
embodied in national accounting statistics (see box, “Measuring Capital’s Share”).
Once these difficulties are overcome, however, we observe a striking empirical
regularity, which Keynes as early as 1930 regarded as the best-established regularity in
all of economic science.
Indeed, Table 2.1 shows that the respective shares of profits and wages in the
national income of three countries with very different histories, particularly in regard to
social matters, remained more or less constant: the wage share never fell below 60
percent and never rose higher than 71 percent, generally hovering around 66–68
percent, and it is impossible to detect any systematic upward or downward trend. It
appears that profits and wages always divide in such a way as to award one-third of
national income to capital and two-thirds to labor.*
 From Share of Value-Added to Household Income
Consider first the relation between this one-third–two-thirds split of income between
capital and labor and the distribution of household income discussed in Chapter 1.
Table 2.1 shows the split of primary income between labor and capital, that is, the sum
of all wages and other compensation, including employer-paid social charges (payroll
taxes), effectively paid by firms to their workers, and the sum of gross profits or
operating surpluses, that is, the residual revenue of firms after paying their workers (see
box). The relation between this and disposable household income is complex. For
example, a significant portion of the sums that appear in the “labor” column of Table 2.1
is paid in the form of social charges, which reappear in the form of pensions and social
transfers in the disposable household income reported in Table 1.1. Furthermore, firms
do not distribute all of their profits to the capitalist households that own their shares or
bonds: a significant portion of firm profits, often more than half, is retained to
compensate for the depreciation of capital (on average nearly 10 percent of value
added) and to pay for new investments without having to seek outside capital.

MEASURING CAPITAL’S SHARE

How does one measure the respective shares of profits and wages in total revenue? What a firm earns by
selling its products to consumers and other firms goes to pay three different types of costs:
The cost of intermediate consumption, that is, of goods and services the firm buys from other firms and
consumes to produce its own goods and services, as opposed to the machinery and equipment that
constitutes the firm’s capital.
The remuneration of workers, which includes net wages paid to workers, employee-paid social
charges deducted from the worker’s pay (gross wage = net wage + employee-paid social charges),
and employer-paid social charges. This aggregate is equal to gross labor income, or, more simply, labor
income.
The remainder of the proceeds from sales (after the first two costs are deducted) is called the gross
operating surplus (GOS). It is generally much higher than the firm’s profit in the strict sense because
the GOS goes to pay not only dividends to shareholders but also interest on loans, taxes on profits, and
the replacement cost of worn machinery and equipment (depreciation of capital, or amortization). This
aggregate represents the gross capital income, or, more simply, capital income.

The firm’s value added is defined as the difference between the revenue from sales and the cost of
intermediate consumption. The value added is thus equal to the sum of labor income and capital income.
When we calculate the profit and wage shares, we calculate the share of capital income and labor income as
percentages of value added. In other words, we deduct the cost of intermediate consumption. This is
perfectly legitimate, since the cost of intermediate goods purchased from other firms goes to remunerate
capital and labor in those firms, and we need to avoid double counting.
In addition to taxes levied directly on capital (such as tax on profits) and labor (such as social charges),
which are included in capital and labor income, firms also pay indirect taxes such as the value-added tax, the
amount of which does not depend directly on how value added is divided between capital and labor. Hence
these taxes cannot be attributed to either gross capital income or gross labor income. In calculating the profit
and wage shares, it is customary to omit these indirect taxes: in other words, we calculate these respective
shares as a percentage of value added net of indirect taxes. Thus the sum of capital’s share and labor’s share
equals 100 percent of net value added, as in Tables 2.1 and 2.2. This makes the results easier to interpret,
since indirect taxes do not depend directly on the capital/labor split.
Finally, another complication has to do with the treatment of self-employment income (farmers, merchants,
doctors, lawyers, and so on). Here, value added goes to pay both the labor of self-employed workers and the
capital they have invested, even though their accounts do not clearly distinguish between wages and profits.
Without correcting for self-employment income, one would find that the wage share of total value added has
increased significantly since the nineteenth century, simply because the percentage of non-self-employed
workers has increased so much (Morrisson, 1996, p. 78). The accounting convention used by the OECD is to
attribute to self-employed workers the same average wage as non-self-employed workers. This convention
 has been applied to adjust the figures in Tables 2.1 and 2.2.

Allowance must also be made for taxes paid by firms on their profits before
distribution to shareholders. This factor is of limited importance, however, because
although the tax on profits in most Western countries is on the order of 40–50 percent,
receipts from this tax generally do not exceed 2.5–3 percent of GDP. In France in the
1990s, the figure was as low as 1.5 percent of GDP, even though the share of capital in
value added was higher than elsewhere (OECD, 1995, p. 78). This is because the notion
of taxable profits is much narrower than the notion of gross operating surplus, since
firms are allowed to deduct not just the estimated depreciation of their capital stock but
also the interest paid to creditors, provisions for anticipated risks, and so on. The tax on
profits is by far the most riddled with loopholes in the whole tax system.
Note, finally, that a significant portion of the sums listed under “wages” in Table 1.1
are in fact paid by government entities out of revenue from the tax on profits or on total
value added (in the form of a value-added tax, or VAT). This tends to increase the share
of wages relative to capital income in total household income compared with the
capital-labor split in the value added by firms. All these factors explain why, given
gross profits on the order of 32–34 percent of value added by firms, the share of capital
income actually received by households is typically only 10 percent of total household
income (see Chapter 1).
What the Constancy of the Profit Share Tells Us
Consider once again Table 2.1. How should we interpret the constancy of the profit
share across time and space? Leaving the matter of retained profits aside, the first
lesson to be drawn from this regularity is that the considerable growth of workers’
purchasing power in the twentieth century cannot be explained by changes in the capital-
labor split. In other words, the reasons why the purchasing power of the French worker
increased fourfold between 1920 and 1990 (see Table 1.6) was not that social conflict
reduced the share of income appropriated by capitalists. Indeed, the share of wages in
total value added was broadly speaking the same two-thirds of national income in 1920
as it was in 1990 (see Table 2.1). Furthermore, although the two world wars and
changes in nomenclature make it difficult to extend statistical series on the profit/wage
split beyond 1920 for France, we can use American statistics to go back all the way to
1869, and these indicate that the variation in the wage share was already limited to
between 66 and 68 percent in the nineteenth century (Duménil and Lévy, 1996, chap.
15). In other words, the profit-wage split has remained almost constant for more than
120 years, even though wages increased tenfold.
To be sure, it is important that capital appropriated a third of value added throughout
this period. If that income had all been distributed to labor, including the portion
devoted to capital depreciation, wages could have been increased by 50 percent, which
would have significantly improved the living conditions of workers in 1870 or even
1990, conditions that were in many cases miserable when compared with the opulence
in which many capitalists lived. At the same time, however, we must admit that this 50
percent increase in wages would have been only half as much as the wage increase that
actually occurred between 1870 and 1910 and less than a quarter as much as the
increase that actually occurred between 1950 and 1994 (Table 1.6). It is difficult,
moreover, to believe that the wage increases of 100 percent between 1870 and 1910
and more than 200 percent between 1950 and 1990 would have occurred if the capital
share had been reduced to zero in 1870 or 1950. Although our knowledge on this point
is limited, it is likely that the supply of capital would have decreased if there had been
that much redistribution, hence that the optimal capital/labor redistribution from the
standpoint of workers would have been much smaller, though surely larger than what
was actually instituted.
Who Pays Social Charges (Payroll Taxes)?
The second lesson to draw from Table 2.1 has to do with the question of fiscal
incidence. In the 1920s and 1930s firms paid little in the way of social charges,
whereas in the 1990s 45 percent of the wage bill consisted of social charges paid by
employers, while workers paid another 20 percent of their gross wages to support the
welfare state (see box). Who actually bore the burden of employer-paid social charges?
Certainly not the employers, since labor’s share of value added, which includes
employer-paid social charges, did not increase between 1920 and 1995. Similarly,
employer-paid social charges were much lower in the United States and United
Kingdom than in France in the 1990s, but labor’s share of value added was no higher in
France than in those two countries—indeed, the opposite was true (see Table 2.1). In
1996, the maximum rate of employer-paid social charges as a percentage of gross
wages was 7.65 percent in the United States and 10.2 percent in the United Kingdom,
with workers paying the same percentage, and total receipts from social charges (paid
by both employers and employees) represented 6–7 percent of GDP, compared with
nearly 20 percent of GDP in France (OECD, 1995, p. 79). If employer-paid charges
were actually paid by employers, we would therefore expect to find that labor’s share
of value added in France was at least 10 percent of GDP higher than in the United States
and United Kingdom.
It is therefore clear that social charges are not paid out of capital income. This is a
crucial fact, since it implies that modern systems of social protection, which are central
to today’s systems of redistribution and which were based on the idea of dividing social
costs between capitalists and workers, have not in fact redistributed income from
capital to labor; labor income has absorbed the full cost. This in no way undermines the
legitimacy of such systems, which do allow for considerable redistribution of labor
income and fulfill an insurance function that private markets are often incapable of
assuming (see Chapter 4). It does, however, profoundly challenge the implicit intention
to redistribute income from capital to labor that in many cases informed the creation of
the welfare state. This vision was closely linked to the classical theory of the capital-
labor split, according to which a better division could be achieved through negotiation,
for example, by setting a higher rate for employer payments than for employee
payments: social benefits were thought of as a supplement to the wage that capitalists
would otherwise have paid to workers.
In fact, the evidence appears to indicate that, as the theory of fiscal incidence would
predict, the only thing one needs to know is how the tax is assessed, that is, how its
amount depends on the level of wages, profits, and so on, and not what the name of the
tax is or who is officially supposed to pay it, that is, who writes the check to the
appropriate collection authority. It does not much matter whether social protection is
paid for by an income tax proportionate to wages or by social charges levied on both
employers and employees. In Denmark, there is no payroll tax, and the generous Danish
system of social protection is financed entirely by income tax (which in practice is
always mainly a tax on wages and social income in view of the limited importance of
capital income). Unsurprisingly, the share of labor income in value added is the same in
Denmark as elsewhere (OECD,1996, p. A27). Danish firms spend as much as French
firms on wages but pay it all to their workers instead of paying social charges; the
workers then pay tax on their income. More generally, the share of social charges in the
financing of European social protection systems varies widely from country to country,
with France and Denmark representing the two extremes, but the share of labor in value
added is more or less the same everywhere. The only relevant parameters are how the
tax rate, whether assessed in the form of an income tax or a social contribution, depends
on the wage level (in other words, how progressive is the assessment) and whether it
also depends on the level of capital income. In particular, the only way to redistribute
income from capital to labor is to tax capital.
A Cobb-Douglas Production Function?
With these lessons in mind, how can we explain the constancy of the share of profits?
The interpretation traditionally favored by economists is that over the past century of
capitalism, the Western economies have been reasonably well described at the
macroeconomic level by a production function of the Cobb-Douglas type, which is to
say, with an elasticity of capital/labor substitution equal to 1. Indeed, only if this
elasticity is 1 can one confidently predict that the respective shares of profits and wages
should remain constant over time regardless of changes in the available quantities of
capital and labor and regardless of political and economic shocks affecting the price of
each factor. An elasticity of 1 would also explain the observed fiscal incidence of
social contributions, whose burden falls on labor and therefore increases its price.
Even with a fixed-coefficient technology, one could of course argue that social and
political conflict always leads to the same result in all countries, with one-third of
income going to capital and two-thirds to labor. As Solow himself has noted, we would
need to know how much variation it would be natural to expect before describing the
constancy of the capital-labor split as surprising (Solow, 1958). Econometricians have
looked at how hiring by individual firms varies in response to variations in the price of
labor, however, and these microeconomic studies show considerable elasticity of
substitution between capital and labor. After comparing the results of several dozen
studies across all Western countries, Hamermesh (1986; 1993) found that most
estimates of the demand elasticity for labor suggest elasticities of substitution between
capital and labor of 0.7 to 1.1, leading to the conclusion that “the Cobb-Douglas
function seems to be a fairly good approximation to reality” (1986, pp. 451–452, 467).
Contrasting experience in various Western countries since the 1970s also suggests
substantial capital/labor substitutability (I will return to this point later). The data thus
seem to confirm the marginalist theory of the capital-labor split and therefore the
superiority of fiscal redistribution over direct redistribution.
Historical Time versus Political Time?
One should be careful not to underestimate the limits of this historical regularity,
however. Although the profit share of income is impressively constant over the long run,
it does vary over the short run, and the long run can seem very long indeed to the
individuals affected. Consider, for instance, the evolution of respective shares of profits
and wages in the OECD countries between 1979 and 1995.
Tables 2.1 and 2.2 show large variations in the profit and wage shares. Although the
wage share tended to increase in the 1970s, it was the profit share that increased in the
1980s and 1990s, in some cases substantially. These variations were widest in France,
where the wage share was 66.4 percent in 1970, rising to 71.8 percent in 1981, then
falling after 1982 to 62.4 percent in 1990 and 60.3 percent in 1995. How can we
explain the fact that more than 5 percent of national income was redistributed from
capital to labor between 1970 and 1982, while 10 percent went from labor to capital
between 1983 and 1995?
It so happens that the first period coincides with a period of substantial wage
increases inaugurated by the Grenelle Accords of 1968. Wages continued to improve in
the 1970s owing to the influence of social movements and substantial increases in the
minimum wage. The last major increment to the minimum wage occurred in 1981. After
1983, wages ceased to be indexed to prices, and increases to the minimum wage were
minimal. In fact, the purchasing power of the average net wage increased 53 percent
between 1968 and 1983 but only 8 percent between 1983 and 1995 (INSEE, 1996a, p.
48). True, GDP increased 44 percent between 1970 and 1983 but only 28 percent
between 1983 and 1995 (INSEE, 1996c, p. 34), and this growth had to finance a
growing burden of pension and health care expenditures, but the reduction of wage
growth relative to national income growth was nevertheless quite real. In other words,
over a period of twenty-five years, the predictions of the classical theory of the capital-
labor split seemed to be accurate: the profit share decreased when social militancy
enabled workers to win important concessions on wages, and it increased when
constraints were imposed on wages. Yet the increased profit share did not lead to the
promised creation of new jobs.
 To be sure, these substantial variations over a twenty-five-year period do not change
the fact that over periods of fifty or a hundred years, wages have always accounted for
roughly two-thirds of added value, so that the increase in workers’ purchasing power of
250 percent since 1950 and 700 percent since 1870 cannot be explained by changes in
the capital-labor split. But why would that matter to the workers who lived through the
twenty-five years from 1970 to 1995? Their standard of living increased sharply from
1968 to 1982 and then stagnated from 1983 to 1995 while output continued to grow.
How could they not associate the improvement in their standard of living with
redistribution from capital to labor? The right-wing view that true improvement in the
standard of living can come only from growth and not from redistribution is valid only
in the long run, and politics, in which workers have a legitimate interest, operates on a
different time scale.
Furthermore, how could workers fail to associate redistribution from capital to labor
with social struggles and wage increases and thus with direct rather than fiscal
redistribution? Indeed, no fiscal redistribution had ever redistributed 10 percent of
national income in so short a period. To give an order of magnitude, the fiscal
redistribution measures adopted by the Socialist government after it came to power in
France in 1981, which were denounced at the time by the right as the height of “fiscal
bludgeoning” and which consisted essentially of a wealth tax and a surtax on top income
brackets, brought in less than 10 billion francs in 1981 (Nizet, 1990, pp. 402, 433), or
0.3 percent of national income. In theory, a government can achieve any level of
redistribution it wants via taxes and transfers, but in practice no transfer of comparable
magnitude has been accomplished in so few years. Inevitably, therefore, workers think
of and experience redistribution primarily in terms of social struggle and wage
increases rather than fiscal reform and transfer payments. It is not so much rejection of
the logic of fiscal redistribution and the price system that sustains left-wing skepticism
of redistribution through taxation; it is rather this historical reality. We will again
encounter this same historical reality and contrast between historical and political time
in the discussion of inequality of labor income in Chapter 3.
This level of variation in the capital-labor split over a period of ten to fifteen years
is not historically unique, moreover, even if certain specific features of French social
and political history helped make the period 1970–1990 particularly noteworthy. For
instance, the share of wages in the value added by American firms decreased from about
65 to 55 percent between 1869 and 1880, then rose again to 65 percent in 1885 and to
66–68 percent in 1890. The average wage rose by only 2 percent between 1869 and
1880 but then increased 27 percent between 1880 and 1885, a period marked by major
strikes and an unusually active trade-union movement (Duménil and Lévy, 1996, chap.
16). Over a period of ten to fifteen years, the marginalist understanding of the capital-
labor split can frequently seem rather absurd in the light of social realities. The same
can be said about the question of fiscal incidence: in the short term, social charges
assessed on employers are indeed paid by the employers and are not immediately
nullified by wage reductions, a fact that inevitably shapes the way many people see
fiscal incidence, even though it is true that over the long run social charges always end
up being paid by labor.
Why Has the Profit Share Not Increased in the United States and United Kingdom?
When it comes to the history of the capital-labor split in the period 1970–2000, things
are not so simple, however. The French pattern was reproduced in Italy, where the
profit share rose from 34.5 percent in 1983 to 42.5 percent in 1995, and to a slightly
lesser extent in Germany, where it rose from 28.2 percent in 1981 to 36 percent in 1995.
Strikingly, however, the United States and United Kingdom seem to have entirely
avoided this increase in the profit share in the 1980s and 1990s: the wage share of US
value added held steady at 66–67 percent throughout this period, while in Britain it
remained between 68 and 71 percent (Table 2.2). It is difficult to compare profit shares
at different points in time between different countries because of numerous differences
in accounting conventions, but comparing trends over time leaves no doubt: the capital
share increased by nearly 10 percent of value added in France, Italy, and Germany,
whereas it did not increase at all in the United States and United Kingdom. In contrast to
what happened with wage inequality, which increased sharply in the United States and
United Kingdom after 1970 (see Chapter 1), the countries where ultraliberalism won out
in the 1980s and 1990s are the only countries in which the profit share did not increase.
How do we explain this?
Part of the explanation is surely that there was catching-up to do: in France, the
profit share had decreased by 5–6 percent of value added in the 1970s owing to very
rapid wage increases. In the United Kingdom, the profit share decrease was much more
moderate, and in the United States it did not occur at all (Table 2.1). This cannot be the
whole story, however: by 1985–1986, the profit share in France had returned to its
1970 level, yet it continued to increase, while the profit share in the United States and
United Kingdom remained stable.
It is hard to avoid relating this to the fact that the United States and United Kingdom
were the only two countries to have created jobs in this period, thus increasing their
total wage bill, while wages stagnated in the other countries. Between 1983 and 1996,
the United States created 25 million jobs, an increase of about 25 percent (from 100.8
million to 126.4 million), while total employment in France increased by barely 2
percent (from 21.9 million to 22.3 million), and GDP in both the United States and
France increased by about 30 percent (OECD, 1996, p. A23). This is surely the best
proof that capital and labor can be combined in different proportions to increase
production by the same amount, hence that there is considerable opportunity for
substitution at the macroeconomic level. Between 1983 and 1996, the French economy
grew by employing skilled labor together with new machinery and infrastructure, while
American growth relied on intensive use of labor, especially relatively unskilled labor
in the service sector (restaurants, commerce. etc.) (Piketty, 1997b). This interpretation
is confirmed, moreover, by available data concerning the growth of the capital stock in
the period 1970–2000, which show that it grew much more rapidly in France and most
other European countries than in the United States (IMF, 1996). The data also show the
degree to which capital/labor substitution can be related to major intersectoral
reallocations (from industry to services, for example), and not just substitution of
workers for machines at the firm or sector level.
The simplest explanation might be that labor was not substituted for capital in
France, and jobs were not created, because wages were too high owing to very rapid
wage growth in the period 1968–1983. This would suggest that the long run, in which
the effects of marginalist theory would manifest themselves, is not as far off as the
individuals involved might have hoped. In order for lower wages to lead to more jobs,
however, the job creation effect would have to outweigh the wage effect, that is, the
elasticity of substitution between labor and capital would have to be greater than 1,
which, as mentioned, is higher than the usual estimates. Furthermore, if the average
wage in the United States rose by barely 5 percent between 1983 and 1996, it increased
by nearly 20 percent in the United Kingdom, compared with less than 12 percent in
France, yet total employment in Britain increased by nearly 10 percent in the same
period (OECD, 1996, A15, A19, A23). In the period 1983–1996, France thus appears
to have lost across the board, since wages and jobs stagnated, leading to an
exceptionally large decrease in labor’s share of value added.
In addition to the average cost of labor, two other factors might explain why the
wage share decreased in France and continental Europe while holding steady in the
United States and United Kingdom. One explanation might be that the variation of wages
with skill level increased in the United States and United Kingdom, and that this was the
sole explanation of employment growth in the 1980s and 1990s (see Chapter 3). A
second possible interpretation is that labor income includes a nonmonetary component
in the form of stability and guaranteed employment, and this component decreased in the
United States and United Kingdom but remained high in France and elsewhere in Europe
(see Cohen et al., 1996 for a comparison of France and the United States). We would
then need to explain why the price of this employment guarantee should have increased
between 1970 and 1995, and compare this to the incontestably high value placed on it
by workers.
 The Dynamics of the Distribution of Capital
Why does capital/labor inequality command so much attention? Not just because capital
claims a significant share of national income. Even more striking is the fact that
capital/labor inequality often reproduces itself or even grows over time. It is this
reproduction that makes capital/labor inequality seem arbitrary, useless, and
incompatible not only with common sense and social justice but also with economic
efficiency: Why should capital-poor individuals or countries be denied the possibility
to invest to the extent their talents permit? In other words, capital/labor inequality
immediately raises the question of efficient (as opposed to pure) redistribution. The
time has come, therefore, to move from the macroeconomic division of total income
between the two factors of production, labor and capital, to the study of the income
distribution at the individual level. What are the dynamics underlying changes in the
income of individual workers and capitalists, and what investment opportunities are
open to them? Does the logic of the market economy lead to an inefficient reproduction
of inequality in the distribution of capital over time? If so, how can it be prevented?

The Theory of Perfect Credit and Convergence

Once again we have dueling theories. The central question has to do with the credit
market. If the credit market were perfectly efficient, that is, if capital were invested
wherever a profitable investment opportunity existed, then any initial inequality in the
distribution of capital should eventually cease to exist. No matter how wealthy a family
or country is initially, all equally enterprising units of labor would then be able to make
the same investments thanks to the credit market. Hence the inequality of initial
endowments would not persist. Of course, even in a perfect credit market, a poor person
or country that borrows to invest must eventually repay the loan and therefore cannot
instantly save enough to close the wealth gap between borrower and lender. In fact, if
the savings rate of low-income earners is lower than that of high-income earners, the
inequality between borrower and lender can persist indefinitely (Bourguignon, 1981).
At the international level, this would correspond to a situation in which per capita GDP
is the same in all countries, since an equal quantity of capital is invested per worker
everywhere, but GNP is lower in poor countries, where the capital is owned by the rich
countries and which must therefore pay a share of their profits every year to capitalists
in other countries. But if borrowers save at the same (or nearly the same) rate as
lenders, they can gradually accumulate the fruits of their labor, borrow less and less,
and eventually catch up. In fact, savings rates are not systematically lower for low-
income countries than for high-income countries: they were above 30 percent in the
Asian “tigers” in the 1950s and 1960s and below 10–15 percent in all Western
countries in the 1980s and 1990s, even though the latter were much wealthier than the
former; that is why these developing countries were able to catch up with the rich
countries (Young, 1995).
Convergence of the rich and poor countries is the principal prediction of the
standard model of growth and capital accumulation (Solow, 1956), which assumes
perfect credit markets. What justifies this assumption? If you believe in market forces,
the answer is simple: competition. Why would a bank or wealthy capitalist not lend to
someone with a profitable investment project who promises to pay a competitive rate of
interest? If the problem faced by poor countries is lack of machinery and infrastructure,
then new investment should enable them to increase their output considerably. What
could prevent rich countries from using their savings to profit from such investment? If
some potential lenders are nervous or hesitant, what is to prevent others from taking
advantage of the opportunity? The perfect credit model thus depends on competition
among savers and financial intermediaries to identify the most profitable investment
opportunities. It follows immediately that redistribution from those well endowed with
capital to those less well endowed can be justified solely on grounds of pure social
justice. If this is correct, then the unequal distribution of capital as such poses no
inherent problem of economic efficiency, because the market would see to it that the
available capital would be invested efficiently (in the Pareto sense, as discussed in the
Introduction). No direct intervention in the production process would be required.
The Question of Convergence between Rich and Poor Countries
The convergence model leads to some particularly striking predictions at the
international level. If investment capability does not vary systematically from one
country to another, we should expect to see a global catch-up phenomenon. Capital
should flow to poor countries, increasing their rate of growth, thus reducing and
eventually eliminating inequality between countries. That is the theory. What do we see
in practice?
A comparison of per capita income with growth rate for a sample of countries in the
period 1960–1990 fails to confirm the theoretical prediction: there is no systematic
relation between the two variables (Mankiw et al., 1992, p. 427). Some Asian countries
that were relatively poor in 1960, such as Taiwan, South Korea, and Singapore, did see
higher growth rates of per capita income than Western countries, but other poor
countries, such as those on the Indian subcontinent or in sub-Saharan Africa,
experienced very low or even negative rates of growth. The convergence model seems
to explain catch-up among the Western countries themselves: for example, the countries
of Western Europe closed the gap with the United States after World War II. It also
accounts for the rapid growth of Asian countries with intermediate levels of per capita
income, which also moved closer to the more advanced economies. But it does not help
us to understand the relation between the richest countries and the poorest or between
the rich countries and the middle-of-the-pack countries of South America, both cases in
which the gap has widened. A similar conclusion would probably hold true over longer
periods—if it were possible, for example, to measure correctly the income gap between
developed and underdeveloped countries since the nineteenth century (Morrisson, 1996,
p. 181). In fact, not only did the rich countries fail to invest massively in poorer
countries, but the exact opposite occurred: broadly speaking, we see a net flow of
capital from the poorest countries to the richest (Lucas, 1990b), because capitalists in
the poor countries invest more in the rich countries than the other way around.
Just because capital has not been invested in the poor countries, which have
remained poor, it does not follow that credit market imperfections are solely
responsible. If, for instance, we control for the “initial stock of human capital” in 1960
(as measured by literacy rates, educational levels, and so on), we find that there is a
negative correlation between initial per capita income in 1960 and average growth rate
in the period 1960–1990. Endogenous growth theorists refer to this as “conditional”
convergence as opposed to the “unconditional” convergence predicted by the Solow
model (Mankiw et al., 1992). For example, South American countries with the same per
capita income as the Asian tigers in 1960 had a much lower stock of human capital,
because large segments of the population had been totally neglected, whereas in Asia
inequality was much lower. These countries grew much more slowly than the Asian
tigers, which were able to catch up with the West. In addition to human capital, initial
inequality also had a negative effect on future growth, whether directly or indirectly
(owing to social and political instability due to inequality) (Benabou, 1996).
Another lesson from the experience of the Asian tigers is the importance of global
market integration. The miraculous growth formula (based on high and relatively
egalitarian investment in human capital on the one hand and economic liberalization and
openness to foreign markets on the other) appears to have spread in the 1980s and
1990s to larger Asian countries. The fact that liberalization has been less successful in
India than in China reminds us, however, of the crucial importance of the egalitarian
element, without which liberalization cannot lead to lasting growth (Drèze and Sen,
1995). Egalitarian educational policies are probably the most basic example of efficient
redistribution (see Chapter 3).
As Robert Lucas (1990b) has shown, moreover, if the difference in per capita
income between the United States and India were explicable solely by different
endowments in machinery, infrastructure, and so on, one would have to conclude that the
marginal productivity of capital in India is 58 times higher in India than in the United
States. With that much of a difference in the return on capital invested in India, it is
difficult to see what kind of credit market imperfection could account for the failure of
Western capital to take advantage of the opportunity. We must therefore yield to the
evidence, which shows that a substantial portion of the inequality between the rich and
poor countries, and indeed of inequality in general, is due not to unequal distribution of
the means of production but to unequal distribution of human capital: the fact that nearly
50 percent of the Indian population is illiterate surely reduces by a substantial amount
the return on an additional unit of capital invested in India (Drèze and Sen, 1995, table
A1).

The Problem of Capital Market Imperfections

Just because other factors influence growth, however, it does not follow that capital
flows from the rich to the poor countries are not important. The absence of substantial
flows from the rich to poor countries reminds us of the chronic weakness of
international capital flows in general. In the West, for example, the available amount of
domestic savings is closely related to the annual level of domestic investment, whereas
we would not expect this to be the case if international financial markets were truly
integrated, allowing domestic investments to be made with imported capital even in the
absence of domestic savings.
In contrast to what the perfect credit market model assumes, a credit operation is
more than just a mechanical placement of capital where there was none, in exchange for
which the lender or investor receives a portion of the profits. In practice, the lender or
investor has to make sure that the proposed project is profitable and that the level of
risk is acceptable. The borrower will of course always claim that it is. The lender will
also want to make sure that the borrower has sufficient incentives to do what needs to
be done over a long period of time even though some of the gains will go to the lender.
Finally, the lender needs to assure himself that the borrower will not simply disappear
with the profits. Investment thus raises a series of problems that economists have
classified as problems of “adverse selection” and “moral hazard.” Such problems arise
wherever there is an “intertemporal” market, that is, a market in which exchanges occur
in different time periods. Credit markets are intertemporal markets, as are social
insurance markets, which we will encounter in Chapter 4. These problems are
particularly difficult in international markets, because the information available about
potential borrowers and investment projects in another country may be of low quality.
This is why international capital flows are much lower than they might be.
Is competition the least harmful way to solve the informational problems that markets
and governments face? In practice, the only thing lenders can really do to make sure they
will recover their principal is to demand that the borrower put up collateral, or,
equivalently, that the borrower invest his own capital initially in order to establish a
credible commitment to the venture and a token of confidence in its viability. Hence the
amount that an individual or firm can borrow to finance an investment depends on his
(or its) initial wealth. As the French saying has it, “on ne prête qu’aux riches” (one
lends only to the rich). This is an efficient mechanism from the standpoint of lenders but
inefficient for society as a whole: total income could be higher if capital were
redistributed in such a way as to enable financing of all profitable investments. The
imperfection of the credit market is a prime example of the kind of market imperfection
that can be used to justify redistribution on grounds of economic efficiency and not just
pure social justice. In principle, it should be possible to allocate resources more
efficiently while achieving a more equitable distribution of wealth (see the
Introduction).
Many critics of capitalism, starting with the socialist theorists of the nineteenth
century, have long been aware of the phenomenon of credit rationing, even if many
considered it so obvious that they saw no need to analyze it or even mention it. It is only
since the 1970s, however, that economic theorists have begun to analyze the basic
reasons for this capital market imperfection and its consequences for redistribution
(Piketty 1994, pp. 774–779). The latter extend beyond the fact that total income could
be increased by redistributing capital. When credit rationing exists, for example, the
choice of what individuals do—whether they choose self-employment or paid labor for
someone else, for example—depends on their initial wealth, so that a redistribution of
wealth can have long-term implications for the occupational structure and for growth, as
shown by a comparison of the relatively egalitarian distribution in France after the
Revolution with the inegalitarian distribution at the time of the Industrial Revolution in
Britain (Banerjee and Newman, 1993).
Possible Public Interventions
What kinds of public intervention would help to combat credit rationing and the
persistence of capital/labor inequality to which it can lead? The principal problem
faced by possible interventions is the same as that which gives rise to credit rationing in
the first place: investing is not simply a matter of placing capital where there is none.
Complex choices have to be made, involving what sectors to invest in, what goods to
produce, and what people should be empowered to make these decisions. The
difficulties are obvious when we look at the radical solution of abolishing private
ownership of capital and collectivizing the means of production. Doing so does nothing
to resolve the problems of creating incentives and allocating resources. Other historical
experiments with efficient redistribution have faced similar difficulties: these include
public investment banks, subsidized loans, and, in poor countries, development banks.
Indeed, the theory of credit rationing tells us that it is just as difficult for a public bank
as for a private bank to make sure that capital is correctly invested when the equivalent
of a market rate of interest is to be deducted from the borrower’s eventual profits. If the
intention is to make a gift to the borrower by charging less than the market rate of
interest, as is often implicitly the case with public investment banks and other
subsidized forms of credit, it is not obvious that a public agency, no matter how well
intentioned, can correctly decide which borrowers to favor, which sectors deserve
additional investment, and so on. These very real difficulties arrive whenever a rich
country seeks to transfer wealth to a poor country. Where should foreign aid go? How
can one make sure that good use is made of it? Capital redistribution is not a matter of
dropping capital from a helicopter into places where there isn’t any. In practice, it is
much easier to allow labor to move from places where capital is in short supply to
places where there is lots of it: unlike capital, labor can integrate itself into the
production process.
In fact, experiments with administered credit have often failed. Many development
banks have made considerable losses without producing a visible return on investment
or increased output. In Western countries, there is great skepticism about the utility of
subsidized loans and public credit, at least since the 1980s.
Agriculture is the only realm in which direct redistribution of capital has met with
tangible success. For example, since 1960, some development banks (such as the
Grameen Bank in Bangladesh) specializing in loans to poor rural households that were
excluded from the traditional banking system have enabled millions of peasants to
obtain equipment and increase their productivity, inspiring similar experiments
throughout the world. Agrarian reforms aimed at redistributing land or securing the
leases of poor peasant farmers have often led to important gains in productivity, in
Bengal for example (Banerjee and Ghatak, 1995).
These significant increases in productivity show the importance of capital market
imperfections: a perfect credit market would have given credit to peasants to enable
them to become landowners and improve their productivity. The problem is obviously
that these productivity gains could not have been achieved if the peasant’s motivation
had been reduced by having to pay back a loan. Redistribution was the only way to
motivate the peasants to improve their productivity. These successful experiments
should be compared with the disastrous results of the collectivization of farming in the
Soviet Union. It is easy to explain why redistribution of privately owned capital worked
well in agriculture, because in the agricultural sector the problems of allocation and
investment are less difficult to solve than elsewhere. It is enough to give each peasant
enough land to increase his incentive to produce and innovate above what it was when
the land was controlled by a landlord or collective farm.
A Flat Tax on Capital?
In order to redistribute capital efficiently where credit rationing exists, one therefore
needs to find tools as transparent and unbiased as possible so as to avoid the pitfalls of
administered credit. Historically, the imposition of progressive taxes on income and
inheritance has greatly contributed to reducing the concentration of capital (as noted in
Chapter 1). More generally, one could envision the imposition of a general tax on
wealth: on coming of age, each citizen would receive a check to be invested in
whatever way he or she deems most profitable. Of course there would obviously be
costs associated with any such permanent redistribution of wealth, since it would
inevitably discourage future accumulation. But those costs should be compared with the
benefits obtained by financing profitable investments that would not otherwise be made.
The traditional argument that the costs of a capital tax in terms of reduced long-term
accumulation always outweigh the benefits ceases to be true when capital markets are
imperfect (Chamley, 1996). Everything therefore depends on the quantitative importance
of the profitable investments that go unfinanced owing to the capital market failure.
Should wealth be taxed at a rate of 1 percent or 5 percent or 0.1 percent? In order to
answer this question, one would need reliable estimates of the quantity of unfinanced
but profitable investment opportunities, estimates that are very difficult to come by.
Furthermore, between zealous proponents of perfect credit markets, for whom the
unequal distribution of capital raises no problem of efficiency, and radical critics of
capitalism, for whom the problem can be solved only by the abolition of private
property, a climate of civil war has long since reigned, and this has discouraged the
accumulation of knowledge in this nevertheless important area.
To sum up, there is clearly no shortage of justifications for a transparent
redistribution of capital and its income. The case can be made purely in terms of social
justice, in order to achieve a true redistribution of income between capital and labor, as
opposed to the illusory redistribution created by employer social charges. Or it can be
made in terms of economic efficiency, in order to combat the negative effects of capital
market imperfections. In practice, efforts at capital/labor redistribution in the twentieth
century have been disastrous not only in countries that tried to abolish private
ownership of capital, where workers’ living standards stagnated while improving
rapidly in the capitalist countries, but also in the West, where a very small fraction of
the tax burden actually falls on capital. These unsatisfactory outcomes show why the
choice of instruments for achieving redistribution is so important. It is not enough
merely to want to redistribute; adequate instruments are also necessary. These historical
experiences, coupled with the objective difficulty of taxing capital income (which is
often hard to monitor because of the many forms of investment and their high degree of
mobility), suggest that a significant improvement could be achieved with an easily
calculated tax designed to end the dramatic hemorrhage of taxable assets that we
observe in practice. This tax should be applied to all forms of capital income over as
broad a geographical area as possible in order to avoid the negative effects of fiscal
competition between states. The instrument I have in mind is a flat tax (that is, a
universal tax with a single rate). Although such a tax is not well suited to fiscal
redistribution of labor income, which requires greater flexibility (see Chapter 4), it
might well be suited to dealing with capital income in today’s world.

* On this apparent regularity, please see the Note to the Reader at the opening of this book; recent research calls it
into question.
 {THREE}

Inequality of Labor Income

Although it is often thought that capital income is very unequally distributed while labor
income is not, the fact is that the lion’s share of income inequality today (and probably
for a long time in the past) is due to labor income inequality. For instance, it is the
increase in labor income inequality that is responsible for the reversal of the Kuznets
curve that has taken place since the 1970s: in the United States, the gap between the top
and bottom 10 percent of the income distribution has increased by nearly 50 percent. In
order to understand inequality as it presently exists and redistribution as it might exist in
the future, we must therefore give up the idea that labor incomes are relatively equal and
that inequality exists primarily between capital and labor. Instead, we need to analyze
the reasons for labor income inequality. The point of such analysis is to determine what
kinds of redistributive instruments might combat it. The goal is no longer to abolish
private ownership of capital, tax profits, or redistribute wealth. The instruments suitable
for dealing with labor income inequality go by other names: taxation of top incomes and
fiscal transfers to those with lower incomes; policies to improve education and training;
minimum wages; and measures to prevent employment discrimination, strengthen unions,
and establish wage schedules, to name a few. Which of these instruments are the most
justifiable? What arguments are invoked to justify (or reject) them, and how are we to
evaluate such arguments?
 Inequality of Wages and Human Capital
The simplest theory of wage inequality is that different workers contribute different
amounts to a firm’s output. The computer specialist who develops a program for
analyzing a firm’s customer records quickly and efficiently is worth more to the firm
than the office worker who processes a certain number of files each day, and that is why
the firm pays the computer specialist more, lest competitors hire her away. Call this the
theory of human capital. It has long met with hostility. Why? No doubt because a theory
that declares that a computer specialist is worth more than an office worker because she
brings more human capital and therefore greater productivity to the firm is often
suspected of suggesting that this inequality of human capital actually measures an
irremediable inequality between two human beings and may even be used to justify a
considerable difference in standards of living. These suspicions are not totally
illegitimate, moreover, since it was Gary Becker and his colleagues at the University of
Chicago, known for their ultrafree market position, who developed and popularized the
theory (Becker, 1964). Of course, economists do not limit themselves to explaining
wage inequality by individual productivity. They also try to explain the origins of human
capital inequality and propose a theory of education and training that leads to rejection
of any ambitious form of public intervention.
It is nevertheless useful to examine these various questions separately, in order to
distinguish, as discussed in the Introduction, between pure redistribution (in the form of
transfers from high earners to low earners) and efficient redistribution (in the form of
interventions in the process of human capital formation). We begin, therefore, by taking
inequality of human capital as a given. Can observed wage inequality be explained
solely in terms of differences of productivity? What does observed wage inequality tell
us about the most efficient way of remedying the unequal standards of living that result
from unequal wages? We will then turn to the question of human capital formation.
Where does human capital inequality come from, and what instruments would enable us
to change it in an efficient way?

The Explanatory Power of the Theory of Human Capital

In its most rudimentary form (that is, ignoring the question of the origins of inequality),
the theory of human capital simply states that labor is not a homogeneous entity and that
for various reasons different individuals are characterized by different endowments of
human capital, that is, by different capacities to contribute to the production of goods
and services demanded by consumers. Given some distribution of human capital within
the population of workers and the demand for various goods and for the human capital
needed to produce them (the demand for labor), then the laws of supply and demand
determine the wages associated with each level of human capital and thus the
distribution of labor income. The notion of human capital is therefore quite general,
because it includes credentials such as diplomas, experience, and, more generally,
individual characteristics that influence a worker’s ability to participate in the
production process for the demanded variety of goods and services. Can this theory
account for the observed inequality of labor income?
Important Historical Inequalities
At this level of generality, the theory of human capital seems inevitable if one wants to
explain the high degree of wage inequality that we observe across time and space. The
fact that the average wage in the developed countries in 1990 was ten times what it was
in 1870, as noted in Chapter 1, cannot be explained solely by the fact that workers in
1990 were so much more skilled than in 1870 that they could produce ten times more in
the same period. What alternative explanation might exist, since we saw that labor’s
share in value added was the same in 1990 as in 1870, so that the increase in wages
was not a result of a decrease in the share of profits (see Chapter 2)? Over the long run,
it is undeniable that the marked increase in the purchasing power of workers was due to
an increase in the productivity of labor.
Similarly, we saw that if one tries to explain why the purchasing power of workers
in less-developed countries is one-tenth what it is in developed countries, the fact that
the vast majority of workers in the developed countries have completed secondary
education, while 50 percent of those in the less-developed countries are illiterate,
surely plays a key role. Other factors, such as the imperfection of the credit market,
which deprives workers in less-developed countries of sufficient investment capital, as
well as closed borders, which prevent them from benefiting from the physical and
human capital in the developed countries, aggravate this inequality even more. Still, it is
significant inequality in the productivity of labor that inevitably accounts for wage
inequality between the North and South.
Supply and Demand
Human capital theory is also indispensable for explaining the less spectacular but still
striking inequalities that exist within a single country over a shorter period of time. For
example, the average wage of a skilled worker to that of an unskilled worker in the
United Kingdom was 2.4 in 1815, rose to 3.8 in 1851, and thereafter decreased steadily
to 2.5 in 1911 (Williamson, 1985). Why was this ratio 60 percent higher in the middle
of the nineteenth century than at the beginning and end? The most convincing
explanation, confirmed by various sources, is that during the first half of the nineteenth
century, the growing mechanization of industry considerably increased the demand for
skilled labor, while at the same time an increase of agricultural productivity led to a
substantial rural exodus that rapidly increased the supply of unskilled labor. In the
second phase, the influx of unskilled labor from the countryside leveled off while
apprenticeships rapidly increased the supply of skilled labor, so that the gap between
the average wages of the two groups began to decrease. Similarly, although on a smaller
scale, it has been shown that the gap between the average wage of workers with only a
high school diploma or less and that of workers who continued their studies after high
school decreased by roughly 15 percent between 1970 and 1980 in the United States
before rising by more than 25 percent between 1980 and 1990 (Murphy and Welch,
1993a,b, p. 106). The decreased gap in the 1970s is all the more striking in that it
occurred at a time when wage inequalities in general were increasing. Now, it so
happens that the growth rate of the population of workers with post–high school
education attained its historic high in the 1970s as the baby boom generation completed
its education and entered the job market; subsequently, this growth rate diminished
substantially.
These two examples are important, because reversals of wage gaps of this magnitude
are seldom seen. In both cases, changes in the relative supply of and demand for
different levels of human capital provide reasonably satisfactory explanations of the
observed changes.
The Rise of Wage Inequality since 1970
Can supply and demand also account for the general increase of wage inequality (and
the other labor-related inequalities discussed in Chapter 1) in a number of Western
countries since 1970? Many economists have sought to explain rising inequality by
shifts in the supply and demand of human capital over the long run. During the first
phase of the Industrial Revolution, they argue, wage inequality increased as industry
demanded more and more skilled labor and large numbers of unskilled laborers
streamed in from the countryside. From the end of the nineteenth century to the 1970s,
wage inequality decreased in all the developed countries. The phase of decreasing
inequality occurred because skill gaps narrowed considerably thanks to rapid
development of mass education and training, and growing demand for industrial
workers with mid-level skills. Since the beginning of deindustrialization in the late
1960s, however, a new phase has ostensibly begun. New sectors (such as business
services, computers, and communications) require workers with very high skill levels,
but much of the population has been unable to acquire these skills through either the
educational system or personal experience. These relatively unskilled workers find
work in low-productivity sectors (such as personal services, the food service industry,
and retailing) or else find themselves unemployed or underemployed. More extreme
versions of the theory take this argument even further, arguing that wage inequality has
risen not simply because the educational system has not kept up with the demand for
human capital from new sectors and new technologies (as was the case in the first half
of the nineteenth century), but rather because technological progress has placed a
premium on individual qualities that have always been unequally distributed. These
innate inequalities went unnoticed earlier, however, because traditional technologies
made more routine demands and did not require these higher-level talents. This
hypothesis goes by the name “skill-biased technological change” (Juhn et al., 1993).
Skill-Biased Technological Change?
At first glance, this theory of the long-term evolution of wage inequality in the West
seems fairly plausible, at least in its less extreme forms. In the United States, the first
country affected by these changes, one does find an increase in wage inequalities linked
to skill: since 1980, the effects of a year of additional study, a higher-level diploma, or
length of professional experience on average pay have increased noticeably. In the
jargon of labor economists, the “return” to skill has increased (Juhn et al., 1993).
The problem is that roughly 60 percent of the overall increase in wage inequality has
occurred within groups of workers who share the same observable characteristics: the
same level of education, length of professional experience, and age (Juhn et al., 1993, p.
431). Furthermore, this inequality within homogeneous groups of workers has increased
since 1970, which explains why the total inequality of the wage distribution (as
measured, say, by the P90/P10 ratio) has been increasing steadily in the United States
since 1970 (as discussed in Chapter 1), even though the return to education decreased
during the 1970s. Similarly, although it is true that unemployment and underemployment
affected less-skilled workers more than highly skilled ones, employment-related
inequality increased among workers of similar skill levels, including the highly skilled.
The theory of skill-biased technological change also implies that unemployment should
have been greater among less-skilled workers in countries where wage inequality did
not increase much, such as France, than in countries where increasing disparities of
productivity were compensated by wage differences, such as the United States. But
while it is true that the unemployment rate among less-skilled workers is much higher in
France than in the United States, the unemployment rate of more-skilled workers is also
higher by roughly the same proportion (Card et al., 1996).
Of course, one shouldn’t underestimate the extreme inadequacy of wage surveys
when it comes to reporting the individual characteristics on which economists rely as
objective measures of individual skills. The significance of the available indicators
varies so widely between countries that any international comparison based on such
data is extremely hazardous: for example, in 1990, less than 25 percent of the active
French population had a baccalauréat or higher degree, whereas more than 85 percent
of the active US population had a high school diploma or higher (Lefranc, 1997, fig. 1),
so that the group described as “unskilled” in the United States was much narrower than
the comparable group in France. Reality clearly contains many more shades of
difference than our mediocre sources suggest: the uneven quality of American high
schools is well known.
The poor quality of available data is also a problem when it comes to studying the
evolution of wage inequality within a particular country. For example, the data
generally include only the total number of years of study and not the level of the
university or the type of degree a worker holds. But every employer has access to this
type of information when hiring and can differentiate between different levels of
education despite the equivalence in number of years, which is the only information
available to the economist. Furthermore, the type of diploma a person holds may also be
used as a “signal” of qualities other than education-related skills (Spence, 1974). If so,
the fact that we can only observe the number of years of study makes it impossible to
measure what really matters to employers. This is one of the traditional limitations of
using observable individual characteristics to explain wage inequality: a considerable
portion of overall inequality remains unexplained. It is plausible to assume that within
groups sharing the same observable characteristics, actual human capital inequality
increased after 1970. For instance, two different degrees requiring the same number of
years of study might equip the student with very different skills.
Proponents of the theory of skill-biased technological change thus rely on very
flexible definitions of human capital. The risk is that the theory then becomes
completely tautological: it is always possible to “explain” any observed wage
distribution by invoking corresponding productivity differences supposedly stemming
from unobservable individual characteristics. Although this theory surely explains a
significant part of the increase in wage inequality and inequality with respect to
employment, to expect it to explain all the observed data given the current state of
knowledge would be unreasonably optimistic.
Wage Inequality and Globalization
Human capital theory has also been used to explain rising wage inequality resulting
from globalization. The rise of North-South trade has allegedly put less-skilled workers
in the North in competition with cheap labor in the South, consequently reducing their
compensation and increasing wage inequality. Although logically plausible, this
explanation encounters a major obstacle: although imports from the Third World have
significantly increased since 1970, they accounted for only 2–2.5 percent of Western
GDP in 1990, or barely 10 percent of the developed countries’ international trade
(Freeman, 1995, p. 16). How can such a small percentage of all goods and services
produced in the West be responsible for such a broad increase in wage inequality? It is
logically possible, of course, that supply and demand for different skill levels cause
rising inequality in a few sectors affected by international trade to diffuse to the rest of
the economy, but this hypothesis needs to be verified empirically. Furthermore, it has
been shown that in the United States and United Kingdom, the segregation of workers
with different human capital endowments in different firms, as measured by the
correlation of wages among employees of the same firm, increased significantly in all
sectors of the economy and not only in sectors affected by international trade (Kremer
and Maskin, 1996). The same kind of segregation has been observed in France
(Kramarz et al., 1995), suggesting a growing divide between different production units,
some being ultraproductive while others produce far less efficiently. In the current state
of knowledge, these findings seem to indicate that the growth of wage inequality stems
from internal structural changes in the production process within the developed
countries and that similar changes would have occurred if those countries had been
closed economies not trading with the rest of the world.

How to Redistribute Labor Income

Suppose that wage inequality is in fact explained by inequality of human capital. What
would the implications be for redistribution? Assume, further, that it is impossible, at
least in the short run, to influence human capital inequality itself, so that the only thing
that can be done is to redistribute incomes spontaneously determined by the market. This
would be a pure redistribution, justified by considerations of social justice alone:
human capital inequality is at least partly a result of factors beyond the control of
individuals, such as social background, natural talent, and unequal initial endowments.
Under such circumstances, what is the best way to redistribute?
As in the case of capital/labor redistribution (discussed in Chapter 2), the key
question is whether substitution between different types of human capital is possible at
the level of the economy as a whole. If the economy is constrained to use fixed
proportions of different types of labor (n workers per computer specialist, say), so that
the total number of employees of each type is fixed, then direct redistribution, to be
achieved by, say, setting a high minimum wage and a low maximum wage, is completely
equivalent to fiscal redistribution, in which the market is allowed to set wages but high
earners are taxed to finance transfer payments to low earners (or to reduce their taxes).
If there is substantial elasticity of substitution between different types of labor (defined
in the same way we defined elasticity of capital/labor substitution in Chapter 2),
however, fiscal redistribution is strictly superior: it allows increasing the income of
relatively unskilled workers by the same proportion as direct redistribution but without
increasing the cost of low-skilled labor to the firm and thus without decreasing the
number of low-skilled jobs. Once again, the superiority of fiscal redistribution comes
from the fact that, unlike direct redistribution, it severs the connection between the price
paid by the firm and the price received by the worker. This argument is quite general
and does not apply only to redistribution between workers of different skill levels. For
example, a system of family allocations financed by a deduction from workers’
paychecks would make it possible to redistribute wages to workers with children
without increasing their cost to firms, in contrast to a direct distribution asking
employers to pay a higher wage to workers with children than to those without.
Again, empirical studies confirm the existence of such substitutability: demand for
low-skilled labor decreases relative to demand for skilled labor when its relative cost
rises, and vice versa. All available econometric studies show that the elasticities in
question are systematically higher than the elasticity of capital/labor substitution
(Krussel et al., 1996; Hamermesh, 1986) (compare Chapter 2), and these findings are
confirmed by historical work on major structural shifts in employment in various
countries and periods. It is easier to replace low-skilled workers with machinery or
skilled workers than to do without skilled workers.
However, the superiority of fiscal transfers and allocation by price is no more
readily accepted in regard to wage redistribution than in regard to capital/labor
redistribution. The left remains skeptical about reducing employer-paid social charges
on low-wage workers. The idea that the payment of unequal wages—possibly quite
unequal wages—to different categories of workers might play a useful allocative role is
hard to accept, as is the idea that wages should therefore be allowed to adjust freely,
while using taxes and transfers to correct the resulting unjust distribution of income. If
wage inequality is unjust, the skeptics argue, why not simply require firms to pay less
unequal wages? The problem here is the same as the one we discussed earlier in
relation to capital/labor redistribution: in a complex world where many different goods
and services are produced, a high price for skilled relative to less-skilled labor may not
be the worst way to encourage firms and consumers to choose goods and services that
make intensive use of less-skilled labor and less-intensive use of high-skilled labor.
Fiscal redistribution makes it possible to preserve the allocative role of the price
system while redistributing income between workers.
A Major Political Issue
As in the case of capital/labor substitution, the political stakes are considerable: if the
rise of wage inequality since 1970 is due to skill-biased technological change and
growing disparities in individual productivity, then the only way to create jobs is to
ensure that prices paid by firms and consumers for different types of labor vary in the
same proportions. The P90/P10 wage ratio in the United States in the 1990s is 4.5,
whereas it is “only” 3.2 in France (Table 1.7). From this one might deduce that in order
to create as many jobs in France as in the United States, and in particular in order to
stop the labor share of value added from decreasing further in France, one ought to
increase the C90/C10 ratio between the 90th and 10th percentiles of wages plus social
charges in France by about 40 percent, which would mean eliminating all employer-
paid social charges on low wages and reassigning them to high wages. This solution
(using social charges to change the C90/C10 ratio rather than seeking to increase the
P90/P10 ratio of wages actually received by workers) is preferable by far: not only is it
more just, since low-wage workers, who are already disadvantaged, should not be
expected to pay the price of skill-biased technological change; it is also the only
solution that avoids the withdrawal of the least-skilled portion of the work force from
the labor market, as seen in the United States (compare Chapter 1).
In fact, all French governments since 1978 have tried to do just this: social charges,
which were previously capped and therefore less of a burden on high earners than on
low earners, were gradually uncapped in 1978–1979 and 1982–1984 for medical
charges and in 1989–1990 for family charges, and then the charges were reduced for
low-wage workers in 1993. This increased the C90/C10 ratio above the P90/P10 ratio
in 1993: the C90/C10 fell from 3.4 in 1970 to 2.9 in 1983 during the “great
compression” of wages in France, before rising to 3.4 in 1995, while the P90/P10 ratio
remained constant (INSEE, 1996a, p. 51). This put France at the level of the United
States in the early 1970s, before wage inequality began to increase (Table 1.7). It is
therefore tempting to conclude that France still has a long way to go to reach the 4.5
ratio that exists in the United States, and that more must be done to shift social charges
from low to high earners (see Chapter 4).
These comparisons should be viewed cautiously, however. For example, the
P90/P10 wage ratio in the United Kingdom was 3.3 at the beginning of the 1990s (Table
1.7), and the C90/C10 ratio was slightly higher owing to the reduction of social charges
on low earners, and this did not prevent British firms and consumers from choosing
more labor-intensive goods and services, whereas the wage share of output continued to
fall in France. Of course, the United Kingdom remains poorer than France, with a lower
average wage, so it may be benefiting from a certain amount of catch-up as per the
convergence model.

Where Does Human Capital Inequality Come From?

Fiscal redistribution makes it possible to limit the effects of human capital inequality on
standards of living without doing anything about the structural causes of inequality. The
cause of human capital inequality remains a central question. If, moreover, Gary Becker
and his Chicago colleagues are staunchly in favor of the free market, it is not so much
because they believe that human capital inequality is the primary cause of wage
inequality as because of their theory of the origins of human capital inequality itself. For
Becker and his colleagues, the acquisition of human capital is much like any other
investment: if the cost of the investment (the price of education, length of study, and so
on) is lower than the “return” on that investment (the additional income available thanks
to the increment of human capital), then the market will find the necessary funds to
finance the profitable investment, just as the market model predicts for investments in
physical capital with perfect credit markets. Similarly, if the experience and on-the-job
training afforded by a certain job yield a significant increment of human capital, then the
worker who is free to choose will accept a lower wage or even pay the employer to
have the job in order to acquire the additional capital.
If this theory is correct, two consequences follow immediately, and it is worth
distinguishing between them. First, the cost of substantial fiscal redistribution of wages
would be considerable, because it would decrease the return on investments in human
capital and thus decrease the incentive for individuals to make such investments, thereby
reducing the number of high-wage jobs to such a degree that low-wage jobs would also
suffer. In other words, if doctors were not allowed to earn ten times as much as workers
to compensate them for their long years of study, there would be fewer doctors to care
for the workers and pay taxes. The theory of human capital investments thus holds that
the elasticity of the supply of human capital (defined in an analogous way to the
elasticity of the supply of capital) is very high. A second, subsidiary argument is
sometimes made as well: it is not only counterproductive but also unjust to redistribute
wages because different individual investments in human capital reflect different
preferences in regard to length of study, arduousness of work, and so on, and the
government has no business judging these different preferences. In practice, however,
the most common argument is that the elasticity of the supply of human capital is high.
But how true is it that wage redistribution runs up against a problem of incentives?
Unfortunately, empirical estimates are much harder to come by than theoretical treatises
in this area. Nevertheless, in the current state of our knowledge, it is fair to say that the
elasticity effects are much smaller than the Chicago theorists imagine (see Chapter 4).
Efficient Inequality?
The second consequence of the theory of efficient investment in human capital is that it
is pointless for the state to intervene to reduce human capital inequality. Because the
free play of market forces and private initiative is supposed to ensure that all profitable
investments in education and training have already been made, any intervention in the
education or labor markets can only do harm. In other words, the theory implies not only
that pure redistribution creates a problem of incentives and must therefore be limited in
scope but also that no efficient redistribution is possible, because the market allocation
of resources is already optimally efficient (in Pareto’s sense, as discussed in the
Introduction).
These recommendations may at first sight seem surprising to anyone accustomed to
reasoning in terms of the ways in which inequality of educational opportunity is
reproduced from generation to generation. One justification for government intervention
in the educational arena is simply that the young students at whom interventionist
policies are directed are generally incapable of judging the profitability of any
particular educational investment, and their parents are not necessarily much better
equipped to do so. Although economists hesitate to invoke such “paternalistic”
arguments, it is hard to deny that if the children of India were to follow the advice of the
Chicago theorists and wait until market forces and parental initiative conspired to teach
them to read, India would very likely remain mired in misery for a long time to come.
Compulsory elementary education is no doubt the most important of all instruments for
efficient redistribution, and research on growth and convergence suggests that the
considerable improvements in standards of living achieved in the West since the
nineteenth century would not have been possible without such schooling.
Another argument against the theory of efficient human capital investments is of
course the imperfection of the credit market, on which I touched earlier. This may
prevent individuals from poorer families from undertaking long courses of study even if
they have the ability to make the investment profitable. The market failure is
compounded by the fact that investments in human capital are long term, but since it is
difficult to make a credible commitment to repay an educational loan, loans are granted
more readily to students whose parents can offer guarantees. On these grounds one can
justify generous financial support to students from modest backgrounds as a remedy for
the inefficiently inegalitarian distribution of human capital endowments.
There are, however, no reliable empirical estimates of the quantitative importance of
this type of credit market failure, and the paternalistic argument does not apply equally
to education at all levels. To be sure, educational achievement varies strongly with
social origin, and children from modest backgrounds generally pursue their studies for a
shorter time than those from wealthier backgrounds with similar test scores at age ten. It
is tempting to conclude that these students would have continued their studies longer in
the absence of imperfect credit markets. But some sociologists suggest that the finding
can also be explained by the hypothesis that students from modest backgrounds are less
motivated to pursue lengthy courses of study because they are not expected to maintain
the same family standing (Boudon, 1973).
This is the sociological version of the “different preferences” argument. It implies
that it is misleading to think that inequality of opportunity could be reduced substantially
by increasing public investment in the education of students from modest backgrounds.
In fact, the influence of social origin on professional success goes well beyond the
problems of credit market failure and access to education: controlling for the level of
educational attainment, the effects of social origin are statistically significant at every
career stage (Goux and Maurin, 1996). More generally, the fact that level of education
explains only a part of overall wage inequality is often invoked to temper the
enthusiasm of those who think they can end inequality by enacting ambitious educational
reforms (Boudon, 1973). If paying for studies were the key explanatory factor,
moreover, one would expect to find a higher level of intergenerational reproduction of
human capital in countries such as the United States, where private financing of
education plays an essential role, than in Europe, where public financing is
predominant. In fact, it seems that intergenerational mobility with respect to level of
education does not vary much at all across time and space (Shavit and Blossfeld, 1993),
nor does intergenerational mobility with respect to level of income (Erikson and
Goldthorpe, 1992).
The Role of the Family and Educational Expenses
Broadly speaking, skeptics of state intervention in education do not deny the importance
of the family in reproducing inequality of human capital but rather seek to show that
inequality inevitably persists because of the family’s central role. Becker’s theories of
the family, as presented in his own works and those of his students (Becker, 1991;
Mulligan, 1996), emphasize the choices that families make when investing in their
children in order to show that any state intervention would risk undermining them. This
is an old intellectual tradition in Chicago. In 1966, the sociologist James Coleman
produced a famous report for the US government on the education of disadvantaged
minorities. The report caused a scandal because it argued that the redistribution of
financial resources to schools in disadvantaged neighborhoods had not yielded any
noticeable improvement in educational outcomes or success in the job market. The
Coleman Report and other work in the same vein concluded that simply spending more
money on education in disadvantaged communities will not improve outcomes because
it is within the family unit and the immediate social environment that inequality
inevitably originates.
Of course, everyone agrees that the factors influencing the transmission of inequality
are far more “environmental” than genetic. Or almost everyone: the psychologist
Richard Herrnstein and the sociologist Charles Murray made front-page news in 1994
when they published The Bell Curve, which many critics accused of defending the idea
that intelligence is to a large extent genetically determined. In fact, Herrnstein and
Murray also recognized that adoption studies showed that children from disadvantaged
sociocultural backgrounds placed at birth in more highly educated families were just as
successful as the biological offspring of those families (Herrnstein and Murray, 1994,
pp. 410–413). But this is not really the central issue. If the key factors have to do with
the family environment, and in particular with the family environment in early childhood
(books in the home, conversations with parents, etc.), so that nothing can really alter
inequality established within the home at this stage of life, then the consequences are not
very different from those of genetically determined inequality. Thirty years after
Coleman, Herrnstein and Murray also stressed the idea that because it is difficult to
detect the effects of investing educational resources in disadvantaged communities, it is
pointless to persist with such policies.
If this theory were valid, there would be no reason to try to alter the unequal
distribution of human capital. It would be better to spend the available funds to reduce
unequal living standards by means of fiscal transfers, within the (potentially strict)
limits defined by the elasticity of the supply of human capital.
The Problem of Inefficient Segregation of Human Capital
These claims stimulated a great deal of debate, especially in the United States after the
Coleman Report was published. More recent work, using better indicators of the effects
of additional spending on education in disadvantaged neighborhoods, shows that the
earlier claims were greatly overstated (Card and Krueger, 1992). Moreover, Coleman’s
findings can be interpreted in different ways. In fact, it is plausible to think that the
effects of educational spending are small not because the family environment is the sole
determinant of scholastic success but because the effects of the school and neighborhood
social environment are greater than those of educational spending per se.
In other words, it is plausible to think that a student’s chance of scholastic success
depends more on the “quality” of his classmates than on that of his teacher, especially at
the primary and secondary level. It is unlikely that scholastic success can be improved
significantly by sending a highly trained teacher into a difficult inner-city neighborhood,
but sending students from that same neighborhood to a highly rated high school or prep
school is much more likely to increase their probability of success. This hypothesis has
been confirmed with the help of a very rich set of intergenerational data from the United
States, the Panel Study of Income Dynamics (PSID), which shows that for a given level
of parental education and income, the likelihood of upward social mobility for the
children varies over a range of two to one depending on the average income of the
neighborhood in which the parents live. These results show that “local externalities,”
which economists have long measured at the microeconomic level of the classroom, can
have a substantial effect on the global dynamics of inequality—of the same order of
magnitude as the effect of parental characteristics themselves (Cooper et al., 1994).
Thus, negative results such as Coleman’s, rather than bolstering the opposition to
state intervention to redistribute financial resources to disadvantaged neighborhoods,
suggest a need for more radical redistributive methods, such as redrawing school zones
to require parents from different groups to send their children to the same schools (given
the impossibility of requiring them to live in the same neighborhoods). Such policies
exist in many countries, but generally on a very small scale: parental choice is often
limited in order to avoid too much social imbalance in the classroom, but geographical
limitations on school rezoning significantly curtail the attainable level of social mixing.
More radical policies such as busing were tried briefly in a number of US cities in the
1960s and 1970s, generally in order to achieve better racial mixing. School busing
marked the culmination of the civil rights era in the United States, but the policy met
with significant parental hostility. This should not have come as a surprise in the United
States, where parents have long been accustomed to local control of schools, including
curricula and teacher hiring.
Nevertheless, parental decisions about where to send children to school have major
consequences on other children, and the anonymity of the price system (in this case, the
price of housing) prevents parents from taking into account the externalities that their
choices imply for others. Thus, even where social integration would have benefited
disadvantaged youngsters far more than it would have cost advantaged ones, researchers
have shown how individual housing choices nevertheless resulted in segregation
(Benabou, 1993). In theory, it is therefore possible that everyone would benefit from
social integration (for example, if the costs of integration for advantaged groups are less
than the tax savings resulting from the scholastic and professional success of the
disadvantaged), but such a social equilibrium cannot be achieved without collective
coercion. Simple rules (such as requiring that average parental income be equal in all
schools in a given community) might therefore be in everyone’s best interest in the long
run.
Discrimination in the Labor Market
Another socioeconomic mechanism leading to inefficient human capital inequality is
labor market discrimination. Initially developed by Phelps (1968) and Arrow (1973) to
explain discrimination against African Americans in the United States, the theory of
labor market discrimination can also be applied to any other group in which employers
are able to observe distinguishing characteristics of group members: women, lower
castes in India, the long-term unemployed, or, more generally, any group whose
members are subject to negative prejudices. The basic idea is simple. Suppose that
employers predict that certain social groups are objectively less likely than others to
contain members with sufficient human capital to qualify for certain jobs. Suppose,
further, that employers cannot observe the qualifications and motivations of each
candidate with perfect accuracy at the time of hiring, so that hiring decisions are made
on the basis of imperfect signals, such as the results of a test, an interview, or a CV.
Since the employers expect that certain groups are less likely to possess the required
human capital, they will hire members of those groups only if their test results, say, are
exceptionally good. In other words, the bar is set higher than for other groups. How will
the groups subject to such discrimination react to this practice on the part of employers?
Since the probability of being hired for a skilled job is low, members of such groups
will be less likely on average to make the necessary investments in human capital and
will do so only if they believe they are likely to perform exceptionally well on the
hiring test or interview. For example, only those with high confidence in their abilities
will undertake long courses of study, prepare intensively for job interviews, and so on.
In other words, their behavior will tend to validate the expectations of the employers,
namely, that the average level of human capital in the group subject to discrimination is
in fact lower than in other groups. It can then be shown that even if the two groups
(people of color and whites, say) are equally capable of acquiring the necessary human
capital, and a fortiori if one of the groups is initially slightly less capable because its
members come from more modest social backgrounds, the employers’ expectations will
influence the behavior of group members in a perverse way to produce a deep and
persistent inequality of human capital and professional achievement (Coate and Loury,
1993).
This human capital inequality is totally inefficient because it is based entirely on a
self-fulfilling prophecy. Economic efficiency requires that groups with identical
capabilities make identical investments in human capital. Inequality due to
discrimination is therefore profoundly perverse. This economic theory of discrimination
is similar in some ways to sociological theories that hold that inequality is often the
result of a dominant discourse that becomes a self-fulfilling prophecy: if it is widely
believed that the members of certain groups are unlikely to succeed, they will be
discouraged from trying (Bourdieu and Passeron, 1964; 1970).
Affirmative Action versus Fiscal Transfers
The political implications of these theories are important. If a significant part of
inequality is in fact due to perverse mechanisms of the sort described, then new
redistributive instruments are needed. For example, the theory of discrimination
suggests that employers should be prohibited by law from discriminating against
minorities. One way to do this is to require employers to show that each hiring and
promotion decision is based on unbiased objective criteria. Another is to impose
affirmative action quotas, requiring employers to hire a certain percentage of minority
workers, in order to break the vicious circle of self-fulfilling prophecies of failure.
Such affirmative action policies became popular in the United States in the 1970s to
protect African Americans, women, and other minorities. Affirmative action, which in
some ways resembles earlier efforts to use labor law to limit employer discretion in
hiring and promotion, is very different from the kinds of policies recommended by
human capital theorists, who say that the best remedy for inequality is to make fiscal
transfers to social groups whose human capital endowments are too low (within the
limits imposed by the elasticity of the supply of human capital), while of course
avoiding any interference in the process of production. Herrnstein and Murray (1994)
challenge the very idea of discrimination and argue that racial inequality persists
because low IQ and low levels of human capital are transmitted from generation to
generation within African-American families.
Can this controversy be settled with data? In the relatively well-documented case of
African Americans in the United States, the observed facts appear to support the theory
of discrimination. Freeman (1973) shows that the only way to explain the reduction of
the wage gap between African Americans and whites after the civil rights movement of
the 1960s is to invoke the progressive erosion of negative prejudices against African
Americans and the discouragement that resulted from such prejudices (see also Bound
and Freeman, 1989). But the best example is surely the impressive improvement in the
situation of women in the labor market since 1950, which can only be explained by
appeal to a theory insisting on the importance of discrimination, prejudice, and
discourse in the production of inequality. In all Western countries, the participation of
women in the labor market rose from a mere 10–20 percent in 1950 to more than 50
percent in the 1980s (OECD, 1985). Progress continued in the 1980s and 1990s: while
wages increased across the board, the average woman’s wage in the United States rose
more than 20 percent relative to the average man’s (Blau and Kahn, 1994). The same
was true in most other developed countries (OECD, 1993, pp. 176–178). No fiscal
transfer could have brought about such a spectacular improvement in the economic
situation of women.
Furthermore, this improvement also occurred in countries with “Mediterranean”
(pro-natalist) tax systems (such as the family quotient in France), which discourage
women’s participation in the labor force, compared with the United States, United
Kingdom, and Scandinavian countries, where individuals are taxed rather than
households.* In short, inequalities based on rank discrimination, such as between
people of color and whites or men and women, are much more susceptible to remedy by
affirmative action and changes in mentality than by any kind of fiscal redistribution.
Unfortunately, the fact that an inequality is based on discrimination does not always
mean that it is easy to eliminate or even reduce. For example, most observers agree that
the results of affirmative action in the United States have been mixed at best. Indeed,
quotas requiring employers to hire a certain percentage of people of color can reinforce
rather than weaken prejudices against African Americans, “who become employable
only when we are forced to employ them,” while at the same time reducing their
incentive to compete for jobs like other citizens, which is precisely the opposite of the
intended goal (Coate and Loury, 1993). Many observers therefore denounce the quota
system. The apparent ineffectiveness of affirmative action contributed greatly to the
conservative reaction against social programs in general in the 1980s and 1990s. In fact,
it is likely that the deterioration of the relative position of African Americans in the
labor market since the 1970s, which fueled this reaction, is more simply explained as a
by-product of the general increase in wage inequality and of deindustrialization, which
hit African American workers hard, especially in the northern United States (Wilson,
1987).
 The Social Determination of Wage Inequality
Some wage inequalities cannot be explained solely in terms of an underlying inequality
of human capital (whether efficient or inefficient). For example, certain economic actors
(such as firms or trade unions) may attempt to manipulate the wage structure resulting
from the supply and demand of human capital to their advantage. Other considerations,
such as the need to motivate workers, may lead employers to look to factors other than
human capital. Even in the absence of overt manipulation of market prices for labor, this
can distort the wage structure in ways not predicted by the theory of human capital. Are
these deviations from competitive prices a good thing or a bad thing? How does the
existence of such deviations affect the question of redistributing labor income?

The Role of Unions in Setting Wages

What do unions do? The traditional economic analysis is simple: unions exercise
monopoly power in the wage-setting process. The legal rights they are granted entitle
them to represent the interests of large numbers of workers, and no worker can offer to
work for a lower wage than that to which the union agrees. Just as a firm with monopoly
power will choose to raise prices even if it means losing a few customers, a union will
use its monopoly power to demand wages higher than would otherwise prevail, even if
it means lowering the overall level of employment. This analysis neglects the fact that
unions generally fight not just for higher wages but also for a certain compression of the
wage hierarchy within each firm. They achieve this second goal by insisting on wage
schedules that limit the difference between the wages paid for different levels of
experience and skill (Freeman and Medoff, 1984).
In any case, the tools that unions use to increase total labor income and decrease
inequality between workers are not tools of efficient redistribution. As noted, wherever
the possibility of capital/labor substitution and/or substitution of one type of labor for
another exists at the level of the economy as a whole, any redistribution that involves
manipulation of the price of labor and/or human capital is inefficient (Chapters 2 and 3).
If labor unions are successful, firms will inevitably use more capital and less labor as
well as more skilled labor and less unskilled labor. The fundamental fact is that it is
always possible to finance the same redistribution more efficiently by way of fiscal
redistribution, that is, by taxing high earners to pay for fiscal transfers to low earners,
because this is the only way of severing the tie between the price paid by the firm for
labor and the amount received by the worker. Hence the question is not how much
redistribution from capital to labor—and redistribution between workers—should exist,
because the answer to this question depends on other factors. It is rather what
instruments should be used for redistribution. Should the wage-setting power of unions
therefore be reduced?
Unions as Substitutes for Fiscal Redistribution?
The first response is that reducing the power of unions can make redistribution more
efficient only if it actually results in replacing the inefficient redistribution achieved by
the unions with efficient fiscal redistribution via the state. In practice, the problem is
obviously that there is no agreement about the proper extent of redistribution. Suppose
that the current government decides that it is just for a low-skilled employee to live on
€760 a month while a highly skilled manager is paid €4,575 a month. The government
might reach such a decision because it believes that this is the only way to preserve the
incentives needed to encourage the manager to acquire the necessary skills. If a union
disagrees because it thinks that the employee should earn €1,525 a month and the
manager only €3,810, then the only way to proceed is to attempt to forcibly impose a
new wage schedule on employers. Of course, it would be better to increase the
manager’s tax by €760 a month and use the proceeds to make a fiscal transfer of €760 to
the low-skilled worker. That way, the firm would not have to pay more to its workers
and less to its managers, which would inevitably lead to hiring fewer workers and more
managers and thus to an increase in unemployment. But unions do not have the power to
levy taxes and make transfers. Historically, the role of unions has been to intervene in
conflicts of this type: when the state fails to play the redistributive role that the unions
believe it should play, they step in and use the resources at their disposal: direct
redistribution through struggle in the workplace.
These instruments are actually quite limited compared with fiscal redistribution.
What is more, the results they achieve are often illusory. Once again we encounter the
conflict between historical time and political time discussed in Chapter 2. As was the
case with redistribution between capital and labor, fiscal redistribution has never
substantially and visibly reduced labor income inequality between workers, although it
has long been technically possible to do so. Historically, major fiscal redistributions
have been rare and have generally taken the form of social expenditures rather than
monetary transfers between workers (see Chapter 4). Even more important, they have
always been put in place very gradually, so that their effects were felt only in the long
run. Such reforms cannot fuel the political imagination on which social and political
struggle thrive. Indeed, the long run may be so long that it makes no sense from the
standpoint of any given generation.
By comparison, the inefficient redistributions achieved by direct manipulation of
wages are much more visible. For example, the purchasing power of the minimum wage
in France rose by about 92 percent between 1968 and 1983, a period in which trade
unions played an essential role; during the same period, the average wage increased
only 53 percent, which reduced the P90/P10 ratio from 4.2 in 1967 to 3.1 in 1983
(compare Chapter 1 and INSEE, 1996a, pp. 44, 48). By the same token, there is no
denying that the two Western countries in which wage inequalities have increased most
since the 1970s, namely, the United States and the United Kingdom, are also the two
countries in which the power of unions has decreased most, in significant part due to
political opposition.
Meanwhile, wage inequalities among employed workers have remained relatively
stable in countries such as Germany and France, where the union coverage rate (that is,
the percentage of workers covered by collective bargaining agreements) has remained
relatively stable, even if the unionization rate (the percentage of workers belonging to
unions) has decreased. This is a major reason for the contrasting evolution of wage
inequalities in the West since the 1970s: it explains 20 to 40 percent of the observed
variance (Card, 1992; Lemieux, 1993). This is totally overlooked by pure human capital
theory and the theory of skill-biased technological change. It is possible that this union-
driven redistribution has not been cost-free in terms of jobs created. But the fact
remains that the United States and United Kingdom have not replaced inefficient union-
driven redistribution by more efficient fiscal distribution; indeed, they have tended to
reduce the latter as well. Under such conditions, unions can play a role as substitutes for
fiscal redistribution.
Do Unions Contribute to Economic Efficiency?
The second response to the question of curtailing union power is that trade unions have
at times served to promote economic efficiency. By representing workers, they facilitate
better communication within the firm. Furthermore, the binding wage schedules
negotiated by unions can under certain conditions be positive forces in themselves. For
instance, human capital theory neglects the fact that a worker’s skills and work habits
are not necessarily qualities universally recognized by the market, which the worker can
sell freely to whatever firm he chooses.
The value of specific forms of human capital is often unique to a particular firm, so
that in practice the human capital market can never be fully competitive. Once a worker
has made the necessary effort and investments to qualify for a specific job, the firm that
hires him may pay a wage well below what his skills are worth, because the worker
cannot make full use of his firm-specific skills in another firm. Anticipating this
expropriation of his investment in human capital, the worker will not invest as fully as if
he were certain of reaping the benefits for himself. Setting a floor wage below which
the firm may not go can therefore resolve this problem and increase economic efficiency
by ensuring that efficient investment opportunities are not neglected. More generally,
setting in advance the wage or wage range that a firm must pay to a worker with specific
skills in a specific job category can provide an incentive for potential workers to
acquire more firm-specific forms of human capital without fear of expropriation by the
employer.
This type of phenomenon is not limited to the case of firm-specific human capital. A
binding wage schedule is, in effect, a commitment not to expropriate, and this can also
allow the firm itself to invest in its workers and reap the benefits. For example, foreign
observers have long been astonished to discover that German firms finance costly
training and apprenticeship centers. Apprentices are generally not required to pay for
the training or commit themselves to working for the firm even though much of the
training they receive is of a general nature and could also be used in other firms. The
most convincing explanation of this is that all firms in a given sector of industry agree to
pay the same entry wage and standardized pay increments, thus ensuring that apprentices
won’t be lured away by competing firms after they are trained (Harhoff and Kane,
1994).
Characteristic features of the wage relation (such as firm-specific human capital and
limited possibility of commitment) may therefore make efficient operation of the labor
market dependent on certain types of collective regulation (such as binding wage
schedules [Piketty, 1994, pp. 788–791]) and government interventions to correct market
failures related to occupational training (Booth and Snower, 1996). In theory, then, the
continued existence of binding wage schedules in some countries is not a costly and
inefficient way of limiting the increase of wage inequality but potentially a means of
encouraging new investments in human capital and thus limiting human capital inequality
in the future. Nevertheless, in the absence of additional empirical evidence, these
arguments obviously cannot be used as systematic justification for centralized wage
schedules. For example, there is no convincing evidence that the rigid wage schedules
that have enabled certain Western countries to avoid rising wage inequality since the
1970s actually encouraged investments that would lead to more jobs and higher wages
in the future.

The Monopsony Power of Employers

Although unions are often said to exercise monopoly power that enables them to raise
wages about the competitive market level, economists are less likely to admit that
employers sometimes exercise similar market power. Among noneconomists, however,
the idea that employers exercise arbitrary power, to which workers respond by
organizing unions, is commonplace. In the jargon of economics, the equivalent of the
monopoly power of unions is the monopsony power of employers. A monopsony exists
when there is only a single possible buyer for a given good, whereas a monopoly exists
when there is only a single possible vendor. A monopolist can demand a price for his
good above the competitive market price (at the risk of inducing customers to buy less
of it), while a monopsonist can insist on paying less than the competitive market price
for the good he wishes to buy (at the risk of inducing suppliers to sell less of it). The
manipulation of market prices thus always decreases the quantity of goods exchanged,
regardless of whether it is to the benefit of the buyer or seller. In the case of the labor
market, a monopsonist employer will oblige his workers to accept a below-market
wage, at the risk of discouraging some workers and thus diminishing the level of
employment.
If monopsony exists, the redistributive implications are significant. For one thing, it
would be inefficient to try to improve the workers’ lot by means of fiscal transfers,
because the employer would take advantage of the transfer to lower his wage offer.
Efficient redistribution would then require an increase in the legal minimum wage in
order to bring the wage paid by the firm closer to the competitive wage. This would
also stimulate the supply of labor and thus increase the total employment level. In
contrast to the usual conclusion, direct redistribution would then be superior to fiscal
redistribution, because it would restore the competitive market equilibrium (after which
fiscal transfers would be a more efficient means of further redistribution). For workers
this would be the best of all possible worlds, since it would be possible both to
improve their standard of living and reduce unemployment without spending a cent of
tax revenue.
What might give employers monopsony power? Monopsony power can arise
because of the existence of firm-specific human capital, which implies that workers are
to some extent able to sell their labor to only one employer. More generally, limited
geographical mobility or lack of information about other jobs may leave some workers
at the mercy of a single employer. More simply, monopsony power can arise if a group
of employers band together to impose a uniform wage schedule on their employees. Do
such coalitions of capitalists actually exist? It is difficult to show empirically. In
particular, it does not appear to be possible to explain increasing wage inequality in the
United States since 1970 in this way. Indeed, the most striking thing about this
phenomenon is that it occurred in an extremely competitive labor market. The
compensation of lawyers, doctors, and managers has exploded since 1970, not because
capitalists collectively decided to divide the labor market but because firms and
individuals competed for their services, constantly hiring desired individuals away
from their competitors by offering ever-increasing salaries. It does not follow that the
ensuing inequality has to be accepted or even that the phenomenon itself is totally
efficient. Nevertheless, the employer monopsony model cannot account for the observed
facts.
When Does a Higher Minimum Wage Increase the Level of Employment?
Although monopsony cannot explain the global evolution of wage inequality, it is still
possible for local monopsonies to exist, in particular in certain markets for unskilled
labor with limited geographic mobility. In the 1990s, several studies of US labor
markets, most notably Card and Krueger (1995), revived this debate. Using the fact that
the legal minimum wage was raised at different dates by different amounts in various
states in the 1980s and 1990s, economists convincingly demonstrated that the effect of
an increase in the minimum wage on the employment level was generally positive and in
any case fairly small. Note, in particular, a famous study of fast-food restaurants in New
Jersey, where total employment rose after an increase in the state’s minimum wage in
1992 (Card and Krueger, 1995, chap. 2). Larry Katz, the author of another of these
studies, served as chief economist in the Department of Labor during the first Clinton
administration, and this body of economic research surely influenced President
Clinton’s decision in 1996 to increase the federal minimum wage from $4.15 an hour to
$5.20, or more than 20 percent, compared with a decline of more than 25 percent in the
purchasing power of the federal minimum wage between 1980 and 1990.
The precise reasons why these minimum wage hikes had positive effects remain
controversial, however. Was it the limited geographic mobility of low-wage workers
that left them subject to a low-wage regime imposed by a local cartel of fast-food
restaurants, so that the increase in the minimum wage did not decrease demand for their
labor but instead increased supply by encouraging additional low-skilled youths to offer
their services—thus exemplifying the theory of monopsony in its purest form? Or was it,
as other studies suggest, that the increase in employment occurred because the higher
minimum wage induced some young people to quit high school and replace less
qualified workers (Neumark and Wascher, 1994)?
In any case, the fact remains that when the legal minimum wage falls as low as it did
in the United States in the late 1980s and early 1990s, low-skilled jobs can become so
unattractive that an increase in the minimum wage can increase the labor supply and the
level of employment. More generally, the potential for local labor market (or firm-
specific human capital) monopsonies is justification enough for a legal minimum wage
in order to make sure that no employer can exploit a monopsony situation beyond a
certain limit.

Efficiency Wages and Fair Wages

If there were no union monopolies, no employer monopsonies, no minimum wage, and
no visible market failures, would the wages paid by firms for different types of labor be
determined solely by supply and demand, as the theory of human capital suggests? The
question might seem absurd, since every labor market we know has unions demanding
the highest wage they think they can obtain for their members, employers seeking to pay
workers as little as possible for their labor, and governments trying to arbitrate between
the two and to achieve some measure of redistribution. A more useful question to ask is
the following: Is the ability of labor unions to distort market prices a result of the legal
rights they are granted (the right to strike, the closed shop, and so on), or would
distorted prices continue to exist (to some extent, at least) without such rights?
Why would employers in a competitive labor market that prevents them from
offering wages lower than those offered by their competitors choose to pay even higher
wages? Because by increasing wages they obtain something of equal value from their
workers. For example, suppose it is impossible for the employers to monitor the
diligence of their workers. A higher wage might be intended to motivate workers to
work harder, because the workers know they might lose something if sacked and forced
to seek work with another employer, who doesn’t pay such a high wage. Indeed, in
sectors where it is difficult to monitor workers, we do observe wage differences that
cannot be explained by differences in human capital (Krueger and Summers, 1988). This
theory, which has often been invoked to explain unemployment in Europe in the period
1980–1990 (see, for example, Phelps, 1994), also implies that if all firms paid higher
wages to motivate their workers, employment would decrease, so that what motivates
workers is really the risk of a period of unemployment between jobs. Another version
of this “efficiency wage” model assumes that in addition to potential loss of wages or
risk of unemployment, workers may be more cooperative if they believe they are being
paid a fair wage. It will then be in the firm’s interest to pay something close to the
perceived fair wage, even if it means decreasing overall employment (Akerlof and
Yellen, 1990). Individual judgments of fairness are frequently quite important in the
wage-setting process (Kahneman et al., 1986; Bewley, 1994). Unemployment can then
be analyzed as the result of a distributional conflict, even in the absence of unions. The
implications of these models for redistribution are obvious: fiscal redistribution should
try to match what is perceived as fair in order to reduce the inefficiency of direct
redistribution by reducing the charges on low earners and shifting them to high earners
and/or corporate profits.
National Traditions and Wage Inequality
More generally, the theory of human capital, even when supplemented by manipulation
of prices by unions or employers, is based on the idea that one can always measure the
contribution of each type of skill to the production process, thus providing an objective,
measurable basis for the return to human capital. As the discussions of discrimination
and firm-specific human capital showed, however, reality is often more complicated.
Measuring the productivity of different types of human capital is not always easy, and
the variance in such measurements often reflects national differences.
For example, as Rothenberg (1996) has shown, wage inequality increases when
workers believe that employers are highly likely to evaluate their productivity correctly,
because those who score low accept their fate, while those who score high will threaten
to quit as a tactic to pressure their employer for higher pay. Such a totally decentralized
process may explain why the United States and United Kingdom, where “faith in
capitalism” has strengthened since the 1970s, are also the countries where wage
inequalities have increased most. In an apparent confirmation of this theory, there has
been a veritable explosion in the compensation of top managers (Goolsbee, 1997;
Feenberg and Poterba, 2000). But it is hard to believe that the actual productivity of
these managers has suddenly increased by a corresponding amount.
Similarly, it is difficult to explain why France, with a P90/P10 ratio of 4.2 in 1967,
was the most inegalitarian country in the Western world (in terms of wages) in the late
1960s and early 1970s without mentioning certain distinctive French attitudes toward
inequality. It is probably not the case that human capital inequality was actually higher
in France at that time than it was elsewhere. In 1976, the French government was
outraged when an OECD report indicated that it was the most inegalitarian country in
the West. In fact, France suffers from an extreme form of “republican elitism”: the
French tend to overestimate the actual difference in productivity between a top manager
who has graduated from an elite school and an ordinary worker. Any difference in pay
is justified by the fact that both had access to l’école républicaine, the supposedly
egalitarian public educational system—perceived as egalitarian despite the fact that the
state spends ten times as much on a graduate of the elite École Polytechnique as on an
average student. The French belief in educational meritocracy is also reflected in the
relative stability of pay differences after graduation, compared with a much higher
degree of variability in Germany (Morrisson, 1996, p. 111). Although the German
system is less inegalitarian, it is probably just as good at offering incentives.
To be sure, the observed variation between countries is quite small compared with
historical differences of inequality, but it is more striking to contemporary observers.
Although some of this variation can be explained by institutional differences between
countries (for example, the German apprenticeship system versus lavish spending on
elite education in France), it is often exaggerated by specific perceptions within each
society. In part this is due to different national histories, which human capital theory
cannot explain and fiscal redistribution can only superficially influence.

* The family quotient results in a tax abatement for large families.—Trans.

 {FOUR}

Instruments of Redistribution

The two previous chapters have tried to show how important it is to understand the
socioeconomic causes of inequality in order to know what instruments of redistribution
are most appropriate. In this chapter I will continue to analyze the most important of
these tools in the light of contemporary experience. As I have done throughout, I will
continue to distinguish between pure and efficient redistribution.
 Pure Redistribution
The primary tool for pure redistribution is fiscal redistribution, which makes it possible
to correct inequality due to unequal initial endowments and market forces while
preserving as much as possible of the allocative role of the price system. I will focus
here on fiscal redistribution of labor income. In Chapter 2, I analyzed some of the
specific problems associated with the redistribution of capital income, which is less
important than labor income.

Average and Marginal Rates of Redistribution

To what extent do governments rely on fiscal redistribution today? In practice, modern
fiscal redistribution depends on a variety of taxes (such as income tax, value-added tax,
social charges), transfers (such as family allowances, unemployment insurance,
guaranteed minimum income, and pensions), and expenses paid directly by the
government (for health, education, and so on). An index often used to measure the extent
of taxes and transfers in a given country is the total of all taxes as a percentage of GDP.
For example, obligatory taxes amount to 30–35 percent of GDP in the United States and
United Kingdom, 45–50 percent in Germany and France, and 60–70 percent in the
Scandinavian countries. This is a poor measure, however, because it tells us nothing
about how taxes, transfers, and expenditures are distributed. Furthermore, different
accounting conventions in different countries make it difficult to compare these numbers.
In Sweden and other Nordic countries, for example, pensions and other state-paid
income is taxed like other income, and this increases the tax burden by nearly 10 percent
of GDP in a totally artificial way. In France, this would be tantamount to financing a
pension increase by increasing social charges on retirees, which obviously has no effect
on actual redistribution.
The only way to measure fiscal redistribution correctly is to use the effective
average and marginal rates of taxation and transfer. The effective average rate for a
given level of income is defined as the sum of all taxes (or transfers) expressed as a
percentage of gross income before taxes and transfers. The rate can be positive or
negative, depending on whether the total tax paid is greater or smaller than the total
transfers received. The effective marginal rate of tax and transfer between two income
levels is defined as the sum of all taxes and transfers that an individual pays or receives
if she transitions from one income level to another, expressed as a percentage of the
difference in gross income. Effective marginal rates are generally positive, since an
increase in gross income generally leads to an increase in taxes paid, but in principle it
could also be negative if a higher income entitled a person to a higher transfer (or lower
net tax).
 These average and marginal rates are effective only to the extent that the numerator
includes all taxes and transfers. In particular, it is essential to include all social charges,
including those paid by employers. As we saw earlier, employer-paid social charges do
not result in redistribution from capital to labor and in the end are always paid by labor
(Chapter 2). Ideally, nonmonetary transfers achieved through government spending
should also be included (I will return to this point). Figure 4.1 shows the average and
effective marginal rates for different deciles of the wage distribution in France in 1996,
taking account of income taxes, social charges, and social benefits (guaranteed minimum
income, housing allowances, and so on) for single individuals (Piketty, 1997a).
The Absence of Redistribution between Workers
Figure 4.1 shows that effective average rates are positive for all wage levels: other than
a very small housing allowance for those earning close to the minimum wage, a worker
with no children receives no direct monetary transfer. The effective average rate is also
positive for low-wage workers with children, because the family allowance received is
always very much smaller than the tax paid, unless the number of children is very high.
Last but not least, Figure 4.1 shows that effective rates do not vary much from one wage
level to another: they are on the order of 45 percent for the bottom two deciles of the
wage distribution and on the order of 55 percent for the top two deciles, with the
majority of average rates around 50 percent. The reason for this is simple: the
proportionality of social charges largely outweighs the progressivity of the income tax.
For example, in 1996, all workers, from minimum-wage employee to CEO, had to pay
6.8 percent of gross wages (approximately 8.5 percent of net) for health insurance. That
8.5 percent of net wages was about what a couple with one child earning ₣20,000 per
month had to pay in income tax in 1996. The 6.8 percent health insurance charge
represented barely 10 percent of total social charges, which amounted to roughly 65
percent of gross wages (20 percent paid by employees, 45 percent by employers).
FIGURE 4.1. Effective marginal and average rates in France in 1996. Source: Piketty, 1997a.

This conclusion also holds for households with children: family allowances of
course account for a higher percentage of supplementary income for low earners, but the
family quotient system (which reduces the tax on large families) results in a larger
decrease in income tax (as percentage of income) for high earners, so that the curve of
average effective tax rates remains fairly flat. To be sure, as the “infinite income” bars
in Figure 4.1 indicate, the effective average rate can go as high as 70 percent for very
high earners who surpass all the ceilings for deductions and family quotients
(applicable to those earning more than about ₣700,000 per year), who were liable to a
top marginal income tax rate of 56.8 percent in 1996. But this affects very few
households, so the actual importance of this top rate is much smaller than the symbolic
importance often attached to it in political debate. In particular, it hardly alters the basic
reality portrayed in Figure 4.1, namely, the absence of any substantial redistribution
between workers. Including other proportional taxes such as the value-added tax, whose
weight is twice that of the income tax, would further reinforce this conclusion.
These observations for France sum up the main feature of the contemporary fiscal
redistribution in all Western countries, in spite of national institutional differences:
there is no significant monetary redistribution between active workers. Taxation of
active workers is broadly proportional and transfers are minimal, so that the variation
in disposable income between active workers is quite close to the variation in wages
paid by employers. As noted in Chapter 1, the countries in which income inequality is
small are the countries in which wage inequality is low and vice versa. It is not the case
that fiscal redistribution between workers reduces initially high wage inequalities. In
addition to traditional expenditures (on justice, defense, highways, and so on), this
broadly proportional tax essentially serves to pay for unemployment insurance,
educational expenses, and above all pensions and health care. In some cases, these
expenditures are advantageous to workers at the low end of the wage scale, but this is
not always the case. I will say more about this later.
The U-Shaped Curve of Marginal Rates
Effective average rates of tax and transfer can measure the degree of redistribution that
actually occurs, but to measure the impact of this redistribution on individual behavior,
effective marginal rates are more useful. Figure 4.1 shows that effective marginal rates
are higher at both the low end and the high end of the income distribution than in the
middle, creating a rather marked U-shaped curve. It is not surprising that marginal rates
are higher at the top end: high earners are in the top income-tax brackets. Marginal rates
are also higher at the low end because a person who goes from zero wage income to
some wage income must not only pay taxes on his pay but also lose certain social
transfer payments available only to those who have no income from work. Consider, for
example, an unemployed person in France who receives €530 per month in guaranteed
basic income and housing allotment but who then finds an employer prepared to pay
€1,370 a month for his labor, presumably because his contribution to the production
process brings in at least that amount. The worker will actually receive just over €760
in net income each month after deducting all social charges. In other words, his gross
income goes from 0 to 1,370, but his disposable income goes from 530 to 760. The
worker’s additional income is just €230 per month, or less than 20 percent, which
yields an effective marginal tax rate of more than 80 percent, as shown in Figure 4.1 for
the first decile of the wage distribution. If the worker has children or was entitled to
unemployment benefits before finding work, the calculation would be slightly different,
but the effective marginal rate would still be on the order of 80–90 percent, and in some
cases more than 100 percent (for details, see Piketty, 1997a).
As it turns out, low-wage workers bear the highest effective marginal rates: a single
worker moving from the ninth to the tenth decile of the wage distribution faces a
maximum marginal rate on the order of 60 percent and 70 percent for an infinite income
in the top income tax bracket (assuming no special deductions), compared with a rate on
the order of 80–90 percent for a person moving from unemployment into the first decile
of the wage distribution. This U-shaped curve of effective marginal rates, with the
highest peaks at the bottom end of the curve, is the second major characteristic of
contemporary fiscal redistribution. Once again, it is a characteristic shared by all
Western countries: reserving social transfers for those who have no income from labor
while excluding low-wage workers is, in appearance at any rate, the least costly way to
fight poverty. It was this thinking that prevailed when national welfare systems were put
in place.

Just Fiscal Redistribution

Are these curves of marginal and average effective rates of redistribution optimal from
the standpoint of social justice? Should they be raised or lowered for different income
groups?
The answers to these questions depend to a large extent on the magnitude of the
negative effects that high redistribution rates may have on incentives to work and to
supply human capital and therefore on redistribution itself. There is a fairly broad
consensus concerning the fundamental purposes of pure redistribution: a redistribution
is just if it does as much as possible to improve the opportunities and living standards
of the least well off. This is John Rawls’s maximin principle, discussed in the
Introduction. Of course, controversy remains as to which individuals are the least well
off. Since individuals differ in multiple dimensions, this is not always easy to decide,
and this difficulty can pose problems when it comes to defining the idea of
responsibility and setting the goals of social justice, as recent work in social justice
theory has shown (Fleurbaey, 1996; Roemer, 1996).
The pragmatic conception of social justice expressed by the maximin principle—
namely, that inequality is tolerable as long as any further redistribution would not be in
the interest of the least well off—has also aroused objections of principle. Some critics
reject the idea that the price system and individual self-interest form a sound basis for
organizing an economy. Surveys show, however, that most people agree that inequalities
due to factors beyond the control of individuals should be corrected insofar as it is
possible to do so. The consensus on this point is even more striking when compared to
the deep agreement that exists as to actions that are within the power of individuals and
therefore the ways in which redistribution may alter incentives to behave in certain
ways (Piketty, 1995). I turn to this conflict next. What is the practical effect of
redistribution on incentives?
Do High Taxes Diminish Revenue?
In the 1980s, an idea gained popularity in the United States: high taxes are a
disincentive to work by high achievers, so much so that tax revenues actually decline
when rates are raised too much, so that high tax rates benefit no one. More broadly, all
Western countries began to wonder whether their redistributive systems had attained
their limits. Taxes, which had risen rapidly in previous decades, began to level off.
These changes were most visible in the United States, where the top marginal income
tax rate was progressively reduced from 70 percent in the late 1970s to 28 percent in
1986.
 Traditional estimates of the elasticity of the supply of labor generally found that it
was quite low, however, on the order of 0.1–0.2 percent (Blundell, 1995, p. 60). Of
course, these studies measured the elasticity of the supply of labor and not of human
capital as such. They considered only the effect of tax rates on the number of hours
worked, which do not vary much for the majority of workers, and not on motivation or
efficiency, nor on incentives to acquire additional human capital or find a more
remunerative job, which are potentially more important. It is hard to imagine that these
incentives are completely insignificant, even if they are difficult to measure. For
example, the proportion of each age cohort entering higher education in the United
States decreased by 5 percent from 1975 to 1980, while the return to a college degree
decreased by 15 percent. Subsequently, the percentage pursuing higher education rose
by about 10 percent in the 1980s (Ehrenberg and Smith, 1994, p. 289). These facts do
not allow us to make a rigorous estimate of the elasticity of the supply of human capital,
however, although they do suggest that the supply will eventually decrease if
appropriate employment opportunities disappear.
One study of high incomes conducted after the 1986 tax reform in the United States
concluded that the cut in top marginal rates had a significant incentive effect, with an
elasticity on the order of 1 (Feldstein, 1995). This estimate applied only to very high
earners, however. In addition, it seems that the bulk of the measured effect was owing to
a transfer of earnings previously taxed as corporate income to earnings taxed as
personal income rather than to actual new earnings due to incentives to work harder and
acquire more human capital (Slemrod, 1995). Such transfers between different
categories of income are common at the top end of the distribution, which therefore
needs to be approached very cautiously. The top marginal income tax rate in the United
States was gradually increased from 28 percent in 1986 to 39 percent in 1993, but this
had no clear effect on the rate of growth of high incomes after 1993 (Goolsbee, 1997).
French data appear to confirm Goolsbee’s results. The ceiling placed on the family
quotient in 1981 amounted to a sharp increase in the marginal tax rate on high-income
households, while the marginal rate on unmarried individuals and childless couples
remained unchanged. This made for a particularly interesting “natural experiment.” A
meticulous examination of the tax data shows that top earners in both groups increased
their income at roughly the same rate, independent of the family quotient and therefore of
differences in their marginal tax rates (Piketty, 1999).
Given the current state of our knowledge, the fact that marginal rates are higher at the
lower end of the wage distribution than in the middle or at the top (Figure 4.1) suggests
that too much attention has been paid to the supposed disincentive effects of higher top
marginal rates. This has not been helpful for judging whether the modern welfare state
has reached the limit of possible redistribution. In many countries estimates have been
made of the effects of high marginal rates on labor market participation by initially
unemployed individuals in various categories (youths, single people, and married
women). All these studies have found much higher elasticities (ranging from 0.7 to 1.2)
for these groups than for those already employed (Blundell, 1995, p. 59; for a study of
recent French data, compare Piketty, 1998). In other words, the prospect of earning a
decent living seems to be more of an incentive for low-income people than the prospect
of a still-higher income for those who are already well off: “poverty traps” are likely to
be more important than “middle-income traps.” The evolution of the Earned Income Tax
Credit (EITC) in the United States since the early 1990s has done much to invigorate
this debate.
The Earned Income Tax Credit in the United States
The EITC, first introduced in 1975, is a tax credit that benefits people at the low end of
the wage spectrum. Since its modest beginnings, it has blossomed into a central element
of the American tax and welfare system. In 1996 the EITC offered a tax credit of 40
percent of earned income on incomes below $9,000 a year, decreasing gradually to 20
percent on incomes up to $29,000 a year. In other words, a worker who earned $9,000
in 1996 received a tax credit of $3,600. This tax credit was “reimbursable,” meaning
that the taxpayer received a check equal to the difference between the credit and the
amount of tax due, which at this level of income was always less than 10 percent, thus
increasing net disposable income by more than 30 percent. With the increase in the
federal minimum wage in 1996, this measure represented an effort by the US
government to make low-paid work more attractive. The fact that the EITC rate was
increased rather suddenly and that the full rate applied only to low-income people with
at least two children made it possible to measure in fairly convincing fashion the
considerable positive effects of the measure on the level of employment in the target
population, revealing an elasticity slightly greater than 1 (Eissa and Liebman, 1996;
Liebman, 1996).
Suppose that the disincentive effects of redistribution are in fact higher at the bottom
end of the income distribution than at the top. Then the U-shaped curve of effective
marginal rates traditionally used to focus redistribution on the poorest workers might
not be the most effective strategy: by eliminating the initial portion of the U curve, that
is, by decreasing charges on low incomes and transferring them to middle and high
incomes, where elasticities are lower, one could finance larger transfer payments to
those who cannot find jobs. This would make the system of redistribution more just by
improving the situation of the least well off. The absence of any substantial fiscal
redistribution between workers would then be equally bad for the jobless. The
experience of the EITC in the United States suggests that a flattening of the initial
portion of the U curve may be a more important priority than the remedy more frequently
recommended by many politicians, namely, reducing marginal tax rates at the top of the
distribution.
Fiscal Redistribution to Reduce Unemployment?
The EITC experience also raises the more general question of the role of fiscal
redistribution in reducing unemployment. Could the positive effects of the EITC on
unemployment be obtained in a country like France? The fact that low wages have
collapsed in the United States and not in France might suggest instead that making low-
wage jobs more attractive and avoiding “poverty traps” are not really issues in France.
But the presumed gap between low wages in France and the United States should not be
overestimated. True, for the first time in American history, bottom-end wages in the
United States have fallen in absolute terms since the 1970s, but they were initially
higher than in France, so that after the US federal minimum wage increase in 1996,
French and American minimum wages were actually quite comparable: on January 1,
1997, the French minimum wage was ₣38 an hour gross or about 29.7 net after
deducting 21.8 percent in social charges, while the federal minimum wage in the United
States was $5.20 an hour (or $4.81 net after deduction of 7.5 percent). At an exchange
rate of 5.50 francs to the dollar, which is probably on the low side given comparative
price levels, and ignoring the EITC, which substantially increased the disposable
income of low-paid American workers, we find that the US minimum wage was
equivalent to ₣26.5 per hour net, compared with 29.7 for the French minimum wage. Of
course, low-paid workers in France receive distributions in kind (health and education)
not available (or less available) in the United States, but the jobless also receive those
benefits, so they don’t matter in regard to work incentives, which is what we are
interested in here. Hence it is not certain that the attractiveness of low-wage jobs and
the avoidance of poverty traps are not issues in France.
The fundamental difference, though, is obviously that the demand for labor is higher
in the United States than in France. One point that is difficult to get around is that
although the net minimum wages in both countries are comparable, the “super-gross”
minimum wages, that is, the wages including all employer-paid charges, are quite
different: US employers paid 7.5 percent in charges in 1996, for a minimum wage (from
the employer’s point of view) of $5.59 or ₣30.7 per hour, compared with charges of
44.8 percent in France in 1993, for a total of ₣55 an hour (reduced to ₣48.1 by the
reduction in social charges on low-paid workers on January 1, 1997).
That is why French strategies for reducing unemployment through structural fiscal
reform since the late 1970s have focused on stimulating demand for low-wage labor
rather than supply: the goal is to reduce the relative cost of low-wage jobs by gradually
shifting employer charges from low-wage to high-wage workers (Chapter 3). This
strategy has made the schedule of charges on active workers slightly more progressive,
although it remains relatively flat (Figure 4.1), and although the reductions in employer
charges from 1993 to 1996 were much too concentrated on the very bottom of the wage
distribution, with the attendant risk of creating “low-wage traps” (because employer
costs increase very rapidly if the employer attempts to increase the worker’s net wage
[Piketty, 1997a]).
Regardless of whether one attempts to influence the supply of labor or the demand,
what quantitative reduction of unemployment can one expect from such structural
reforms? If the restructuring of employer charges is systematic and universal (rather than
limited to specific categories of employers and employees), all available studies
suggest that the long-term effects will be “significant.” What does “significant” mean?
The problem is that the elasticities of supply and demand on which these strategies are
based, while not negligible, unfortunately tend to be about 1 or less, which means that
the shift in taxes required to create one job is close to the market value of that job, that
is, the cost to the employer per job. It is therefore tempting to conclude that the
government could bear the cost itself by creating public-sector jobs, as the
Scandinavian countries did to reduce unemployment in the 1980s. Or else that the
government could impose job sharing by reducing working hours and compensating for
the reduced purchasing power of low-paid workers. All these strategies would cost
about the same in terms of francs invested per job created. Of course, one important
difference is that a job in the private sector reflects consumer demand, whereas the
contribution of a public-sector employee to output is sometimes more dubious.
Furthermore, job creation through work sharing implicitly assumes that the currently
unemployed can perform the same tasks just as well as the currently employed, which
may be true in some cases but is not universally true. In any case, the effects of fiscal
redistribution on unemployment have been unimpressive enough that the question is not
yet settled. In addition, the private-sector jobs that one might hope to create this way, in
the food service, retail, and similar sectors, are considered less valuable than the
industrial jobs of the postwar boom years.
Negative Income Tax and Basic Income
A radical fiscal redistribution proposal gained a great deal of attention in the 1960s and
1970s. Milton Friedman proposed that each adult citizen should receive a certain lump
sum every month, regardless of income or labor market status. The original idea was to
interfere as little as possible with the virtuous mechanism of the market and to replace
other forms of “social protection” with a single modest allocation (Friedman, 1962).
Subsequently, Friedman’s proposal was taken up by proponents of a more substantial
redistribution known as “basic income” (Van Parijs, 1995). In its initial form, this
universal transfer was conceived as part of a negative income tax. All existing forms of
fiscal redistribution were to be replaced by taxing earned income at a single marginal
rate and using the proceeds to finance the basic income transfer. The higher the desired
universal transfer, the higher the marginal tax rate would have to be. George McGovern,
the 1968 Democratic presidential candidate, learned this lesson to his dismay when
voters realized that the flat 33.3 percent marginal rate he proposed would not be nearly
enough to finance the transfer he envisaged. The basic income is thus a very different
instrument from the EITC, because the latter fits into the existing fiscal redistribution
system with a negative marginal rate of minus 40 percent. It does not alter existing taxes
and transfers, and the additional transfer it provides is not universal but cuts off at a
certain level of income.
The negative income tax and basic income proposals may at first sight seem
surprising in light of the foregoing discussion. Obviously, the amount of any “universal”
transfer must be smaller than the taxes levied on those earning above a certain level of
income, since the transfer to those lower down the income scale has to be paid for. Why
pay a basic income to everyone if it means raising taxes on a portion of the population
by an amount greater than the basic income they receive? If the goal is to reduce the
marginal effective rate on low-income people while preserving the same level of
transfer for those who earn nothing, it would seem simpler to cut taxes (such as payroll
taxes) on low earners or pay them directly while shifting the burden in the middle of the
income distribution or above. Anything that the basic income can do to combat “poverty
traps” can also be accomplished with existing fiscal instruments with such a universal
transfer.
In fact, however, a basic income offers certain subtle advantages. It can guarantee no
loss of income to welfare recipients willing to take a job, for example, and thus
improve their incentive to work. If the job is subsequently lost, the worker is
nevertheless guaranteed to receive at least the basic income, whereas under the existing
French system, a worker who loses a job has to requalify for the RMI, with the attendant
administrative delays and social stigma (Van Parijs, 1995). More generally, “left
libertarians” argue that a universal transfer allows for a less inquisitorial social policy,
since it is not based on the recipient’s social or marital status.
 Efficient Redistribution
In many situations, inequality calls for a collective effort of redistribution not only
because it contradicts our sense of social justice but also because it implies a
formidable waste of human resources that could be better utilized for the benefit of all.
As noted previously, typical examples include discrimination and monopsony power in
the labor market. It is not only insulting but also inefficient to offer fiscal compensation
to the victim of unjust discrimination or exploitation by an employer. Inequality of this
kind requires redistributive instruments capable not only of redistributing income but
also of correcting the market failures responsible for them. Such instruments include
affirmative action, minimum wage legislation, and other direct labor market
interventions. Education and training policies can also serve as powerful instruments of
efficient redistribution by modifying the structure of labor income inequality. We
discussed these instruments in Chapter 3. Two other types of efficient redistribution
have been important historically and politically: social insurance and Keynesian
demand management.

Redistribution and Social Insurance

The imperfection of the credit market (or the mere fact that “people lend only to the
rich”) is responsible for the most transparent form of inefficient inequality. Instruments
for combating it effectively have not always been easy to come by, however (see
Chapter 2). In fact, the incentive and information problems that are responsible for
credit rationing exist in any intertemporal market, including, in particular, the market for
insurance. This may be the reason why the market has proved incapable of providing
adequate social insurance, thus justifying the need for the compulsory government-
provided social protection schemes that form the heart of the modern welfare state.
Efficient Social Insurance
For example, no private insurance company has ever offered a policy covering the risk
of temporary unemployment and providing a suitable replacement income, despite the
obvious utility of such a policy and the large number of people who would be willing to
pay for one. An obvious explanation for this market failure is that it may be difficult to
observe what a person actually earns, and the individual in question always has an
interest in understating his income in order to pay a lower premium. The government’s
advantage in this respect is that, over time, it has acquired the legal and administrative
capacity to monitor what employers actually pay workers and thus to establish each
worker’s claim to unemployment insurance.
Adverse selection is also an important factor. Companies have an interest in
attracting low-risk clients, but the worker is better informed than the company about his
level of risk. Companies will therefore tend to offer policies designed to separate low-
risk workers from high-risk ones. Such policies will be inefficient. For instance, they
may propose high deductibles or cover only minor risks. Adverse selection can be
particularly problematic in the health insurance market, where individuals often have a
good deal of private information about their risks. In such situations, competition may
be totally inefficient, not only for high-risk individuals who may find themselves
excluded from the market but also for low-risk individuals who are offered inefficient
contracts. Making insurance compulsory may then be in everyone’s interest even though
low-risk individuals will have to subsidize high-risk ones (Rothschild and Stiglitz,
1976). The same phenomenon may also justify government regulation of other insurance
markets, such as the market for automobile insurance (Henriet and Rochet, 1988).
There is also another justification for government-provided health insurance: only
the seller (the health care provider) can accurately judge the cost of the service, which
may lead to unjustifiably high prices and unnecessary procedures (Arrow, 1963). This
argument is often used to explain why health care costs are so high in countries where
private health insurance predominates, such as the United States, and to justify
government regulation of health-care expenditure in public health-care systems.
The imperfection of intertemporal markets may also justify public pension systems.
Adverse selection exists, since a pension is also a kind of “survival insurance.” Indeed,
the market for converting savings into lifetime annuities is far from perfect. Still, the
problem of private information about one’s own life expectancy is surely less important
in regard to pensions than in regard to unemployment and health risks. The primary
justification for government pensions is simply the imperfection of markets for
transferring income from working years to retirement years. Low-income individuals
have limited access to the kinds of investments needed to accumulate adequate
retirement savings, and a state-guaranteed public pension system can compensate for
this deficiency.
In these and other situations, competitive markets are often incapable of providing
the goods and services that consumers value most. A compulsory public system can fill
this role efficiently. It is then justifiable to treat public expenditures for these purposes
separately from other government expenditures and to exclude the charges levied to
finance them from the calculation of average and marginal effective rates of fiscal
redistribution: if a worker goes from earning ₣5,000 a month to ₣10,000, he also
doubles his claim on retirement income, and any additional charges he is required to
pay for these additional rights should be counted as deferred income. If we take such
deferred income into account, we have to reduce the average and marginal “effective”
rates in Figure 4.1 by 15–20 points (Piketty, 1997b). A pure system, in which all social
insurance would take the form of charges to purchase deferred income, would not
involve redistribution at all but simply correct market imperfections in an efficient
manner. Social charges would simply be payments for the coverage that individuals
would desire if the market were capable of providing it.
Is Social Insurance an Instrument of Fiscal Redistribution?
Nevertheless, not all social insurance expenditures are neutral with respect to
redistribution of labor income. Health insurance is the simplest case: it is financed in
France by a payroll tax proportional to income, whereas most reimbursements for care,
hospital fees, and the like are the same for all. Absent any efficiency justification, this
redistributive effect would suffice to legitimize the system. In view of the impossibility
of achieving genuine redistribution from capital to labor, compulsory public health
insurance (as implemented in France) is simply a form of negative tax that provides for
lump-sum transfers paid for by a proportional payroll tax. Even though this is a form of
pure redistribution, there is no reason why health insurance should not be accounted
separately from other forms of fiscal redistribution and paid for by a specific charge so
that each individual can measure its cost relative to other expenses. Educational
expenditures could be financed in a similar way.
Since there is no significant direct redistribution between active workers, as we
have seen, these two forms of social expenditure, on health and education, are two main
forms of contemporary redistribution. Redistribution is achieved through government
spending rather than monetary transfers. Both are lump-sum transfers from which
everyone benefits equally, no matter what his or her level of income (at least in regard
to primary and secondary schooling), and they are financed by charges that increase
proportionately or slightly progressively with income. The level of in-kind
redistribution is what distinguishes countries that redistribute less from countries that
redistribute more, since monetary transfers between active workers are of negligible
importance everywhere. For example, a French worker paid the minimum wage is paid
roughly the same amount as an American worker on minimum wage, but the American
worker must pay for his own health insurance and for the education of his children,
which can be quite costly. Because of this fiscal redistribution, a minimum-wage
worker in France is incontestably better off than a minimum-wage worker in the United
States (unless he or she is young, in good health, and childless, as minimum-wage
workers often are).
The expenditures of public pay-as-you-go (PAYGO) pension systems, which account
for by far the lion’s share of all social spending, are quite different. The proportional
payroll taxes paid during a worker’s working life entitle him or her to receive transfer
payments upon retirement—payments that are also proportional to past income. One
might think, therefore, that the redistributive effect would be nil. In fact, inequality in
retirement is primarily a matter of unequal life expectancies: broadly speaking, low-
wage workers have markedly shorter life expectancies than high-income workers, so
they receive pension payments for a shorter time period. The available studies that look
at the French pension system in its entirety indicate that for every franc of payroll tax
paid during a worker’s working life, top managers receive pension payments that are
more than 50 percent greater than the pension payments received by ordinary workers
(Chassard and Concialdi, 1989, p. 76). In other words, pension payments redistribute
upward: a substantial portion of the payroll tax paid by workers goes to finance the
pensions of top managers. One must obviously bear in mind that a pension system based
on private saving might leave workers without any pension at all if their savings were
lost in speculation and in the imperfection of intertemporal markets. The disastrous
experiences of pension funds in the interwar wars justified the establishment of PAYGO
government pension plans by demonstrating the inability of the financial markets to pay
defined benefits on the basis of a given savings rate. Times have changed, however: the
financial products available in the 1990s offered more opportunity for collective
investments with guaranteed returns, even for small savers (such as minimum-wage
workers, who paid payroll taxes combining employer and employee charges of ₣1,500 a
month in 1996).
Moving to a private pension plan would not really solve the problem of unequal life
expectancies, however, because investors in these collective plans would include both
high- and low earners with highly unequal life expectancies, just as the government
PAYGO plans do. In any case, the main problem is obviously that the transition from
one plan to another must be very gradual. It would be unjust to deprive workers near the
end of their working lives of the pensions they have been promised, even if those
pensions were financed in an antiredistributive manner.
Of course, public pension plans always set a floor on pensions such as le minimum
vieillesse in France. Every retiree receives this minimum pension, even if he or she has
not paid a sufficient amount in payroll taxes during working years. These pensions are
certainly redistributive. What is more, it is these minimum limits on pension benefits
that have made it possible to eradicate the once pervasive problem of “elder poverty”
and considerably reduced household income inequality throughout the Western world.
Such transfers account for only a very small portion of total pension expenses in France
and elsewhere in Europe, however. Furthermore, similar minimum benefit limits exist
and play the same beneficial role in countries where the public pension system was
created primarily for this purpose, as in the United States and United Kingdom.
The mixed results of public pension systems also illustrate the dangers of believing
the myth of “the insurance society” (Rosanvallon, 1995). If one thinks of redistribution
in terms of social insurance in a society all of whose members are subject to the same
“risks,” against which they must be ensured collectively, one may well fail to
redistribute where redistribution is needed, for example, between active workers, while
establishing “redistributions” that fail to redistribute, as is the case with many pension
plans and certain redistributions in kind, such as public higher education, which often
become mechanisms for redistributing from low earners to high earners, as in France.

Redistribution and Demand

“Keynesian” demand management is an effective redistributive mechanism that occupies
an important place in the contemporary image and practice of state intervention. In the
public mind, what Keynes taught was that an increase in wages can stimulate demand
for goods and services and thereby increase output and raise the level of employment.
As far as redistribution is concerned, this is the best of all possible worlds, because
everything improves at the same time and nobody has to pay for it. It is a powerful idea,
but its conceptual and empirical foundations are relatively fragile. Why should a
redistribution of demand increase the output of the economy? If it is simply a matter of
transferring purchasing power from firms and capitalists to workers (which may be an
excellent thing to do from the standpoint of social justice), why would overall demand
increase, unless we assume that firms and capitalists are not spending their full
purchasing power on either consumption or investment goods? In fact, there is little or
no “dormant” purchasing power: income that is not consumed immediately is always
invested somewhere in one form or another: in government bonds, for example. One
possible interpretation is that if the total demand for goods and services remains
constant, the composition of the consumption basket nevertheless changes, and this
stimulates more output. For example, unconsumed income may not have been invested in
the most useful way possible, and redistribution toward workers or mobilization of idle
resources through government spending might make for a more efficient use of savings.
Another classic argument is that redistribution of purchasing power shifts demand
toward goods that can be produced efficiently only on a very large scale, thus
stimulating output from larger firms at the expense of smaller-scale productive units.
For instance, a high level of inequality in purchasing power may prevent or delay
industrialization, because the poor are too poor to generate sufficient demand for
industrial goods, while the rich demand mainly imported goods and domestic services
(Murphy et al., 1989; Piketty, 1994, pp. 791–794).
Clearly, then, there is no shortage of arguments to justify the idea that a redistribution
of purchasing power can both diminish inequality and stimulate activity for the benefit
of all. There is, however, no reason to believe that all the conditions necessary for these
virtuous mechanisms to operate are satisfied simultaneously. Each case must be
evaluated on its merits. Furthermore, economists nowadays usually explain Keynesian
stimulus quite differently: the argument is that prices and wages do not adjust
sufficiently rapidly in the short term, for example, because nominal wages are sticky, so
that the only way to raise the level of employment is to lower real wages by way of
inflationary stimulus. The idea that inflation can “lubricate” the gears of the economy
and eliminate rigidities takes us a long way away from the best of all possible
redistributive worlds, where the argument was rather that it was an increase in the
purchasing power of workers that was supposed to stimulate the economy. A stimulus
policy may also have other consequences, such as increasing the stock of public debt,
which inevitably increases the demand for and therefore the return on capital—an
unwanted redistributive effect. In any case, stimulus policies generally have short-term
effects, and their long-term redistributive consequences are difficult to assess,
especially when compared with the powerful structural instruments analyzed
previously.
The example of Keynesian redistribution also shows how fruitless it may be always
to seek an efficient redistribution that can solve every problem at once. The danger was
already clear in the myth of the “insurance society,” but the point is more general. For
example, it is misleading and counterproductive to argue that all human capital
inequality is a consequence of discrimination or that low wages are always a
consequence of employer monopsony. Although it is essential to identify efficient
redistribution wherever it exists, it is pointless to denounce every inequality as a sign of
gross inefficiency that the right policy can eliminate. To do so is to delegitimize the
taxes needed to finance fiscal transfers, which may not eliminate every imagined
inequality but nevertheless help to attenuate very real inequalities in standards of living.
References
 Abbreviations
AER American Economic Review
JPE Journal of Political Economy
QJE Quarterly Journal of Economics

Adelman, I. and S. Robinson. 1989. “Income distribution and development.”Handbook of Development Economics,
vol. 2. New York: North-Holland.
Akerlof, G. and J. Yellen. 1990. “The fair wage-effort hypothesis and unemployment.” QJE 105: 255–283.
Arrow, K. 1963. “Uncertainty and the welfare economics of medical care.” AER 53: 941–973.
______ 1973. “The theory of discrimination.” In O. Ashenfelter and A. Rees, eds.,Discrimination in Labor
Markets. Princeton: Princeton University Press.
Atkinson, A. 1983. The Economics of Inequality. Oxford: Clarendon Press.
Atkinson, A., L. Rainwater, and T. Smeeding. 1995. Income Distribution in OECD Countries. Paris: OECD.
Atkinson, A. and J. Stiglitz. 1980. Lectures on Public Economics. New York: McGraw-Hill.
Banerjee, A. and M. Ghatak. 1995. Empowerment and Efficiency: The Economics of Tenancy Reform. Cambridge,
MA: MIT.
Banerjee, A. and A. Newman. 1993. “Occupational choice and the process of development.” JPE 101: 274–299.
Becker, G. 1964. Human Capital. New York: Columbia University Press.
______ 1991. A Treatise on the Family. Cambridge, MA: Harvard University Press.
Benabou, R. 1993. “Workings of a city: Location, education, production.” QJE 108: 619–652.
______ 1996. “Inequality and growth.” NBER Working Paper 5658, inNBER Macroeconomics Annual 1996,vol.
11.
Bewley, T. 1994. “A field study on downward wage rigidity.” New Haven: Yale University.
Blau, F. and L. Kahn. 1994. “The impact of wage structure on trends in US gender wage differentials.” NBER
Working Paper 4748.
Blundell, R. 1995. “The impact of taxation on labour force participation and labour supply.” OECD Jobs Study
Working Papers, 8. OECD Publishing. http://dx.doi.org/10.1787/576638686128.
Booth, A. and D. Snower. 1996. Acquiring Skills: Market Failures, Their Symptoms and Policy Responses.
Cambridge: Cambridge University Press.
Boudon, R. 1973. L’Inégalité des chances. Paris: Armand Colin.
Bound, J. and R. Freeman. 1989. “Black Economic Progress: Erosion of Post-1965 Gains in the 1980s.” In S. Shulman
and W. Darity Jr., eds., The Question of Discrimination: Racial Equality in the U.S. Labor Market,pp. 32–49.
Middletown, CT: Wesleyan University Press.
Bourdieu, P. and J.-C. Passeron. 1964. Les Héritiers. Paris: Minuit.
______ 1970. La Reproduction. Paris: Minuit.
Bourguignon, F. 1981. “Pareto-superiority of unegalitarian equilibria in Stiglitz’ model of wealth distribution with convex
savings function.” Econometrica 49: 1469–1475.
Bourguignon, F. and M. Martinez. 1996. Decomposition of the Change in the Distribution of Primary Family
Incomes: A Microsimulation Approach Applied to France, 1979–1989. Paris: DELTA.
Card, D. 1992. “The effect of unions on the distribution of wages: Redistribution or relabelling?” NBER Working
Paper 4195.
Card, D. and R. Freeman. 1993. Small Differences that Matter: Labor Markets and Income Maintenance in
Canada and the United States. Chicago: University of Chicago Press.
Card, D., F. Kramarz, and T. Lemieux. 1996. “Changes in the relative structure of wages and employment: A
comparison of the United States, Canada and France.” NBER Working Paper 5487.
Card, D. and A. Krueger. 1992. “Does school quality matter?” JPE 100: 1–40.
______ 1995. Myth and Measurement: The New Economics of the Minimum Wage. Princeton: Princeton
University Press.
Chamley, C. 1996. Capital Income Taxation, Income Distribution and Borrowing Constraints. Paris: DELTA.
Chassard, Y. and P. Concialdi. 1989. Les Revenus en France. Paris: La Découverte, “Repères.”
Coate, S. and G. Loury. 1993. “Will affirmative action eliminate negative stereotypes?” AER 83: 1220–1240.
Cohen, D., A. Lefranc, and G. Saint-Paul. 1996. “French unemployment: A transatlantic perspective.”Economic
Policy 12 (25): 265–292.
Coleman, J. 1966. Equality of Educational Opportunity. Washington, DC: US Dept. of Health, Education, and
Welfare.
Cooper, S., S. Durlauf, and P. Johnson. 1994. “On the transmission of economic status across generations.”ASA
Papers and Proceedings, pp. 50–58.
CSERC. 1996. Les Inégalités d’emploi et de revenu. Paris: La Découverte.
Davis, S. 1992. “Cross-country patterns of change in relative wages.” NBER Macroeconomics Annual 1992.
Douglas, P. 1976. “The Cobb-Douglas production function once again: Its history, its testing and some new empirical
values.” JPE 84: 903–915.
Drèze, J. and A. Sen. 1995. India: Economic Development and Social Opportunity. Delhi: Oxford University
Press.
Ducamin, R. 1995. Rapport de la commission d’études des prélèvements fiscaux and sociaux pesant sur les
ménages. Paris: Ministère de l’Économie et des Finances.
Duménil, G. and D. Lévy. 1996. La Dynamique du capital: un siècle d’économie américaine. Paris: PUF.
Ehrenberg, R. and R. Smith. 1994. Modern Labor Economics. New York: Harper Collins.
Eissa, N. and J. Liebman. 1996. “Labor supply response to the earned income tax credit.” QJE 111 (2): 605–637.
Erickson, C. and A. Ichino. 1995. “Wage differentials in Italy.” In R. Freeman and L. Katz, eds.,Differences and
Changes in Wage Structure, pp. 265–306. Chicago: University of Chicago Press.
Erikson, R. and J. Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies.Oxford:
Clarendon Press.
Feenberg, D. and J. Poterba. 2000. “The income and tax share of very high income households.”AER 90 (2): 264–
270.
Feldstein, M. 1995. “The effect of marginal tax rates on taxable income: A panel study of the 1986 Tax Reform Act.”
JPE 103 (3): 551–572.
Fleurbaey, M. 1996. Théories économiques de la justice. Paris: Economica.
Freeman, R. 1973. “Changes in the labor market status of Black Americans, 1948–1972.” Brookings Papers on
Economic Activity, no. 1, pp. 67–120.
______ 1995. “Are your wages set in Beijing?” Journal of Economic Perspectives 9 (3): 15–32.
______ 1996. Disadvantaged Young Men and Crime. Cambridge, MA: Harvard University Press.
Freeman, R. and J. Medoff. 1984. What Do Unions Do? New York: Basic Books.
Friedman, M. 1962. Capitalism and Freedom. Chicago: University of Chicago Press.
Goldin, C. and R. Margo. 1992. “The great compression: The wage structure in the United States at mid-century.”
QJE 107: 1–34.
Goolsbee, A. 1997. “What happens when you tax the rich? Evidence from executive compensation.” NBER Working
Paper 6333.
Gottschalk, P. 1993. “Changes in inequality of family income in seven industrialized countries.” AER 83 (2): 136–142.
Goux, D. and E. Maurin. 1996. “Meritocracy and social heredity in France: Some aspects and trends.”European
Sociological Review 13 (2): 159–177.
Hamermesh, D. 1986. The demand for labor in the long run. In O. Ashenfelter and R. Layard, eds.,Handbook of
Labor Economics, pp. 429–471. Amsterdam: North Holland.
______ 1993. Labor Demand. Princeton: Princeton University Press.
Harhoff, D. and T. Kane. 1994. “Financing apprenticeship training: Evidence from Germany.” NBER Working Paper
4557.
Henriet, D. and J.-C. Rochet. 1988. “Équilibres et optima sur les marchés d’assurance.” In Mélanges économiques
en l’honneur d’Edmond Malinvaud. Paris: Economica.
Herrnstein, R. and C. Murray. 1994. The Bell Curve: Intelligence and Class Structure in American Life.New
York: The Free Press.
IMF. 1996. World Economic Outlook.
INSEE. 1994. “Un siècle de données macroéconomiques.” INSEE Résultats, nos. 303–304.
______ 1995. “Revenus et patrimoine des ménages, édition 1995.” INSEE Synthèses, no. 1.
______ 1996a. “Séries longues sur les salaires.” INSEE Résultats, no. 457.
______ 1996b. “Revenus et patrimoine des ménages, édition 1996.” INSEE Synthèses, no. 5.
______ 1996c. “Rapport sur les comptes de la nation 1995.” INSEE Résultats, nos. 471, 472, 473.
______ 1996d. “L’évolution des salaires jusqu’en 1994.” INSEE Synthèses, no. 4.
______ 2002. “Les salaires dans l’industrie, le commerce et les services en 2000.” INSEE Résultats Sociétés, no. 7.
Judd, K. 1985. “Redistributive taxation in a simple perfect foresight model.” Journal of Public Economics 28: 59–83.
Juhn, C., K. Murphy, and B. Pierce. 1993. “Wage inequality and the rise in returns to skill.” JPE 101: 410–442.
Juhn, C., K. Murphy, and R. Topel. 1991. “Why has the natural rate increased over time?” Brookings Papers on
Economic Activity 2: 75–142.
Kahneman, D., J. Knetsch and R. Thaler. 1986. “Fairness as a constraint on profit seeking.” AER 76: 728–741.
Katz, L., G. Loveman, and D. Blanchflower. 1995. “A comparison of changes in the structure of wages in four OECD
coutries.” In R. Freeman and L. Katz, eds.,Differences and Changes in Wage Structure. Chicago: University of
Chicago Press.
Kolm, S. C. 1972. Justice et équité. Paris: Éditions du CNRS.
Kramarz, F., S. Lollivier, and L. Pelé. 1995. “Wage inequalities and firm-specific compensation policies in France.”
Document de travail INSEE-CREST 9518.
Kremer, M. and E. Maskin. 1996. “Wage inequality and segregation by skill.” NBER Working Paper 5718.
Krueger, A. and L. Summers. 1988. “Efficiency wages and the interindustry wage structure.” Econometrica 56: 259–
293.
Krussel, P., L. Ohanian, J. V. Rios-Rull, and G. Violante. 1996. “Capital-skill complementarity and inequality.”
University of Rochester.
Kuznets, S. 1955. “Economic growth and economic inequality.” AER 45: 1–28.
Landais, C. 2007. “Les hauts revenus en France (1998–2006): une explosion des inégalités?” PSE Working Paper.
Lefranc, A. 1997. “Évolutions des marchés du travail français et américains: quelques éléments d’analyse
comparative.” Revue Economique 48 (5): 1041–1060.
Lemieux, T. 1993. “Unions and wage inequality in Canada and in the United States.” In D. Card and R. Freeman,
eds., Small Differences that Matter. Chicago: University of Chicago Press.
Lhomme, J. 1968. “Le pouvoir d’achat de l’ouvrier français au cours d’un siècle: 1840–1940.” Le Mouvement social
63: 41–70.
Liebman, J. 1996. “Essays on the earned income tax credit.” PhD diss., Harvard University.
Lollivier, S. and D. Verger. 1996. “Patrimoine des ménages: déterminants et disparités.” Économie et Statistique
296–297: 13–32.
Lucas, R. 1990a. “Supply-side economics: An analytical review.” Oxford Economic Papers 42, pp. 293–316.
______ 1990b. “Why doesn’t capital flow from rich to poor countries?” AER 80: 92–96.
Mankiw, G., D. Romer, and D. Weil. 1992. “A contribution to the empirics of economic growth.” QJE 107: 407–437.
Meyer, C. 1995. “Income distribution and family structure.” PhD diss., MIT.
Morrisson, C. 1991. “L’inégalité des revenus.” In M. Lévy-Leboyer and J.-C. Casanova, eds.,Entre l’État et le
marché: l’économie française de 1880 à nos jours. Paris: Gallimard.
______ 1996. La Répartition des revenus. Paris: PUF.
Mulligan, C. 1996. Parental Priorities and Economic Inequality. Chicago: University of Chicago Press.
Murphy, K., A. Shleifer, and R. Vishny. 1989. “Income distribution, market size and industrialization.” QJE 104: 537–
564.
Murphy, K. and F. Welch. 1993a. “Inequality and relative wages.” AER 83 (2): 104–109.
______ 1993b. “Occupational change and the demand for skill.” AER 83 (2): 122–126.
Neumark, D. and W. Wascher. 1994. “Employment effects of minimum and subminimum wages: Reply to Card, Katz
and Krueger.” Industrial and Labor Relations Review 48: 497–512.
Nizet, J.-Y. 1990. Fiscalité, économie et politique: l’impôt en France, 1945–1990. Paris: LGDJ.
OECD. 1985. The Integration of Women in the Economy. Paris: OECD.
______ 1993. Perspectives de l’emploi, July. Paris: OECD.
______ 1995. Statistiques des recettes publiques des pays membres de l’OCDE, 1965–1994. Paris: OECD.
______ 1996. Economic Outlook no. 59, June. Annual Projections for OECD Countries.http://stats.oecd.org/ Paris:
OECD.
______ 2000. Taux de chômage standardisés (www.oecd.org). Paris: OECD.
Phelps, E. 1968. “The statistical theory of racism and sexism.” AER 62: 659–661.
______ 1994. Structural Slumps: The Modern Equilibrium Theory of Unemployment, Interest and Assets.
Cambridge, MA: Harvard University Press.
Piketty, T. 1994. “Inégalités et redistribution.” Revue d’économie politique 104: 769–800.
______ 1995. “Social mobility and redistributive politics.” QJE 110: 551–584.
______ 1997a. “La redistribution fiscale face au chômage.” Revue française d’économie.
______ 1997b. “Les créations d’emploi en France et aux États-Unis: ‘services de proximité’ contre ‘petits boulots?’ ”
Notes de la fondation Saint-Simon 93 (December). See also Économie et Statistique 318 (1998–8): 73–
99.
______ 1998. “L’impact des incitations financières au travail sur les comportements individuels: une
estimation pour le cas français.” Économie et Prévision 132–133 (Jan.-Mar.): 1–35.
______ 1999. “Les hauts revenus face aux modifications des taux marginaux supérieurs de l’impôt sur le revenu en
France, 1970–1996.” Économie et Prévision 138–139 (Apr.-Sept.): 25–60.
______ 2001. Les Hauts Revenus en France au XXe siècle. Inégalités et redistributions 1901–1998. Paris:
Grasset.
Piketty, T. and E. Saez. 2003. “Income inequality in the United States, 1913–1998.” QJE 118: 1–39.
Rawls, J. 1972. A Theory of Justice. Oxford: Clarendon Press.
Roemer, J. 1996. Theories of Distributive Justice. Cambridge, MA: Harvard University Press.
Rosanvallon, P. 1995. La Nouvelle Question sociale. Paris: Seuil.
Rothemberg, J. 1996. “Ideology and the distribution of income.” Cambridge, MA: MIT.
Rothschild, M. and J. Stiglitz. 1976. “Equilibrium in competitive insurance markets.” QJE 90: 629–650.
Shavit, Y. and H. P. Blossfeld. 1993. Persistent Inequality: Changing Educational Attainment in 13 Countries.
Boulder: Westview.
Slemrod, J. 1995. “Income creation or income shifting? Behavioral responses to the Tax Reform Act of 1986.” AER 85
(2): 175–180.
Solow, R. 1956. “A contribution to the theory of economic growth.” QJE 70: 65–94.
______ 1958. “A skeptical note on the constancy of relative shares.” AER 48: 618–631.
Spence, M. 1974. Market Signalling: Informational Transfer in Hiring and Related Screening Processes.
Cambridge, MA: Harvard University Press.
Topel, R. 1993. “What have we learned from empirical studies of unemployment and turnover?” AER 83: 110–115.
Van Parijs, P. 1995. Real Freedom for All: What (If Anything) Can Justify Capitalism? Oxford: Clarendon Press.
Williamson, J. 1985. Did British Capitalism Breed Inequality? Boston: Allen & Unwin.
Williamson, J. and P. Lindert. 1980. American Inequality: A Macroeconomic History. New York: Academic Press.
Wilson, W. J. 1987. The Truly Disadvantaged: The Inner City, the Underclass and Public Policy. Chicago:
University of Chicago Press.
Wolff, E. 1992. “Changing inequality of wealth.” AER 82 (2): 552–558.
Young, A. 1995. “The tyranny of numbers: Confronting the statistical realities of the East Asian growth miracles.”
QJE 110: 641–680.
 Contents in Detail

Note to the Reader

Introduction
1. The Measurement of Inequality and Its Evolution
Different Types of Income
Wage Inequality
International Comparisons
Income Inequality
International Comparisons
Inequalities in Time and Space
The Historical Evolution of Inequality
From Laws of History to Uncertainties
From Wages to Incomes
Inequality with Respect to Employment
2. Capital-Labor Inequality
The Share of Capital in Total Income
The Question of Capital/Labor Substitution
What Capital/Labor Substitution Means
Redistribution: “Fiscal” or “Direct”?
The Elasticity of Substitution between Capital and Labor
The Elasticity of Capital Supply
Are Capitalists and the Price System Necessary?
A Compromise between Short-Term and Long-Term Theories?
From Share of Value-Added to Household Income
What the Constancy of the Profit Share Tells Us
Who Pays Social Charges (Payroll Taxes)?
A Cobb-Douglas Production Function?
Historical Time versus Political Time?
Why Has the Profit Share Not Increased in the United States and United Kingdom?
The Dynamics of the Distribution of Capital
The Theory of Perfect Credit and Convergence
The Question of Convergence between Rich and Poor Countries
The Problem of Capital Market Imperfections
Possible Public Interventions
A Flat Tax on Capital?
3. Inequality of Labor Income
Inequality of Wages and Human Capital
The Explanatory Power of the Theory of Human Capital
Important Historical Inequalities
Supply and Demand
 The Rise of Wage Inequality since 1970
Skill-Biased Technological Change?
Wage Inequality and Globalization
How to Redistribute Labor Income
A Major Political Issue
Where Does Human Capital Inequality Come From?
Efficient Inequality?
The Role of the Family and Educational Expenses
The Problem of Inefficient Segregation of Human Capital
Discrimination in the Labor Market
Affirmative Action versus Fiscal Transfers
The Social Determination of Wage Inequality
The Role of Unions in Setting Wages
Unions as Substitutes for Fiscal Redistribution?
Do Unions Contribute to Economic Efficiency?
The Monopsony Power of Employers
When Does a Higher Minimum Wage Increase the Level of Employment?
Efficiency Wages and Fair Wages
National Traditions and Wage Inequality
4. Instruments of Redistribution
Pure Redistribution
Average and Marginal Rates of Redistribution
The Absence of Redistribution between Workers
The U-Shaped Curve of Marginal Rates
Just Fiscal Redistribution
Do High Taxes Diminish Revenue?
The Earned Income Tax Credit in the United States
Fiscal Redistribution to Reduce Unemployment?
Negative Income Tax and Basic Income
Efficient Redistribution
Redistribution and Social Insurance
Efficient Social Insurance
Is Social Insurance an Instrument of Fiscal Redistribution?
Redistribution and Demand
 Index

Adoption studies, education outcomes and, 82

Adverse selection: credit markets and, 60–61; pension systems and, 115–116; social insurance and, 115
Affirmative action, 86–88, 114
Africa: income inequality, 15; rates of growth, 58
African Americans, discrimination against, 85–88
Aggregate production function, of Solow, 30
Agriculture, direct redistribution of capital and, 63–64
Allocative role, of price system, 30–33, 37–40, 100
Arrow, Kenneth, 85
Asia: growth rate per capita, 58, 59; income inequality, 15; savings rates, 57
Average and marginal rates of redistribution, 100–102, 102f; absence of redistribution between workers, 102–
104; compulsory public systems and, 116; Earned Income Tax Credit in US and, 108–109; negative income
tax and basic income, 112–113; social justice and, 105–106; taxes and revenue, 106–108; unemployment and,
109–112; U-shaped curve of marginal rates, 104–105, 109

Basic income, guaranteed, 3, 23, 104, 112–113

Becker, Gary, 67, 78, 81
Behavioral differences, wealth inequality and, 13
Belgium, 14
Bell Curve, The (Herrnstein and Murray), 82
Bernstein, Eduard, 18
Binding wage schedules, 92–93
Busing, school integration and human capital, 84

Cambridge capital controversy, 30, 39

Canada, and income inequality, 10, 14, 23
Capital: flat tax on, 64–65; unequal ownership of, 26. See also Distribution of capital
Capital income: income inequality and, 12, 13; share received by households, 5–8, 6t, 44–45; social charges and,
34, 46–48; source of, 42–43; taxation of, 36–37, 43, 64–65. See also Income, share of capital in
Capitalism: capital-labor substitution, 39; Cobb-Douglas production function, 48; critics of and credit rationing,
61–62, 64; increases in inequality and, 17–18, 98
Capital-labor inequality, 26–27; capital-labor substitution, 27–40; classic and marginalist theories, 40–55;
dynamics of distribution of capital, 55–65
Capital-labor substitution, 27–30; elasticity of, 32–35, 37–39, 48–49, 51–54, 75–76; elasticity of capital supply,
35–37, 39; fiscal and direct redistribution, 30–32; market economy and price system, 37–40
Card, David, 95–96
Classical and marginalist theories, of capital-labor split, 29–30, 39–40, 41t; Cobb-Douglas production function
and, 48–49; economic value added and, 42–43; household income distribution, 41, 44–45; political and
historical time and, 41t, 49–53, 50t; profit share constancy and, 41t, 45–46; profit share in United Kingdom
and United States, 41t, 53–55; share of social charges, 46–48
Clinton, Bill, 96
Cobb-Douglas production function, 33–34, 48–49
Coleman, James, and report by, 82, 83–84
Collectivization, of means of production, 39, 62, 63–64
Competition: credit markets and, 57, 61; social insurance and, 115; taxes and, 37, 65
Compulsory education, human capital and, 80–81
Conditional convergence, 59
Convergence, between rich and poor countries, 57–60
C ratios, labor inequality and, 76–77
Credit markets: convergence between rich and poor countries, 57–60; imperfection of, 58–65, 69, 80–81, 114;
perfect credit market theory and, 56–57, 78
Credit rationing, 61–62, 64

Deferred income, social insurance and, 116

Demand management, redistribution and, 114, 119–121
Denmark: social protection, 47–48; wage inequality, 10
Development banks, as possible intervention in credit market, 62–63
Direct redistribution: agriculture and, 63–64; elasticity of substitution between capital and labor, 30–35, 48–49;
fiscal redistribution and, 28, 31, 98; inequality of labor income and, 75–76; unions and, 90–91
Disadvantaged communities, human capital and, 82–83
Discrimination, in labor market, 85–88, 113–114, 121
Distribution of capital, dynamics of, 55–56; capital market imperfections, 60–65; perfect credit and
convergence, 56–60
D ratios: sources of household income, 6t; wage inequality, 8–10, 9t

Earned Income Tax Credit (EITC), in US, 108–109, 112

Economic efficiency: distribution of capital, 55, 57, 61, 65; human capital, 86; unions and, 92–94
Economic value added: capital income and labor income, 41t, 42; capital share of (1979–1995), 50t;
complications of calculating, 43
Education: human capital and, 67, 69–70, 79–84, 92–93; redistribution and, 59–60, 117; wage inequality and, 72–
73, 99
Efficiency wages, 97–99
Efficient redistribution, 35, 62–63, 67, 113–114; demand management and, 119–121; egalitarian education policy
and, 59, 80; minimum wage and, 94–95; Pareto efficiency and, 2–3, 79; social insurance, 114–119. See also
Human capital entries
Elasticity of substitution between capital and labor, 32–35, 37–39, 48–49, 51–54, 75–76
Elasticity of supply of capital, 35–37, 39
Elasticity of supply of human capital, 78–79, 82–83, 87, 107
Employment: inequality with respect to, 23–25; job creation and elasticity of substitution between capital and
labor, 51, 53–55. See also Unemployment
Estate tax, progressive, 19, 64

Failure, human capital and self-fulfilling prophecies, 84–85

Fair wages, 97–99
Family, human capital and role of, 81–83
Family quotient system, in France, 87–88, 103, 107–108
Fiscal incidence, of taxes and social charges, 32, 46–48, 52
Fiscal redistribution: direct redistribution and, 28, 31, 98; and elasticity of substitution between capital and labor,
30–35, 48–49; inequality of labor income and, 75–76; unions as substitutes for, 90–92. See also Taxes
Flat tax, on capital, 64–65
France: attitudes toward inequality, 99; average and marginal rates of redistribution, 102, 102f; capital and labor
shares of value added, 41t; effective minimum wage, 110–111; Generalized Social Contribution, 34, 38;
household income sources, 6t; income inequality, 12–15, 12t, 24–25; income inequality (1870–1994), 16–17,
16t; income inequality, historical evolution of, 18–19, 22; minimum wage, 91; percentage of obligatory taxes,
 101; profit share in, 49–55, 50t; social charges, 46–48; taxes receipts as percent of GDP, 44; unions in, 91–
92; wage inequality, 8–11, 11t, 72–73
Freeman, Richard, 87–88
Friedman, Milton, 1, 3, 112

Generalized Social Contribution (CSG), in France, 34, 38

Geographic mobility, wage inequality and human capital, 96
Germany: employment, 24, 25; income inequality, 14; percentage of obligatory taxes, 101; profit share, 53;
unions, 91–92; wage inequality, 10, 91, 99
Gini coefficient, 10
Globalization: market integration, 59–60; wage inequality, 73–74
Goolsbee, Austen, 107
Grameen Bank, 63
Grenelle Accords (1968), 49
Gross operating surplus (GOS), 42
Guaranteed basic income, 3, 23, 104, 112–113

Hamermesh, Daniel, 49
Health insurance: adverse selection and, 115; justifications for compulsory, 115–117; as percentage of social
charges in 1966 France, 103
Herrnstein, Richard, 82, 87
Hidden underemployment, 25
Household size, income and, 12, 14, 22
Human capital: elasticity of supply of, 78–79, 82–83, 87, 107; measuring types of productivity and, 98; unequal
distribution of, 58–60; wage inequality and, 88–89, 92–99
Human capital, structural causes of inequality, 78–79; affirmative action versus fiscal transfers, 86–88;
discrimination in labor market and, 85–86; efficiency and, 79–81; inefficient social integration and, 83–84;
role of family and education expenses, 81–83
Human capital theory, 66–68; globalization and wage inequality, 73–74; historical inequalities and, 68–69; rise of
wage inequality since 1970, 70–71; skill-biased technological change and, 71–73, 76–77, 92; supply and
demand and, 69–70

Incentives: basic income and, 113; credit markets and, 60, 62, 114; effects of redistribution on, 105–110; of
households to save and invest, 35; human capital and investments, 78–88, 90, 93; of owners to accumulate
capital and invest, 28–29
Income: distribution by deciles and centiles, 5–8, 6t; household size and, 12; inequality of, 12–16, 12t, 15t;
inequality of, historical evolution, 17–25, 19f, 21t; left-right debate about inequality of, 1–3; types and
distribution of, 5–8, 6t. See also Capital income; Wage entries
Income, share of capital in: capital-labor substitution, 27–40; classic and marginalist theories, 40–55. See also
Capital income
Income tax, progressive, 19, 48, 64, 102–103, 106
India: human capital, 59–60, 80, 85; inequality, 16, 17
Individual self-interest, economic organization and, 39, 106
Inflation, as stimulus, 121
Information: credit markets and problems of, 61; insurance markets and, 114–115; prices and, 38
Inheritance of capital, 26; progressive tax on estates, 19, 64
Insurance society, myth of, 119, 121
Integration of schools, human capital and, 84
Intergenerational mobility, human capital and, 81, 83
Intermediate consumption, as share of firms’ costs, 42
Intertemporal markets, 60–61, 114, 115–116, 118
Italy: historical evolution of inequality, 21; labor participation rate, 25; profit share, 53; unemployment, 24

Katz, Larry, 96
Keynes, John Maynard, 40; demand management, 114, 119–121
Kolm, Serge-Christophe, 2
Krueger, Alan, 95–96
Kuznets curve, 18–20, 22, 66
Kuznets, Simon, 18

Labor income: as share of firms’ costs, 42; ways of redistributing, 74–77. See also Elasticity of substitution
between capital and labor; Wage entries
Labor market participation rate, 25, 108
Left-right conflict, about inequality and redistribution, 1–3
Lollivier, Stéfan, 13
Lucas, Robert, 59
Luxembourg Income Study (LIS), 14, 22

Marginalist theory, of capital-labor split. See Classical and marginalist theories, of capital-labor split
Marginal productivity of labor, capital-labor substitution and, 29–30, 34, 59
Marginal rates of redistribution. See Average and marginal rates of redistribution
Market economy. See Price system
Marx, Karl, 26, 30, 39; proletarianization thesis of, 17–18
Maximin principle, of Rawls, 2, 35, 106
McGovern, George, 112
Means of production, collectivizing of, 39, 62, 63–64
Minimum wage: EITC and, 109; health insurance and, 103; monopsony power of employers, 96; raising of, and
effect on level of employment, 95–96; redistribution and, 75, 94; unions and, 91; in US and France, 50, 110–
111, 117; wage distribution and, 8
Monopoly power, of unions, 89, 94
Monopsony power, of employers, 94–96, 113–114, 121
Moral hazard, credit markets and, 60–61
Murray, Charles, 82, 87

Negative income tax, 1, 3, 112–113

Nonwage compensation, 6t, 8, 12, 13. See also Self-employment compensation
Norway: historical evolution of inequality, 22; income inequality, 14; wage inequality, 10

OECD countries: evolution of shares of profits and wages, 49–53, 50t; historical evolution of inequality, 21;
income inequality, 14–15, 15t; wage inequality, 10–11, 11t

Panel Study of Income Dynamics (PSID), 83

Pareto efficiency, 2–3, 57, 79
Part-time work, income inequality and, 25
Pay-as-you-go (PAYGO) pension systems, 117–118
Payroll taxes. See Social charges
Pension plans: private, 118; public, 115–119
Phelps, Edmund, 85
Poverty traps, human capital and, 108, 110, 113
P ratios: income inequality, 12–14, 12t, 15t, 16–17, 23–25, 76–77; inequality’s historical evolution, 20–23, 21t;
minimum wage, 91; P defined, 7; sources of household income and, 6t; wage inequality, 8–11, 77
Price system: allocative role of, 30–33, 37–40, 100; elasticity of substitution and, 32–40; housing and educational
outcomes, 84; role in capital-labor share of total income, 27–30, 32; social justice and, 106
Primary distribution, 28
Prison population, underemployment and, 24
Private sector jobs, unemployment and fiscal redistribution, 111–112
Profit share: constancy of, 41t, 45–46; historical and political time and, 49–53; in US and UK, 53–55
Progressive estate tax, 19, 64
Progressive income tax, 19, 48, 64, 102–103, 106
Public investment banks, as possible intervention in credit market, 62–63
Public-sector jobs: pensions and, 115–119; unemployment and fiscal redistribution, 111–112; wages, 10
Purchasing power, of workers: changes in twentieth century, 45, 50–51, 68–69, 91, 96, 111; inequality in time
and space, 16–17, 16t; redistribution of, 120–121
Pure redistribution, 32, 55, 67; absence of redistribution between workers, 102–104; average and marginal rates
of redistribution, 100–102, 102f; Earned Income Tax Credit, in US, 108–109; fiscal redistribution to reduce
unemployment and, 109–112; fundamental purposes of, 105–106; high taxes and revenue, 106–108; negative
income tax and basic income, 112–113; Pareto efficiency and, 2–3; U-shaped curve of marginal rates, 104–
105, 109

Rawls, John, 2, 35, 106

Redistributive policy, left-right conflict about, 1–3. See also Average and marginal rates of redistribution; Direct
redistribution; Efficient redistribution; Fiscal redistribution
Rentier society, 19–20, 19f
Retained earnings, elasticity of capital supply and, 36–37
Retirees: income distribution and, 5, 6t, 7; social charges and, 101. See also Pension plans
Ricardo, David, 30
RMI, in France, 113
Rothenberg, Jerome, 98

Savings: behavioral differences and, 13; credit markets and, 56–57, 60; elasticity of capital supply and, 35–37;
Marx and, 39; pensions and, 115–116, 118
Self-employment: calculation of value added, 43; compensation as percentage of household income, 5, 6t; credit
rationing and, 62; income inequality and, 12
Self-interest, and economic organization, 39, 106
Singapore, 58
Skill-biased technological change, 71–73, 76–77, 92
Social charges: average and marginal rates of redistribution, 100–102, 108; employer-employee shares of, 46–
48, 104–105; fiscal incidence and, 52–53; as political issue, 76–77; price of labor and, 32, 42–44; shifting
burden of from low- to high-wage jobs, 34, 76–77, 111, 113, 119; unemployment and, 110–111
Social income: household income and, 5, 6t, 7, 105; income inequality and, 12, 23; taxes and, 47
Social insurance: efficient redistribution and, 114–116; as instrument of fiscal redistribution, 116–119; markets as
intertemporal markets, 61
Social integration, housing and educational outcomes and human capital, 83–84
Social justice: left-right consensus about fundamental principles of, 1–2; left’s skepticism about taxes and, 38,
52, 76; methods of redistribution and economic efficiency, 2–3, 32, 35–37, 55–57, 61, 64–65, 74, 105–106
Social origin, influence on human capital, 80–81
Solow, Robert, 30, 48, 59
South America: growth rate per capita, 58; human capital, 59; income inequality, 15
South Korea, 58
Stimulus programs, redistribution and, 121
Subsidized loans, as possible intervention in credit market, 62–63
Supply and demand: human capital and wages, 68, 69–70, 73–74, 88–89; unemployment and social charges,
 111–112. See also Demand management
Sweden: historical evolution of inequality, 22; income inequality, 14; percentage of obligatory taxes in, 101; wage
inequality, 10

Taiwan, 58
Taxes: average and marginal rates of redistribution, 100–105; capital income and, 28–32, 36–37, 43–44, 64–65;
Earned Income Tax Credit, in US, 108–109, 112; effect on disposable income, 14, 22–23; Left’s skepticism
of redistribution through, 38, 52, 76; negative income tax, 1, 3, 112–113; progressive, on estates, 19, 64;
progressive, on income, 19, 48, 64, 102–103, 106; receipts as percent of GDP, 44; redistribution methods, 31–
35, 75–76; revenue and, 106–108. See also Social charges
Theil and Atkinson indices, 10
Time and space, income inequality and, 16–17, 16t

Underemployment, 24–25, 71–72

Unemployment: efficiency wages and, 97–98; elasticity of substitution between capital and labor, 34; fiscal
redistribution and, 94–95, 109–112; inequality with respect to, 13, 23–25; skill-biased technological change
and, 71–72; unemployment insurance, 100, 104, 114–115; unions and, 90; U-shaped curve of marginal rates,
104–105
Unions: capital-labor split and, 28, 52; economic efficiency and, 92–94; monopoly power of, 94, 97; role in
setting wages, 88–90; as substitutes for fiscal redistribution, 90–92; tax reforms of 1980s and 1990s and, 38
United Kingdom: capital and labor shares of value added, 41t; changes in wages, 69; constancy of profit share,
41t, 53–55; faith in capitalism, 98; income inequality and, 14, 18, 20–25; percentage of obligatory taxes in,
101; power of unions, 91; social charges, 46–48; wage inequality, 10, 74, 77
United States: capital and labor shares of value added, 41t; changes in wages, 70; constancy of profit share, 41t,
53–55; effective minimum wage, 110–111; EITC in, 108–110; faith in capitalism, 98; income inequality and,
14, 15, 18, 22–25; marginal income tax rate, 106, 107; percentage of obligatory taxes, 101; power of unions,
91; prison population, 24; social charges, 46–48; wage inequality, 10, 11, 72–73, 74
U-shaped curve of marginal rates, 104–105, 109

Value-added tax (VAT), 44–45, 103. See also Economic value added
Verger, Daniel, 13

Wage inequality, 8–11, 9t, 11t; historical evolution of, 20–21; human capital and, 66–74; human capital and,
structural causes of, 78–88; unemployment and underemployment, 23–25; in US and UK, 53
Wage inequality, social determination of, 88–89; efficiency wages and fair wages, 97–98; monopsony power of
employers, 94–96; national traditions and, 98–99; role of unions in setting wages, 89–94
Wages: binding wage schedules, 92–93; capital-labor substitution and, 30–32; labor income and capital share,
42; as percentage of household income, 5–7, 6t. See also Wage inequality entries
Welfare state: basic income and, 113; EITC and, 108–109; limits of, 108; redistribution and intent of, 47, 105,
114. See also Social charges
Women, labor market improvements and, 87–88