AEA DISTINGUISHED LECTURE - Public Economics and Inequality: Uncovering Our Social Nature
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
AEA Papers and Proceedings 2021, 111: 1–26
https://doi.org/10.1257/pandp.20211098
AEA DISTINGUISHED LECTURE
Public Economics and Inequality: Uncovering Our Social Nature†
By Emmanuel Saez*
The standard economic model is based on people.2 Indeed, the large increase in income
rational and self-interested individuals who concentration in the United States and a num-
interact through markets, yet it is obvious that ber of other advanced economies since 1980
humans are also social beings who care about has attracted a lot of attention (see for example,
and act within groups such as families, work- Piketty 2014’s best-seller success) and figures
places, communities, or nations. prominently in the policy debate.
In the standard model, individuals care about The evolution of inequality is illustrated on
only their own consumption independently of panel A of Figure 1, which depicts the share of
social context. Taken literally, the model says income earned by the top 10 percent of adults
that a person struggling at the poverty thresh- since the early twentieth century in the United
old today gets as much utility as a successful States and France using comparable method-
professional two centuries ago when income ology. In this paper, I will contrast the United
per capita was less than one-tenth of what it is States (often described as a small government
today.1 Therefore, economic growth should beat free-market economy) with France (with a larger
inequality concerns in the long run. As Robert government and more regulated economy). Let
Lucas once put it, “Of the tendencies that are us focus first on the pretax series before taxes
harmful to sound economics, the most seduc- and transfers. Both countries show dramatic
tive, and in my opinion the most poisonous, is variations over the century. The United States
to focus on questions of distribution” (Lucas experienced a sharp—and lasting—compres-
2004). Yet, in spite of this extraordinary income sion in inequality exactly during World War II
growth, concerns about inequality and poverty at a time of deep government involvement in the
remain alive and well in our advanced econo- war economy. Around 1980, coinciding with the
mies, implying that relative positions matter to Reagan revolution, inequality started to increase
and has now reverted back to its pre-World
War II levels. France also experienced a sharp
reduction in inequality during the first half of
the twentieth century but did not embrace the
Reaganian revolution and experienced only
* University of California, Berkeley (email: saez@ a much more modest increase in inequality in
econ.berkeley.edu). This paper was presented as the AEA recent decades. This figure strongly suggests
Distinguished Lecture in January 2021, available online
at https://www.aeaweb.org/conference/2021-live-stream. that political developments play a large role in
I thank Ulrike Malmendier, Pascal Michaillat, Thomas shaping pretax inequality, over and above tradi-
Piketty, Stefanie Stantcheva, and Gabriel Zucman for help- tional economic forces of technology or global-
ful comments and discussions. Akcan Balkir provided out- ization (Alvaredo et al. 2018).
standing research assistance. I acknowledge funding from
the Berkeley Center for Equitable Growth and the Berkeley
Perhaps the most striking fact in modern
Stone Center on Wealth and Income Inequality. economies illustrating both our social nature and
†
Go to https://doi.org/10.1257/pandp.20211098 to visit concerns for inequality is the size of government
the article page for additional materials and author disclo-
sure statement.
1
The Maddison Project gathers long-term growth statis-
2
tics and shows that advanced economies experienced a more A wide literature has documented relative well-being
than tenfold increase in GDP per capita since the industrial effects (see, e.g., Luttmer 2005 and most recently Hvidberg,
id-nineteenth century (Maddison 2007).
revolution of the m Kreiner, and Stantcheva 2020 for systematic surveys).
12 AEA PAPERS AND PROCEEDINGS MAY 2021
Panel A. Top 10 percent income shares in the income and are used to fund not only public
United States and France, 1910–2018 goods needed for the functioning of the economy
50% but also a wide array of transfers back to indi-
viduals, both in cash and in kind. Even though
45%
modern economies generally allocate the fruits
40% of production to workers and owners through
35% a capitalistic market system with well-defined
30%
property rights, as societies, a significant frac-
tion of market incomes, typically between one-
25%
third and one-half, is shared (that is, effectively
20% “socialized”) through government. Panel B in
Figure 1 shows how the tax burden (including
10
20
30
40
50
60
70
80
90
00
10
20
20
19
19
19
19
19
19
19
19
19
20
20
US pretax France pretax
all taxes at all levels of government) is dis-
US posttax France posttax tributed across income groups in 2018 in the
United States and France. In both countries, the
Panel B. Average tax rates by income group in tax system is approximately a proportional tax
2018: United States versus France
that takes the same percentage of pretax income
60%
France from each group, with some progressivity over
50%
most of the distribution and some regressivity at
40% the top. In France, tax rates are almost 20 points
United States
30% higher than in the United States throughout the
20% distribution, but it is worth reemphasizing that a
10% tax rate of almost 30 percent in the United States
0% is still a significant share of the economy.
While a proportional tax does not affect
P1 –10
P2 –20
P3 –30
P4 –40
0 0
P6 –60
P7 –70
P8 –80
0 0
P9 95– 5
P9 99. –99 9
9. 9– .9
9. 9.9 9
9– 9
0
P5 –5
P9 –9
P –9
P 9 9
P9 9–9 9.9
99 9
10
inequality, government spending does reduce
P0
0
0
0
0
0
0
0
9 9
inequality substantially. Many government
transfers, such as universal health insurance or
public education, are allocated on a per-person
Figure 1. Inequality and Government basis. Others are targeted to lower earners (such
as Medicaid health insurance in the United
Notes: This figure illustrates the evolution of inequality and States). Panel A in Figure 1 illustrates the direct
the direct impact of government through taxes and transfers
in the United States and France. Panel A depicts the share
equalizing impact of taxes and government
of total national income earned by the top 10 percent of spending on inequality by depicting the top 10
adults (aged 20 and above) from 1910 to 2018 on a p retax percent income share after subtracting all taxes
and posttax basis. Income within married couples is equally and adding all transfers. This share is substan-
split. P
retax income is before taxes and excluding govern- tially below the pretax income share, especially
ment transfers. Posttax income is after subtracting all taxes
and including all government transfers and spending. Panel over the past 50 years when government is big-
B depicts the average tax rate (as a percent of pretax income) ger. Even though US inequality is almost as high
in 2018 by decile (P0–10 is the bottom decile, etc.), with now as it was in the early twentieth century on
a finer breakdown within the top decile. It includes taxes a pretax basis, it is still substantially lower on a
at all levels of government. Series for both countries fol-
low the same distributional national account methodology
posttax basis. For France, inequality on a p osttax
(Alvaredo et al. 2020). basis today is close to its a ll-time low. The total
equalizing effect of government is much wider
Sources: Piketty, Saez, and Zucman (2018), September 2020
update for the United States; Bozio et al. (2020) for France
than depicted in the figure, as public policies
such as universal public education also have a
large positive and equalizing impact on p retax
and the large direct impact it has on the distribu- incomes.
tion of economic resources. In advanced modern The standard economic model of self-
economies, we pool a large fraction of the eco- interested agents who interact through markets
nomic output we produce through government. generates efficiency under classical competitive
In the richest countries today, taxes generally assumptions. Given the technology and resource
raise between 30 and 50 percent of national constraints, there is no way to reorganize fromVOL. 111 AEA DISTINGUISHED LECTURE 3
a market equilibrium to make everybody bet- before knowing your position on the economic
ter off. The economy that arises from such a ladder—behind the veil of ignorance, as Rawls
complex web of market interactions generates (1971) famously put it. The tax and transfer
the illusion of sociality even though sociality system effectively provides such insurance.
is not part of the model. As Margaret Thatcher However, as we shall see, the redistribution done
(1987) put it for a wide audience, “Who is soci- by government is primarily in kind or targeted
ety? There is no such thing! There are individ- to specific groups and therefore quite different
ual men and women, and there are families, and from the across-the-board redistribution pre-
no government can do anything except through dicted by standard utilitarian optimal tax theory.
people, and people look to themselves first.” In this paper, I want to argue that the social
Public economists have worked primarily within nature of humans, absent from the standard eco-
this standard framework to explain (or criticize) nomic model, is crucial for understanding our
why we nonetheless observe such large govern- large modern governments and why concerns
ments in practice. Government intervention is about inequality are so pervasive. A social solu-
traditionally justified in two domains. tion arises when a situation is resolved at the
First, there can be market failures if the stan- group level—rather than the individual level.
dard competitive assumptions do not hold, for For example, providing classical public goods
example, because of externalities, market power, such as national defense funded through tax-
or asymmetric information. Such situations can ation of private incomes is a social solution.
often be addressed with a government interven- Generally, a social solution requires coopera-
tion that can restore efficiency and sometimes tion (such as obeying tax laws) and fair distri-
even create a Pareto improvement. For exam- bution of the resulting surplus (accepting how
ple, corrective taxation can be used to properly the public good will be financed through taxes).
price externalities and get the economy back to Humans have been shaped, through many mil-
market efficiency. While such government inter- lennia of evolution, to work together for the
ventions to address market failures are broadly benefit of the group. This extraordinary ability
supported by economists, they can hardly justify to cooperate and find social solutions is perva-
the very large size of governments we observe. sive even outside government, which is just the
Second, public economists justify direct most obviously visible form of social coopera-
redistribution with taxes and transfers as a way tion in our modern economies. I will illustrate
to increase social welfare, generally measured this through a number of examples taken from
as the straight sum of individual utilities—the various fields of economics: public economics,
famous utilitarian criterion. Because marginal labor economics, behavioral economics, and
utility of consumption decreases with income, lab experiments. The ideas presented here draw
redistributing resources from high earners to upon large academic literatures in many fields
low earners increases total utility, a point orig- of economics and the social sciences, of which
inally made by Edgeworth (1897). But because I know and have cited but a small subset, pri-
individuals care about only themselves, they marily the readings that have influenced me the
view taxes as reducing their gain from work most. The empirical evidence shown is volun-
and hence might work less. This creates an tarily illustrative rather than comprehensive so
equity-efficiency trade-off that is resolved by the as to present ideas in the simplest way.
optimal income tax analysis that Mirrlees (1971) In human societies, childcare and education
launched. This second aspect is more controver- for the young, retirement benefits for the old,
sial among economists because it requires intro- health care for the sick, and income support for
ducing a social welfare function that is outside those in need are to a large extent resolved at
the standard framework. In most interpretations, the social level rather than the individual level.
this social welfare function reflects the views This was traditionally done informally through
of society embodied into a government that the community and family and is now achieved
chooses the tax system. Another interpretation, through the modern social state in advanced
easier to embed in the standard framework, is economies. Even though an individual solu-
that self-interested individuals would want some tion through markets is theoretically possible, it
redistribution as insurance against the risk of does not work well in practice without signif-
being poor if such insurance could be obtained icant institutional or government help. Human4 AEA PAPERS AND PROCEEDINGS MAY 2021
societies are good at providing education, health through history to what it is today, revealing
care, and retirement and income support even our social nature. In Section II, I describe how
though individuals are not. However, such the modern social state works. In Section III, I
social aspects are quite different from a gen- emphasize that labor supply choices have a very
eral willingness to redistribute out of utilitarian large social component through social state reg-
principles. In all advanced economies, income ulations. In Section IV, I show that this social
support is primarily targeted to groups unable to nature permeates beyond government and helps
work and delivered in kind rather than in cash. explain why inequality concerns are so perva-
Income support to groups expected to work and sive. Section V concludes.
support themselves always raises concerns and
hence is generally paired with help or a push to I. How Social Redistribution Evolved to What It
find work. Societies dislike having to face the Is Today
equity-efficiency trade-off and hence try to cir-
cumvent it rather than embrace it, as standard A. Prehistory: Hunter-Gatherers
optimal tax theory posits.
The standard concern is that the social state Homo sapiens is 200,000 years old, and up
with its large taxes and transfers might discour- to 12,000 years ago, humans have been orga-
age labor supply. However, the social state also nized in hunter-gatherer societies (Harari 2014).
intentionally reduces labor supply by design Therefore, the hunter-gatherer form of social
through various regulations: child labor prohi- organization covers 94 percent of human his-
bitions and compulsory education limit work by tory and over 99.4 percent of homo history
the young, retirement benefits sharply reduce if one considers our hominin hunter-gatherer
work in old age, and overtime hours-of-work ancestors that go back at least two million years
regulations and mandated paid vacation (for (Tattersall 2012). A handful of hunter-gatherer
example, five weeks in France) reduce work societies remain to this day and have been stud-
across the board. This implies that labor supply ied extensively by anthropologists, with various
should be seen partly as a social choice, with studies analyzing specifically their economic
society having disutility of labor for the very organization.3
young, the old, and very long hours with no Hunter-gatherer societies are generally
vacation break. small—typically less than 100 people. They
Social solutions are common even outside are also fairly egalitarian, as they have minimal
government, including within private firms, private wealth and leaders with limited power.
and play a significant role in the distribution Private wealth is minimal because the land
of pretax market incomes. Almost any work and its natural resources—animals for hunting
activity requires cooperation in production that and plants for gathering—is communally held.
cannot be mediated solely through markets. Furthermore, nomadism drastically limits the
Situations with cooperation in production and accumulation of private housing, goods, and
ensuing distribution of surplus are common in tools (Woodburn 1982). Leaders have lim-
all human societies. This has shaped us to be ited power because the rank and file vigilantly
both good at cooperation and very sensitive to keep leaders from becoming despotic through
inequality. Even in modern economies, rigid a reverse dominance strategy, as hypothesized
compensation rules for distributing the fruits of by Boehm (1999, p. 105). As he summarizes it,
production are a pervasive way to resolve the “All men seek to rule, but if they cannot rule they
distributional issue. This also means that there prefer to be equal.” A good analogy familiar to
is more scope to address inequality at the p retax the reader is the modern academic department
level than economists generally believe. Pretax
distribution of production between workers and
owners has indeed been historically the place 3
Malinowski (1922), Mauss (1925), and Firth (1939)
where inequality is shaped, with government are classic studies. Economic anthropology is an active field
setting the rules of the game and hence having a of research within anthropology/sociology but with a mod-
est impact (so far) on mainstream economics. Thurnwald
very large but indirect impact. (1932) and Sahlins (1972) are influential economic studies
The paper is organized as follows. In Section I, of early societies; see Wilk and Cliggett (2007) for a modern
I discuss how government redistribution evolved textbook on economic anthropology.VOL. 111 AEA DISTINGUISHED LECTURE 5
in universities: faculty are protective of their c ollection, and cooking. Assuming that one-third
independence and do not tolerate a chair with of work is hunting, where product is pooled,
excessive power. this adds 25 percent of output to the shared
Community cooperation and sharing is com- pool.6 Similar to nations today, there is pool-
mon for many tasks. As is well known, humans ing of resources within the group but not across
were able to hunt (sometimes to extinction) big groups. Instead of pooling resources, different
and dangerous game through group cooperation groups trade goods and sometimes fight over the
(Diamond 1997). There are sharing norms for control of natural resources (Thurnwald 1932).
the produce of hunting through customs and rec- How are hunters motivated to hunt if the
iprocity rather than markets. The rationale for produce is communally shared? From an evo-
such norms is that this avoids distributional con- lutionary perspective, if cooperation gave hom-
flicts. Distribution among the full community is inids and then humans an edge, it makes sense
common and also makes sense in a setting with that motivation cannot be solely self-interested
limited storage. In contrast, gathering is an indi- and that individuals value working for the
vidual task and is typically only shared within group (Gintis et al. 2003, Henrich et al. 2004).7
the family, not the community (Thurnwald Laziness is indeed seen as an a ntisocial behav-
1932, p. 266). Therefore, there is generally ior (Thurnwald 1932). As Polanyi (1944,
norm-based sharing for cooperative work but p. 270) summarized the anthropological work of
not for individual work. Malinowski, Thurnwald, and Firth on primitive
Humans’ life cycle means that the young, old, societies, “The usual incentives to labor are not
and sick cannot support themselves and hence [individual] gain but reciprocity, competition,
need support from others. In hunter-gatherer joy of work, and social approbation.” Again,
societies, they are taken care of through a mix these motivations for work are familiar to the
of family and community support. Presence of modern academic researcher. Monetary gain is
the elderly has been documented in the fossil not our immediate motivation, while reciprocity
record among ancestors of Homo sapiens and when working in a team with coauthors often
exploded over the last 30,000 years (Caspari is. Competition with other teams to produce the
and Lee 2004). This implies that the old were most influential research is clearly a motivator.
cared for in prehistorical human societies. Joy of work is needed to be able to sustain long
They also helped with child-rearing (the grand- hours of work for many years. Social appro-
mother hypothesis proposed by Williams 1957). bation in the form of reputation among peers
Children are generally raised in village groups. clearly matters to us as well.
Children play and learn and do not start working
until adolescence.4 B. History: The Coercive State
Therefore, the pooling of economic resources
among hunter-gatherers is fairly high, proba- About 12,000 years ago, agriculture trans-
bly around 50 percent. This rough estimate can formed human societies into sedentary com-
be obtained as follows. Wealth in the form of munities that sometimes became socially
land and its natural resources is communal and stratified, as a ruling class could take control
hence so is the implicit capital income, esti- (Thurnwald 1932). In the most fertile areas—
mated to be 25 percent of output by analogy to Mesopotamia, the Nile Valley, the Yangtze River
preindustrial economies.5 Labor income, 75 per- in China—formal city-states arose about 6,000
cent of the remaining output, is generally pooled years ago. They were organized as despotic
for hunting but typically not gathering, firewood
6
This is naturally a very rough approximation in such
a vast anthropological context. Besides food production,
4
For example, Biesele and Howell (1981) analyze the there are other labor tasks, some of which are communal
contemporaneous !Kung hunter-gatherers from South West like warfare and some of which are private such as making
Africa and discuss aging and child-rearing aspects. household artifacts.
5
Piketty and Zucman (2014) and Piketty (2014) show 7
Social species have evolved to cooperate in production.
that capital income and wealth were mostly derived from In the most extreme cases, such as ants or bees, each individ-
land but were quantitatively as important as in our advanced ual works for the group, implying an almost 100 percent tax
economies in eighteenth-century France and the United rate. Humans are in between purely social species and purely
Kingdom (relative to the size of the economy). individualistic species (Wilson 2012).6 AEA PAPERS AND PROCEEDINGS MAY 2021
kingdoms, and they invented taxes and writing In coercive states, social support for the
as administrative tools for the new formal state. young, elderly, and sick shrunk down to the
Writing defines the beginning of history and was family rather than the community, although
initially invented for the administration and, in some institutions, most notably the church, did
particular, the tax administration of the state. provide some education, health care, and indi-
They also used forced labor, with various grada- gent support for those with no family support.
tions from serfdom to slavery. The goal was to For the vast majority of the population, educa-
serve and enhance the power of the state—often tion was minimal. For the wealthy few, educa-
identified with the ruler or ruling family—rather tion was generally paid for by parents. Child
than the welfare of the community. Taxes and labor became very common. People usually kept
forced labor funded the construction of mon- working in old age, even when their productiv-
umental cities and infrastructure, defense and ity—and hence earnings—fell (Minois 1989).
warfare, law and order, and the other functions When they could no longer support themselves,
of the administrative state, that is, the regalian they were generally supported by their children,
public goods. Health declined due to the risk of as savings and accumulation of wealth was lim-
relying on a single crop and infectious diseases ited to a small elite. For the few countries for
in dense populations. This made the early city- which we have data, the bottom 90 percent of
states fragile (Scott 2017). the population owned very little (Piketty 2020).
These despotic communities are much more Health care was rudimentary, and the sick gen-
unequal both politically and economically than erally relied on their families to support them
earlier hunter-gatherer communities (Boehm while they could not work.
1999). Indeed, wealth and political power are Forced labor slowly decreased and was abol-
closely correlated, as chiefs or kings rule and ished in most Western countries a few centuries
have control of the land and infrastructure.8 ago, although slavery lasted into the nineteenth
Coercive states slowly overtook the century in their colonies and in the newly inde-
hunter-gatherer and sedentary agricultural
pendent United States.
communities in the Western world as well as The level of taxes—even when including
in many parts of Asia and South America and local taxes—was typically low, less than 10 per-
became the most common form of social orga- cent of output in all western countries for which
nization up until modern democracies started we have data all the way until the beginning
emerging a few centuries ago, such as the United of the twentieth century (Piketty 2020, chapter
States and France in the late eighteenth cen- 9). Such a low level of taxation can only fund
tury. Democratization itself is a slow process, regalian public goods (administration, law and
as many groups are initially excluded. Women order, defense, infrastructure) but not the social
could not vote until 1920 in the United States state. Furthermore, the tax system was typically
and 1945 in France. Many countries excluded regressive, as taxes were either taxes on con-
the poor from voting with poll taxes or literacy sumption, rudimentary poll taxes and property
or property ownership requirements. Emerging taxes on real estate, or flat taxes on agricultural
democracies could also be extremely coercive output (Ardant 1971).
against specific groups, either internally, such Throughout this long history, the challenge of
as Native Americans as the frontier expanded despotic states is to extract revenue to enhance
or enslaved African-Americans until 1865, or the power of the state without generating tax
externally through colonies—the French and revolts (Webber and Wildavsky 1986). As Jean-
British empires undemocratically ruled over Baptiste Colbert, the finance minister of Louis
about 30 percent of the world population in the the XIV of France, put it, “The art of taxation
early twentieth century (see Piketty 2020 for an consists in so plucking the goose as to procure
extended analysis). the largest quantity of feathers with the least
possible amount of hissing.” Taxes were low
because enforcement was difficult in a mostly
8 informal economy, and tough enforcement could
To this day in autocratic states, wealth and political
power are often closely aligned, as the autocrat has the backfire in the form of tax revolts. This view of
power to redistribute wealth toward himself and his family taxes funding a coercive state still has resonance
(Acemoglu and Robinson 2012). today among libertarians, particularly in theVOL. 111 AEA DISTINGUISHED LECTURE 7
United States, which they view as a democracy Panel A. The rise of the fiscal state
founded through a tax revolt against the British 60
Tax revenue/national
crown: “no taxation without representation.”9
income (percent)
50
40
C. The Rise of the Social State in the
30
Twentieth Century
20
An extraordinary transformation of our soci- 10
eties took place during the twentieth century. 0
In advanced economies, the size of the gov-
18 0
80
19 0
19 0
10
20
19 0
19 0
19 0
19 0
19 0
80
20 0
20 0
10
18
7
9
0
3
4
5
6
7
9
0
20
18
18
19
19
19
ernment measured by tax revenue to national
Sweden Germany United States
income increased from less than 10 percent to
France Britain
levels between 30 and 50 percent.10 Panel A in
Figure 2 (taken from Piketty 2020) illustrates Panel B. The rise of the social state in Europe
this for a few countries—the United States, the 50
Percent of national income
United Kingdom, France, and Sweden. In all 6%
these countries, the tax ratio was low (below 10 40 5%
percent) and flat until World War I, increased 9%
30
until around the late 1970s, and then has been
roughly stable thereafter. The exact timing of the 20
11%
tax increases and the final level of the ratio differ 6%
across countries, with France and Sweden stabi- 10
1% 2%
10%
lizing around 50 percent and the United States 6% 8%
around 30 percent (and the United Kindgom
0
18 0
18 0
19 0
00
10
20
30
19 0
19 0
60
70
19 0
90
00
10
around 40 percent).
7
8
9
4
5
8
20
19
18
19
19
19
19
19
20
Panel B in Figure 2 (also created in Piketty Other social spending Retirement + disability benefits
2020) depicts the evolution of the composition Cash social transfers Education
of government spending (relative to national Health care Regalian public goods
income) in Europe (the average of France,
Germany, Sweden, and the United Kingdom) Figure 2. The Rise of Government in the Twentieth
since 1880. Up to the early twentieth century, Century
the bulk of government spending was devoted to
regalian public goods, as discussed previously. Notes: Panel A depicts total tax revenue as a percent of
national income by decade in five countries. Taxes at all
The growth of government over the twentieth levels of government are included. This is reproduced
century is almost entirely due to the growth from Piketty (2020, figure 10.14) and updated so the last
in the social state that provides education and point is 2018. Panel B depicts the composition of govern-
childcare support for the young, health care for ment spending by decade in Europe (average for Germany,
France, Britain, and Sweden). This is reproduced from
the sick, retirement benefits for the old, and an Piketty (2020, figure 10.15). Regalian public goods include
array of income support programs for groups in defense, law and order, administration, and infrastructure.
need, such as the disabled, unemployed, or poor. Cash social transfers include unemployment benefits, family
Essentially, the social state provides support benefits, and m eans-tested benefits. Other social spending
for those who cannot provide for themselves includes in-kind spending such as public housing.
because they are young, old, sick, or otherwise Sources: Piketty (2020, figures 10.14 and 10.15), created
from OECD statistics and earlier historical statistics
9
As we mentioned above, America was at first an incom- unable to earn a living. Let us review the histori-
plete democracy. Before then, American colonies had a lot cal development of each pillar in turn.
of autonomy in setting their tax systems with regressive
taxes in the South but incredibly progressive taxes—for the Education.—Tax-funded and compulsory
time—in the North (Einhorn 2006).
10
Flora (1983) gathers the most extensive historical sta- education is the pillar of the social state that
tistics, and OECD (2020c) covers the contemporary period develops first. The motivation is in part nation
since 1965. building and hence fits within the framework8 AEA PAPERS AND PROCEEDINGS MAY 2021
of the coercive state building its power. But the Panel A. School enrollment at ages 5–14, 1830–1930
social aspect of universal human development 100%
and opportunity is also part of the motivation.
Prussia and Austria were the first to adopt 80%
compulsory schooling in the mid-eighteenth 60%
century (Van Horn Melton 1988). America was United States
also a precursor in mass education but follow- 40% Slavery and school US Blacks
ing a decentralized process (instead of nation 20%
prohibition end Prussia
France
building). Massachusetts was the first US state
to enact compulsory education in 1852 (and had 0%
0
0
0
0
0
0
0
0
10
20
30
already achieved a high level of schooling when
3
4
5
6
7
8
9
0
19
18
18
18
18
18
18
18
19
19
19
it was still a British colony). Mississippi was
the last state to adopt compulsory schooling in Panel B. Primary school enrollment in Russia,
1917. Korea, and Indonesia Korea compulsory
Historically, mass education is always gov- primary education
100%
ernment driven through a combination of gov-
ernment funding (at all levels including higher 80% Russia compulsory
education) and compulsory schooling (for pri-
primary education
60%
mary and then secondary education). Meyer,
Ramirez, and Soysal (1992) provide an empir- 40% Indonesia
compulsory
ical analysis of the development of mass educa- 20% primary education
tion through nation-states.
Let us illustrate this in Figure 3. Panel A in 0%
Figure 3 shows that in the early nineteenth cen-
0
0
10
20
30
40
50
60
70
80
90
9
0
19
18
19
19
19
19
19
19
19
19
19
tury, Prussia and the United States already had
school enrollment rates at age 5–14 of around Figure 3. The Rise of Mass Education
two-thirds, substantially higher than other coun-
tries. France caught up slowly over the nine- Notes: Panel A depicts the fraction of children aged 5–14
teenth century. Slavery in America also imposed enrolled in school (public or private) in the United States,
school prohibitions on the enslaved so that the among Blacks in the United States, in Prussia, and in France
from 1830 to 1930. Enslaved children in the United States
school enrollment rate of Black children was were prohibited from attending schools. Panel B depicts the
minuscule before the Civil War. After the Civil fraction of children enrolled in primary school (public or pri-
War ended, Black children enrollment rates shot vate) in Russia, Korea, and Indonesia from 1890 to 1990 and
up to over 50 percent by 1880, one of the fast- flags when compulsory primary education was introduced in
each country. In each of these three countries, compulsory
est increases ever seen, showing the enormous schooling leads to a large increase in primary school enroll-
power of the state in restricting or promoting ment in the following years.
education. Black children enrollment increased
Sources: Panel A: Lindert (2004) and US Census Bureau
more slowly in the Jim Crow period of discrim- (1975, series H435); panel B: Lee and Lee (2016)
ination after 1880 and did not catch up to white
enrollment until after World War II.
Panel B depicts various twentieth century
experiences showing that compulsory schooling
can have dramatic impacts quickly. The most century, also had very proactive education pol-
extreme case is Soviet Russia, which achieved icies (with mass education likely fueling eco-
quasi-universal enrollment by 1940 from a level nomic growth, Easterlin 1981). Panel B shows
of around 40 percent in 1920. Education was a sharp increase for Korea in the late 1940s and
a way to both indoctrinate young minds and Indonesia in the 1970s after they finally adopted
develop the economy. The communist expe- universal compulsory schooling (but starting
riences mix the coercive state—perhaps in its from preexisting high bases of 60 percent). Mass
most extreme form under Stalin in the Soviet education requires mobilization to build schools
Union and in North Korea still today—and and train new teachers, a process that takes years
the social state. Asian countries, which experi- if not decades after enactment (see Duflo 2001
enced fast economic growth over the twentieth for an analysis of the Indonesian experience).VOL. 111 AEA DISTINGUISHED LECTURE 9
Retirement.—Retirement support is the pil- Income Support.—Income support pro-
lar of the social state that generally develops grams have a long history of tension between
second. Retirement programs first developed offering support to those in need and concerns
privately through negotiations between large about discouraging work. As a result, many
employers and unions, an indication that work- programs start as narrow programs, targeting
ers need institutional help to provide for their groups deemed unable to support themselves
retirement.11 Public retirement programs some- such as widows with children (at a time when
times grew out of such private retirement pro- mothers were not expected to work outside the
grams through a general mandate. Germany household), the elderly or the disabled (before
was the first country to introduce such a general retirement programs existed), or the unem-
mandatory public retirement system in 1889 ployed. Programs supporting those out of work
under Otto von Bismarck. Public retirement often had a “workfare” component, requiring
programs sometimes developed as a retirement recipients to do some work for the government
benefit aside from or on top of private retirement to receive support, such as in the Poor Laws in
programs such as Social Security in 1935 in the England (Lindert 2004).
United States (Costa 1998). A large literature has discussed why the social
Before public retirement programs existed, state emerged in the twentieth century. It is likely
a large fraction of the elderly was working (80 that democratization replacing earlier coercive
percent of men aged 65 or older were gainfully authoritarian states led the population to demand
employed in the United States in the late nine- help through the social state (see for example,
teenth century, see Figure 5). The elderly who Acemoglu and Robinson 2000 and Lindert
could no longer work enough to support them- 2004). Looking at history (since the beginning
selves had to rely on family support. Public of writing), the modern social state appears as
retirement systems were a way to provide social a unique and very recent historical develop-
insurance through the state instead of relying on ment (150 years old), while the coercive state
self-insurance or family insurance. is the norm throughout history (the preceding
3,000–5,000 years), with a number of autocratic
Health Care.—Like retirement benefits, states still existing to this day. However, extend-
health-care benefits start with private arrange- ing the time frame back into our long prehistory
ments between employers and employees and its social communities, the modern social
that then get mandated by the state. The ear- state starts to look more like a homecoming—
liest program started again in Germany in adapted to the modern world and economy, to
1883 under Bismarck to cover workers. Such be sure—rather than a radically new, and hence
employer-related systems developed in most perhaps fragile, development.
Western countries in the early twentieth century.
Universal health insurance expanding coverage II. The Four Pillars of the Social State Today
to nonworkers started after World War II, first
with the United Kingdom launching its National The social state raises a puzzle for the stan-
Health Service in 1948, then spreading quickly dard economic model. Rational individuals in a
among almost all advanced economies, the market economy with functioning credit markets
United States being the notable exception. In the should be able to largely manage on their own.
early twentieth century, health care was a small The young (or their parents) can borrow to pay
fraction of the economy. But today, health care for their education if this is a worthy investment.
is about 10 percent of the economy in OECD Health care is largely a private good for which
countries and about 18 percent in the United people can buy insurance. Workers can save
States (OECD 2020b). Therefore, health care is for their retirement, anticipating that their work
now a very large component of the social state ability will decline with age. Finally, people can
(Figure 2, panel B). also dip into their savings whenever they face a
temporary income loss such as unemployment.
Economists traditionally justify social insur-
11
The Laibson (2018) AEA Richard T. Ely Lecture
was devoted to this phenomenon of private paternalism of ance by focusing on market failures such as
employers on behalf of their employees from a behavioral asymmetric information leading to credit fail-
economics perspective. ure or insurance failure. However, the resulting10 AEA PAPERS AND PROCEEDINGS MAY 2021
solutions—guaranteed student loans so that tax funding is approximately proportional to
anybody can borrow for education, mandates income. Education also has enormous indirect
for health insurance to deal with adverse-selec- redistributive effects by giving opportunities to
tion death spirals, or mandatory annuitization of succeed economically to children from disad-
retirement savings to make sure retirees never vantaged backgrounds.
run out of savings—are much lighter interven- Higher education is not (yet) compulsory, but
tions than what actual social states do. More its capacity is built through public universities,
radically, the field of behavioral economics has as private universities serve only a small frac-
shown that individuals fail to behave as in the tion of students in advanced economies (OECD
standard model, particularly in circumstances 2020a). Even in the United States, where private
that involve the time dimensions that are crucial higher education is large, three quarters of stu-
for education investment, health insurance, or dents attend public institutions (National Center
old-age or buffer stock savings (see Thaler 2015 for Education Statistics 2020). Therefore, the
for a description of the emergence of this young government essentially controls the supply side.
field). Higher education is also highly subsidized, as
Looking back at the deep history we have tuition costs paid by students are only a fraction
sketched in the previous section, it is easy to see of the real costs of higher education (and in many
the common theme between community support countries in Europe such as France, tuition costs
of hunter-gatherer societies, family support in are almost zero). Even though higher education
coercive state societies, and the social state of is an individual choice, at the aggregate level,
modern social democracies. It is worth review- it looks like a government choice. One striking
ing briefly the current structure of the four pil- example comes from the US GI Bill after World
lars of the social state to understand its logic and War II, which paid for the higher education of
why the standard economic model solution does veterans and dramatically boosted college edu-
not work well in practice. cation of men (using women as a control group)
(Stanley 2003).
A. Education Why not have students or their families pay
directly for education or borrow with student
As discussed above, mass education of mod- loans as a standard economic model would rec-
ern times is always government driven through ommend? The experience of student loans in
a combination of compulsory schooling (for the United States shows that they become an
primary and then secondary education) and gov- unbearable burden for a significant fraction of
ernment funding (at all levels including higher borrowers. There are several reasons for this.
education), generally taking the form of direct Attending college is not a guarantee to get a
government provision of schools and teachers. degree, let alone a well-paying job afterward.
In OECD countries, education is always primar- Furthermore, handling debt is challenging for
ily government funded and often overwhelm- many, as behavioral economics has shown. This
ingly so (OECD 2020a). forces the government to provide relief e x post
If education is compulsory, government (Baum 2016).
funding must follow, as low-income families Aside from funding, why not have for-profit
would not be able to afford it at full cost. This education instead of having public institutions
is because education requires highly qualified provide education? The United States has
labor—teachers. This remains the case today indeed experienced a surge in for-profit higher
as it was in the nineteenth century, as teachers education institutions as state funding for pub-
are always skilled workers and the technology lic institutions has retreated. The evidence sug-
of education has not changed much. If a teacher gests that students can be lured into high-cost,
is paid three times the average working-class low-quality for-profit schools (Deming, Goldin,
earnings and can teach 20 students, the cost and Katz 2012). This problem is exacerbated
per school-age child is 300/20 = 15 percent of when f or-profit schools put more weight on prof-
working-class earnings, which is prohibitive for its than service after take-up by private equity
many families, especially large ones. Therefore, (Eaton, Howell, and Yannelis 2020). This is a
mass education is highly redistributive on a striking failure of the market that arises because
direct basis: every child gets an education, while students are not able to assess perfectly the valueVOL. 111 AEA DISTINGUISHED LECTURE 11
and cost of education offers (Akerlof and Shiller standard life cycle savings model, need a heavy
2015 develop this aspect of “consumer failures” institutional hand to steer individuals in the right
more generally). direction.
In sum, education is largely decided at the Therefore, the problem of retirement is also
social level, not the individual level. resolved at the social level, not the individual
level.
B. Retirement Benefits
C. Health Care
The elderly lose their ability to work and
hence support themselves with their earnings. All advanced economies provide universal
The standard economic solution is that workers health insurance, with the United States being
should save for retirement (the famous life cycle the unique exception in leaving about 10 per-
model of Modigliani). A large body of work in cent of its population uninsured. Health-care
behavioral economics shows that, in contrast costs have become large (10 percent of national
to what the standard model posits, individu- income on average in the richest countries)
als are not able to save on their own and invest due to enormous progress in medicine (OECD
wisely (see Thaler and Sunstein 2009 for an 2020b). Hence, tax funding is the norm, as low-
overview). Indeed, before retirement programs er-income families would not be able to afford
existed, family (or community) support—not the full cost. Therefore, universal health insur-
saving—was the main source of support, as dis- ance creates significant redistribution by income
cussed previously. and also, of course, by health and health-risk
Public retirement programs are mandatory status.
and funded by taxes. They are typically intro- One important question is why health-care
duced as pay-as-you-go systems where taxes quality is the same for all in such universal
on workers immediately pay for the pensions health-care systems (at least as a principle, not
of retirees, replacing the former p ay-as-you-go always realized in practice). Why isn’t health
family-based system where children take care insurance offered in grades, with cheap insur-
of their elderly parents. Instead of having to ance covering only the most cost-effective treat-
support their elderly parents, adult children pay ments? Probably because humans are willing to
taxes to fund retirement benefits. Social insur- spend a lot of resources to save a specific life,
ance also allows for pooling risk much more that is, an actual person with a condition that can
effectively than family insurance. be treated.12 This is likely a consequence of our
Benefits are generally related to lifetime social nature shaped by evolution: taking care
earnings so that the public retirement system of the sick or injured was helpful for group sur-
is generally not highly redistributive from a vival. This makes withholding treatment to the
lifetime perspective (see Brown, Coronado, poorly insured socially unbearable. In the United
and Fullerton 2009 for a US analysis), but it is States, hospitals are obliged to offer emergency
highly redistributive from a c ross-sectional per- care to all patients, even those without insurance
spective: elderly retirees with no earnings get and unable to pay. 13 Of course, on top of this,
support from workers with earnings. adverse selection provides a strong standard
Even the most radical privatization rationale for mandating health insurance.
reforms—as in Chile—maintain mandatory Even in the United States, where private insur-
contributions. Even private employer pension ance covers slight more than half of the popula-
plans are either mandatory (such as the tradi- tion, it is primarily offered through employers,
tional defined-benefit employer pension plans which are now mandated to offer it (if they have
in the United States) or highly encouraged
through enormous price incentives (such as the
401(k) employer matches in the United States) 12
Economists have noted that societies are willing to
or defaults (automatic 401(k) plan enrollment spend a lot more resources to save an actual life than to save
upon hiring, which has an enormous impact on a statistical life (such as reducing accident risk through bet-
ter safety). See, e.g., Gruber (2016, chapter 12).
participation even in the medium term, Madrian 13
In practice, health insurance cost in US private insur-
and Shea 2001). Therefore, retirement programs, ance varies primarily based on deductibles and copays and
even when they come closest to mimicking the not quality of care.12 AEA PAPERS AND PROCEEDINGS MAY 2021
50 or more full-time employees). This man- (formerly food stamps) benefits a single adult
date is economically equivalent to forcing each with no dependent in the United States can get
insured worker to pay the full cost of the insur- (the only form of quasi-cash means-tested trans-
ance regardless of earnings. This is a crushing fer available to this group).
and unbearable burden for low-paid and insured Therefore, there is a widespread social view
workers (Saez and Zucman 2019 describe it as that people who are expected to work and sup-
a privatized poll tax, and Case and Deaton 2020 port themselves should not be supported by the
discuss the labor market impacts). community. Everywhere, there is strong social
A recent literature has also shown strong reprobation against “free loaders” who could
evidence of behavioral effects in health-care work and support themselves but decide to live
choices, particularly in the US context where off government support (Lindert 2004 provides
choice is most extensive (see Chandra, Handel, historical context; Saez and Stantcheva 2016
and Schwartzstein 2019 for a recent survey). present survey evidence). This is why income
People make mistakes in health-care utiliza- support is concentrated among groups unable
tion and treatment choices. Copayments and or unexpected to work, such as the unemployed,
deductibles lead consumers to reduce demand the disabled, and the elderly.
for high-value care. This may explain why uni- This feels very different from a utilitarian
versal health-care systems have low copays optimal tax and transfer system that explicitly
and deductibles and why health-care decisions trades off equity (redistribution to the poor) with
for patients are made primarily by health-care efficiency (some loss of output due to reduced
professionals. Like for education, the difficulty incentives to earn; see Piketty and Saez 2013 for
for users to understand and navigate health-care a survey). Instead, income support tags recipi-
choices implies that the market does not nec- ents who are not expected to work to avoid the
essarily deliver efficiency. In sum, the problem equity-efficiency trade-off (Akerlof 1978).
of health care is also primarily resolved at the But our innate aversion to free loaders can
social level rather than the individual level. be interpreted as a way to detect the presence
of behavioral responses and adjust redistribution
D. Income Support with better targeting or a better design (more
incentives to work, help finding work, or less
Income support programs are obviously a generous benefits). In other words, the public
social-level form of help for people in need. processes efficiency costs through a fairness lens
They are targeted to specific groups such as (“free loaders take advantage of the system,”
the unemployed with unemployment benefits, “beneficiaries would be destitute without help”).
the disabled with disability benefits, the elderly Stantcheva (2020) shows, indeed, that distri-
poor with minimum o ld-age benefits, and chil- butional and fairness considerations are more
dren in poor families (with family benefits such important than efficiency considerations when
as the refundable tax credits and traditional the public reasons about taxes and transfers.
welfare for single parents in the United States).
Unconditional means-tested support is generally E. Social Group Scope
modest and most often in kind (such as hous-
ing or nutrition support) and combined with While we have seen that pooling of resources
job-training help. For example, even in France, through taxes and transfers is very large at the
with its generous social state, a single adult with level of the nation (and subnational govern-
no resources would receive €560/month in cash ments in decentralized countries), it is striking
as of 2020 (Revenu de Solidarité Active).14 This how small transfers are across countries. Direct
cash amount is only about 17 percent of average foreign aid from rich countries toward develop-
income per adult in France. But it is still over ing countries is modest (around 0 .2 percent of
three times as high as the meager $200/month GDP in the United States and always below 1
in Supplemental Nutrition Assistance Program percent anywhere else). It is targeted to crises
(such as disaster relief), security (defense spend-
ing in the context of alliances), and development
14
This can be combined with various in-kind support for aid (a substantial portion of which runs through
housing, utility costs, or public transportation. international organizations such as the WorldVOL. 111 AEA DISTINGUISHED LECTURE 13
Bank or the International Monetary Fund). In groups less attached, such as the elderly (Gruber
US public opinion polls, foreign aid is often the and Wise 1999), secondary earners—especially
least popular item of government spending (for when women were less attached to the labor
example, Pew Research Center 2019). Even in force (Blau and Kahn 2007)—or single parents
the European Union, an old and deep commu- (Meyer and Rosenbaum 2001).
nity of independent nations, the common bud- Taking a broader view though, social deter-
get is only 1 percent of the EU economy (and minants of labor supply become readily visible.
hence minuscule relative to the government First, the social state has indeed reduced labor
budgets of each nation member). Direct trans- supply along various dimensions (youth labor,
fers across EU countries—even in the context of old-age labor, and long hours), but it has done
severe crisis, such as the financial crisis of 2009 so intentionally by design and regulation and
when Germany directly bought Greek public not as an unintended consequence. Second,
debt—quickly become controversial. social norms also affect labor supply (for exam-
It has also been noted that the social state is ple, whether mothers should work outside the
smaller in countries fractionalized along ethnic household).
or religious lines (see Alesina and La Ferrara To give a broad-level view, Figure 4 depicts
2005 for a survey on the economic effects of the employment rate (people working divided
fractionalization). For example, Alesina and by population) by five-year age bins in the
Glaeser (2004) argue that this is the main reason United States and France in 2019 using OECD
why the United States has a smaller social state statistics for men in panel A and women in panel
than European countries. Alesina, Miano, and B.15 Employment rates are similarly high in
Stantcheva (2018) show through surveys that both countries in prime age (25–59) and actu-
the public dislikes redistribution toward immi- ally slightly higher in France for women. But
grants and that emphasizing the presence of employment rates are substantially lower among
immigrants reduces support for redistribution. the young and old in France. This strongly sug-
All this evidence shows that the scope of the gests that differences in labor supply along the
social group matters greatly. Humans are will- extensive margin in France versus the United
ing to pool resources with the social group they States are driven by education and labor regu-
identify with but typically not others, another lations or social norms regarding work among
striking piece of evidence demonstrating our the young, retirement decisions for the old, and
social nature (as opposed to universal utilitarian perhaps differences in family norms or policies
principles). The nature of the group can vary regarding female market work, not by the taxes
depending on situations and is also malleable. needed to fund the larger social state in France.
For example, the rise of the social state was in Let us examine each in turn.
large part replacing family support by nation-
state support in specific domains. Youth Labor.—Child labor was prevalent
before the rise of the social state and mass
III. The Social State and Labor Supply education (Basu 1999 provides a survey from
a developing country perspective). It was
The main critique leveled at the large mod- common in most US states up until the Great
ern social state is that it might discourage work Depression (Moehling 1999). US census data
and hence depress economic activity. Indeed, show that almost 20 percent of children ages 10
in the standard economic model, labor sup- to 15 worked for pay at the end of the nineteenth
ply reduces utility, and the sole motivation for century, and this declined in the early part of
work is individual gain. Therefore, taxes and the twentieth century (US Census Bureau 1975,
transfers can reduce labor supply. An enormous series D80). Compulsory education mechani-
empirical literature has shown that this model cally reduced child labor supply. But labor reg-
has indeed some relevance (see Pencavel 1986 ulations were also enacted to further reduce and
and Blundell and Macurdy 1999 for classic
surveys). Estimated labor supply responses are 15
This follows Blundell, Bozio, and Laroque (2013),
generally—but not uniformly—fairly modest who present such highly insightful graphs for the United
for groups strongly attached to the labor force States, the United Kingdom, and France at the annual level
(such a prime-age males) but can be large for (instead of five-year age bins).You can also read
