The Past, Present, and Future of Economics: A Celebration of The 125-Year Anniversary of The JPE and of Chicago Economics

The Past, Present, and Future of Economics:
A Celebration of the 125-Year Anniversary

of the JPE and of Chicago Economics
Introduction
John List
Chairperson, Department of Economics
Harald Uhlig
Head Editor, Journal of Political Economy
The Journal of Political Economy is celebrating its 125th anniversary this

year. For that occasion, we decided to do something special for the JPE
and for Chicago economics. We invited our senior colleagues at the de-
partment and several at Booth to contribute to this collection of essays. We
asked them to contribute around 5 pages of final printed pages plus ref-
erences, providing their own and possibly unique perspective on the var-
ious fields that we cover.
There was not much in terms of instructions. On purpose, this special
section is intended as a kaleidoscope, as a colorful assembly of views and
perspectives, with the authors each bringing their own perspective and
personality to bear. Each was given a topic according to his or her spe-
cialty as a starting point, though quite a few chose to deviate from that,
and that was welcome. Some chose to collaborate, whereas others did not.
While not intended to be as encompassing as, say, a handbook chapter,
we asked our colleagues that it would be good to point to a few key papers
published in the JPE as a way of celebrating the influence of this journal
in their field. It was suggested that we assemble the 200 most-cited papers
published in the JPE as a guide and divvy them up across the contribu-
tors, and so we did (with all the appropriate caveats). Some chose to fea-
[ Journal of Political Economy, 2017, vol. 125, no. 6]

© 2017 by The University of Chicago. All rights reserved. 0022-3808/2017/12506-0037$10.00
1723
This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
1724 journal of political economy
ture these centrally, some picked a few, and others ignored that list alto-
gether. Some have chosen to provide an overview, some have chosen to
give their perspective on the field or added a wish list of their own, some
have zoomed in on one contribution in particular, and one of the authors
actually added a new theorem, generalizing a previous key contribution.
There was a large canvas to draw on, and we are pleased that our col-
leagues have explored all its corners, in their own and unique way, in the
spirit of the multifaceted mosaic that our faculty represents within the
subject of economics.
We asked that their contribution be about what the field has accom-
plished or about where the field might or should be going in the future.
It is probably the nature of the beast that all chose a largely backward-
looking perspective, providing an overview of how the field has devel-
oped over time and how the JPE helped this process along by publishing
some of the key ideas and key contributions. But hop on board and start
reading! Because of the intended density of the summary, it then actually
becomes quite a forward-looking piece. The speed at which the ideas fly by
makes it hard to stop that train of thought at the present time and fron-
tier: one immediately yearns to proceed and to do and learn more, once
this flow has stopped. This then is the ultimate goal: to encourage new
thinking and excellent, thought-provoking, and paradigm-shifting research,
which will pave the way for a future and better understanding of economic
phenomena and to have the JPE continue to participate in publishing those
key contributions. When we were PhD students or young researchers, we
wish we had had access to a dense overview of this sort. If you belong to
that group, or if you are simply young at heart, and seek to read a research-
stimulating piece, then this is our present to you.
There is no good and certainly no linear way of ordering these contri-
butions. Printing technology still demands that the pieces be presented
in some order: so here is ours, with all due apology. We have chosen to
open with the most general, most bird’s-eye, and most long-run of per-
spectives to then gradually “zoom in” toward the deep foundations and
the essays discussing assignment problems as well as experimental ap-
proaches. We can hear the outcry already: wait, isn’t economics all about as-
signing goods to agents? We share that concern. The reader is thus en-
couraged to browse, pick, read, and reorder to her or his heart’s content.
Indeed, what could possibly be more general than general equilibrium
theory? Hugo Sonnenschein reflects on the role of “Chicago and the Or-
igins of Modern General Equilibrium” in his essay, juxtaposing the Arrow-
Debreu Walrasian analysis to the excess demand function approach and
discussing the role of preference orderings. Ufuk Akcigit writes about “Eco-
nomic Growth,” its crucial impact on the welfare of nations, and the cru-
cial role of innovations for understanding the growth process itself. David
Galenson complements this with the “Economic History” perspective, ex-

past, present, and future of economics 1725
amining in particular the role of human creativity for innovations. He ar-
gues that conceptual innovators are young, while experimental innovators
are, shall we say, experienced. Long-run growth requires the assignment of
property rights and effective institutions for contract enforcement, issues
intimately tied to the development of political institutions and political
decision making, as Roger Myerson discusses in his essay on “Political
Economics in the Journal of Political Economy.” A crucial aggregate di-
mension of political choice is “Aggregative Fiscal Policy,” as summarized
by Nancy Stokey, highlighting in particular the role of government debt
financing.
Zooming in a bit more on the shorter-run aggregate fluctuations and
the international trade dimension is the topic in “Business Cycles and In-
ternational Trade” by Harald Uhlig, emphasizing the role of rational ex-
pectations in particular. Greg Kaplan points out that the richer aspects
of “Inequality, Heterogeneity, and Consumption” are “today . . . front and
center in macroeconomics,” building up to the analysis of general equi-
librium models with heterogeneous households and aggregate shocks.
These aggregate shocks expose households to risks, which need proper
analysis with “Time-Series Econometrics in Macroeconomics and Finance,”
as Lars Hansen highlights, emphasizing in particular the role of consump-
tion, permanent income, and asset pricing. The asset pricing aspects are
then further elaborated on in the essay on “Asset Pricing: Models and
Empirical Evidence” by George Constantinides, highlighting the role of
preferences, and the piece by Eugene Fama on “Finance at the University
of Chicago,” narrating the transition to modern finance and the Chicago-
led research on market efficiency, factor models, and the economics of
organizations. Richard Thaler offers a different approach to thinking about
financial markets, savings, and consumption in his discussion of “Behav-
ioral Economics.” He contrasts it with “price theory à la Chicago School
led by the intellectual giants Gary Becker, Milton Friedman, and George
Stigler,” which indeed plays a central role in many of the other contribu-
tions here. He then questions whether the stock market can add or sub-
tract, why there are noise traders, and why optional but then mandatory
savings plans can help to increase savings.
The issue of “Corporate Finance” receives a more detailed treatment
by Robert Vishny and Luigi Zingales in their contribution, organizing it
around the Modigliani-Miller irrelevance proposition, agency costs, and
the market for corporate control. Particularly important for all these fi-
nancial dimensions is the banking and monetary system. Thus, Douglas
Diamond, Anil Kashyap, and Raghuram Rajan write about “Banking and
the Evolving Objectives of Bank Regulation,” emphasizing the importance
of liquidity provision and aggregate liquidity shortages for thinking about
financial regulation. Liquidity is the topic at the heart of “Monetary Eco-
nomics,” which Fernando Alvarez summarizes, organized around the three

traditional functions of money. Robert Lucas reflects on his “Memories of
Friedman and Patinkin,” their perspectives on quantity theory as well as its
juxtapositionorlackthereoftothe Keynesianparadigmand its liquidity trap
unemployment equilibrium.
This is naturally followed by a more in-depth examination of “Labor
Markets” by Robert Shimer, taking a mostly macroeconomic perspective
on the Phillips curve, search, labor supply elasticities, sectoral shifts, as
well as contractual issues and their frictions. The more microeconomic
perspective is the topic of “Chicago Labor Economics” by James Heckman,
emphasizing the rigorous interplay between data and theory. He formu-
lates the guiding principles of Chicago tradition that “theory is used to
interpret data. Data are used to test theory. Understanding the mecha-
nisms . . . is essential.” This interplay between data and theory is likewise
the key for Stephane Bonhomme and Azeem Shaikh, arguing to “Keep
the ECON in Econometrics,” an intended pun on Leamer’s admonition
to take the con out of econometrics. Showcasing that interaction, they
discuss structural econometric models of the labor market, the marriage
market, and partial identification. Derek Neal returns to the finer theoret-
ical details of the labor market for understanding “Life Cycle Wage Dynamics
and Labor Mobility,” examining issues of life cycle investment in general
human capital, life cycle wage growth, and the search for good matches.
This naturally leads to the examination of “The Human Capital Approach
to Intergenerational Mobility” by Magne Mogstad, giving central impor-
tance to two papers by Becker and Tomes, the first of which was published
in the JPE. Human capital also plays a central role in Robert Topel’s essay
on “Health Economics,” taking the perspective that health can be viewed
as human capital. He notes that “net of medical expenditures, the value
of increased life expectancy between 1970 and 2000 in the United States”
equaled “a flow of about $2 trillion per year,” thus emphasizing the quan-
titative importance of this topic. Providing a different perspective on the
issues arising in labor market contracts or contractual issues more gener-
ally is the aim of “Agency Issues” by Canice Prendergast. He starts from a
theoretical perspective, relating pay to performance, points out that
“most people do not get paid this way,” and then moves to discuss other
motivations such as career concerns. Effectively echoing the guidelines
by James Heckman, Stephane Bonhomme, and Azeem Shaikh, he be-
moans the “paucity of empirical work” to complement these theoretical in-
sights. Long-term contracts often are solutions in environments, featur-
ing asymmetric information and associated incentives. “Information
Economics” is the theme of Emir Kamenica’s essay. He points out that ad-
ditional issues such as information acquisition and communication arise
and that information design is poised to complement mechanism design,
when thinking about arrangements between agents.

Turning from labor markets to goods markets, Michael Greenstone traces
“The Continuing Impact of Sherwin Rosen’s ‘Hedonic Prices and Implicit
Markets: Product Differentiation in Pure Competition,’” emphasizing the
power of the paradigm as well as the substantial empirical challenges in uti-
lizing it, met with increasing, but as-of-yet incomplete, success by the
“credibility revolution” of exploiting quasi-experimental variation, start-
ing in the late 1990s. Price systems ultimately serve the goal of assigning
goods to individuals. As Phil Reny shows in his piece on “Assignment
Problems,” these can be quite thorny, in particular if no monetary transfers
are allowed or if individual preferences over goods exhibit complemen-
tarities. For that, he discusses an important result by Eric Budish, pub-
lished in the JPE in 2011, and then extends it to allow for arbitrary pref-
erences and both divisible and indivisible goods, providing a new theorem
and its proof. A particularly important mechanism of assigning goods to
buyers is auctions. Ali Hortaçsu therefore summarizes the literature on
“Auctions in the Journal of Political Economy, 1894 –2017.” He highlights
the literature on the auctioning of government land, the wireless spec-
trum, Treasury securities, and electricity, pointing to the issues arising
from complementarities and potential collusion. A decidedly less desira-
ble procedure for assigning goods results from criminal activity. Thus, the
underlying economic motives as well as methods for deterrence are the
focus in “The Economics of Crime,” described by Steven Levitt. He empha-
sizes how that literature has taken an empirical turn in recent years, ex-
ploiting natural experiments, and that many questions remain open. A
common theme in this as well as many of the other essays is that data
are sorely needed to examine and test existing theories or to develop
these theories further, but that it may be hard to find such data in the first
place. Conducting experiments, in particular experiments in the field, to
generate such data should then prove particularly compelling, as John List
argues in his essay “Experimental Economics in the Journal of Political
Economy.” He discusses how experiments have already proven useful to ex-
amine market institutions and individual choice, and how these experi-
ments moved from the lab to the field and a natural setting, deepening
our understanding of the underlying phenomena.
The Journal of Political Economy and the University of Chicago are proud
to have been active participants in these exciting lines of inquiries over
the years. The essays make clear that much has been accomplished but
that much remains to be done. It is our hope and desire that the JPE will
remain at the forefront of these new and exciting developments and that
authors will continue to consider the journal as their outlet of choice for
their path-breaking contributions. In that spirit, we are looking forward
to the next 125 years.

Chicago and the Origins of Modern

General Equilibrium
Hugo F. Sonnenschein
University of Chicago
I.
In the early 1950s, with substantial time together at the University of Chi-
cago, Kenneth Arrow and Gerard Debreu (1954) provided a first mathe-
matically rigorous proof of the existence of equilibrium in a general Wal-
rasian model.1 Before that time, aside from some specialized modeling,
the equality between the number of equations and unknowns in the Wal-
rasian model was taken to be the basis for believing that every Walrasian
system contains at least one equilibrium.2
With the result of Arrow and Debreu (1954), we learn that the Walra-
sian model is fully able to determine the prices of all commodities as a
function of tastes, technologies, and the distribution of wealth. Their find-
ings made rigorous the interdependencies that are essential to the deter-
mination of prices. We have, for the first time, an adequate general equi-
librium theory of value. Their result is a cornerstone of the foundation of
multimarket price theory. It demonstrates that in every well-specified Wal-
rasian economy there is always at least one set of relative prices that bal-
ance supply and demand in all markets.
The Arrow-Debreu existence theorem is noteworthy in other important
dimensions. First, it provides a first modern statement of the Walrasian
model. It also promoted a standard in economics for precision of formu-
lation and mathematical rigor. In addition, the approach can be seen to
explicitly unite the issue of the existence of equilibrium for situations of
imperfect competition, as in the Cournot model, with the issue of the ex-
istence of Walrasian equilibrium and perfect competition. Finally, with
the benefit of hindsight, the approach has had a remarkable ability to an-
ticipate and facilitate several important advances in the existence theory.
The perspective on the Debreu and Arrow-Debreu contributions presented here was de-
veloped in joint work with Wayne Shafer. A central point of this note is communicated in
Duffie and Sonnenschein (1989). Philip Liang provided most helpful research assistance.
1
The article “a” is carefully chosen. Lionel McKenzie (1954) presented a general proof
of the existence theorem at precisely the same time.
2
Wald’s contributions (1935, 1936a, 1936b) do more than count equations and un-
knowns. They are a milestone in the development of a satisfactory general existence the-
orem but have features that make his analysis quite specialized by contemporary standards.

This celebratory issue of the JPE is an ideal place to remind us of the
extent to which the origins of modern general equilibrium theory were
at the University of Chicago.
I will argue further that the Arrow-Debreu approach has many advan-
tages when compared to the more customary excess demand approach
as found in Debreu (1959) and Arrow and Hahn (1971).
II.
From the point of view of this presentation, it is the work of Debreu (1952)
that is most central to the Arrow-Debreu existence theorem. In terms of
the mathematics, the essence of Arrow-Debreu is an application of De-
breu (1952), who did this work as a member of the Cowles Commission
at the University of Chicago. However, there is apparently no good record
of the origin of the idea for Debreu’s study. It is clear that the existence
problem was much in the air at Cowles in Chicago in the early 1950s and
that the Kakutani fixed-point theorem (1941) was well understood and
available to play a role. Arrow and Debreu were in close communication,
and each had made definitive contributions to the modern framing of
the Walrasian model that are essential to establishing a general equilib-
rium existence theorem.3
III.
Most textbook treatments of modern general equilibrium theory follow
Debreu (1959) and approach the existence of equilibrium via the con-
struction of excess demand functions. The Arrow-Debreu approach is fun-
damentally different and comes with a number of advantages. As I have al-
ready suggested, Debreu (1952) provides a mathematical treatment that
applies to both the problem of equilibrium existence in the Walrasian
model and the problem of existence in Cournot markets in which agents
have the power to influence prices. Furthermore, Debreu’s theorem pro-
3
For a particularly important example, see Arrow (1951a), where the treatment of the
second welfare theorem reveals boundary issues that must be dealt with in equilibrium ex-
istence. Arrow’s first appointment as assistant professor was at the University of Chicago,
but he was a “time-splitter,” notably with responsibilities at Rand and Columbia, where
he was a graduate student. In 1949, he left Chicago to accept a position as acting assistant
professor at Stanford. Arrow (1951a) was reprinted as a Cowles Commission paper, and it is
reasonable to assume that the Chicago environment, and in particular the collaboration
with Debreu, were of substantial importance. But a mystery remains: who saw Debreu
(1952) as a key to the Arrow-Debreu existence theorem? Yes, it is a natural extension of
Nash (1950), but who had the idea of framing it and using it for Walrasian existence? I
am unable to find the answer to this question in the writings of Arrow and Debreu.

vides an entry into the Cournot theory that includes the case in which
there are externalities in production. As we will see, it also applies to Wal-
rasian economies with significant externalities. It is surely a benefit of the
excess demand function approach that the excess demand functions of
each agent are independent of the behavior of other agents both in and
out of equilibrium. In this situation one simply sums equations to deter-
mine the aggregate excess demand whose zeros define equilibrium prices.
However, the Arrow-Debreu approach, which directly confronts the fact
that the prices faced by agents depend on the choices of other agents,
has advantages in terms of the range of behaviors and models that can be
covered by an equilibrium existence theorem.
In the formal presentation that follows, I will demonstrate that the
Arrow-Debreu approach does not accommodate only externalities and
price-dependent preferences. I will also show how, with a slight adjust-
ment, it goes a substantial way toward accommodating preferences that
are not orders, as in Mas-Colell (1974). I assume that the reader has some
background with the modern literature and is comfortable with the stan-
dard textbook approaches to the equilibrium existence theorem via ex-
cess demand functions.4 To simplify matters I emphasize the case of pure
exchange. Debreu (1952), with a slight emendation, is a starting point
for the formal commentary.
Debreu defined an abstract economy or generalized n-person game G
Q for each agent, a choice set Xi,Qa constraint correspondence Ai :
by,
Xj → Xi , and a utility function Ui : Xj → R. One slightly extends this
formulation by making the utility of each agent i state dependent where
the state includes the position of all agents including one’s self. Let X 5
Q
Xj . The utility function of i is Ui : Xi X → R.
One interprets the ith agent’s objective as that of choosing for each
x ∈ X a zi ∈ Ai ðxÞ that maximizes Ui(zi, x) subject to zi ∈ Ai ðxÞ. The
vector of actions x 5 ðx1 , x2 , : : :, xn Þ is an equilibrium for G if each xi max-
imizes Ui(, x) subject to zi ∈ Ai ðxÞ for each i. This is Debreu’s extension
of Nash (1950) for noncooperative n-person games, but here we have
the emendation that preferences vary with one’s own position. Debreu
writes, “the [constraint set] is actually independent of the ith compo-
nent of [x], but . . . we find it more convenient to [include the ith com-
ponent]” (1982, 702). The assumption that Ui depends on the state, in-
cluding one’s own allocation, and then again on one’s choice, is similar
in spirit, although more substantive. It captures the idea that perspective
influences preferences over choices and one’s perspective includes one’s
own position in a state. This idea is particularly congenial to a behavioral
economics interpretation.
4
Debreu (1959, 1982) and Mas-Colell, Whinston, and Green (1995) are particularly use-
ful starting points.

If each Xi is a compact, convex, and nonempty subset of R , if each Ai n
is continuous, nonempty, and convex-valued, and if each Ui is continu-

ous on X and quasi-concave in its first variable, then the above emenda-
tion is without mathematical consequence and Debreu (1952) shows that
an equilibrium exists. The proof applies the fixed-point theorem of Ka-
kutani (1941) to a best-response correspondence, which is upper hemi-
continuous by the Berge theorem (1959).
I turn now to the manner in which the Debreu theorem is applied in
the Arrow-Debreu theorem on the existence of equilibrium. An economy
E is specified by n 2 1 consumers, each with consumption
Q sets Yi ∈ Rl
and with initial endowments ql ∈ R1 . Let Y 5 Yj and D 5 fp ∈ Rl1 :
l
opi 5 1g be the set of normalized prices. Each consumer has a real-

valued utility function defined on Yi Y D, so that the utility of a choice
depends on one’s perspective, which is conditioned on the consump-
tion state and prices. A Walrasian free-disposal equilibrium is a point (y,
p) such that oyi ≤ oqi , p yi 5 p qi for each i, and yi maximizes state-
dependent utility on the set of points yi ∈ Y that satisfy p yi ≤ p qi .
To prove the existence of equilibrium, Arrow and Debreu associate with
E a generalized n-person game in the following manner. The first n 2 1
agents correspond to the consumers above with choice sets Yi and the
nth agent is a fictitious “market player” who chooses from the price sim-
plex D.5 Thus, the state variable is a pair (y, p). The constraint correspon-
dences for the first n 2 1 agents are the usual budget correspondences,
which depend on p and initial endowments. The constraint correspon-
dence for the nth agent is a constant and in each state allows him to pick
any price vector in D. The utility function for the first n 2 1 agents is as
specified in the definition of the economy E. Given the state (y, p), the
market player chooses q ∈ D to maximize qðoyi 2 oqi Þ. Provided that
each Yi is compact and convex, each qi is interior to Yi, and each Ui is con-
tinuous and quasi-concave in its first argument, this generalized game
will satisfy the sufficient conditions mentioned above for an equilibrium.
It is straightforward to verify that this equilibrium is a Walrasian equilib-
rium for E provided that each consumption is not a local (unconstrained)
maximum in its first argument.
One notes that the preceding approach explicitly allows for consumer
preferences that depend on the consumption choices of other consum-
ers. It similarly allows for externalities in economies with production, in-
cluding the case in which the choices of producers affect the utility of
consumers. The manner in which government actions, such as optimal
taxation, are incorporated into an existence theorem requires a bit more
5
The “market player” is not to be thought of as an actual agent. It is added in the same
spirit as one adds a fictitious player to adjust prices in Walrasian tatonnement.

care, but again, the right way to do this is through the approach of the
Debreu theorem on abstract economies.6
IV.
I now demonstrate that the Arrow-Debreu approach, minimally extended,
sheds light on a beautiful contribution of Mas-Colell (1974), which gener-
alized in a fundamental manner the demand side of equilibrium theory.
Rather than defining preferences as a binary relation, usually transitive
and complete, Mas-Colell demonstrated that what is really essential to
the equilibrium existence theory is the assumption that each consumer
knows what he prefers to each possible consumption, with the obvious re-
quirement that he does not prefer any bundle to itself! To understand
why this development was very surprising, consider the case of finite-choice
spaces with no room for convexity and various other topological require-
ments that are essential in virtually all approaches to equilibrium theory.
For constraint sets with three elements x, y, and z, with x preferred to y, y
preferred to z, and z preferred to x, it is not coherent for any choice to be
made. Similarly, with a two-element constraint set composed of x and y,
with x in the set of bundles preferred to y and y in the set of bundles pre-
ferred to x, there can be no coherent choice. But there turns out to be
“magic” in the continuity and convexity that are essential to general equi-
librium existence, and Mas-Colell employed these to redo the equilibrium
existence theorems so that “better than sets” rather than transitive and
complete preferences are all that is required for the existence of equilib-
rium prices.7
So how does the Arrow-Debreu method, minimally extended, apply to
the existence of equilibrium in these behavioral worlds without ordered
preferences? When utility functions are state-dependent and depend on
the consumption state of an agent, the preferred sets of each agent play
much the same role as the preference correspondences of Mas-Colell: the
6
There is some important literature that is designed to accommodate such government
activities (see, e.g., Sontheimer 1971; Shoven 1974); however, with the benefit of hindsight
these efforts demonstrate the benefits of approaches that are not excess demand based.
See Gale and Mas-Colell (1975) and Shafer and Sonnenschein (1975, 1976).
7
In the mid-1960s, I circulated a manuscript that proved that equilibrium existence
could be established in the Debreu (1959) model without the requirement of preference
transitivity, and this manuscript was widely circulated and discussed with Arrow, Debreu,
and McKenzie. I am fortunate that the lengthy delay in the publication of the volume Pref-
erences, Utility, and Demand (Chipman et al. 1971) did not have any bad effects on my career.
It should also be pointed out that the Mas-Colell advance, which has one replace prefer-
ence relations with preferred set correspondences, is not only a deep substantive and con-
ceptual breakthrough but unlike my analysis also requires one to abandon the excess de-
mand approach to equilibrium existence. To understand this point, see the appendix of
Mas-Colell (1974).

points on the boundaries of these sets, which can be thought of as “behav-
ioral indifference sets,” may cross, and it is possible for y to be preferred
to x when the base choice is x and x to be preferred to y when the base
choice is y. For similar reasons transitivity fails.8 But again, this emenda-
tion of Arrow and Debreu is easy to put forth with the benefit of hindsight,
and in particular with Mas-Colell’s approach to the representation of
preferences firmly in hand.9
V.
I have written about one of those moments when the giants who walked
the halls of my university were particularly productive. It is the story of
Chicago’s role in the origins of general equilibrium. The particular mo-
ment that I chose, and the subsequent applications that I cited, closely
follow my particular interests. But general equilibrium is a very large area,
even when one pays particular attention to the existence literature. The
work of the early 1950s has been followed by fundamental contributions
by Aumann, Scarf, and a long list of others. Chicago and the JPE have
played their part in these subsequent developments, and I want to cite
four papers from the JPE in general equilibrium, welfare economics,
and consumer choice applicable to general equilibrium that have influ-
enced my own thinking and teaching.
Kenneth J. Arrow, “A Difficulty in the Concept of Social Welfare”
(1950). The timing suggests a substantial Cowles-Chicago influence, and
this predates the publication of his monumental Social Choice and Individ-
ual Values (1951b).
Kelvin Lancaster, “A New Approach to Consumer Theory” (1966). This
is among the most-cited JPE contributions, and perhaps the most cited in
general equilibrium broadly interpreted.
David Cass and Menahem Yaari, “A Re-examination of the Pure Con-
sumption Loans Model” (1966). This is a basic work for opening a discus-
8
This point is made in Duffie and Sonnenschein (1989).
9
One should note that Mas-Colell’s continuity assumption on the preference corre-
spondence is an open graph condition. This is less demanding than the assumption that
the state-dependent utility functions Ui are continuous. Shafer and Sonnenschein (1975),
motivated by Mas-Colell’s work, showed that Debreu (1952) can be strengthened to allow
for preference correspondences with open graphs as opposed to continuous utility func-
tions, but the proof is more than emendation of Debreu’s arguments. With their result
in hand, one can use the Arrow-Debreu method to prove equilibrium existence theorems
of the Mas-Colell type. This further supports the argument that Debreu (1952) and Arrow
and Debreu (1954) did more than provide rigorous foundations for the theory of value. It
also provides a mathematical approach to the existence theorem that is adequate for deal-
ing with a variety of important generalizations. And again, much of this work was done at
the University of Chicago.

sion of the general equilibrium welfare theorems in models with infini-
ties.
David Cass and Karl Shell, “Do Sunspots Matter?” (1983). The paper is
a milestone in the general equilibrium approach to asset theory.
I conclude with some remarks that contrast the excess demand ap-
proach to equilibrium existence with the approach via abstract economies
as pioneered by Debreu (1952) and Arrow and Debreu (1954). Because
of my work on “anything goes,” sometimes referred to as Sonnenschein,
Mantel, and Debreu, I have something at stake here, and one might con-
jecture that I am a fan of the excess demand approach.10
This is not at all my view. I am pressed to think of any significant result
on equilibrium existence that cannot be achieved as well via the abstract
economy approach. The approach via excess demand has the advantage
that it sometimes separates the behavior of agents in a descriptively con-
venient manner: in the absence of externalities, one’s excess demand
function does not depend on the actions of other agents. But from the
point of view of expanding applications, this is also sometimes the weak-
ness of the approach. There are benefits in directly confronting the fact
that the prices faced by agents depend on the choices of other agents.
Furthermore, perhaps the continued absence of a foundation for using
tatonnement to define laws of motion for Walrasian economies renders
the excess demand construct less interesting? I regard Scarf (1960, 1973)
to be milestones in helping economists to think through important is-
sues of how one might compute the equilibrium prices of an economy.
Also, I do not question the relevance of computing Walrasian equilibria;
however, it is less than clear that, even for this purpose, excess demand
functions are the way to go.
Finally, I do not believe that the “anything goes theorem” suggests the
end of economic theory. Rather, it tells us that the empirically relevant
restrictions that are generated by theory are likely to depend on specific
assumptions regarding the form of preferences, technology, and the dis-
tribution of income.
I close by observing that Debreu (1982) follows closely his 1952 approach
and the method of Arrow-Debreu. For me, Arrow and Hahn (1971) gives
too little appreciation to the Arrow-Debreu method. But this is not the
place to quibble with my teachers. In summary, the Arrow-Debreu ap-
proach, with Debreu (1952) replaced by the Mas-Colell-inspired strength-
ening presented in Shafer and Sonnenschein (1975), is a good way to go.
10
For a summary of “anything goes,” see Shafer and Sonnenschein (1982).

References
Arrow, K. J. 1950. “A Difficulty in the Concept of Social Welfare.” J.P.E. 58 (4):
328–46.
———. 1951a. “An Extension of the Basic Theorems of Classical Welfare Eco-
nomics.” In Second Berkeley Symposium on Mathematical Statistics and Probability,
507–32. Berkeley: Univ. California Press.
———. 1951b. Social Choice and Individual Values. New York: Wiley.
Arrow, K. J., and G. Debreu. 1954. “Existence of an Equilibrium for a Competi-
tive Economy.” Econometrica 22:265–90.
Arrow, K. J., and F. H. Hahn. 1971. General Competitive Analysis. San Francisco:
Holden-Day.
Berge, C. 1959. Espaces topologiques. Stuttgart: Macmillan.
Cass, D., and K. Shell. 1983. “Do Sunspots Matter?” J.P.E. 91 (2): 193–227.
Cass, D., and M. E. Yaari. 1966. “A Re-examination of the Pure Consumption
Loans Model.” J.P.E. 74 (4): 353–67.
Chipman, J. S., et al., eds. 1971. Preferences, Utility, and Demand: A Minnesota Sym-
posium. New York: Harcourt Brace Jovanovich.
Debreu, G. 1952. “A Social Equilibrium Existence Theorem.” Proc. Nat. Acad. Sci.
38 (10): 866–93.
———. 1959. Theory of Value: An Axiomatic Analysis of Economic Equilibrium. Cowles
Foundation Monograph no. 17. New Haven, CT: Yale Univ. Press.
———. 1982. “Existence of Competitive Equilibrium.” In Handbook of Mathemat-
ical Economics, vol. 2, edited by K. J. Arrow and M. Intriligator, 697–743. Amster-
dam: Elsevier.
Duffie, D., and H. F. Sonnenschein. 1989. “Arrow and General Equilibrium The-
ory.” J. Econ. Literature 27 (2): 565–98.
Gale, D., and A. Mas-Colell. 1975. “An Equilibrium Existence Theorem for a
General Model without Ordered Preferences.” J. Math. Econ. 2 (1): 9–15.
Kakutani, S. 1941. “A Generalization of Brouwer’s Fixed Point Theorem.” Duke
Math. J. 8 (3): 457–59.
Lancaster, K. J. 1966. “A New Approach to Consumer Theory.” J.P.E. 74 (2): 132–
57.
Mas-Colell, A. 1974. “An Equilibrium Existence Theorem without Complete or
Transitive Preferences.” J. Math. Econ. 1 (3): 237–46.
Mas-Colell, A., M. D. Whinston, and J. R. Green. 1995. Microeconomic Theory. Vol. 1.
New York: Oxford Univ. Press.
McKenzie, L. W. 1954. “On Equilibrium in Graham’s Model of World Trade and
Other Competitive Systems.” Econometrica 22:147–61.
Nash, J. F. 1950. “Equilibrium Points in n-Person Games.” Proc. Nat. Acad. Sci. USA
36 (1): 48–49.
Scarf, H. 1960. “Some Examples of Global Instability of the Competitive Equilib-
rium.” Internat. Econ. Rev. 1 (3): 157–72.
———. 1973. The Computation of Economic Equilibria. New Haven, CT: Yale Univ.
Press.
Shafer, W. J., and H. F. Sonnenschein. 1975. “Equilibrium in Abstract Economies
without Ordered Preferences.” J. Math. Econ. 2 (3): 345–48.
———. 1976. “Equilibrium with Externalities, Commodity Taxation, and Lump
Sum Transfers.” Internat. Econ. Rev. 17:601–11.
———. 1982. “Market Demand and Excess Demand Functions.” In Handbook of
Mathematical Economics, vol. 2, edited by K. J. Arrow and M. Intriligator, 671–93.
Amsterdam: Elsevier.

Shoven, J. B. 1974. “On the Computation of Competitive Equilibrium on Inter-
national Markets with Tariffs.” J. Internat. Econ. 4 (4): 341–54.
Sontheimer, K. C. 1971. “The Existence of International Trade Equilibrium with
Trade Tax-Subsidy Distortions.” Econometrica 39:1015–35.
Wald, A. 1935. “Über die eindeutige positive Lösbarkeit der neuen Produk-
tionsgleichungen.” Ergebnisse eines mathematischen Kolloquiums (6): 12–20.
———. 1936a. “Über die Produktionsgleichungen der ökonomischen Wert-
lehre.” Ergebnisse eines mathematischen Kolloquiums (7): 1–6.
———. 1936b. “Über einige Gleichungssysteme der mathematischen Ökono-
mie.” Zeitschrift für Nationalökonomie 7:637–70. Translated as “On Some Systems
of Equations of Mathematical Economics.” Econometrica 19 (1951): 368–403.
Economic Growth: The Past, the Present, and the

Future
Ufuk Akcigit
University of Chicago, National Bureau of Economic Research, and Centre for Economic
Policy Research
Is there some action a government of India could take that

would lead the Indian economy to grow like Indonesia’s or
Egypt’s? If so, what, exactly? If not, what is it about the “na-
ture of India” that makes it so? The consequences for hu-
man welfare involved in questions like these are simply stag-
gering: Once one starts to think about them, it is hard to
think about anything else. (Lucas 1988, 5)
Introduction
These words by the Nobel laureate Chicago economist Robert Lucas Jr.
summarize why so many great scholars found it hard to “think about any-
thing else” and spent their careers trying to understand the process of
economic growth. Economies are complex systems resulting from the ac-
tions of many actors. This complexity makes it challenging, but also infi-
nitely interesting, to understand the determinants of economic growth.
What are the roles of human capital, fertility, ideas, basic science, and pub-
lic policy for growth? These are just some of the important questions that
I would like to thank my colleagues and current and former students for very valuable
feedback and discussions. The paper reflects solely my own views.

were posed by many highly influential studies featured in the issues of
the Journal of Political Economy over the years. Indeed, this journal has been
the platform to diffuse many of the brilliant ideas and start important de-
bates in the field of economic growth. In this short paper, my goal is to
revisit some of those seminal papers, briefly describe some of the more re-
cent contributions, and end with some thoughts about the future direc-
tion of the field. The reader should note in advance that the list of work
covered here is by no means exhaustive and mostly targets work that has
been featured in issues of the JPE.
The Past
In this section, I discuss some of the past seminal contributions to the field
of economic growth. The papers will be grouped under the following sub-
topics: (i) capital accumulation, (ii) innovation, (iii) technology adoption,
and (iv) human capital and fertility.
Physical Capital Accumulation, Spillovers, and Growth

For a long time, the workhorse to study economic growth was the neo-
classical growth model (Ramsey 1928; Cass 1965; Koopmans 1965). The
main difference from the classic Solow-Swan model (Solow 1956; Swan
1956) is that the neoclassical model explicitly endogenizes the savings rate
through a utility-maximizing household. One of the key components of
both models is the production function
Yt 5 AK at Lt12a , (1)
where Yt is aggregate output at time t, K t is the capital stock, Lt is the
labor, A is the level of productivity, and a ∈ ð0, 1Þ.1 A major implication
of the neoclassical and the Solow-Swan models is that capital accumulation
can serve as a source of short-run economic growth but cannot be a driver
of long-run growth because of decreasing returns through a < 1. These
models predict “convergence in per capita income,” whereby poorer coun-
tries grow faster until they catch up with richer countries.
The JPE featured many influential empirical and theoretical works on
economic growth. On the empirical side, Barro and Sala-i-Martin (1992)
tested the convergence result of neoclassical theory. In this work, Barro
and Sala-i-Martin studied the growth rates of 48 US states between 1840
and 1963 as a function of their initial per capita income and found signif-
icant empirical evidence for convergence at the state level. Barro and Sala-i-
1
This is the particular specification used to facilitate the discussion in this paper. This
production function can be written in a more general constant returns form.

Martin also analyzed the convergence hypothesis in the cross-country data.
This time their results were more mixed. The authors identified conver-
gence only after controlling for school enrollment rate and the ratio of
government consumption to GDP. They thus provide some evidence for
“conditional convergence” at the country level.
A production function with decreasing returns was the main reason
why endogenous growth did not occur in the neoclassical growth model.
In one of the seminal JPE papers, Romer (1986) overcame the problem of
decreasing returns and generated endogenous long-run growth by intro-
ducing spillovers that led to increasing returns in the production func-
tion. In this model, growth resulted from the combination of capital ac-
cumulation and the associated spillovers. More specifically, Romer assumed
productivity to be a linear function of the capital stock:
At 5 gKt :
In this model, just like the neoclassical model, markets are perfectly com-
petitive and knowledge creation is simply a by-product of capital accumu-
lation. Therefore, one can also interpret productivity formation as com-
ing from learning by doing. Stokey (1988) was another great JPE paper
that showed how economywide learning by doing could lead to sustained
long-run growth. It was not until Romer’s later work, which I will describe
below, that agents in the economy had explicit incentives to create new
ideas. While Romer (1986) introduced productivity spillovers through cap-
ital accumulation, Lucas (1988) introduced similar spillovers through hu-
man capital externalities. In their interesting JPE paper, Glaeser et al.
(1992) empirically study the existence of knowledge spillovers within
and across industries.
On the theory side, Rebelo (1991) is one of the well-known JPE papers
that generated endogenous long-run growth by eliminating the decreas-
ing returns from the neoclassical production function (1). This paper con-
sidered the so-called “AK model,” in which the production technology does
not feature labor and is linear in capital with a 5 1:
Yt 5 AK t :
In this model, the linear structure prevents capital accumulation from run-
ning into decreasing returns and can generate long-run growth. This trac-
table framework allowed Rebelo to also study the impact of public policy on
economic growth, which was not possible in the Ramsey-Cass-Koopmans
or Solow-Swan models since they did not generate long-run endogenous
growth.
It is now widely accepted in the literature that the world economy has
experienced persistent technological progress over the past 200 years and

R&D and innovation have played a central role in the advancement of the
world technology frontier during this period (Acemoglu 2008). Unfortu-
nately, endogenous growth models based on the neoclassical framework
were insufficient and had little to say about this aspect of economic growth.
Therefore, the early 1990s saw the rise of innovation-based growth mod-
els, as I explain next.
Innovation-Based Growth
New technologies emerge as a result of costly R&D efforts by individ-
uals and companies. The new technologies eventually introduce into the
market a new product variety as in Romer (1990) or a better version of an
existing product or technology that makes the earlier version obsolete
through Schumpeterian creative destruction as in Aghion and Howitt
(1992) and Grossman and Helpman (1991). An entrepreneur’s or firm’s
incentive to undertake these costly R&D efforts is to gain market power.
These explicit efforts and market incentives were missing in the endoge-
nous growth models based on a neoclassical framework.
The new ingredient in the innovation-based endogenous growth mod-
els is the production function for ideas. The number of new ideas—the
_
change in productivity A—is assumed to be a function of the existing knowl-
edge stock At and the number of researchers R who spent time producing
them:
A_ t 5 dAt R, (2)
where d > 0 captures the research productivity. In these models, agents face
an occupational choice. Individuals can work either as production work-
ers (L) and earn the production wage (wt) or as research workers (R) who
produce new ideas and receive the return to their innovation (Vt). The
key equation is the free-entry condition into research, which determines
the allocation of the work force to the production and research sectors:
wt 5 Vt dAt :
This equilibrium split of the workforce determines the current level of pro-
duction through L and the rate of growth of knowledge (and hence per
capita income) through R.
Romer’s model views each innovation as the introduction of a new prod-
uct variety that becomes a permanent part of the economy and the inven-
tor of which becomes its permanent producer. Thus, the model abstracts
from competition, firm exit, and firm turnover. Schumpeterian models,
on the other hand, prioritized the industrial organization aspect of eco-
nomic growth. Through the notion of creative destruction, Schumpeterian

models introduced firm entry-exit and competition into the endogenous
growth literature (e.g., Aghion and Howitt 1992; Grossman and Helpman
1991). This feature has been essential to map these models to the firm- and
inventor-level microdata and estimate them, as I explain below. Moreover,
these features enable Schumpeterian models to generate richer policy
implications. Since the Romer-style product variety models feature uninter-
nalized spillovers through (2), they generate major underinvestment in
R&D and call for research subsidies. Schumpeterian models, on the other
hand, brought in a competitive force through the so-called “business steal-
ing” effect whereby new entrants try to replace incumbents through new
innovations. If the spillover associated with each innovation is not big
enough, the social return from it would be less than the private return,
which, in equilibrium, could make firms overinvest in R&D. Therefore,
Schumpeterian models call for empirical estimates of the spillovers and
business stealing externality to inform optimal policy.
In the innovation-based growth models, the R&D production function
in (2) has been taken as a reduced-form representation, in which some
inputs (either human capital or R&D dollars) turn into innovations. How-
ever, a few papers have studied this production function in detail and tried
to understand the steps that lead to innovation. One of the key findings
of this literature has been that there are at least two steps in creating a
practical innovation. First, universities, public research labs, and occasion-
ally private firms invest in theory-oriented and abstract “basic research” to
produce fundamental, essential, first-stage background knowledge in the
form of theories and equations. In the second step, profit-seeking innova-
tors and companies invest in more familiar, data-oriented or end product–
oriented “applied research” to produce practical and patentable findings.
For instance, Wallace Carothers’s basic research led to the invention of
the famous Carothers equation that formed the basis for the applied re-
search by DuPont that resulted in the invention of “nylon.” What are the
incentives to do basic research? What are the roles of government and uni-
versities in innovation and economic growth? These were the key ques-
tions that Richard Nelson investigated in his highly influential JPE article
(Nelson 1959). He questioned the sources of spillovers from basic research
and discussed the reasons why some firms would invest more in basic re-
search that is more uncertain, with less clear goals or less close ties to spe-
cific practical problems or to the creation of a specific object. Nelson’s an-
swer to this question was that successful scientific advances through basic
research often have many practical applications that are not predictable
ex ante. These scientific findings form the key inputs for many subsequent
applied research projects, which can then lead to practical and patentable
findings. Because of this nature of basic research, firms are unlikely to uti-
lize all the economic value through patents. Further, a research finding

that has applications in one field or sector might have applications in many
others. He noted that “it is for this reason that firms which support research
toward the basic science end of the spectrum are firms that have fingers in
many pies” (302). Even though this is a fundamental problem for innova-
tion policy, the split between basic and applied research and the explicit
role of the universities in the growth process are still understudied.2
Relatedly, in his well-known JPE article, Chad Jones (1995) questioned
the R&D production function (2). The basis for Jones’s critique was that
one of the key implications of the endogenous growth models that used
(2) is that increased population size is associated with an increased steady-
state growth rate. This was due to the fact that a larger population raises
the return to innovation through the so-called “market size effect” and also
increases the supply of potential researchers to do R&D. Jones argued that
these predictions were empirically not plausible since larger countries do
not necessarily grow faster, and even though the United States increased
its R&D effort over the years, its growth rate did not go up. He proposed a
“semi-endogenous” growth model by modifying the R&D production func-
tion of (2) as follows:
A_ 5 dAf R l ,
where f, l ∈ ð0, 1Þ. This modification removed the impact of the popu-
lation level and led to the following long-run growth rate: g 5 ln=ð1 2 fÞ,
where n is simply the rate of population growth. A strong prediction of
Jones’s specification is that long-run growth is affected only by the exoge-
nous population growth rate n and is invariant to any government policy.
This paper started the “scale effect” debate in the literature.
During the 1990s, the JPE was the stage for the “scale effect” debate.
Soon after Jones’s influential paper, the JPE published two other interest-
ing articles by Alwyn Young (1998) and Peter Howitt (1999), who proposed
“product proliferation” as a remedy to the “scale effect” problem. Howitt’s
solution, which builds on Young’s proliferation idea, is to propose a mech-
anism whereby, as the number of product varieties in an economy grows,
the effectiveness of research effort on each variety decreases as the popula-
tion gets spread more thinly over a larger number of varieties. This makes
the expected growth rate in each variety independent of the overall pop-
ulation level while preserving the role of policy for economic growth.
2
A notable exception is the recent work by Akcigit, Hanley, and Serrano-Velarde (2016),
which shows that even after controlling for firm size, firms that operate in more industries
are more likely to invest in basic research. This finding provides empirical support for Nel-
son’s “fingers in many pies” hypothesis of basic research.

Technology Adoption and Growth
While countries at the world technology frontier grow through innova-
tions, nonfrontier countries grow mostly through imitation or technology
adoption. Why do some countries adopt new technologies while others do
not? In their seminal JPE paper, Murphy, Shleifer, and Vishny (1989) study
this question. They theoretically formalize the famous “big-push” hypothe-
sis of industrialization of Rosenstein-Rodan (1943). In their analysis, there
is an underlying market size effect that works through the complementar-
ities between different sectors in the economy. When certain firms or in-
dustries invest in new technologies, these investments also raise demand
in other industries. The investing firm receives only a fraction of their con-
tribution to the profits in the overall economy. Therefore, each individual
firm might not find it profitable to adopt a new technology, whereas a co-
ordinated investment by all firms could justify such a technology adop-
tion. As a result, in those economies where actions can be coordinated
through public policies, firms might be willing to invest in new technol-
ogies and can generate a big push that leads to industrialization.
In another well-known JPE paper, Stephen Parente and Edward Pres-
cott (1994) studied the macroeconomic implications of barriers to tech-
nology adoption. One of the major shortcomings of the neoclassical growth
model is its inability to generate empirically plausible income differences
across countries. Parente and Prescott fixed this problem by extending the
neoclassical growth model to incorporate firm investment for technology
adoption from the world frontier. Country-specific barriers to technology
adoption, due to weak property rights or other institutional aspects, could
hinder the flow of technologies, leading to the observed technology and in-
come gaps across countries.
Relatedly, in another important JPE paper, Katz and Shapiro (1986) stud-
ied the importance of network externalities in technology adoption. This
paper showed that technology adoption could depend on whether a tech-
nology is “sponsored,” that is, whether an entity owns the right to that tech-
nology and therefore has the incentives to promote it. An important as-
pect in this analysis is the dynamic consideration of technology adoption.
For instance, this paper highlights the concept of “penetration pricing,” a
strategy of offering low prices today to build up a network and influence
the expectations of the customer about the future size of the network. In
this case, it is possible that an inferior technology might get adopted be-
cause it is sponsored by an entity that strategically affects the size of the
network.
Empirical studies on technology adoption have been relatively rare. The
influential JPE article by Foster and Rosenzweig (1995) studied the “Green
Revolution” in India, during which farmers in some Indian regions re-
ceived an opportunity to adopt new technologies. The paper finds that in-

formational frictions and the lack of knowledge about how to use the new
seeds were major impediments to technology adoption. It also identifies
significant and quantitatively important learning spillovers; that is, farm-
ers learn from neighbors who already use the new seeds.
Human Capital, Fertility, and Growth

A large literature has studied the link between fertility and economic growth.
The Malthusian model (Malthus 1798) is a well-known framework accord-
ing to which population grows faster once per capita income rises. How-
ever, this prediction is at odds with the empirical evidence that fertility
has gone down as income grew around the world, especially in Western
economies. Neoclassical growth models circumvented this contradiction
by focusing on endogenous capital accumulation rather than on the fer-
tility choice. Nobel laureate and Chicago economist Gary Becker (1960)
formulated the quantity-quality trade-off that parents face when deciding
on the number of children and the investment in each child’s human cap-
ital. Subsequently, Becker and Barro (1988) and Barro and Becker (1989)
introduced fertility choice into growth models. The seminal JPE paper by
Becker, Murphy, and Tamura (1990) is the first paper that introduced fer-
tility choice into an “endogenous growth” framework. The model implies
that the rate of return on human capital relative to those on children is high
when human capital in the society is abundant. Therefore, the model pre-
dicts large families and little investment in each child when the society
has limited human capital.
The Present
The early models of endogenous growth set the stage for a fruitful litera-
ture to come, and the JPE was their leading outlet. What was missing in
the early literature was firm- or individual-level heterogeneity that would
provide a better connection between growth theory and micro-level data.
Providing solid micro-level foundations would lead to stronger macro-
growth models that generate more accurate positive predictions and rel-
evant normative implications.
A major step in that direction was another important JPE paper by
Klette and Kortum (2004), which built a novel Schumpeterian growth model.
The earlier endogenous growth models predict that innovations come
only from new entrants, new entrants and young firms are the largest firms
in the economy, and exit rates would be uncorrelated with firm age or size.
While these predictions are at odds with the firm-level data, the Klette-
Kortum model fixed these problems by defining a firm as a collection of
production units. In this framework, firms could grow by introducing better-

quality versions of the products of other firms and thus add those prod-
ucts to their portfolio. Klette and Kortum could thus map the benchmark
Schumpeterian models to a more realistic firm dynamics setting. In a more
recent JPE article, Akcigit and Kerr (forthcoming) extended the Klette-
Kortum framework by allowing firms to improve not only other firms’ prod-
ucts through “external innovations” but also their own products through
“internal innovations.” Innovation and firm size heterogeneities allow the
model to generate a close fit to the firm- and innovation-level data from the
US Census Bureau and the US Patent Office. The estimated model shows
that small firms are expending disproportionately more effort to do ex-
ternal innovations, and the spillovers associated with external innovations
are significantly larger. These findings suggest that R&D policies that aim
to correct for underinvestment in R&D should take into account the dif-
ferential spillovers generated by different-sized firms in the economy. As
it is exemplified by these papers, a close dialogue between endogenous
growth theory with heterogeneity and microdata allows researchers to
quantify certain mechanisms and study the impact of counterfactual indus-
trial policies.
Heterogeneity across individuals has also been the key feature of two re-
cent JPE papers. Jaimovich and Rebelo (2017) showed that taxation has a
nonlinear impact on growth once talent heterogeneity is taken into ac-
count. Jones and Kim (forthcoming) showed that focusing on differential
innovation efforts by entrants and incumbents can be important in un-
derstanding the rise in top income inequality.
Heterogeneity in ideas has been the essential component of the surging
endogenous growth models that incorporate the importance of human
interactions for human capital accumulation and growth (e.g., Alvarez,
Buera, and Lucas 2008; Lucas 2009). The JPE recently published two pa-
pers, Lucas and Moll (2014) and Perla and Tonetti (2014), that show that
growth can also occur when less knowledgeable agents in an economy in-
teract with more knowledgeable agents and imitate them to improve their
own productivities.
The JPE continues to promote frontier research on various additional
aspects of economic growth. Trade and innovation (e.g., Atkeson and Bur-
stein 2010), growth miracles (e.g., Young 2012), value of life (e.g., Jones
2016), environment (e.g., Acemoglu et al. 2016; Aghion et al. 2016), and
institutions (e.g., Acemoglu, Robinson, and Verdier 2017) are just a few
of the exciting facets of economic growth featured in recent issues of the
JPE.
The Future
The future of the field of economic growth is more exciting than ever. Com-
puters are becoming more powerful. Many countries are making their

firm- and individual-level microdata sets available to researchers. In addi-
tion, thanks to optical character recognition techniques, more and more
large-scale historical records are being digitized to be used in economic
research.3 My recommendation to young researchers is to take notice of
the rapid change in the field of economic growth and invest not only in
their theoretical skills but also in their computational and empirical knowl-
edge. I also recommend reading the seminal papers—some of which I de-
scribed here. They are full of stimulating ideas, many of which could not
be investigated empirically because of a lack of data or developed quantita-
tively because of a lack of computational power. Now could be the time to
revisit those ideas.
The University of Chicago’s JPE has been the host of many seminal con-
tributions, and it will, without any doubt, continue to promote the frontier
research in the field. Have we answered the questions that Robert Lucas
posed in the opening quote? Certainly not entirely. Yet the field has made
significant progress toward that goal over the years. This is a call for young
researchers to come and think harder about growth-related issues. The
field has so many interesting open questions that they will surely, once
they start, have a hard time “thinking about anything else.”
References
Acemoglu, D. 2008. Introduction to Modern Economic Growth. Princeton, NJ: Prince-
ton Univ. Press.
Acemoglu, D., U. Akcigit, D. Hanley, and W. Kerr. 2016. “Transition to Clean
Technology.” J.P.E. 124 (1): 52–104.
Acemoglu, D., J. Robinson, and T. Verdier. 2017. “Asymmetric Growth and Insti-
tutions in an Interdependent World.” J.P.E. 125 (5): 1245–1305.
Aghion, P., A. Dechezleprêtre, D. Hémous, R. Martin, and J. Van Reenen. 2016.
“Carbon Taxes, Path Dependency, and Directed Technical Change: Evidence
from the Auto Industry.” J.P.E. 124 (1): 1–51.
Aghion, P., and P. Howitt. 1992. “A Model of Growth through Creative Destruc-
tion.” Econometrica 60 (2): 323–51.
Akcigit, U., J. Grigsby, and T. Nicholas. 2017. “The Rise of American Ingenuity:
Innovation and Inventors of the Golden Age.” Working Paper no. 23047, NBER,
Cambridge, MA.
Akcigit, U., D. Hanley, and N. Serrano-Velarde. 2016. “Back to Basics: Basic Re-
search Spillovers, Innovation Policy and Growth.” Discussion Paper no. 11707,
Centre Econ. Policy Res., London.
Akcigit, U., and W. R. Kerr. Forthcoming. “Growth through Heterogeneous In-
novations.” J.P.E.
Alvarez, F. E., F. J. Buera, and R. E. Lucas Jr. 2008. “Models of Idea Flows.” Work-
ing Paper no. 14135, NBER, Cambridge, MA.
3
For instance, the digitized historical patent files matched to census records by Akcigit,
Grigsby, and Nicholas (2017) could provide a unique opportunity for researchers to shed
light on Golden Age innovators, inventions, and economic growth more broadly.

Atkeson, A., and A. T. Burstein. 2010. “Innovation, Firm Dynamics, and Interna-
tional Trade.” J.P.E. 118 (3): 433–84.
Barro, R. J., and G. S. Becker. 1989. “Fertility Choice in a Model of Economic
Growth.” Econometrica 57 (2): 481–501.
Barro, R. J., and X. Sala-i-Martin. 1992. “Convergence.” J.P.E. 100 (2): 223–51.
Becker, G. S. 1960. “An Economic Analysis of Fertility.” In Demographic and Eco-
nomic Change in Developed Countries, 209–40. New York: Columbia Univ. Press
(for NBER).
Becker, G. S., and R. J. Barro. 1988. “A Reformulation of the Economic Theory of
Fertility.” Q.J.E. 103 (1): 1–25.
Becker, G. S., K. M. Murphy, and R. Tamura. 1990. “Human Capital, Fertility, and
Economic Growth.” J.P.E. 98, no. 5, pt. 2 (October): S12–S37.
Cass, D. 1965. “Optimum Growth in an Aggregative Model of Capital Accumu-
lation.” Rev. Econ. Studies 32 (3): 233–40.
Foster, A. D., and M. R. Rosenzweig. 1995. “Learning by Doing and Learning from
Others: Human Capital and Technical Change in Agriculture.” J.P.E. 103 (6):
1176–1209.
Glaeser, E., H. D. Kallal, J. A. Scheinkman, and A. Shleifer. 1992. “Growth in Cit-
ies.” J.P.E. 100 (6): 1126–52.
Grossman, G., and E. Helpman. 1991. Innovation and Growth in the World Economy.
Cambridge, MA: MIT Press.
Howitt, P. 1999. “Steady Endogenous Growth with Population and R&D Inputs
Growing.” J.P.E. 107 (4): 715–30.
Jaimovich, N., and S. Rebelo. 2017. “Nonlinear Effects of Taxation on Growth.”
J.P.E. 125 (1): 265–91.
Jones, C. I. 1995. “R&D-Based Models of Economic Growth.” J.P.E. 103 (4): 759–84.
———. 2016. “Life and Growth.” J.P.E. 124 (2): 539–78.
Jones, C. I., and J. Kim. Forthcoming. “A Schumpeterian Model of Top Income
Inequality.” J.P.E.
Katz, M. L., and C. Shapiro. 1986. “Technology Adoption in the Presence of Net-
work Externalities.” J.P.E. 94 (4): 822–41.
Klette, T. J., and S. Kortum. 2004. “Innovating Firms and Aggregate Innovation.”
J.P.E. 112 (5): 986–1018.
Koopmans, T. C. 1965. “On the Concept of Optimal Economic Growth.” In The
Econometric Approach to Development Planning, 225–300. Amsterdam: North-Holland
(for Pontificia Acad. Sci.).
Lucas, R. E., Jr. 1988. “On the Mechanics of Economic Development.” J. Monetary
Econ. 22 (1): 3–42.
———. 2009. “Ideas and Growth.” Economica 76 (301): 1–19.
Lucas, R. E., Jr., and B. Moll. 2014. “Knowledge Growth and the Allocation of
Time.” J.P.E. 122 (1): 1–51.
Malthus, T. R. 1798. An Essay on the Principle of Population. London: Lawbook Ex-
change.
Murphy, K. M., A. Shleifer, and R. W. Vishny. 1989. “Industrialization and the Big
Push.” J.P.E. 97 (5): 1003–26.
Nelson, R. R. 1959. “The Simple Economics of Basic Scientific Research.” J.P.E.
67 (3): 297–306.
Parente, S. L., and E. C. Prescott. 1994. “Barriers to Technology Adoption and
Development.” J.P.E. 102 (2): 298–321.
Perla, J., and C. Tonetti. 2014. “Equilibrium Imitation and Growth.” J.P.E. 122
(1): 52–76.
Ramsey, F. P. 1928. “A Mathematical Theory of Saving.” Econ. J. 38 (152): 543–59.

Rebelo, S. 1991. “Long-Run Policy Analysis and Long-Run Growth.” J.P.E. 99 (3):
500–521.
Romer, P. M. 1986. “Increasing Returns and Long-Run Growth.” J.P.E. 94 (5):
1002–37.
———. 1990. “Endogenous Technological Change.” J.P.E. 98, no. 5, pt. 2 (October):
S71–S102.
Rosenstein-Rodan, P. N. 1943. “Problems of Industrialisation of Eastern and
South-Eastern Europe.” Econ. J. 53 (210/211): 202–11.
Solow, R. M. 1956. “A Contribution to the Theory of Economic Growth.” Q.J.E. 70
(1): 65–94.
Stokey, N. L. 1988. “Learning by Doing and the Introduction of New Goods.”
J.P.E. 96 (4): 701–17.
Swan, T. W. 1956. “Economic Growth and Capital Accumulation.” Econ. Record 32
(2): 334–61.
Young, A. 1998. “Growth without Scale Effects.” J.P.E. 106 (1): 41–63.
———. 2012. “The African Growth Miracle.” J.P.E. 120 (4): 696–739.
Economic History
David W. Galenson
University of Chicago and Universidad del CEMA
The field of economic history was largely transformed during the 1960s,
when scholars trained in economics departments began to apply economic
theory and econometrics to historical questions. The resulting quantitative
research, often called the new economic history or cliometrics, has now
produced significant revisions of some issues that had previously been stud-
ied primarily by historians, as well as novel results on long-run relationships
of interest to economists. Much of this research has been based on com-
puter analysis of micro-level data sets collected from historical archives.
The first major debate within the new field was initiated by a 1958 ar-
ticle by Alfred Conrad and John Meyer, “The Economics of Slavery in the
Antebellum South.” The authors’ conclusion that slavery was profitable in
the nineteenth century directly contradicted the contention of some his-
torians that the Civil War was not necessary to eliminate slavery, because
the institution would have disappeared even in the absence of legal inter-
vention. Few issues in American history are as contentious as slavery, and
this initial paper prompted dozens of empirical investigations of the de-

mography and economics of the slave economy. The most comprehensive
of these was Time on the Cross, published by Robert Fogel and Stanley Enger-
man in 1974, which used micro-level evidence from the federal census of
1860 to establish that southern farms were more efficient than their north-
ern counterparts and that there were significant economies of scale within
southern agriculture. Further analysis of these scale effects led to identi-
fication of the most important source of slavery’s economic efficiency (Fogel
and Engerman 1977). From the finding that scale economies appeared
above a threshold plantation size of 15 slaves—which typically implied the
presence of 10 adult slaves, or enough to form one slave gang—Fogel and
Engerman inferred that it was the gang system that made slavery partic-
ularly efficient. The economic advantage of slave plantations was greatest
for crops that were physically suited to rapid and routinized cultivation and
least for more delicate crops that were damaged by rapid handling: sugar,
rice, and cotton could be produced efficiently by gangs, whereas tobacco
and grain could not. Slave gangs were equivalent to assembly lines in the
fields, as the labor of cultivation could be clearly divided among the mem-
bers of a gang, and the pace of work could be controlled by a driver who
could readily monitor each slave’s work. This research not only provided
a more precise understanding of the economics of slavery but also un-
derscored the magnitude of the achievement of the abolitionists, whose
moral crusade triumphed over a wealthy and powerful slave economy, not
a weak and failing one.
Quantitative investigations have also produced a new understanding of the
colonial economy and the origins of slavery in North America. Seventeenth-
century English colonizers had knowledge only of an Old World in which
land was scarce and expensive, and labor abundant and cheap, and they
consequently anticipated that ownership of vast amounts of land would
yield vast wealth. Instead, they were shocked to experience a veritable so-
cial revolution, as the reversal of factor scarcity produced radically higher
economic and social mobility. Colonizers now had to recognize that the
greatest economic problem in the New World was to recruit and control
a supply of labor to exploit the abundant land (Galenson 1996). The plant-
ers’ initial response was to bring to America the young men and women
who worked on English farms. Indentured servitude allowed these work-
ers to afford the high cost of transportation: prospective servants signed
contracts in England promising to work for colonial planters in return for
passage to the colonies and maintenance there during their terms. These
terms were necessarily longer than the annual contracts customary in En-
gland, as servants were bound for at least 4 years and often 7 or 8. More
productive servants could repay the fixed cost of passage more quickly, and
econometric analysis of large collections of individual contracts reveals a
strong inverse relationship between term length and expected individual
productivity, as older servants and skilled craftsmen received shorter terms.

The contracts also show compensating differentials for less desirable des-
tinations, as servants willing to travel to the West Indies, where mortality
rates were higher and opportunities after servitude worse, received shorter
terms than those bound for mainland colonies (Galenson 1981).
Over time rising English wages raised the cost of servants to colonial
planters, and falling world sugar prices concurrently lowered the price of
African slaves brought to the Americas. These trends together produced
a dramatic shift in colonial labor forces: in the last quarter of the seven-
teenth century, for example, the price of servants relative to slaves recorded
in Maryland probate inventories increased by more than 50 percent, and
the inventories’ holdings of bound labor fell from four servants per slave
to nearly the reverse, more than three slaves per servant. The rise of black
slavery in North America was thus not caused by any preference for slaves,
but by the rising cost of white labor (Menard 1977).
The settlement of the nineteenth-century Midwest has been an active
topic of historical research ever since Frederick Jackson Turner’s celebrated
1896 speech on the role of the frontier in American history. Quantitative
studies have confirmed Turner’s contention that the frontier was a place
of equality and opportunity for both the native-born and immigrants and
have added to our knowledge of the mechanisms that caused this. Stud-
ies of persistence—done by tracing residents enumerated in a community
in the manuscript schedules of successive decennial censuses—have shown
that settlers did not typically settle for long: few midwestern counties or cit-
ies had 10-year persistence rates as high as 50 percent (Curti 1959; Thern-
strom 1973). Stage migration to the frontier, with a series of short moves,
was most common, so the farther a settler was from his place of birth, the
older he tended to be (Bogue 1963). These investments in migration pro-
duced positive returns: one study found that Utah families who had changed
residence in the preceding decade had, on average, lower wealth than fami-
lies otherwise alike who had not migrated, but higher incomes; the foreign-
born similarly had lower wealth, but higher incomes, than their native-
born counterparts. The timing of migration was important: early arrival
in a new frontier community had a substantial positive effect on a house-
hold’s income and wealth. Although inequality was low in early frontier
communities, concentration of both income and wealth appears to have
increased steadily with duration of settlement (Pope 1989).
Within the past two decades, new theories of economic development
have been based on historical evidence. Stanley Engerman and Kenneth
Sokoloff (2012) observed that the economic history of European coloni-
zation in the Americas posed the puzzle that the wealthiest areas of settle-
ment in the colonial era, in the Caribbean and Latin America, fared worse
in the long run than the North American regions that were initially mar-
ginal economically. They argued that the key difference lay in the regions’
initial factor endowments, which led to major differences in the structure

of their economies and consequently in inequality. The ability of the south-
ern regions to produce staple crops on large plantations led to great eco-
nomic inequality. The wealthy elites that dominated these economies cre-
ated political institutions that limited land ownership and education, and
these deficiencies subsequently hampered modern economic growth. In
contrast, the predominance of family farms in the northern regions led
to much lower inequality and, consequently, to the development of polit-
ical institutions that fostered policies toward education, land ownership,
and immigration that proved much more favorable to economic growth
in the modern era. Daron Acemoglu and James Robinson (2012) have
based an even wider-ranging theory of development on extensive analysis
of historical evidence, arguing that political institutions are the key deter-
minant of the long-run economic success or failure of nations. In their
scheme, extractive institutions allow narrow elites to capture a society’s re-
sources for their own benefit, whereas inclusive institutions spread eco-
nomic benefits more widely. In the long run, extractive institutions deter
innovation and economic growth, whereas inclusive institutions foster edu-
cation and technology, and consequently growth.
There is widespread agreement among economists who study economic
growth that technological change is its most important long-run source.
And economists have recognized that talented individuals account for most
innovations (Arrow 1962). In view of this, Simon Kuznets (1962) declared
that “we need far more empirical study than we have had so far of the uni-
verse of inventors” (32). Historical studies have now responded to this ap-
peal, with new findings about innovators.
A major revision concerns the life cycle of human creativity. A long-
standing belief has been that creativity is predominantly a prerogative of
youth. Yet this belief is wrong. One recent study of the careers of nearly
3,000 physicists concluded that the timing of a scientist’s most important
publications was randomly distributed within the scientist’s body of work
(Sinatra et al. 2016). And analysis of the careers of innovators in a wide
variety of activities has shown why this is so. In nearly every intellectual
activity, there are two distinctly different types of creativity, each of which
is associated with a very different pattern of discovery over a career. Con-
ceptual innovators make sudden breakthroughs by formulating new ideas,
creating syntheses of concepts that had not previously been considered to
be related. The most radical of these new ideas usually occur early in in-
novators’ careers, when they are least constrained by acquired habits of
thought. In contrast, experimental innovators work inductively and grad-
ually. Their greatest contributions generally arrive after long periods of
research, when they have accumulated great knowledge of their subject.
Among important conceptual innovators in the modern era, Albert Ein-
stein, Herman Melville, Pablo Picasso, T. S. Eliot, Orson Welles, Andy War-

hol, and Bob Dylan all made their landmark contributions between the
ages of 24 and 36, whereas the great experimental innovators Charles Dar-
win, Mark Twain, Paul Cézanne, Frank Lloyd Wright, Robert Frost, Irving
Berlin, and Alfred Hitchcock made their greatest contributions between
the ages of 48 and 76 (Galenson 2006).
The field of economic history is extremely broad in its coverage of both
time and space, and scores of additional areas of research could be added
to those summarized briefly above. Yet these serve as significant examples
of the gains that have resulted from the systematic application of economic
theory and econometrics to large bodies of historical evidence, both in
the precision of our knowledge of the past and in the confidence we can
attach to generalizations about economic change.
References
Acemoglu, Daron, and James Robinson. 2012. Why Nations Fail: The Origins of
Power, Prosperity, and Poverty. New York: Crown Bus.
Arrow, Kenneth. 1962. “Economic Welfare and the Allocation of Resources for
Innovation.” In The Rate and Direction of Inventive Activity: Economic and Social
Factors, edited by Richard Nelson, 609–25. Chicago: Univ. Chicago Press.
Bogue, Allan. 1963. From Prairie to Cornbelt: Farming on the Illinois and Iowa Prairies
in the Nineteenth Century. Chicago: Univ. Chicago Press.
Conrad, Alfred, and John Meyer. 1958. “The Economics of Slavery in the Ante-
bellum South.” J.P.E. 66 (2): 95–130.
Curti, Merle. 1959. The Making of an American Community: A Case Study of Democracy in
a Frontier County. Stanford, CA: Stanford Univ. Press.
Engerman, Stanley, and Kenneth Sokoloff. 2012. Economic Development in the
Americas since 1500: Endowments and Institutions. Cambridge: Cambridge Univ.
Press.
Fogel, Robert, and Stanley Engerman. 1974. Time on the Cross: The Economics of
American Negro Slavery. 2 vols. Boston: Little, Brown.
———. 1977. “Explaining the Relative Efficiency of Slave Agriculture in the An-
tebellum South.” A.E.R. 67 (3): 275–96.
Galenson, David. 1981. “The Market Evaluation of Human Capital: The Case of
Indentured Servitude.” J.P.E. 89 (3): 446–67.
———. 1996. “The Settlement and Growth of the Colonies: Population, Labor,
and Economic Development.” In The Cambridge Economic History of the United States,
vol. 1, edited by Stanley Engerman and Robert Gallman, 135–207. Cambridge:
Cambridge Univ. Press.
———. 2006. Old Masters and Young Geniuses: The Two Life Cycles of Artistic Creativ-
ity. Princeton, NJ: Princeton Univ. Press.
Kuznets, Simon. 1962. “Inventive Activity: Problems of Definition and Measure-
ment.” In The Rate and Direction of Inventive Activity: Economic and Social Factors,
edited by Richard Nelson, 19–51. Chicago: Univ. Chicago Press.
Menard, Russell. 1977. “From Servants to Slaves: The Transformation of the
Chesapeake Labor System.” Southern Studies 16 (4): 355–90.
Pope, Clayne. 1989. “Households on the American Frontier: The Distribution of
Income and Wealth in Utah, 1850–1900.” In Markets in History, edited by David
Galenson, 148–89. Cambridge: Cambridge Univ. Press.

Sinatra, Roberta, et al. 2016. “Quantifying the Evolution of Individual Scientific
Impact.” Science 354 (6312): aaf5239-1–8.
Thernstrom, Stephan. 1973. The Other Bostonians: Poverty and Progress in the Amer-
ican Metropolis, 1880 –1970. Cambridge, MA: Harvard Univ. Press.
Political Economics in the Journal of Political

Economy: Six Landmark Papers
Roger Myerson
When the Journal of Political Economy was founded, the term “political
economy” was commonly used to denote the academic field that we
now call “economics,” where most scholars focus more on the study of
markets than on the study of politics and government. But the JPE has
published important papers that apply the theoretical and empirical
methodologies of modern economics to the study of politics and politi-
cal institutions, a subfield which may be called “political economics.”
Duncan Black’s 1948 JPE paper “On the Rationale of Group Decision-
Making” was written with the explicit goal of providing a basis for the de-
velopment of a pure science of politics. Black analyzed the fundamental
problems of public decision making by considering the problems of vot-
ing in a committee that must choose among some given set of alterna-
tives. He found that, if the alternatives are points on a line and each voter
prefers alternatives that are closer to his or her ideal point, then the me-
dian of the voters’ ideal points can be identified as the choice that is pre-
ferred by a majority over any other alternative. In other cases in which this
one-dimensional structure is lacking, however, Black recognized that the
committee’s decision could depend on the order in which the alterna-
tives are considered. That is, in the general case, the outcome of social
decision making can depend on details about who gets to set the agenda.
These fundamental problems of social choice theory were revisited in
Kenneth Arrow’s 1950 JPE paper on “A Difficulty in the Concept of Social
Welfare,” which proved an early version of Arrow’s famous possibility the-
orem. In this paper, Arrow listed some basic consistency conditions for
how a social preference ordering should depend on individual prefer-
ences, and he showed that if there are three or more alternatives for so-

ciety to rank, then collective decision making cannot satisfy these consis-
tency conditions unless one individual is a dictator.
For an introductory example to illustrate the potential problems for
consistent social decision making, Arrow considered the question of
how to choose among three alternatives {A, B, C } when there are three
voters: one voter who ranks A best and C worst, one voter who ranks B
best and A worst, and one voter who ranks C best and B worst. In this sim-
ple society, a majority of the voters (two out of three) prefer A over B, but
another majority of the voters prefer B over C, and yet another majority
of the voters prefer C over A. Duncan Black’s 1948 paper also discussed
this paradoxical example, which is now known as the Condorcet cycle, al-
though Condorcet’s original version was rather more complicated. A
general social choice rule that treats alternatives symmetrically (neutral-
ity) and treats voters symmetrically (anonymity) could not identify a sin-
gle chosen alternative for this example.
In 1970, Amartya Sen published a JPE paper on “The Impossibility of a
Paretian Liberal,” which extended Arrow’s impossibility results by show-
ing a potential incompatibility between liberalism and Pareto’s criterion
for social efficiency. Sen illustrated his impossibility theorem with a sim-
ple paradoxical example that involves two individuals and a social choice
about which of them should read a certain pornographic book. A basic
axiom of liberalism might stipulate that, when society faces a question of
whether a particular book should be read by a particular individual i or
by nobody, the social choice should be determined according to i’s own
preferences. In Sen’s example, we suppose that at most one individual
can read the book. Individual 1, who is a prude, would most prefer that
nobody should read the book; but his second choice would be that he
(individual 1) should read it, and he would consider individual 2 read-
ing the book as the worst alternative. Individual 2 would prefer reading
the book herself rather than having nobody read it, but 2’s first prefer-
ence would be that individual 1 (the prude) should read the book. Then
liberalism would require that society should accept 1’s ranking that no-
body reading the book is better than individual 1 reading it, and liberal-
ism would also require that society should accept 2’s ranking that individ-
ual 2 reading the book is better than nobody reading it. But individual 2
reading the book is Pareto-dominated by individual 1 reading the book,
as that is the one comparison that both individuals agree about. Thus, lib-
eralism can lead to a Pareto-dominated social choice.
These impossibility theorems of social choice theory warn us that,
when a group must choose among three or more alternatives, we cannot
generally expect to predict what their collective choice will be (or should
be) only on the basis of the members’ individual preferences over these
alternatives. The outcome must also depend on the process by which the
group makes collective decisions. That is, public policies must depend

not only on the economic fundamentals that determine individuals’
preferences over these policies but also on the structure of the political
institutions in which public policy decisions are actually made. Every so-
ciety has constitutional rules that delegate substantial social decision-
making powers to a smaller group of individuals, including leaders who
can set the agenda for social decision making (with powers to determine
which alternatives will be considered and in what order) and legislative
representatives who are empowered to vote on policies for their constit-
uents who elected them.
Thus, since 1980, the literature of political economics has increasingly
focused on questions about how the quality of social choices will depend
on the structure of political institutions. We want to understand how the
structure of political institutions can affect the conduct of political lead-
ers and the performance of government in public policy making.
For example, a 1981 JPE paper by Barry Weingast, Kenneth Shepsle,
and Christopher Johnsen examined reasons why elected local represen-
tatives in a legislature may have incentives to choose inefficiently high
levels of public spending. Each locally elected representative cares pri-
marily about the benefits of public spending in his own district, while
the costs of public spending are spread among all districts, and so sys-
tematic biases may be introduced by the basic structure of representative
democracy. Even on a public policy question that has the kind of simple
one-dimensional structure that Duncan Black assumed, the selected pol-
icy might be determined not by the preferences of the median voter in
the nation but by the median among all voters who are at the median of
their respective districts.
The 1988 JPE paper by Barry Weingast and William Marshall considers
the problem of creating a market for Pareto-improving transactions in a
legislature that has responsibility for determining multidimensional
public policies that it can revise by a majority vote at any time. A simplis-
tic analogy with economic markets might suggest a market for dynamic
vote trading, where one representative might sell his vote on some policy
dimensions in exchange for others’ votes on a dimension that he cares
more about. But even if such exchanges were legal and enforceable, ef-
ficient vote trades for dynamic policy making would generally require
that promises of future votes should be dependent on future political
conditions that may be too complex to be stipulated in practical con-
tracts.
Weingast and Marshall argue that a legislature can solve this problem
by allocating agenda control over different policy dimensions to differ-
ent legislative committees and then letting legislators bid for seats in
their most-preferred committees. Thus, a committee system allows that
an individual legislator can have some confidence of his ability to pre-

vent adverse changes of policy on a dimension that is particularly impor-
tant to him by preventing such changes from coming to a vote in the leg-
islature. On the other hand, the fact that committee-approved legislation
cannot be enacted without ratification by a majority of the legislature ef-
fectively constrains the committee’s power to change policy except in po-
litical conditions in which many other legislators would be receptive to
such changes. Thus, rules that vest legislative agenda control in a strong
committee system, as in the US Congress, can be understood as a practi-
cal institutional adaptation to the difficulties of trading votes in a dynamic
policy-making process. Weingast and Marshall also observe that, in coun-
tries with strongly disciplined political parties, transactions for Pareto im-
provement in dynamic legislative policy making may instead be facilitated
by party leaders, who serve as trusted intermediaries for exchanges of
promises of political support.
Weingast and Marshall’s 1988 JPE paper is exemplary of research in
political economics papers that applies concepts of economic analysis
for better understanding of political institutions. This strand in the liter-
ature treats political institutions as analytically similar to economic mar-
kets, in that both can be understood as systems of competitive interac-
tions among rational agents. But the literature in political economics
also includes many papers in which the focus is on how political institu-
tions affect the performance of economic markets themselves. The 2005
JPE paper by Daron Acemoglu and Simon Johnson is an outstanding
example of this part of the literature.
Economic markets depend on political and legal institutions for the
enforcement of property rights and contracts. Acemoglu and Johnson
suggest a way to distinguish these two effects empirically in nations that
were formerly European colonies. Former European colonies have gen-
erally adopted different legal mechanisms for contract enforcement ac-
cording to whichever country in Europe was the colonizing power. But
the quality of property rights enforcement in different colonies has tended
to depend more on the degree to which Europeans could expect to make
long-term settlements in the colony, which in turn depended on mortality
rates for potential European settlers and population density around 1500.
Thus, Acemoglu and Johnson (2005, 971) argue that “the way in which
countries were colonized, but not who colonized them, is a robust determi-
nant of property rights institutions, whereas who colonized, but not the
details of colonization strategy, shapes contracting institutions.”
With these instruments, Acemoglu and Johnson find that the histori-
cal development of property rights institutions has had a significant ef-
fect on long-run economic growth, investment, and financial develop-
ment. In contrast, they find that institutions for contract enforcement
appear to have mattered mainly for the relative importance of different

forms of financial intermediation (debt or equity). To understand these
results theoretically, Acemoglu and Johnson observe that any adverse ef-
fects of their government’s weakness in contract enforcement could be
mitigated by citizens organizing social networks with their own private
systems of contract enforcement. But ordinary citizens would have no
such remedy if institutions for property rights failed to protect private
investments from expropriation by those who control the government,
and so millions of people would lose any incentive to invest in improving
their economic situation. This basic recognition, that failures of economic
development can be caused by dysfunctional political institutions, is a fun-
damental motivation for ongoing research in political economics, apply-
ing economic analysis to the comparative study of political institutions.
References
Acemoglu, Daron, and Simon Johnson. 2005. “Unbundling Institutions.” J.P.E.
113 (5): 949–95.
Arrow, Kenneth. 1950. “A Difficulty in the Concept of Social Welfare.” J.P.E. 58
(4): 328–46.
Black, Duncan. 1948. “On the Rationale of Group Decision-Making.” J.P.E. 56
(1): 23–34.
Sen, Amartya. 1970. “The Impossibility of a Paretian Liberal.” J.P.E. 78 (1): 152–
57.
Weingast, Barry R., and William J. Marshall. 1988. “The Industrial Organization
of Congress; Or, Why Legislatures, like Firms, Are Not Organized as Markets.”
J.P.E. 96 (1): 132–63.
Weingast, Barry R., Kenneth A. Shepsle, and Christopher Johnsen. 1981. “The
Political Economy of Benefits and Costs: A Neoclassical Approach to Distrib-
utive Politics.” J.P.E. 89 (4): 642–64.
Aggregative Fiscal Policy

Nancy L. Stokey
I. Introduction
There is some truth in the old adage that in economics the questions
never change, only the answers. This note looks at three questions in ag-

gregative fiscal policy that have been the subject of recent JPE contribu-
tions and also have long prior histories: the use of debt finance, the role
of commitments about future policy, and the size of government.
II. Debt Finance

What are the effects of financing government spending with debt rather
than contemporaneous taxation?
David Ricardo (1817) raised the issue, discussing both ordinary ex-
penses of the state and unusual expenses like wars. He asserted that bor-
rowing by the government “is a system which tends to make us less thrifty—
to blind us to our real situation.” As an example, Ricardo considered an
individual who must pay £100 for the expense of a war, saying that “he
would endeavor, on being at once called upon for his portion, to save
speedily the £100 from his income. By the system of loans, he is called
on to pay only the interest of this £100, or £5 per annum, and considers
that he does enough by saving this £5 from his expenditure, and then de-
ludes himself with the belief that he is as rich as before” ([1817] 2004,
chap. 18, 163).
In “Are Government Bonds Net Wealth?” (1974), Robert Barro stud-
ied the effect of debt finance on consumption and saving in a setting
in which it is later generations who will repay the debt. To this end he
uses a variant of the Samuelson (1958)–Diamond (1965) overlapping
generations model. Each individual works when young and uses his earn-
ings for consumption and asset accumulation. When old he may receive
a (nonnegative) inheritance and uses his assets to finance consumption
and, perhaps, a bequest to his heir.
Generations are altruistically linked through preferences: each indi-
vidual values the utility of his heir. Thus, one can readily construct “dy-
nastic preferences” over the consumptions of all the members and a “dy-
nastic budget constraint” augmented by an additional set of constraints
representing the fact that bequests must be nonnegative.
Barro asks about the effect of a one-time government transfer to the
old, financed with bonds. In every subsequent period the government
uses a lump-sum tax to finance the interest on the debt and rolls over
the principal. It is easy to show that the bond injection leaves the dynastic
budget constraint unchanged. Consequently, if, with no transfer, debt, or
taxes, all of the bequests would have been strictly positive, then they will
remain so: each bequest will be increased by the size of the next genera-
tion’s tax liability. The dynasty’s consumption plan will be unchanged,
and the bond-financed transfer has no effect.
As Buchanan (1976) pointed out, Ricardo had noticed this argument
about the capitalization of future tax obligations. However, as Buchanan

suggested and O’Driscoll (1977) argued more forcefully, Ricardo did not
accept the argument as a description of behavior. O’Driscoll quotes Ri-
cardo as saying “but the people who pay the taxes never so estimate them,
and therefore do not manage their private affairs accordingly” (208). Evi-
dently Ricardo himself was not a Ricardian.
In his subsequent paper “On the Determination of the Public Debt”
(1979), Barro noted that even if one accepts the hypothesis about Ri-
cardian equivalence, it holds only if taxes are lump-sum. If revenue must
be raised with distorting taxes, a second consideration appears.
Since the deadweight loss from a distorting tax is convex in the tax rate,
given a fixed revenue requirement, the total distortion is reduced by levy-
ing lower taxes on many commodities instead of a high tax on a single
commodity. Thus, if government expenditure is uneven over time, the to-
tal distortion is lower if the tax rate remains approximately constant and
the government issues debt and/or acquires assets, borrowing and lend-
ing as required, to keep the tax rate smooth. Financing extraordinary ex-
penditures like wars is an obvious example, but the principle applies more
broadly.
Of course, this idea also has an earlier antecedent, in Frank Ramsey’s
(1927) classic study. Ramsey looked at a static problem, where the issue
is taxing various commodities at possibly different rates. But his idea ap-
plies as well to a dynamic setting in which the commodities are dated
consumption goods.
III. The Role of Commitment
Can government policies that seem attractive in the short run be detri-
mental in the long run? Can policy ever be improved by taking options
away from a benevolent government?
Soon after the adoption of the American Constitution, the new Trea-
sury Secretary, Alexander Hamilton, argued in his First Report on Public
Credit (1790) for assumption and repayment by the federal government
of all outstanding debt that the states had issued during the Revolution-
ary War, asserting that full repayment would establish the nation’s rep-
utation with credit markets and allow it to borrow in the future at af-
fordable interest rates. Thomas Jefferson and James Madison took the
opposite side of the debate, arguing that much of the debt had been
bought up by speculators at less than face value and that those creditors
should be paid less. In the end Hamilton prevailed, and all of the debt
was repaid.
In “Rules Rather than Discretion: The Inconsistency of Optimal Plans”
(1977), Finn Kydland and Edward Prescott identified a fundamental lim-
itation in using the tools of dynamic optimization to formulate govern-

ment policy. The key issue is that “current decisions of economic agents
depend in part upon their expectations of future policy actions” (474; em-
phasis added).
Consequently, a government at one date may announce/suggest a pol-
icy for a later date, because of the beneficial effect on current private-
sector behavior. But the (benevolent) policy maker at the later date—if it
can—may choose to ignore the suggestion and do something else instead.
Of course, if agents in the private sector anticipate this change, it under-
mines the beneficial effect the announcement was intended to have.
What Kydland and Prescott call a rule is a limitation, through a consti-
tutional provision or legislation or some other means, on what a policy
maker can do. By contrast, a regime of discretion allows the policy maker
wide latitude in deciding how to respond. In many situations a rule may
be strictly preferred—by all—to a regime with discretion.
As an example, they describe the problem faced by a central bank. Un-
der a regime with discretion, it may be tempted to announce a low infla-
tion target and then exceed it, with the goal of reducing unemployment.
But rational agents will anticipate the additional inflation, so unemploy-
ment is unaltered and inflation is higher than desired. A rule would tie
the central bank’s hands.
Similarly, in the short run, capital income seems to be an excellent tar-
get for taxation, since it is supplied inelastically. But in the longer run,
such a tax discourages investment. A government could tax capital in-
come heavily in the short run and at the same time promise very low taxes
in the future. But is the promise credible?
And finally, defaulting on public debt, either explicitly or implicitly—
through inflation—reduces the need to levy distorting taxes, a pure ben-
efit. But if potential lenders anticipate a risk of default, the cost of sub-
sequent borrowing rises. Perhaps Jefferson had a change of heart about
Hamilton’s position when, during his presidency, debt was issued to fi-
nance the Louisiana Purchase, and Madison during his presidency, when
it was used to finance the War of 1812.
IV. The Size of Government

What determines the size of government?
Alexis de Tocqueville (1835) argued that it depends on which of three
classes is in control of making the laws. If it is the rich, “probably it will be
little enough concerned about economizing on public funds, because a
tax that strikes a considerable fortune takes away only from the surplus
and produces little sensible effect.” If the middle classes make the laws,
“they will not be prodigal with taxes because there is nothing so disas-
trous as a large tax striking a small fortune.” Finally, if those with “little
or no property” make the laws, public costs will be higher, since “all the

money that is expended in the interest of society seems able only to profit
them without ever harming them,” and they are capable of finding “the
means of assessing the tax in a manner that strikes only the rich and prof-
its only the poor” (2000, vol. 1, pt. 2, chap. 5, 200).
Since the “different categories can be more or less numerous,” in a de-
mocracy the group sizes and the extent of the franchise determine which
group is decisive.
In “A Rational Theory of the Size of Government” (1981), Allan Melt-
zer and Scott Richard expanded on this idea, developing a model in
which the only role of the government is to redistribute income, and its
only instrument is a flat-rate tax on earnings, used to finance lump-sum
transfers. The government’s budget is balanced, and the size of govern-
ment—the share of tax revenue in total income—is determined by voting.
Individuals are identical in terms of their preferences over consump-
tion and leisure but heterogeneous in terms of labor productivity. Given
the tax system in place, each individual chooses how much time to spend
working.
As voters, individuals realize that a higher tax rate, up to a point, fi-
nances a bigger transfer, although they rationally forecast the disincen-
tive effects of taxation: there is no “fiscal illusion.” Hence preferences
over the tax rate are monotone in the individual’s own productivity, with
all those below some threshold preferring to implement the revenue-
maximizing rate and doing no work.
The political equilibrium is determined by the median voter, so it de-
pends on the distribution of productivity across individuals and the ex-
tent of the franchise. For a fixed distribution of income, expanding voting
rights to groups with lower incomes increases the demand for redistribu-
tion. Once suffrage is universal, the growth of government depends only
on changes in the distribution of income. If economic growth increases in-
come inequality, then the model predicts that growth also increases the
demand for redistributive taxation.
V. Conclusion
Are the formal models of modern economics only an embellishment of
older ideas? They are more than that, for two reasons. First, they clarify
exactly what is being asserted, providing a more solid base from which
further theoretical arguments can proceed. In addition, they provide a
guide for empirical work, suggesting what kinds of data should be gath-
ered and how they should be used to examine the competing hypotheses.
The old adage is only partly correct: the questions get sharper and clearer,
even if entirely satisfactory answers remain elusive. Unsurprisingly, re-
search on these three questions has continued: all of these papers are
among the most highly cited in the JPE over its 125 years.

References
Barro, Robert J. 1974. “Are Government Bonds Net Wealth?” J.P.E. 82 (Novem-
ber/December): 1095–1117.
———. 1979. “On the Determination of the Public Debt.” J.P.E. 87, no. 5, pt. 1
(October): 940–71.
Buchanan, James M. 1976. “Barro on the Ricardian Equivalence Theorem.” J.P.E.
84 (April): 337–42.
Diamond, Peter A. 1965. “National Debt in a Neoclassical Growth Model.” A.E.R.
55 (December): 1126–50.
Hamilton, Alexander. 1790. First Report on Public Credit. Washington, DC: Trea-
sury Dept.
Kydland, Finn E., and Edward C. Prescott. 1977. “Rules Rather than Discretion:
The Inconsistency of Optimal Plans.” J.P.E. 85 ( June): 473–92.
Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of
Government.” J.P.E. 89 (October): 914 –27.
O’Driscoll, Gerald P., Jr. 1977. “The Ricardian Nonequivalence Theorem.” J.P.E.
85 (February): 207–10.
Ramsey, Frank P. 1927. “A Contribution to the Theory of Taxation.” Econ. J.
37:47–61.
Ricardo, David. [1817] 2004. The Principles of Political Economy and Taxation. Re-
print. Mineola, NY: Dover.
Samuelson, Paul A. 1958. “An Exact Consumption-Loan Model of Interest With
or Without the Social Contrivance of Money.” J.P.E. 66 (December): 467–82.
Tocqueville, Alexis de. [1835] 2000. Democracy in America. Translated and edited
by Harvey C. Mansfield and Delba Winthrop. Chicago: Univ. Chicago Press.
Business Cycles and International Trade

Harald Uhlig
My essay will examine two rather separate topics, though there is a bit of
a connection. One concerns business cycles. The other concerns inter-
national trade and exchange rates. With all due apologies and very few
exceptions, I shall focus on the most highly cited papers published in
the Journal of Political Economy.
Business Cycles
The 1970s and early 1980s saw a revolution in our thinking about mac-
roeconomics generally and business cycles specifically. Central to this de-

velopment was a revolutionary paradigm shift in how the expectations of
agents regarding the future should be taken into account in their cur-
rent choices, notably for consumption, mostly insisting that these expec-
tations should be rational and agree with the formulation of probability
theory. Before that revolution, it was customary to assume a consumption
function, according to which aggregate consumption rises in somewhat
less than a proportional manner and according to the marginal propen-
sity to consume, when current aggregate income rises, regardless of con-
cerns about developments of these incomes in the future. That older
paradigm is still remarkably alive in quite a number of undergraduate
textbooks on macroeconomics and policy discussions, but it has been en-
tirely upended by the rational expectations revolution as far as the scien-
tific analysis and thinking about business cycles and other economic phe-
nomena are concerned: only in special cases then does the old paradigm
still work. From the beginning and in particular in the work by Sargent,
the hypothesis of rational expectations was connected to the issue of how
rational expectations or other forms of expectations can be learned (see
Uhlig 2012a).
Perhaps the most seminal contribution in this (then) new thinking
about consumption is the paper by Lucas (1978). He examined the op-
timization problem of an agent with time-separable preferences, who can
freely trade assets with stochastic returns R t11 in period t on one unit of
resources invested at t. He derived what is now typically referred to as the
Lucas asset pricing equation:
u 0 ðct11 Þ
1 5 Et ½Mt11 R t11 , where Mt11 5 b ,
u 0 ðct Þ
where u0 (ct) is the period t felicity for an agent, consuming ct; b is the dis-
count factor; Et denotes the conditional expectation, given all available
information at time t ; and Mt11 is the resulting stochastic discount factor.
The first part of the equation also holds for far more general preference
formulations and has given rise to a substantial literature on asset pricing,
as other essays in this issue discuss.
Here I shall focus on the macroeconomic implications and (mostly)
keep to the separable formulation. Assuming R t11 5 1 1 r to be a con-
stant and safe return, it then follows that detrended marginal utility
½bð1 1 r Þt u 0 ðct Þ is a random walk. If, moreover, utility is quadratic, then
consumption likewise detrended is itself a random walk with drift. These
are the celebrated results in Hall (1978), who writes that therefore “con-
sumption is unrelated to any economic variable that is observed in earlier
periods. In particular, lagged income should have no explanatory power
with respect to consumption” (972). Hall proceeds to test and to then
confirm these permanent-income predictions of the theory, while Sar-

gent (1978) instead obtains a rather decisive rejection. Flavin (1981) rec-
onciles these two apparently conflicting findings. She rejects the joint
rational expectations–permanent income hypothesis and finds that con-
sumption exhibits excess sensitivity to current income.
For that exercise, it is ultimately crucial to estimate the revision in per-
manent income and the persistent reaction of income due to current
news. Cochrane (1988) estimates the permanent reaction to be fairly small.
The literature on persistence, unit roots, and cointegration since then has
grown to impressive size.
Hall (1988) assumes that uðcÞ 5 c 121=j so that j is the intertemporal
elasticity of substitution: a popular specification in much of macroeco-
nomics. Exploiting time variation in R t11 , he calculates various estimates
of j and generally finds them to be small, near zero, or even negative. The
macroeconomic literature since then has tended to assume j to be be-
tween 0.5 and 1, and sometimes as low as 0.2, as well as allowed for ex-
tensions such as habit formation and borrowing-constrained or hand-to-
mouth consumers. That literature furthermore typically assumes the log
of total factor productivity (TFP) to exhibit short-run fluctuations around
a time trend or to be a random walk with drift. These are then ingredients
for building more substantial business cycle models.
The revolution in thinking about business cycles was to view them as
equilibrium phenomena, where agents optimally react to shocks and pol-
icy changes, utilizing rational expectations. The program was laid out in
Lucas (1975), though that paper did not yet feature preference-based
optimizing behavior of agents. The program was completed in particular
in the seminal contribution of Kydland and Prescott (1982), extending
the stochastic neoclassical growth theory and giving rise to real business
cycle theory. The contribution by Long and Plosser (1983) allows for a
rich industry structure: a theme that recently has received considerable
renewed attention in the production-network-based analysis of macro-
economic fluctuations. Real business cycle theory postulates that aggre-
gate fluctuations are driven by exogenous fluctuations in TFP rather than,
say, exogenous fluctuations in “aggregate demand” (which now would
need to be derived from exogenous fluctuations in preference parame-
ters) or policy. Prices and wages are assumed to be flexible and markets
are assumed to clear. Many now dismiss such flexibility out of hand as
unrealistic, and the literature has since moved to typically imposing a
range of other frictions. Then again, one may argue that “reality” provides
a better rationale in its favor than may be apparent at first (see Uhlig
2012b).
The real business cycle paradigm has since been extended and criti-
cally examined in a variety of ways. Backus, Kehoe, and Kydland (1992)
extend the paradigm to the international realm, providing a connection

between the two sections here. Empirically, Hamilton (1983) argues that
oil prices provide a substantial source of aggregate fluctuations, with Mork
(1989) arguing that the effect is asymmetric and much stronger for oil
price increases than oil price decreases. Basu and Fernald (1997) is an im-
portant paper, examining the intricacies of measuring the exogenous
component of TFP and the challenges in utilizing it as a driving force.
More recent versions of business cycle theories enrich them with a consid-
erably larger set of shocks and frictions. In particular, the assumption of
sticky prices is appealing to many and has become a standard ingredient
of most of the business cycle analysis in the recent decade or so. One im-
portant example is the framework by Christiano, Eichenbaum, and Evans
(2005), which, together with the related Smets and Wouters (2003) model,
has become the blueprint for many workhorse models used in central
banks around the world for policy analysis. In the wake of the financial cri-
sis of 2008, these models have recently become extended by paying greater
attention to financial intermediation and the role of the financial frictions
more generally. At its core, all these models still feature a real business cycle
engine, albeit modified and extended in considerable ways.
International Trade and Exchange Rates

Balassa (1964) together with Samuelson (1964) is the classic source for
the well-known Balassa-Samuelson effect, that the purchasing power par-
ity or consumer price level is higher in richer countries.
Dornbusch (1976) develops his classic exchange rate overshooting re-
sult for exchange rates. He assumes perfect foresight: the companion to
rational expectations, if there are no further stochastic disturbances in
the future. He considers a monetary expansion in a model of perfect cap-
ital mobility and slow adjustments of goods markets. He demonstrates
that the initial and immediate depreciation of the exchange rate is then
followed by a gradual appreciation of the exchange rate, to compensate
for the ensuing inflation differential. Lothian and Taylor (1996) use unit
root econometric methods, freshly developed in the decade prior to the
publication of their paper, and demonstrate that the dollar-sterling and
the franc-sterling real exchange rates are stationary, an important issue
for the construction of international trade models.
The study of international trade and exchange rates has undergone
profound paradigm shifts in the last few decades. It has incorporated the
macroeconomic paradigm shift toward rational expectations and general
equilibrium analysis described in the first section. The new trade theory
furthermore views trade as arising from imperfect competition between
possibly multinational firms, each producing its own variety.
Helpman (1984) provides a simple theory of international trade with
multinational corporations, building on then-recent advances in analyz-

ing vertical integration and international trade in differentiated products.
For production of a specific variety, he distinguishes between a general-
purpose input possibly produced elsewhere, such as management, distri-
bution and product-specific R&D, and local labor. Multinational corpo-
rations with entrepreneurial centers and subsidiaries together with their
location decision then arise endogenously, explaining the simultaneous
existence of intersectoral trade, intraindustry trade, and intrafirm trade.
More recently, Antràs and Helpman (2004) examine the issue of global
sourcing and the choice of organizational form for firms in international
trade and relate sectoral productivity dispersion and headquarter inten-
sity to the degree of integration and input imports.
Backus et al. (1992) have extended the real business analysis described
in the previous section to a two-country setting. While they do not feature
firm heterogeneity or sticky prices, they emphasize in particular the role
of the capital stock and capital investment. More recent trade models of-
ten abstract from physical capital accumulation, though it may remain
fruitful to include such forces as well.
Obstfeld and Rogoff (1995) critically reexamine the Dornbusch over-
shooting result as well as a number of other classic predictions in a new
two-country model. Their model marries global macroeconomic dynam-
ics to a supply framework based on monopolistic competition and sticky
nominal prices, thereby providing novel insights into the dynamics of ex-
change rates and current accounts. It has become a benchmark and work-
horse model in this field of inquiry. The latest generation of international
trade models builds on the seminal contributions of Eaton and Kortum
(2002) and Melitz (2003), focusing on matters such as firm entry and exit
as well as trade costs, which the JPE unfortunately missed out on publish-
ing: at least Samuel Kortum was on the faculty at the University of Chi-
cago for a number of years. The field has been moving forward quickly
in recent years, and these developments will be exciting to watch or to
participate in.
References
Antràs, Pol, and Elhanan Helpman. 2004. “Global Sourcing.” J.P.E. 112 (3): 552–
80.
Backus, David K., Patrick J. Kehoe, and Finn E. Kydland. 1992. “International Real
Business Cycles.” J.P.E. 100 (August): 745–75.
Balassa, Bela. 1964. “The Purchasing-Power Parity Doctrine: A Reappraisal.”
J.P.E. 72 (December): 584 –96.
Basu, Susanto, and John G. Fernald. 1997. “Returns to Scale in U.S. Production:
Estimates and Implications.” J.P.E. 105 (April): 249–83.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. “Nom-
inal Rigidities and the Dynamic Effects of a Shock to Monetary Policy.” J.P.E.
113 (February): 1–45.

Cochrane, John H. 1988. “How Big Is the Random Walk in GNP?” J.P.E. 96 (Oc-
tober): 893–920.
Dornbusch, Rudiger. 1976. “Expectations and Exchange Rate Dynamics.” J.P.E.
84 (December): 1161–76.
Eaton, Jonathan, and Samuel Kortum. 2002. “Technology, Geography, and Trade.”
Econometrica 70 (5): 1741–79.
Flavin, Marjorie A. 1981. “The Adjustment of Consumption to Changing Expec-
tations about Future Income.” J.P.E. 89 (October): 974–1009.
Hall, Robert E. 1978. “Stochastic Implications of the Life Cycle–Permanent In-
come Hypothesis: Theory and Evidence.” J.P.E. 86 (December): 971–87.
———. 1988. “Intertemporal Substitution in Consumption.” J.P.E. 96 (April):
339–57.
Hamilton, James D. 1983. “Oil and the Macroeconomy since World War II.” J.P.E.
91 (April): 228–48.
Helpman, Elhanan. 1984. “A Simple Theory of International Trade with Multi-
national Corporations.” J.P.E. 92 ( June): 451–71.
Kydland, Finn E., and Edward C. Prescott. 1982. “Time to Build and Aggregate
Fluctuations.” Econometrica 50:1345–70.
Long, John B., and Charles I. Plosser. 1983. “Real Business Cycles.” J.P.E. 91 (Feb-
ruary): 39–69.
Lothian, James R., and Mark P. Taylor. 1996. “Real Exchange Rate Behavior: The
Recent Float from the Perspective of the Past Two Centuries.” J.P.E. 104 ( June):
488–509.
Lucas, Robert E., Jr. 1975. “An Equilibrium Model of the Business Cycle.” J.P.E.
83 (December): 1113–44.
———. 1978. “Asset Prices in an Exchange Economy.” Econometrica 46 (6): 1429–
45.
Melitz, Marc J. 2003. “The Impact of Trade on Intra-industry Reallocations and
Aggregate Industry Productivity.” Econometrica 71:1695–1725.
Mork, Knut Anton. 1989. “Oil and the Macroeconomy When Prices Go Up and
Down: An Extension of Hamilton’s Results.” J.P.E. 97 ( June): 740–44.
Obstfeld, Maurice, and Kenneth Rogoff. 1995. “Exchange Rate Dynamics Re-
dux.” J.P.E. 103 ( June): 624–60.
Samuelson, Paul A. 1964. “Theoretical Notes on Trade Problems.” Rev. Econ. and
Statis. 46 (2): 145–54.
Sargent, Thomas J. 1978. “Rational Expectations, Econometric Exogeneity, and
Consumption.” J.P.E. 86 (August): 673–700.
Smets, Frank, and Raf Wouters. 2003. “An Estimated Dynamic Stochastic General
Equilibrium Model of the Euro Area.” J. European Econ. Assoc. 1 (5): 1123–75.
Uhlig, Harald. 2012a. “Agents as Empirical Macroeconomists: Thomas J. Sar-
gent’s Contribution to Economics.” Scandinavian J. Econ. 114 (December):
1055–81.
———. 2012b. “Economics and Reality.” J. Macroeconomics 34:29–41.

Inequality, Heterogeneity, and Consumption

in the Journal of Political Economy
Greg Kaplan
Today, inequality and heterogeneity are front and center in macroeconom-

ics. Most macroeconomists agree that the distribution of household-level
variables, in particular consumption and wealth, matters for the dynamics
of macroeconomic aggregates and that macroeconomic shocks affect the
distribution of consumption and wealth across households. However, it
was not always this way. Getting to this point has been a long road, along
which papers published in the JPE have been essential guideposts.
I will discuss six influential papers from the JPE that have helped shape
the way in which inequality is studied by economists today. I have orga-
nized my discussion along three strands that have each contributed to
the introduction of heterogeneity into macroeconomic models: the mi-
croeconomics of consumption behavior, the use of structural models of
precautionary savings for insights about policy, and finally, the develop-
ment of general equilibrium models with heterogeneous households and
aggregate shocks. Within each of the three strands, I have chosen two de-
fining JPE papers that roughly correspond to the trajectory that this intel-
lectual journey has followed.
First I will examine two empirical papers—Zeldes (1989) and Attanasio
and Weber (1995)—that provided new insights into the microeconomics
of consumption behavior. Both are empirical analyses that are closely
guided by theory. These and other related empirical papers paved the
way for a class of structural models of consumption—precautionary sav-
ings models—that later became the workhorse models of heterogeneous
agent macroeconomics. I will next highlight two early examples of how
structural models of consumption can be used to explore the ways in
which economic policies shape the distribution of household outcomes.
Hubbard, Skinner, and Zeldes (1995) and Rosenzweig and Wolpin (1993)
illustrate two alternative ways in which structural models of consumption
can be disciplined by data and turned into quantitative laboratories; the
first calibrated their model of the United States using mostly external
sources, while the second estimated their model of rural India using max-
imum likelihood. In both models, precautionary motives drive household
responses to changing government policies. Finally, I will discuss two early

heterogeneous agent models with aggregate shocks: Imrohoroglu (1989),
one of the earliest partial equilibrium models, and Krusell and Smith
(1998), a now-seminal general equilibrium model. These took previously
stand-alone precautionary savings models and cast them in an equilibrium
framework exposed to aggregate shocks, thus opening the door for fully
fledged macroeconomic models of inequality.
Zeldes (1989) was one of the first papers to provide convincing evi-
dence using micro panel data that liquidity constraints are indeed impor-
tant for household-level consumption. The importance of his contribu-
tion is reflected in the fact that today, virtually every paper being written
about consumption, at either the individual or aggregate level, some-
where addresses the implications of binding liquidity constraints. The pa-
per is notable because rather than simply rejecting the implications of
consumption models that abstract from liquidity constraints, as had been
done by much of the preexisting literature, he carefully derives testable
implications of a model that includes liquidity constraints. Zeldes suggests
three novel tests for the importance of these constraints. The first test is
beautiful for its simplicity. Because the model with liquidity constraints
predicts that consumption growth should be sensitive to income for low-
wealth households but not for high-wealth households, Zeldes splits his
sample between these two groups of households and measures the sensitiv-
ity of consumption growth to income growth for each. Using data on food
consumption from the Panel Study of Income Dynamics (PSID), he con-
firms the predictions of the model: that low-wealth households have a
much lower sensitivity of consumption growth to income than high-
wealth households. Refined versions of this test for the presence of liquid-
ity constraints are still the go-to approaches in empirical analyses of con-
sumption behavior today. Many of these recent studies (and there are
many) essentially repeat Zeldes’s analysis using better-identified income
shocks and much larger and higher-quality data sets, reaching the same
conclusion.
Zeldes (1989) also derives two additional tests implied by the consump-
tion model with liquidity constraints: (i) that Lagrange multipliers on li-
quidity constraints should be positive for constrained households and
(ii) that the multipliers should be negatively related to current income.
He estimates Lagrange multipliers for low-wealth households by using
the residuals from their Euler equations when the remaining parameters
are estimated in the sample of high-wealth households. He finds that the
estimated multipliers are indeed positive and are negatively related to in-
come, as predicted. One leaves the paper with the sense that any model of
household consumption should treat liquidity constraints seriously. The
structural literature that followed did.
Another beautifully executed empirical analysis of consumption be-
havior that is also carefully motivated by theory—this time without liquid-

ity constraints—is Attanasio and Weber (1995). Using the Consumer Ex-
penditure Survey (CEX), the authors pioneered the approach of working
with synthetic cohorts (i.e., data moments for groups of households from
the same birth cohort with similar demographic characteristics) to over-
come the limitation that the CEX has only a short panel component, un-
like the PSID. This enabled them to exploit the comprehensive consump-
tion data in the CEX, rather than the PSID, which at the time contained
data on only food consumption.
Attanasio and Weber (1995) make three points. First, they show that
looking only at food consumption can be misleading because prefer-
ences are nonseparable between food consumption and other consump-
tion categories. Partly for this reason, the vast majority of the structural
consumption literature that has followed favors comprehensive measures
of consumption or explicitly models this nonseparability. Second, they
illustrate the pitfalls of using aggregate data to test models of hetero-
geneous households. By aggregating microdata in exactly the way pre-
scribed by theory, they show that there can be large differences between
the dynamics of the log of mean consumption (the focus of representa-
tive agent models and what is measured in aggregate data) and the dy-
namics of the mean of log consumption (the focus of heterogeneous
agent models and that can be constructed only with household-level data).
Third, they show that it is easy to spuriously reject frictionless life cycle
models of consumption if one ignores predictable changes in either house-
hold composition or labor supply of individual household members. They
explain how the hump-shaped age profiles for family size and female la-
bor supply would lead to a hump-shaped age profile for consumption—
an observation that had frequently been cited as evidence against the fric-
tionless model.
All three points lead one to rethink the numerous previous studies that
had seemingly shown deviations from frictionless consumption models,
including Zeldes (1989). Attanasio and Weber (1995) never actually claimed
that liquidity constraints and precautionary motives were not important,
only that the then-existing tests were much more fragile than one might
have thought. Reading both of these papers today, one is struck by the
careful connection between empirics and theory. Both papers carefully
explain their null and alternative hypotheses, build up their estimating
equations from precisely specified models, and go to great lengths to
spell out the assumptions required to go from their model to their regres-
sion equations. These are classic qualities of empirical analyses in the JPE.
Attanasio and Weber (1995) essentially highlight the limitations of ex-
amining theories of consumption through the lens of only a small subset
of a model’s predictions. By exposing these limitations, they drove later
papers to use a larger set of predictions that are obtained by explicitly com-
puting consumption and savings decisions under alternative parameteri-

zations. These structural papers that followed moved beyond testing mod-
els to using models to quantify the effects of public policies on household
consumption and savings behavior. Two JPE-published papers represent
some of the best early examples of how to utilize a quantitative structural
model of consumption for effective policy analysis.
Hubbard et al. (1995) is a classic example of the power of a calibrated
structural model. The authors observe that many households with low
lifetime incomes accumulate little or no wealth over their lifetimes. Even
just before retirement, when life cycle models of precautionary savings
predict that households should hold substantial wealth, many such house-
holds are essentially hand to mouth. According to life cycle consumption
theory, having low lifetime income, even in the presence of liquidity con-
straints, is no excuse for not saving for retirement: households should
smooth consumption, albeit at a low level. While liquidity constraints
might explain why households do not borrow, it does not explain why
they do not save. Hubbard et al. suggest a possible reason for the lack
of saving: asset-based means-tested public insurance programs, which they
model as a consumption floor, reduce the incentives for households to
save. The presence of a consumption floor not only reduces households’
exposure to consumption fluctuations—lowering their incentive to save
for precautionary reasons—but also implies an effective tax rate of 100 per-
cent on assets in the states of the world where the consumption floor binds.
The authors use a simple two-period model as an elegant theoretical
proof of concept. But the calibrated life cycle model, which is the meat
of the paper, provides two additional benefits. First, it acts as a quantita-
tive proof of concept, which, in my opinion, is one of the most valuable
benefits of quantitative structural analysis. It is one thing to show that asset-
based means-tested public insurance programs can distort savings deci-
sions; it is another to show that in empirically plausible settings these dis-
tortions are large enough to have an economically important effect on
observed savings. To do so, the authors choose parameter values that they
argue reflect US data, of which the most important are the stochastic pro-
cesses for earnings risk and medical expense risk, the level of the consump-
tion floor, and the degree of risk aversion. They then simulate their model
economy and show that it generates the aforementioned patterns of life cy-
cle wealth accumulation by lifetime income. Second, the calibrated model
can be used to evaluate the implications of alternative versions of means-
testing public programs for wealth accumulation, which the authors show
can be substantial.
Another example of how a quantitative structural model of precaution-
ary savings can be used to evaluate public policies is Rosenzweig and
Wolpin (1993). The authors consider the consumption-savings problem
of farmers in India whose only mechanism for smoothing consumption is
the accumulation of bullocks. Bullocks are also an input used in agricul-

tural production, making this paper one of the first examples of a struc-
tural model in which households save in a productive asset in the face of
idiosyncratic risk. The authors confront their model with panel data from
the International Crops Research Institute for the Semi-arid Tropics. They
construct the likelihood over sequences of farmers’ assets and profits and
use a two-stage maximum likelihood procedure to estimate preference pa-
rameters, prices of bullocks and other inputs, and production parame-
ters. Even today, Rosenzweig and Wolpin’s paper remains one of the
few examples of maximum likelihood estimation of a precautionary sav-
ings model of consumption using microdata. Their parameter estimates
imply underinvestment in bullocks on the part of farmers, as a result of
borrowing constraints and the inability of farmers to accumulate precau-
tionary savings in a financial asset. Through a series of counterfactual ex-
periments, the authors evaluate the relative merits of alternative interven-
tions. They find that the provision of actuarially fair weather insurance
would have little effect on farmer welfare, whereas access to assured in-
come streams would have a large effect on welfare. These are quantitative
conclusions that can be obtained only with a suitably parameterized model.
There are important senses in which the models of neither Hubbard
et al. (1995) nor Rosenzweig and Wolpin (1993) are “macroeconomic.”
First, in neither model is the return on savings determined as an equilib-
rium outcome. Second, in neither paper do the authors explore how ag-
gregate disturbances affect the economy. I will finish by discussing two
papers in the JPE that contributed to the transition toward developing re-
alistic models of consumption that are macroeconomic in this sense.
Imrohoroglu (1989) was a significant early paper that recognized the
potential importance of precautionary motives in the presence of aggre-
gate shocks. The paper was motivated by Lucas’s (1987) famous costs of
business cycles calculation. He had shown that in representative agent
economies the welfare costs of business cycles are small both because
fluctuations in aggregate income are themselves small and because these
fluctuations have only a second-order effect on welfare. It was natural to
conjecture that in heterogeneous agent economies with incomplete mar-
kets this quantitative conclusion might be overturned, both because fluc-
tuations in individual income can be substantial and because the pres-
ence of liquidity constraints means that for some households these
fluctuations have a first-order effect on welfare.
Imrohoroglu (1989) set out to evaluate this conjecture. She examines a
consumption-savings model with liquidity constraints in which house-
holds face unemployment risk that varies stochastically with macroeco-
nomic conditions. It is interesting to note how our understanding (and
expectations) of what it means for a macroeconomic model to be labeled
as “general equilibrium” has evolved. Despite describing her environ-
ment as general equilibrium, most macroeconomists today would describe

her model as a partial equilibrium environment because all prices—interest
rates, wages, job destruction rates, and job finding rates—are exogenous.
She finds that when aggregate shocks change the extent of unemployment
risk faced by households, the welfare cost of business cycles can be four to
five times larger than in a corresponding representative agent economy.
Perhaps the most influential macroeconomic model with heteroge-
neous agents and incomplete markets is that of Krusell and Smith (1998).
They study an infinite-horizon consumption-savings problem in which
ex ante identical households are subject to idiosyncratic unemployment
risk. As in the other models I have discussed, households can self-insure
this risk through a single risk-free asset. Krusell and Smith’s innovation
was to embed this precautionary savings problem in a stochastic version
of the neoclassical growth model. As in Aiyagari (1994), they interpret
the savings instrument as capital that is used by a representative firm as
input to a constant returns to scale production function. The interest rate
earned by households is thus determined in equilibrium as the marginal
product of capital. However, they differ from Aiyagari in that they allow
for the possibility that the production function is disturbed by exogenous
stochastic productivity shocks.
Krusell and Smith (1998) wanted to understand how the equilibrium
business cycle dynamics of macroeconomic variables in this heteroge-
neous agent economy compare to those in a corresponding representa-
tive agent economy—an important open question at the time. If the mac-
roeconomic dynamics of the two economies were not too different, it
would provide some justification for the common practice of studying
macroeconomics through the lens of a single representative agent. An-
swering this question, however, required solving their model, which raised
substantial challenges. Even before Krusell and Smith’s study, it was well
understood that the relevant state variable in this type of economy is an
infinite-dimensional object—the endogenous cross-sectional distribution
of households’ employment states and holdings of capital.
The best word to describe Krusell and Smith’s (1998) approach to this
challenge is “chutzpah.” Perhaps, they thought, all the information con-
tained in the distribution of household wealth is overkill. What if we look
for an equilibrium in a smaller space by summarizing the distribution with
only a finite-dimensional set of moments? What if we use just one mo-
ment, the mean? Lo and behold, it worked, in a very precise sense. They
showed that using only the mean of the distribution of capital holdings,
households could forecast future interest rates extremely accurately, which
are what matter for consumption decisions. Thus, Krusell and Smith could
approximate the equilibrium with a much smaller and computationally
feasible set of state variables.
Krusell and Smith (1998) labeled this finding “approximate aggrega-
tion.” It arises because in precautionary savings models optimal savings

decisions are extremely close to linear, except for households with very
little capital. But since the savings decisions of households with little cap-
ital matter little for the dynamics of aggregate capital, the dynamics of
aggregate capital (and hence the interest rate) depends approximately
on only the level of aggregate capital, not on the distribution of capital
across households.
Using this computational strategy, Krusell and Smith (1998) simulate
the dynamics of aggregate output, consumption, and investment in a
plausibly calibrated version of their model. They find that the dynamics
of these variables are virtually indistinguishable from the dynamics of a
similarly calibrated representative agent economy. It is important to re-
member that their finding of indistinguishability between the aggregate
dynamics of the heterogeneous agent and representative agent econo-
mies is conceptually different from their finding of approximate aggrega-
tion. It is relatively easy to construct economies in which approximate ag-
gregation holds but in which aggregate dynamics look different in the
corresponding heterogeneous agent and representative agent econo-
mies. For example, they show that when the model is modified to better
match the empirical distribution of wealth (in part by exploiting the ideas
in Hubbard et al. [1995]), the comovement of consumption and income
looks very different from the corresponding representative agent econ-
omy.
The lasting influence of Krusell and Smith (1998) is remarkable. It has
turned out that approximate aggregation is far more applicable than one
might have thought and has been used in a number of other contexts
in papers published in the JPE. For example, a variant of the Krusell and
Smith algorithm was used by Khan and Thomas (2013) in the context of
an economy with heterogeneous firms and by Favilukis, Ludvigson, and
Van Nieuwerburgh (2017) in the context of a model with fluctuating ag-
gregate house prices.
The JPE has played an essential role in fostering the growth of the study
of macroeconomics with heterogeneity. I hope, and predict, that the jour-
nal will continue to play such a role in the future.
References
Aiyagari, S. Rao. 1994. “Uninsured Idiosyncratic Risk and Aggregate Savings.”
Q. J.E. 109 (3): 659–84.
Attanasio, Orazio P., and Guglielmo Weber. 1995. “Is Consumption Growth Con-
sistent with Intertemporal Optimization? Evidence from the Consumer Ex-
penditure Survey.” J.P.E. 103 (6): 1121–57.
Favilukis, Jack, Sydney C. Ludvigson, and Stijn Van Nieuwerburgh. 2017. “The
Macroeconomic Effects of Housing Wealth, Housing Finance, and Limited
Risk Sharing in General Equilibrium.” J.P.E. 125 (1): 140–223.
Hubbard, R. Glenn, Jonathan Skinner, and Stephen P. Zeldes. 1995. “Precaution-
ary Saving and Social Insurance.” J.P.E. 103 (2): 360–99.

Imrohoroglu, Ayse. 1989. “Cost of Business Cycles with Indivisibilities and Li-
quidity Constraints.” J.P.E. 97 (6): 1364–83.
Khan, Aubhik, and Julia K. Thomas. 2013. “Credit Shocks and Aggregate Fluctu-
ations in an Economy with Production Heterogeneity.” J.P.E. 121 (6): 1055–
1107.
Krusell, Per, and Anthony A. Smith Jr. 1998. “Income and Wealth Heterogeneity
in the Macroeconomy.” J.P.E. 106 (5): 867–96.
Lucas, Robert E., Jr. 1987. Models of Business Cycles. London: Blackwell.
Rosenzweig, Mark R., and Kenneth I. Wolpin. 1993. “Credit Market Constraints,
Consumption Smoothing, and the Accumulation of Durable Production As-
sets in Low-Income Countries: Investments in Bullocks in India.” J.P.E. 101
(2): 223–44.
Zeldes, Stephen P. 1989. “Consumption and Liquidity Constraints: An Empirical
Investigation.” J.P.E. 97 (2): 305–46.
Time-Series Econometrics in Macroeconomics

and Finance
Lars Peter Hansen
I. Introduction
Ninety years ago, Slutsky (1927) and Yule (1927) opened the door to the
use of probability models in the analysis of economic time series. Their
vision was to view economic time series as linear responses to current and
past independent and identically distributed impulses or shocks. In dis-
tinct contributions, they showed how to generate approximate cycles with
such models. Each had a unique background and perspective. Yule was an
eminent statistician who, in the words of Stigler (1986), among his many
contributions, managed “effectively to invent modern time series analysis”
(361). Yule constructed and estimated what we call a second-order model
and applied it to study the time-series behavior of sunspots. Slutsky wrote
his paper in Russia in the 1920s motivated by the study of business cycles.
Much later, his paper was published in Econometrica, but it was already on
the radar screen of economists, such as Frisch. Indeed Frisch was keenly
aware of both Slutsky (1927) and Yule (1927) and acknowledged both in

his seminal paper (1933) on the impulse and propagation problem. Build-
ing on insights from Slutsky and Yule, Frisch pioneered the use of impulse
response functions in economic dynamics. His ambition was to provide
explicit economic interpretations for how current-period shocks alter
economic time series in current and future time periods.1 The Journal
of Political Economy provided an important platform for research that con-
fronts Frisch’s ambition in substantively interesting ways.
II. Rational Expectations Econometrics

A stumbling block for implementing Frisch’s (1933) ambition was how
to capture people’s beliefs about the future. Investment and other deci-
sions are in part based on people’s views of the future. Constructing pru-
dent economic policy depends in part on how private agents will respond
in the future. Once economic decision makers are included in formal dy-
namic economic models, their expectations come into play and become
an important ingredient to the model specification. Thus the time-series
econometrics research agenda grounded in economics had to take a stand
on how people inside economic models made forecasts.2 The rational ex-
pectations approach pioneered by Muth (1961) and Lucas (1972a, 1972b)
provided a coherent and model-consistent way to capture people’s beliefs.
It has been implemented in different ways in econometric practice. One
way is to exploit the resulting rational expectations equilibrium by fully
specifying the underlying economic model. The resulting model solution
then determines the beliefs of the economic agents inside the economic
model. Empirical evidence comes into play because econometricians face
uncertainty about the underlying parameters of the rational expectations
equilibrium and use data to infer their values. This vision is well articulated
in Sargent’s (1981) JPE treatise on interpreting economic time series. The
restrictions are sometimes implemented with two-step shortcuts whereby
parameters of processes for exogenous dynamics are estimated and plugged
into econometrically derived relationships. Additional parameters are es-
timated in a second step. Other approaches start with partially specified
models and then use historical time-series evidence to impose rational ex-
pectations without fully solving for the dynamic equilibrium. Econometric
support for this approach was provided in my 1982 paper with an initial
application in the JPE (Hansen and Hodrick 1980). This second paper
1
Sims (1980) and others advanced this idea by developing tractable multivariate time-
series methods and engaging in the identification of interpretable shocks in the multivar-
iate setting.
2
See Hansen (2014) for more discussions of modeling challenges for econometricians
and economic agents inside the models that they build.

added a new perspective to the empirical link between forward and spot
exchange rates.
III. Consumption and Permanent Income

Friedman’s (1957) famed permanent income model has implications
for both macroeconomic time series and microeconomic cross-sectional
data. Its rational expectations counterpart is perhaps most simply depicted
with a quadratic utility function, uncertain labor income, and a subjective
rate of discount equal to the rate of return on assets. Insights have broader
implications, but in this simplest setup, consumption is a martingale. This
observation was featured by Hall (1978) in his well-known JPE paper on
consumption and income dynamics. Permanent income theory in this
guise illustrates how even transient implications for income can have per-
manent consequences for consumption while maintaining Friedman’s ba-
sic insight that the permanent shocks to income are absorbed much more
prominently into the consumption responses. The impact of the transient
shocks is mitigated through savings behavior.
The Hall approach is a stark example of a partially specified model ex-
ploiting rational expectations. The martingale implications for consump-
tion can be tested, as was done by Hall (1978), without having to specify
correctly the income dynamics. Flavin (1983), also published in the J PE,
completed the model specification and discussed the implied cross-
equation restrictions of the type featured in Sargent (1981) to represent
the excess sensitivity of consumption to transitory income. To derive the
cross-equation restrictions as implied by a rational expectations equilib-
rium requires specifying the information about income used by consum-
ers. For instance, information other than lagged income could be perti-
nent in predicting future income suggesting that the correct equilibrium
may include other state variables. Testing the predictability of the first dif-
ference of consumption, however, does not require this complete specifi-
cation. In this Hall-Flavin setup, the first difference of consumption reveals
a news component in the information set of consumers (abstracting from
measurement error). As Hansen, Roberds, and Sargent (1991) empha-
sized, this news component should be present-value neutral and offset
by future income responses to this same shock. This gives a testable restric-
tion on the corresponding impulse response function of income to the
consumption news.
IV. Consumption and Asset Pricing

While the original Hall-Flavin research featured aggregate (and micro)
implications with constant interest rates, the JPE published a variety of

papers that explored the empirical challenges that allowed for time var-
iation in these rates. In addition, this literature built links between the
macroeconomy and asset pricing with the aim of explaining empirical
heterogeneity in the cross section of financial returns. As an outcome
of this research, macroeconomists have featured the so-called equity pre-
mium, the observed gap between expected aggregate equity returns and
Treasury bill returns, but the observed heterogeneity is much more perva-
sive. The Hall (1978) style reasoning turned out to be directly extendable
to “Euler equation” representations of multiple assets, not just bonds and
aggregate equity returns. Such representations support the equilibrium
representation of asset prices using so-called stochastic discount factors
that both discount the future and adjust for risk whereby the stochastic dis-
count factors are explicitly linked to the macroeconomy through variables
such as consumption. This stochastic discount factor approach provided a
platform for empirical analysis. The conceptual underpinnings for this
line of research were supported by theoretical derivations in Rubinstein
(1976), Lucas (1978), and Breeden (1979).
The JPE published several important papers that explored empirical
evidence related to this research. Hansen and Singleton (1983) used a
linear time model to depict the implied linkages between consumption
and returns. It featured restrictions across the predictable component of
the time series and could accommodate a small cross section of returns.
The linear time series approach had nice pedagogical value, making the
overidentifying restrictions transparent, but it required a lognormal as-
sumption without any scope for stochastic volatility. This linear time se-
ries approach was in contrast to the approach used by Hansen and Sin-
gleton (1982), who avoided the distributional assumption by studying
alternative conditional moment restrictions. Both papers allowed for econo-
metricians to understate the information used by economic agents. More-
over, both papers are among a collection of papers that document the em-
pirical challenge posed by a representative or stand-in consumer model
with time-separable power utility preferences as was commonly used in
the macroeconomics literature. The power utility specification led to a sto-
chastic discount factor that was a simple function of consumption growth.
While many refer to this as the equity premium puzzle, it really is a more
general phenomenon pertaining to the pricing of a heterogeneous cross
section of returns.
In a later JPE paper, Hansen and Jagannathan (1991) provided a fur-
ther characterization of the puzzle by stripping away the parametric struc-
ture of the stochastic discount factor. In the absence of arbitrage, there
exist valid stochastic discount factors; however, they may possess different
properties than what are implied by models with more parametric struc-
ture. Allowing for a much larger class of stochastic discount factors elim-
inated the possibility of fully identifying the stochastic discount factor

process from data and changed the econometric challenge to character-
izing the set of potential stochastic discount factors that are consistent
with empirical evidence. Specifically, Hansen and Jagannathan derived
sharp bounds on the implied mean–standard deviation trade-off for sto-
chastic discount factors that are consistent with the evidence from finan-
cial markets. Subsequent research extended and refined this analysis in
a variety of ways. Empirical puzzles are well defined only relative to a family
of models, and the bounds in the Hansen and Jagannathan paper and its
extensions provided a more general way to pose puzzles with the aim of
suggesting what is needed to construct models with better empirical under-
pinnings.
In response in part to the empirical challenges, the JPE has published
several innovative papers that explored different specifications of investor
preferences. For instance, Constantinides (1990) built a fully specified
model in which investors have preferences that display habit persistence.
Investors’ period utilities depend not only on current-period consumption
but also on that consumption relative to a habit stock of past consump-
tions. In effect, the habit stock provides a reference point for current con-
sumptions.3 Campbell and Cochrane (1999) altered these preferences in
two ways. They featured a model in which the habit stock contributes a so-
cially determined reference point based on past social consumptions. In ad-
dition, the counterpart to the habit stock has a nonlinear evolution equa-
tion. The Campbell and Cochrane paper, in particular, featured a model
in which the market compensation for the exposure to macroeconomic risk
is larger in bad macroeconomic times than in good ones. They provided an
endogenous mechanism for this variation. While stylized, their analysis was
supported by some empirical evidence, much more so than its counterpart
with a power utility function. Others have extended and refined this as an
empirically relevant asset pricing model.
A different strand of empirical research explored an alternative spec-
ification of investor preferences based on a recursive utility formulation.
Such preferences, by design, feature investor concerns about the inter-
temporal composition of risk. This research built on theoretical underpin-
nings provided in Kreps and Porteus (1978) and Epstein and Zin (1989)
and was prominently represented in two important JPE papers: Epstein
and Zin (1991) and Campbell (1996). The stochastic discount factor in re-
cursive utility models depends on the next-period continuation value rel-
ative to a risk-adjusted counterpart. This continuation value, familiar from
recursive methods in economic dynamics, encodes investor perceptions
about the future consumption prospects. Thus recursive utility preferences
bring in a forward-looking contribution into the valuation of even short-
3
Becker and Murphy (1988), also published in the JPE, used a similar formulation in a
microeconomic analysis of “rational addiction.”

term returns. The empirically oriented Epstein and Zin and Campbell pa-
pers accommodated this forward-looking perspective in different ways. The
Epstein and Zin (1991) research followed an econometric approach similar
to that in Hansen and Singleton (1982) modified by using a clever measure-
ment scheme. Under their parametric specification, the return on wealth
reveals the relevant information about the continuation value contribution
to the stochastic discount factor. Campbell (1996) used a time-series formu-
lation with forward-looking restrictions of a type that is common in linear
rational expectations models but applied to financial variables. Campbell
cleverly avoided using consumption data and instead featured the time-
series properties of the market return, including its predictability. One rea-
son to avoid using aggregate consumption data, as in Campbell, is that only
a limited segment of the population participates in security markets. There
has been a variety of subsequent empirical work that has built on these ini-
tial empirical contributions and their insights. Many of the resulting papers
have demonstrated that the forward-looking channel added by recursive
utility could have an important impact in asset pricing. Bansal and Yaron
(2004) is a prominent example. In a related JPE contribution, Hansen,
Heaton, and Li (2008), like Campbell, used linear time-series methods
and rational expectations restrictions. Specifically, Hansen et al. character-
ized and measured long-term risk components that are simultaneously in
the macroeconomic time series and in the cash flows from broad-based
portfolios of equities.
V. Imperfections in Financial Markets

Markets are not fully complete, and there are limits as to how much risk
they can share. The presumed market structure also alters the predicted
equilibrium pricing of financial securities. For instance, suppose that
consumers/investors face idiosyncratic components to labor income risk
that cannot be fully diversified in financial markets. Two prominent ex-
amples of papers that took this as a starting point are Krusell and Smith
(1998) and Constantinides and Duffie (1996). Both were published in
the JPE. Krusell and Smith featured dynamic models for which the im-
pact of market incompleteness was relatively benign in the sense that
simple averages could be used to summarize distributional impacts for rep-
resenting the evolution of the macroeconomy. In contrast, Constantinides
and Duffie featured models in which idiosyncratic shocks to labor income
have permanent components. Moreover, they presumed that there are
macroeconomic impacts on the distributions of these idiosyncratic shocks.
In their model, the equilibrium stochastic discount factor inherits these
macroeconomic impacts. Both papers have interesting benchmark econo-
mies, and their contributions have had a remarkable impact on subsequent

research. In a J PE paper related to Constantinides and Duffie (1996),
Heaton and Lucas (1996) probed into the microeconomic evidence and
explored the quantitative implications of market incompleteness for as-
set pricing.
One rationale for why financial markets cannot fully diversify labor in-
come risks is that idiosyncratic shocks are private information. The
Kocherlakota and Pistaferri (2009) JPE paper took this perspective and
presumed that the observed cross-sectional allocations are Pareto opti-
mal after taking account of the private information. They derived the cor-
responding asset pricing implications and contrasted them with the ones
implied by the incomplete market formulation of Constantinides and
Duffie (1996) and others. Different attributes of the cross-sectional distri-
bution of shocks come into play for the private information economy.
Kocherlakota and Pistaferri exposed some of the resulting measurement
challenges for asset pricing.
VI. Conclusion
Journal of Political Economy publications have played a prominent role in
the study of macroeconomics and finance using time-series methods.
The research disseminated by this journal delivered on Frisch’s (1933)
and others’ ambition to use economic dynamic models to interpret time-
series evidence. The published research characterized empirical challenges
and explored implications of new models designed to confront these chal-
lenges.
References
Bansal, Ravi, and Amir Yaron. 2004. “Risks for the Long Run: A Potential Reso-
lution of Asset Pricing Puzzles.” J. Finance 59 (4): 1481–1509.
Becker, Gary S., and Kevin M. Murphy. 1988. “A Theory of Rational Addiction.”
J.P.E. 96 (4): 675–700.
Breeden, Douglas T. 1979. “An Intertemporal Asset Pricing Model with Stochastic
Consumption and Investment Opportunities.” J. Financial Econ. 7 (3): 265–96.
Campbell, John Y. 1996. “Understanding Risk and Return.” J.P.E. 104 (2): 298–
345.
Campbell, John Y., and John H. Cochrane. 1999. “By Force of Habit: A
Consumption-Based Explanation of Aggregate Stock Market Behavior.” J.P.E.
107 (2): 205–51.
Constantinides, George M. 1990. “Habit Formation: A Resolution of the Equity
Premium Puzzle.” J.P.E. 98 (3): 519–43.
Constantinides, George M., and Darrell Duffie. 1996. “Asset Pricing with Hetero-
geneous Consumers.” J.P.E. 104 (2): 219–40.
Epstein, Larry G., and Stanley E. Zin. 1989. “Substitution, Risk Aversion and the
Temporal Behavior of Consumption and Asset Returns: A Theoretical Frame-
work.” Econometrica 57 (4): 937–69.

———. 1991. “Substitution, Risk Aversion, and the Temporal Behavior of Con-
sumption and Asset Returns: An Empirical Analysis.” J.P.E. 99 (2): 263–86.
Flavin, Marjorie A. 1983. “Excess Volatility in the Financial Markets: A Reassess-
ment of the Empirical Evidence.” J.P.E. 91 (6): 929–56.
Friedman, Milton. 1957. “The Permanent Income Hypothesis.” In A Theory of the
Consumption Function, 20–37. Princeton, NJ: Princeton Univ. Press.
Frisch, Ragnar. 1933. “Propagation Problems and Impulse Problems in Dynamic
Economics.” In Economic Essays in Honour of Gustav Cassel, 171–205. London:
Allen & Unwin.
Hall, Robert E. 1978. “Stochastic Implications of the Life Cycle–Permanent In-
come Hypothesis: Theory and Evidence.” J.P.E. 86 (6): 971–87.
Hansen, Lars Peter. 1982. “Large Sample Properties of Generalized Method of
Moments Estimators.” Econometrica 50 (4): 1029–54.
———. 2014. “Nobel Lecture: Uncertainty Outside and Inside Economic Mod-
els.” J.P.E. 122 (5): 945–87.
Hansen, Lars Peter, John C. Heaton, and Nan Li. 2008. “Consumption Strikes
Back? Measuring Long-Run Risk.” J.P.E. 116 (2): 260–302.
Hansen, Lars Peter, and Robert J. Hodrick. 1980. “Forward Exchange Rates as
Optimal Predictors of Future Spot Rates: An Econometric Analysis.” J.P.E.
88 (5): 829–53.
Hansen, Lars Peter, and Ravi Jagannathan. 1991. “Implications of Security Mar-
ket Data for Models of Dynamic Economies.” J.P.E. 99 (2): 225–62.
Hansen, Lars Peter, William Roberds, and Thomas J. Sargent. 1991. “Observable
Implications of Present-Value-Budget Balance.” In Rational Expectations Econo-
metrics. Boulder, CO: Westview.
Hansen, Lars Peter, and Kenneth J. Singleton. 1982. “Generalized Instrumental
Variables Estimation of Nonlinear Rational Expectations Models.” Econometrica
50:1269–86.
———. 1983. “Stochastic Consumption, Risk Aversion, and the Temporal Behav-
ior of Asset Returns.” J.P.E. 91 (2): 249–65.
Heaton, John, and Deborah J. Lucas. 1996. “Evaluating the Effects of Incomplete
Markets on Risk Sharing and Asset Pricing.” J.P.E. 104 (3): 443–87.
Kocherlakota, Narayana, and Luigi Pistaferri. 2009. “Asset Pricing Implications
of Pareto Optimality with Private Information.” J.P.E. 117 (3): 555–90.
Kreps, David M., and Evan L. Porteus. 1978. “Temporal Resolution of Uncertainty
and Dynamic Choice.” Econometrica 46 (1): 185–200.
Krusell, Per, and Anthony A. Smith. 1998. “Income and Wealth Heterogeneity in
the Macroeconomy.” J.P.E. 106 (5): 867–96.
Lucas, Robert E. 1972a. “Econometric Testing of the Natural Rate Hypothesis.”
In The Econometrics of Price Determination, edited by O. Eckstein, 50–59. Wash-
ington, DC: Board Governors, Fed. Reserve System.
———. 1972b. “Expectations and the Neutrality of Money.” J. Econ. Theory 4 (2):
103 –24.
———. 1978. “Asset Prices in an Exchange Economy.” Econometrica 46 (6): 1429–45.
Muth, John H. 1961. “Rational Expectations and the Theory of Price Move-
ments.” Econometrica 29 (3): 315–35.
Rubinstein, Mark. 1976. “The Valuation of Uncertain Income Streams and the
Pricing of Options.” Bell J. Econ. 7:407–25.
Sargent, Thomas J. 1981. “Interpreting Economic Time Series.” J.P.E. 89 (2):
213–48.
Sims, Christopher A. 1980. “Macroeconomics and Reality.” Econometrica 48 (1):
1–48.

Slutsky, Eugen. 1927. “The Summation of Random Causes as the Source of Cyclic
Processes.” In Problems of Economic Conditions, vol. 3. Moscow: Conjuncture Inst.
Reprinted in Econometrica 5 (April 1937): 105–46.
Stigler, Stephen M. 1986. The History of Statistics: The Measurement of Uncertainty be-
fore 1900. Cambridge, MA: Harvard Univ. Press.
Yule, George Udny. 1927. “On a Method of Investigating Periodicities in Dis-
turbed Series, with Special Reference to Wolfer’s Sunspot Numbers.” Philosoph-
ical Trans. Royal Soc. London. Ser. A, Containing Papers of a Mathematical or Physical
Character 226:267–98.
Asset Pricing: Models and Empirical Evidence

George M. Constantinides
Hansen and Singleton (1982, 1983), Shiller (1982), Mehra and Prescott
(1985), and Weil (1989) pose a major challenge in economics in the con-
text of a Lucas (1978) exchange economy. Hansen and Singleton (1982)
reject the Euler equations of per capita consumption at any level of rel-
ative risk aversion (RRA). Mehra and Prescott (1985) show that the aver-
age premium of the stock market over the risk-free rate cannot be ratio-
nalized in a calibrated standard economy and coin the “equity premium
puzzle.” Mehra and Prescott (1985) and Weil (1989) further show that
the puzzle is a dual one: as the assumed RRA is increased to rationalize
the equity premium, the implied risk-free rate becomes too high. More
generally, the challenge is to simultaneously explain the moments of ag-
gregate consumption and dividend growth, risk-free rate, market return,
market price-dividend ratio, and the term structure of interest rates in
the context of an economy with rational economic agents. The volumi-
nous research effort to address this challenge continues to this day and
includes explorations of preferences for early resolution of uncertainty,
absence of complete consumption insurance, uncertainty about the state
of the economy, habit persistence, macroeconomic crises resulting in a ca-
tastrophic drop in consumption, uncertainty about the economic model
and its parameters, borrowing constraints, and deviations from rational-
ity. In this essay I describe some of these explorations without providing
an exhaustive review of the literature. My apologies to authors who are not
being cited here.

Kreps and Porteus (1978) introduce a class of non–von Neumann–
Morgenstern preferences that capture preference for early resolution of
uncertainty. These preferences are sensitive to low-frequency but persis-
tent innovations in the state variables, thereby addressing many asset pric-
ing puzzles. Epstein and Zin (1989, 1991) adopt a conveniently homo-
thetic form of these preferences (commonly referred to as “Epstein-Zin”
preferences) and point out that these preferences disentangle the RRA
coefficient and the elasticity of intertemporal substitution (EIS) in con-
sumption. These preferences have the potential to resolve the dual equity
premium and risk-free rate puzzles because the equity premium is driven
by the RRA coefficient while the risk-free rate is driven by the EIS. Further-
more, they obtain in closed form the marginal rate of substitution (MRS),
thereby making these preferences tractable. Hansen, Heaton, and Li
(2008) elucidate properties of these preferences.
Bansal and Yaron (2004) are the first to adopt the Epstein-Zin prefer-
ences to address the asset pricing agenda. They build a model in which
the mean and variance of consumption and dividend growth depend on
a state variable with low frequency but persistent innovations and suc-
cessfully address many targets of the asset pricing agenda. Constantinides
and Ghosh (2011) and Beeler and Campbell (2012) point out that the
model implies too high autocorrelation of the aggregate consumption
growth rate and, therefore, excessive predictability of consumption growth
by the price-dividend ratio. Despite these limitations, the coupling of pref-
erence for early resolution of uncertainty with low-frequency but persis-
tent innovations in state variables is a significant breakthrough that will
continue to influence research in the foreseeable future. Two models dis-
cussed next, one on uninsurable household income shocks and the other
on learning about the state of the economy, owe an intellectual debt to
Epstein-Zin preferences and Bansal and Yaron (2004).
Cochrane (1991), Attanasio and Davis (1996), Blundell, Pistaferri, and
Preston (2008), and others provide empirical evidence that consumption
insurance is incomplete: households face a substantial amount of uninsur-
able idiosyncratic labor income risk. Constantinides (1982) highlights the
pivotal role of complete consumption insurance, showing that the equilib-
rium of such an economy with households that have heterogeneous en-
dowments and von Neumann–Morgenstern preferences is isomorphic to
the equilibrium of a homogeneous-household economy. Mankiw (1986)
shows that, in a two-period economy with incomplete consumption insur-
ance, the concentration of aggregate shocks among the population is an
important determinant of the level of the equity premium. Constantinides
and Duffie (1996) further show that, in the absence of complete consump-
tion insurance, given the aggregate income and dividend processes, any
given (arbitrage-free) price process can be supported in the equilibrium

of a heterogeneous-household economy with judiciously chosen persis-
tent and countercyclical idiosyncratic income shocks.
Brav, Constantinides, and Geczy (2002) present empirical evidence
that the equity and value premia are consistent with a stochastic discount
factor (SDF) obtained as the average of individual households’ MRS with
low and economically plausible values of the RRA coefficient. Since these
premia are not explained with an SDF obtained as the per capita MRS
with low values of the RRA coefficient, the evidence supports the incom-
plete consumption insurance hypothesis. They further show that the
countercyclical skewness of the idiosyncratic income shocks plays a key
role in driving prices, a property confirmed by Guvenen, Ozkan, and Song
(2014) using a very large data set from the US Social Security Administra-
tion.
Being framed in terms of economies in which households are endowed
with power utility, neither of these papers allows the RRA coefficient and
the EIS to be disentangled, a step that appears to be important in address-
ing the level and time-series properties of the risk-free rate, price-dividend
ratio, and market return. By introducing recursive preferences, Constan-
tinides and Ghosh (2017) provide empirical evidence that negatively
skewed, persistent, and countercyclical household consumption shocks
explain the moments of aggregate consumption and dividend growth,
risk-free rate, aggregate market return, and market price-dividend ratio
and explain the cross section of size-sorted, book-to-market-equity-sorted,
and industry-sorted portfolio returns.
The second model that owes an intellectual debt to Epstein-Zin pref-
erences and Bansal and Yaron (2004) is a model of uncertainty about the
state of the economy. An overload of worldwide macroeconomic, busi-
ness, and political news inundates investors. Little is known as to how
investors cope with this vast amount of information and, in particular,
which subset of information they pay attention to. In the macro-finance
literature, researchers typically model investors as focusing on the histo-
ries of a limited number of macroeconomic variables, typically consump-
tion and GDP growth, and applying a filter to rationally extract relevant
information about the economy. These models fare poorly in explaining
several features of stock market data, including the high average level of
the equity premium, the low level of the risk-free rate, the high variability
of the price-dividend ratio, and the low predictability of consumption
growth by the price-dividend ratio.
Ghosh and Constantinides (2017) establish that two broad categories
of publicly available macroeconomic information are the most highly
correlated with the marketwide price-dividend ratio. The first category
consists of price levels, including the Consumer Price Index for All Ur-
ban Consumers (CPI-U) and the Producer Price Index. The second cat-
egory consists of labor market variables, including average hourly earn-

ings, average hours of production, and numbers of employees in private
nonfarm payrolls in different sectors. These are the two classes of macro
variables that, according to FactSet, Bloomberg users pay the most atten-
tion to. On the other hand, contrary to the implications of learning mod-
els in which investors are assumed to learn from the consumption and
GDP histories alone, the price-dividend ratio has negligible correlation
with the contemporaneous consumption and GDP growth or a weighted
average of current and lagged consumption and GDP growth rates.
Motivated by the above evidence, Ghosh and Constantinides (2017)
model investors as learning about the latent state of the economy from
either the CPI or earnings per hour histories. The model provides a good
fit to the sample moments of consumption and dividend growth, market
return, marketwide price-dividend ratio, and risk-free rate. In contrast, an
alternative nested model in which the investors learn from the consump-
tion history alone fails along a number of dimensions: it implies essentially
zero volatility of the price-dividend ratio, thereby failing to explain the
excess volatility puzzle; it fails to generate the high persistence in the mar-
ketwide price-dividend ratio—one of the most robust features observed
in the data; and the estimated consumption growth in the second regime
is 25.7 percent, something that has not been observed in US history even
during the Great Depression of 1929.
The high persistence in the beliefs process in the main model, com-
bined with the preference for early resolution of uncertainty, yields a high
equity premium and low risk-free rate, consistent with the data. In contrast,
in the alternative model the low persistence of the beliefs process yields a
low equity premium. Finally, consistent with the data, the main model gen-
erates strong time variation in the conditional mean and variance of the
market return. Perhaps more impressive is the observation that it does so
without relying on countercyclical heteroscedasticity of the consumption
growth rate or the additional signal (the volatilities of consumption growth
and the signal are set to be equal in the two states)—a phenomenon for
which there is limited empirical evidence. Instead, the model generates
time variation in the conditional moments of the market return from the
heteroscedasticity of the beliefs process.
Early papers that recognize habit persistence through non-time-
separable von Neumann–Morgenstern preferences include Marshall
(1920), Duesenberry (1949), Pollak (1970), and Ryder and Heal (1973).
Habit preferences have had some success in addressing the asset pricing
agenda because habit tracks the business cycle. In a recession consump-
tion is low relative to habit (average past consumption), the equity pre-
mium is high because of the induced high RRA, and the risk-free rate is
low because of the precautionary demand for savings.
Constantinides (1990) models an economy in which the preferences
of the representative consumer exhibit linear internal habit; that is, the

consumer takes into account the effect of current consumption on future
habit. The calibrated model resolves the equity premium and risk-free
rate puzzles with a low RRA coefficient and a low EIS while matching
the mean and variance of consumption growth, albeit implying higher
autocorrelation of aggregate consumption growth than the autocorrela-
tion observed in the data. Ferson and Constantinides (1991) do not re-
ject the model on a system of assets consisting of size-sorted equity port-
folios, a Treasury bill, and a Treasury bond.
Campbell and Cochrane (1999) modify the Constantinides (1990)
model in two ways. First, they model the habit as nonlinear, thereby ren-
dering the interest rate volatility low or zero even when consumption
growth is serially uncorrelated. Second, they model habit as external; that
is, the consumer does not take into account the effect of current con-
sumption on future habit. The calibrated model resolves the equity pre-
mium and risk-free rate puzzles while matching the mean, variance, and
autocorrelation of consumption growth. However, the assumption that
habit is external renders the RRA coefficient implausibly high and fluctu-
ating between 60 and a few hundred. Ljungqvist and Uhlig (2015) point
out that government interventions that occasionally destroy part of the ag-
gregate endowment can lead to substantial welfare improvements.
In both the Constantinides (1990) and Campbell and Cochrane (1999)
models the only source of innovation is consumption growth, and these
models counterfactually imply that the correlation between the market
price-dividend ratio and consumption growth is about 50 percent, unlike
the correlation of close to zero in the data. Nevertheless, models with habit
preferences and additional sources of innovation are free from this criti-
cism. Chen and Ludvigson (2009) compare the Constantinides (1990)
and Campbell and Cochrane (1999) models and conclude that habit is
better described as nonlinear and internal rather than linear and exter-
nal, and the model performs well in explaining the cross section of equity
returns. Havranek, Rusnak, and Sokolova (2017) collect and examine es-
timates of habit from a wide range of studies and establish that habit is
significant but varies widely across studies.
Rietz (1988) points out that the possibility of a macroeconomic crisis
resulting in a catastrophic drop in consumption in principle resolves the
equity premium puzzle because the MRS becomes very high at the cata-
strophic state, adding the caveat that the size of the required drop in
consumption has never been observed in US history. Barro (2006),
Barro and Ursùa (2008), Gabaix (2012), Nakamura et al. (2013), Wachter
(2013), and others argue that rare disasters explain the equity premium
and related puzzles. Barro (2006) and Barro and Ursùa (2008) present do-
mestic and international evidence that macroeconomic crises are associated
with a large and sustained drop in aggregate consumption. As Constan-
tinides (2008) points out, Barro’s calibrated model treats the peak-to-trough

decline in aggregate consumption during macroeconomic crises (which
last, on average, 4 years) as if this decline occurs in 1 year, thereby mag-
nifying by a factor of four the size of the observed annual disaster risk.
Similar ad hoc magnification of the annual aggregate consumption de-
cline during macroeconomic crises is employed in a number of papers
that follow Barro (2006). Using an econometric methodology that allows
the probabilities attached to different states of the world to differ from
their sample frequencies and is therefore robust to the rare events prob-
lem in the data, Julliard and Ghosh (2012) reject the rare events explana-
tion for the equity premium puzzle. They show that to explain the equity
premium puzzle one should be willing to believe that economic disasters
occur every 6.6 years, on average. Moreover, Backus, Chernov, and Martin
(2011) demonstrate that options imply smaller probabilities of extreme
outcomes than the probabilities estimated from international macroeco-
nomic data.
Recent literature by Hansen and Sargent (2001), Epstein and Schneider
(2003), Klibanoff, Marinacci, and Mukerji (2005), Maccheroni, Marinacci,
and Rustichini (2006), and Johannes, Lochstoer, and Mou (2016), among
others, addresses the uncertainty about the economic model and its param-
eters. This literature argues that a more challenging high-dimensional
learning problem confronts investors where they need to learn not only
about the current state but also about the true underlying model and its
parameters and that such a learning problem plays an important role in
enhancing the empirical performance of these models. This important
class of issues is discussed in the essay by Lars Hansen, a major contribu-
tor to this literature.
Borrowing constraints address the equity premium and risk-free rate
puzzles and provide a partial explanation for the limited participation
of young consumers in the stock market and the demand for bonds in
the context of households in different stages of their life cycle as young,
middle-aged, and old. Constantinides, Donaldson, and Mehra (2002)
consider an overlapping generations economy in which consumers live
for three periods. In the first period, a period of human capital accumu-
lation, consumers receive a relatively low income. In the second period,
consumers are employed and receive income subject to great uncertainty.
The consumers consume part of this income and save the rest by investing
in stocks and bonds. In the third period, consumers consume the assets
accumulated during the second period. The key feature is that the bulk
of the future income of young consumers comes from wages that they
earn during the second period, while the income of the elderly primarily
comes from their savings in stocks and bonds during their middle age.
Young people would like to invest in equities, given the observed high
equity premium. However, they are reluctant to reduce their current con-
sumption in order to save by investing in stocks, because the bulk of their

lifetime income comes from their wages in their middle age. They want to
borrow against their future income, but the borrowing constraints pre-
vent them from doing so. Human capital alone cannot be used as collat-
eral for large loans in modern economies for reasons of moral hazard and
adverse selection. The model explains why many consumers do not par-
ticipate in the stock market when they are young. Middle-aged consumers
earn income that they partly consume and partly save by purchasing eq-
uities and bonds. The old earn no income and consume their savings.
Therefore, the risk of stock and bond ownership is concentrated in the
hands of middle-aged consumers who save. This concentration of risk
generates the high equity premium and the demand for bonds, in addi-
tion to the demand for shares by the middle-aged. The model acknowl-
edges and addresses at the same time the issue of the limited participa-
tion in the stock market and the demand for bonds.
References
Attanasio, Orazio P., and Steven J. Davis. 1996. “Relative Wage Movements and
the Distribution of Consumption.” J.P.E. 104:1227–62.
Backus, David, Mikhail Chernov, and Ian Martin. 2011. “Disasters Implied by Eq-
uity Index Options.” J. Finance 66:1967–2009.
Bansal, Ravi, and Amir Yaron. 2004. “Risks for the Long Run: A Potential Reso-
lution of Asset Pricing Puzzles.” J. Finance 59:1481–1509.
Barro, Robert J. 2006. “Rare Disasters and Asset Markets in the 20th Century.”
Q. J.E. 121:823–66.
Barro, Robert J., and José F. Ursùa. 2008. “Macroeconomic Crises since 1870.”
Brookings Papers Econ. Activity (Spring): 255–335.
Beeler, Jason, and John Y. Campbell. 2012. “The Long-Run Risks Model and Ag-
gregate Asset Prices: An Empirical Assessment.” Critical Finance Rev. 1:141–82.
Blundell, Richard, Luigi Pistaferri, and Ian Preston. 2008. “Consumption In-
equality and Partial Insurance.” A.E.R. 98:1887–1921.
Brav, Alon, George M. Constantinides, and Christopher C. Geczy. 2002. “Asset
Pricing with Heterogeneous Consumers and Limited Participation: Empirical
Evidence.” J.P.E. 110:793–824.
107:205–51.
Chen, Xiaohong, and Sydney C. Ludvigson. 2009. “Land of Addicts? An Empir-
ical Investigation of Habit-Based Asset Pricing Models.” J. Appl. Econometrics
24:1057–93.
Cochrane, John H. 1991. “A Simple Test of Consumption Insurance.” J.P.E. 99:
957–76.
Constantinides, George M. 1982. “Intertemporal Asset Pricing with Heteroge-
neous Consumers and without Demand Aggregation.” J. Bus. 55:253–67.
———. 1990. “Habit Formation: A Resolution of the Equity Premium Puzzle.”
J.P.E. 98:519–43.
———. 2008. “Comment on Barro and Ursùa.” Brookings Papers Econ. Activity
(Spring): 341–50.
Constantinides, George M., John B. Donaldson, and Rajnish Mehra. 2002. “Ju-
nior Can’t Borrow: A New Perspective on the Equity Premium Puzzle.” Q.J.E.
117:269–96.

Constantinides, George M., and Darrell Duffie. 1996. “Asset Pricing with Hetero-
geneous Consumers.” J.P.E. 104:219–40.
Constantinides, George M., and Anisha Ghosh. 2011. “Asset Pricing Tests with
Long-Run Risks in Consumption Growth.” Rev. Asset Pricing Studies 1:96–136.
———. 2017. “Asset Pricing with Countercyclical Household Consumption
Risk.” J. Finance 73:415–59.
Duesenberry, James S. 1949. Income, Saving, and the Theory of Investment Behavior.
Cambridge, MA: Harvard Univ. Press.
Epstein, Larry G., and Martin Schneider. 2003. “Recursive Multiple-Priors.”
J. Econ. Theory 113:1–31.
Epstein, Larry G., and Stanley Zin. 1989. “Substitution, Risk Aversion, and the
Temporal Behavior of Consumption and Asset Returns: A Theoretical Frame-
work.” Econometrica 57:937–69.
———. 1991. “Substitution, Risk Aversion, and the Temporal Behavior of Con-
sumption and Asset Returns: An Empirical Analysis.” J.P.E. 99:263–86.
Ferson, Wayne E., and George M. Constantinides. 1991. “Habit Persistence and
Durability in Aggregate Consumption: Empirical Tests.” J. Financial Econ.
29:199–240.
Gabaix, Xavier. 2012. “Variable Rare Disasters: An Exactly Solved Framework for
Ten Puzzles in Macro-Finance.” Q. J.E. 127:645–700.
Ghosh, Anisha, and George M. Constantinides. 2017. “What Information Drives
Asset Prices.” Working paper, Univ. Chicago.
Guvenen, Fatih, Serdar Ozkan, and Jae Song. 2014. “The Nature of Countercy-
clical Income Risk.” J.P.E. 122:621–60.
Hansen, Lars Peter, John Heaton, and Nan Li. 2008. “Consumption Strikes Back:
Measuring Long-Run Risk.” J.P.E. 116:260–302.
Hansen, Lars Peter, and Thomas J. Sargent. 2001. “Robust Control and Model
Uncertainty.” A.E.R. 91:60–66.
Hansen, Lars Peter, and Kenneth J. Singleton. 1982. “Generalized Instrumental
Variables Estimation of Nonlinear Rational Expectations Models.” Econometrica
50:1269–86.
———. 1983. “Stochastic Consumption, Risk Aversion, and the Temporal Behav-
ior of Asset Returns.” J.P.E. 91:249–65.
Havranek, Tomas, Marek Rusnak, and Anna Sokolova. 2017. “Habit Formation
in Consumption: A Meta-Analysis.” European Econ. Rev. 95:142–67.
Johannes, Michael, Lars Lochstoer, and Yiqun Mou. 2016. “Learning about Con-
sumption Dynamics.” J. Finance 71:551–600.
Julliard, Christian, and Anisha Ghosh. 2012. “Can Rare Events Explain the Equity
Premium Puzzle?” Rev. Financial Studies 25:3037–76.
Klibanoff, Peter, Massimo Marinacci, and Sujoy Mukerji. 2005. “A Smooth Model
of Decision Making under Ambiguity.” Econometrica 73:1849–92.
Kreps, David M., and Evan L. Porteus. 1978. “Temporal Resolution of Uncertainty
and Dynamic Choice Theory.” Econometrica 46:185–200.
Ljungqvist, Lars, and Harald Uhlig. 2015. “Comment on the Campbell-Cochrane
Habit Model.” J.P.E. 123:1201–13.
Lucas, Robert E., Jr. 1978. “Asset Prices in an Exchange Economy.” Econometrica
46:1429–45.
Maccheroni, Fabio, Massimo Marinacci, and Aldo Rustichini. 2006. “Ambiguity
Aversion, Robustness, and the Variational Representation of Preferences.”
Econometrica 74:1447–98.
Mankiw, N. Gregory. 1986. “The Equity Premium and the Concentration of Ag-
gregate Shocks.” J. Financial Econ. 17:211–19.
Marshall, Alfred. 1920. Principles of Economics: An Introductory Volume. 8th ed. Lon-
don: Macmillan.

Mehra, Rajnish, and Edward C. Prescott. 1985. “The Equity Premium: A Puzzle.”
J. Monetary Econ. 15:145–61.
Nakamura, Emi, Jón Steinsson, Robert J. Barro, and José F. Ursùa. 2013. “Crises
and Recoveries in an Empirical Model of Consumption Disasters.” American
Econ. J.: Macroeconomics 5:35–74.
Pollak, Robert A. 1970. “Habit Formation and Dynamic Demand Functions.”
J.P.E. 78:745–63.
Rietz, Thomas A. 1988. “The Equity Risk Premium: A Solution.” J. Monetary Econ.
22:117–31.
Ryder, Harl E., Jr., and Geoffrey M. Heal. 1973. “Optimum Growth with Inter-
temporal Dependent Preferences.” Rev. Econ. Studies 40:1–33.
Shiller, Robert J. 1982. “Consumption, Asset Markets and Macroeconomic Fluc-
tuations.” Working paper, NBER, Cambridge, MA.
Wachter, Jessica A. 2013. “Can Time-Varying Risk of Rare Disasters Explain Ag-
gregate Stock Market Volatility?” J. Finance 68:987–1035.
Weil, Philippe. 1989. “The Equity Premium Puzzle and the Risk-Free Rate Puz-
zle.” J. Monetary Econ. 24:401–21.
Finance at the University of Chicago

Eugene F. Fama
Research in finance has two main areas: (i) corporate—theory and em-
pirical work on optimal investment and financing decisions by firms (the
demand side of capital formation)—and (ii) asset pricing—portfolio the-
ory and related models of risk and expected return (the supply side of
capital formation). The University of Chicago Booth School (formerly the
Graduate School of Business) has long been front and center in corpo-
rate finance, and it is joined by the Chicago Economics Department as
an asset pricing ringleader.
Corporate finance was kick-started by the “irrelevance of capital struc-
ture” theorems of Franco Modigliani and my longtime Chicago Booth
mentor, colleague, and friend, Merton Miller (Modigliani and Miller
1958; Miller and Modigliani 1961; the latter published, like many semi-
nal papers in finance, in the now defunct Journal of Business). I have
The comments of George Constantinides and many years of interaction on the topics of
this paper with Kenneth French and John Cochrane are gratefully acknowledged

worked some in corporate finance, but more of my work is in asset pric-
ing, and I largely focus here on Chicago’s role in asset pricing research. I
give my (business school) perspective, which has a strong eye toward ap-
plications. This is in contrast to the macro-finance perspective on asset
pricing, which is more concerned with integrating asset pricing and mac-
roeconomics. Lars Hansen (my 2013 co–Nobel laureate along with Rob-
ert Shiller) is a massive contributor to macro-finance. Aside from a few
comments, I leave that area to him and George Constantinides in this
anniversary edition of the JPE.
I emphasize that this is a perspective piece, biased toward Chicago’s
contributions, not a literature review meant to give everyone just due.
For readers offended by my Chicago-centric overview, mea culpa in ad-
vance.
The asset pricing research I discuss has two main poles: (i) work on
market efficiency—the proposition that asset prices reflect all available
information—and (ii) models of market equilibrium—the nature of risk
and the relation between risk and expected return. In my Nobel lecture
(Fama 2014) I call these the Siamese twins of asset pricing because, as
outlined below, they are inseparable pieces of asset pricing models based
on rational behavior. I begin by discussing the work on market efficiency
and then turn to models of market equilibrium.
Market Efficiency
Research on the behavior of commodity prices and prices of financial as-
sets has a long history going back at least to Bachelier (1900), but the
coming of computers in the late 1950s produced an explosion of work,
primarily on the behavior of stock prices and stock returns. Interest was
concentrated at the University of Chicago’s Booth School and the Mas-
sachusetts Institute of Technology Economics Department and Sloan
School. At Chicago, the early players were Larry Fisher (creator of the
now-ubiquitous Center for Research in Security Prices [CRSP] data files),
Merton Miller, Harry Roberts, and Lester Telser, with Benoit Mandelbrot
as an occasional visitor. At MIT, Sydney Alexander, Paul Cootner, Franco
Modigliani (Merton Miller’s longtime coauthor), and Paul Samuelson
carried the ball.
When I finished PhD prelims and it came time to write a thesis in 1962,
I was twice a father and anxious to finish quickly. A PhD student could get
faculty attention with a thesis on the behavior of stock prices, which ex-
plains Fama (1964), published in the Journal of Business. More of my thesis
tests Mandelbrot’s hypothesis that stock returns conform better to the
nonnormal (fat-tailed) class of symmetric stable distributions than to the
normal distribution, but about a third of it is on what I later dub market
efficiency.

Market efficiency is the hypothesis that asset prices reflect all available
information. But what are the testable implications? The early answer is
that in an efficient market, prices follow random walks and future returns
are unpredictable from currently available information. Mandelbrot
(1966), another seminal paper in the Journal of Business, and Samuelson
(1965) show that the random walk hypothesis is too strong. In their mod-
els, prices in an efficient market are submartingales. In simple terms, the
presumption is that the price of an asset at time t, pt, is set to deliver an
equilibrium expected return at t 1 1, EðR t11 jft,m Þ, where ft,m is the infor-
mation embedded in the asset’s time t price. If ft,m is all available infor-
mation, ft, then
Eð pt11 jft Þ 5 pt ½1 1 EðR t11 jft,m Þ, (1)
and
EðR t11 jft Þ 2 EðR t11 jft,m Þ 5 0:0: (2)
Equation (2) implies that R t11 2 EðR t jft,m Þ, the deviation of the return
from the equilibrium expected return, EðR t jft,m Þ, is unpredictable from
information available at t. But (1) and (2) hold only when ft,m 5 ft , that
is, when the information embedded in the price pt is all available infor-
mation. If some information is missed in setting prices at t, then R t11 2
EðR t jft,m Þ is predictable from the broader information set ft.
Implicit in the submartingale models of Samuelson (1965) and Man-
delbrot (1966) is what I call the joint hypothesis problem: tests of market
efficiency are conditional on an assumed model for equilibrium prices
and expected returns, which means that tests of efficiency are joint tests
of efficiency and the assumed asset pricing model (Fama [1970], spelled
out better in chap. 5 of Fama [1976]). Though not commonly acknowl-
edged, the reverse is also true: asset pricing models based on rational be-
havior implicitly or explicitly assume a strong form of market efficiency:
investors agree on the joint distributions of future asset payoffs and they
get them right, which means that prices reflect all available information.
In short, market efficiency and models of market equilibrium are the Si-
amese twins of asset pricing (Fama 2014).
The early work (1950s and 1960s) on market efficiency focuses on the
time-series properties of stock returns. Worried that the newly minted
CRSP data would not find their way into academic research, James Lorie,
the founder of CRSP, suggested I do a study of the adjustment of stock
prices to stock splits. The resulting paper, Fama et al. (1969) coauthored
with Lawrence Fisher and two of the all-time best Chicago finance PhD
students, Michael Jensen and Richard Roll, is the first event study. Event

studies subsequently play a major role in tests of market efficiency in the
finance and accounting literatures, and they are often used to assess dam-
ages in court cases.
An important issue in the market efficiency literature is whether pro-
fessional investors have private information about asset prospects that
they use to enhance returns. The most complete data are for mutual funds,
and there is a large literature on the performance of actively managed
funds, that is, funds that attempt to enhance returns by choosing under-
priced stocks. The seminal early study is Michael Jensen’s (1968) University
of Chicago PhD thesis. He uses the capital asset pricing model (CAPM) of
Sharpe (1964) and Lintner (1965) as the model for equilibrium expected
returns, and as in many studies thereafter, he finds that actively managed
funds generally underperform the expected return predictions of the
CAPM.
There is a large literature on mutual fund performance with many ex-
cellent papers. I cannot resist touting the one I like best, Fama and
French (2010). In this paper, we use the stochastic properties of individ-
ual fund returns from 1984 to 2006 to simulate the cross-section distribu-
tion of risk-adjusted average returns when the market is efficient and ac-
tive fund managers have no private information. The models for risk
adjusting are the CAPM, the three-factor model of Fama and French
(1993), and Carhart’s (1997) four-factor extension. The results are strik-
ing. About 3 percent of actively managed funds seem to have enough pri-
vate information to cover the costs (reported management fees and ex-
penses) imposed on investors, which means that going forward, we
expect them to perform like low-cost passive funds. The remaining 97 per-
cent do not produce returns that suggest they have enough private infor-
mation to cover costs. When we examine fund returns before costs, there
is a near-perfect balance of funds in the extreme right tail of risk-adjusted
average returns that do better than would be expected by chance and los-
ers in the extreme left tail that do worse. The aggregate portfolio of ac-
tive funds matches the overall US market return before costs, and its
monthly return is correlated .991 with the market return. In other words,
in aggregate, active mutual funds are an expensive way to hold the mar-
ket portfolio.
Though mutual funds are not the entire active investor universe, our
mutual fund results are in line with Sharpe’s (1991) observation that active
investing is a zero-sum game before costs: winners eat losers. The logic is
that an active investor holds an unbalanced portfolio, overweighting some
assets relative to their market caps and underweighting others. This means
that in aggregate, other investors have to take offsetting positions, under-
weighting the assets overweighted by the active investor and overweighting
the assets underweighted. But passive investors hold cap-weight portfolios,

either of the entire market or of subsets (value stocks, growth stocks, en-
ergy stocks, etc.). Thus, the unbalanced portfolio of an active investor has
to be balanced by offsetting positions of other active investors—a zero-sum
game before costs. After costs, active investing is a negative-sum game.
French (2008) provides estimates of the costs.
Asset Pricing
Finance as an area of scientific research had its birth at the University of
Chicago. The parting shot is Harry Markowitz’s Economics Department
PhD thesis on portfolio theory, subsequently published as a journal arti-
cle (Markowitz 1952) and then as a magnificent Cowles Foundation mono-
graph (Markowitz 1959). When I started teaching investments in 1963, the
textbooks of the day focused on the futile task of teaching students to pick
stocks. In my early teaching years, Markowitz (1959) was the main reading
in my investments course.
The mean-variance-efficient set of Markowitz’s portfolio model is the
foundation of the first formal asset pricing model, the CAPM of Sharpe
(1964) and Lintner (1965). The CAPM provides the first rigorous anal-
ysis of asset risk and the equilibrium relation between risk and expected
return. The CAPM is a one-period model. It gets a tour de force multi-
period extension in Robert Merton’s (1973a) intertemporal CAPM, the
ICAPM, which offers a theoretical framework for recent multifactor mod-
els, for example, the three-factor model of Fama and French (1993).
The CAPM predicts that market b, the slope in the regression of an
asset’s return on the market return, suffices to describe the cross section
of expected asset returns. The initial tests of the model are cross-section
regressions of average asset returns on estimates of their b’s and other
variables. The model predicts that the slopes for other variables are in-
distinguishable from zero. Black, Jensen, and Scholes (1972) argue that
ordinary least squares standard errors for slope estimates from such re-
gressions are too low because they do not adjust for cross-correlation of
the regression residuals (return correlation beyond that associated with
regression explanatory variables).
Fama and MacBeth (1973) provide a simple cure. Instead of cross-
section regressions of average monthly returns on b estimates and other
variables, the regressions are run month by month. Averages of monthly
slopes and t-statistics for the averages are used to test the CAPM predic-
tion that b suffices to describe expected asset returns. This approach is in
effect repeated sampling in which the time-series variation in month-by-
month regression slopes picks up the effects of cross-correlation of resid-
uals without requiring an estimate of the residual covariance matrix. The
approach has been used so much and so long in tests of asset pricing

models it is often referenced as Fama-MacBeth, without citation of the
JPE source. The approach is generally applicable in panel regressions,
balanced and unbalanced, when it is sensible to weight periods (rather
than observations) equally.
Asset pricing, as represented by the CAPM and the ICAPM, is rather di-
vorced from other areas of economics. This changes with the consumption-
based CCAPM of Economics Department standout Robert Lucas (1978)
(see also Breeden 1979). The CCAPM, with its simple but powerful eco-
nomic insights, took macroeconomics by storm and gave birth to a large
body of empirical work and theoretical extensions by the most talented
macro-finance researchers. Excellent and much-cited JPE examples are
Epstein and Zin (1991) and Campbell and Cochrane (1999). Lucas’s No-
bel citation is for his pathbreaking work in macro models with rational ex-
pectations, but on the basis of its impact on research in the intersection
of macroeconomics and finance, the CCAPM is as important.
Work on asset pricing models is currently in a bit of a bind. The CAPM
had a 25-year run when it was widely accepted as the ruling paradigm. The
slow accumulation of anomalies—patterns in average returns that violate
the model’s predictions—led to the general conclusion that the model is
an empirical failure. (The death knell is Fama and French [1992].) Despite
the efforts of many talented macro-finance empiricists, the consumption-
based CCAPM rests in empirical limbo, and we are unaware of anyone
who suggests that it can be useful in applications (e.g., evaluating portfo-
lio performance or determining the cost of capital).
The recent alternatives offered by asset pricing research in finance (as
opposed to macro-finance) are factor models, which expand the CAPM
by including factors beyond the CAPM market factor. The three-factor
model of Fama and French (1993) had a 201-year run, but in the face of
accumulating anomalies, it now has a five-factor extension (Fama and
French 2015). There are other examples. Merton’s (1973a) ICAPM pro-
vides a ready-made theoretical framework that can accommodate factor-
based asset pricing models. But until research identifies and empirically
validates the state variables captured by model factors, the motivation for
factor models remains empirical: the factors are chosen to capture pat-
terns in average returns.
An important question (posed forcefully by John Cochrane) is, What is
the discipline that prevents factor models from degenerating into mind-
less data dredging? The answer offered by Fama and French (1993, 2012,
2015, 2016, 2017) has three parts. (1) Robustness: A model should com-
pete well on samples for different periods and different markets. (2) Par-
simony: Though the guidelines remain to be drawn, other things equal,
models with fewer factors are better. (3) Factors should have some moti-
vation from theory, even if the theory is somewhat loose. For example, the

three-factor and five-factor models of Fama and French (1993, 2015) are
consistent with the dividend discount valuation model widely used in fi-
nance and accounting. In the future we can hopefully do better. Time will
tell.
Time-Varying Expected Returns

The expected return on a security is the expected compensation for
holding the security. It is a price, and like other prices, it almost surely
varies through time. Beginning with Fama (1975), research that attempts
to measure variation in expected returns typically focuses on time-series
regressions of returns on forecasting variables. For stocks, for example,
regressions of returns on lagged dividend yields (ratios of annual divi-
dends to price) are common. Fama and French (1988a, 1989) are exam-
ples, and Cochrane (2011) provides an insightful summary.
Fama and French (1988b) take a different tack. They observe that if ex-
pected stock returns are highly autocorrelated but slowly mean-reverting,
stock returns will have negative autocorrelation that increases with the re-
turn horizon. We estimate that 30–40 percent of the variation of 3–5-year
returns is due to time-varying expected returns, but the small samples for
long horizons mean that the estimates are imprecise. Estimate impreci-
sion is a general plague in the literature on time-varying expected returns.
The Economics of Organizations

Spurred by the pathbreaking paper of Jensen and Meckling (1976), my
research in the late 1970s took a detour into agency theory. Earlier pa-
pers emphasize agency problems. I was interested in how competitive
forces lead to mechanisms to mitigate agency problems. The first paper,
Fama (1980), argues that managerial labor markets, inside and outside
of firms, act to control managers faced with the temptations created by
diffuse residual claims that reduce the incentives of individual residual
claimants to monitor managers.
Michael Jensen and I then collaborated on two papers in a JPE sister
journal (Fama and Jensen 1983a, 1983b) that study more generally how
competition leads to different mechanisms to mitigate agency problems as-
sociated with separation of management and residual risk bearing, and
how an organization’s activities, and the special agency problems they
pose, affect its residual claims and control mechanisms. For example, we
argue that the redeemable residual claims of a financial mutual (e.g., an
open-end mutual fund) provide strong discipline for its managers, but re-
deemability is cost-effective only when assets can be sold quickly with low
transactions costs. We also argue that the nonprofit form, in which no

agents have residual claims to net cash flows, is a response to the agency
problem associated with activities in which there is a potential supply of
donations that might be expropriated by residual claimants.
Options Pricing and the JPE

Finally, though derivatives pricing is not my focus here, the options pric-
ing papers of Black and Scholes (1973) and Merton (1973b), which re-
sulted in Nobel Prizes for Merton and Scholes in 1997, warrant acknowl-
edgment. These papers are familiar to students of finance and economics
more generally, and they underpin a vibrant derivatives industry. One is
hard-pressed to find other papers with such a combination of academic
and applied impact. Myron Scholes is a member of the outstanding cohort
of Chicago PhD students of the late 1960s and early 1970s, he and Fischer
Black were Booth faculty for extended periods, and Robert Merton has an
honorary Chicago PhD (granted before his Nobel). Black and Scholes
(1973) is one of many fundamental JPE papers in finance.
References
Bachelier, L. J. B. A. 1900. Theorie de la speculation. Paris: Gauthier-Villars. Reprinted
in The Random Character of Stock Market Prices, edited by Paul H. Cootner. Cam-
bridge, MA: MIT Press, 1964.
Black, Fischer, Michael C. Jensen, and Myron S. Scholes. 1972. “The Capital Asset
Pricing Model: Some Empirical Tests.” In Studies in the Theory of Capital Markets,
edited by Michael C. Jensen. New York: Praeger.
Black, Fischer, and Myron S. Scholes. 1973. “The Pricing of Options and Corpo-
rate Liabilities.” J.P.E. 81 (3): 637–54.
Breeden, Douglas T. 1979. “An Intertemporal Asset Pricing Model with Stochas-
tic Consumption and Investment Opportunities.” J. Financial Econ. 7 (3): 265–
96.
107:205–51.
Carhart, Mark M. 1997. “On Persistence in Mutual Fund Performance.” J. Finance
52:57–82.
Cochrane, John H. 2011. “Discount Rates: American Finance Association Presi-
dential Address.” J. Finance 66:1047–1108.
Epstein, Larry G., and Stanley Zin. 1991. “Substitution, Risk Aversion, and the
Temporal Behavior of Consumption and Asset Returns: An Empirical Analy-
sis.” J.P.E. 99:263–86.
Fama, Eugene F. 1964. “The Behavior of Stock-Market Prices.” J. Bus. 38:34 –105.
———. 1970. “Efficient Capital Markets: A Review of Theory and Empirical
Work.” J. Finance 25:383–417.
———. 1975. “Short-Term Interest Rates as Predictors of Inflation.” A.E.R.
65:269–82.
———. 1976. Foundations of Finance. New York: Basic Books.

———. 1980. “Agency Problems and the Theory of the Firm.” J.P.E. 88:288–307.
———. 2014. “Two Pillars of Asset Pricing.” A.E.R. 104:1467–85.
Fama, Eugene F., Lawrence Fisher, Michael Jensen, and Richard Roll. 1969. “The
Adjustment of Stock Prices to New Information.” Internat. Econ. Rev. 10:1–21.
Fama, Eugene F., and Kenneth R. French. 1988a. “Dividend Yields and Expected
Stock Returns.” J. Financial Econ. 22:3–25.
———. 1988b. “Permanent and Temporary Components of Stock Prices.” J.P.E.
96:246–73.
———. 1989. “Business Conditions and Expected Returns on Stocks and
Bonds.” J. Financial Econ. 25:23–49.
———. 1992. “The Cross-Section of Expected Stock Returns.” J. Finance 47 (2):
427–65.
———. 1993. “Common Risk Factors in the Returns on Stock and Bonds.” J. Fi-
nancial Econ. 33 (1): 3–56.
———. 2010. “Luck versus Skill in the Cross-Section of Mutual Fund Returns.” J.
Finance 65 (5): 1915–47.
———. 2012. “Size, Value, and Momentum in International Stock Returns.” J.
Financial Econ. 105 (3): 457–72.
———. 2015. “A Five-Factor Asset Pricing Model.” J. Financial Econ. 116 (1): 1–
22.
———. 2016. “Dissecting Anomalies with a Five-Factor Model.” Rev. Financial
Studies 29:70–103.
———. 2017. “International Tests of a Five-Factor Asset Pricing Model.” J. Finan-
cial Econ. 123:441–63.
Fama, Eugene F., and Michael C. Jensen. 1983a. “Agency Problems and Residual
Claims.” J. Law and Econ. 26:327–49.
———. 1983b. “Separation of Ownership and Control.” J. Law and Econ. 26:301–
25.
Fama, Eugene F., and James D. MacBeth. 1973. “Risk, Return, and Equilibrium:
Empirical Tests.” J.P.E. 81 (3): 607–36.
French, Kenneth R. 2008. “The Cost of Active Investing.” J. Finance 63:1537–73.
Jensen, Michael C. 1968. “The Performance of Mutual Funds in the Period
1945–1964.” J. Finance 23:389–416.
Jensen, Michael C., and William H. Meckling. 1976. “Theory of the Firm: Mana-
gerial Behavior, Agency Costs and Ownership Structure.” J. Financial Econ.
3:305–60.
Lintner, John. 1965. “The Valuation of Risk Assets and the Selection of Risky In-
vestments in Stock Portfolios and Capital Budgets.” Rev. Econ. and Statis.
47:13–37.
Lucas, Robert E., Jr. 1978. “Asset Prices in an Exchange Economy.” Econometrica
46 (6): 1429–45.
Mandelbrot, Benoit. 1966. “Forecasts of Future Prices, Unbiased Markets, and
Martingale Models.” J. Bus. 39 (suppl.): 242–55.
Markowitz, Harry. 1952. “Portfolio Selection.” J. Finance 7:77–91.
———. 1959. Portfolio Selection: Efficient Diversification of Investments. Cowles Foun-
dation Monograph no. 16. New York: Wiley.
Merton, Robert C. 1973a. “An Intertemporal Capital Asset Pricing Model.”
Econometrica 41:867–87.
———. 1973b. “Theory of Rational Option Pricing.” Bell J. Econ. 4 (1): 141–83.
Miller, Merton H., and Franco Modigliani. 1961. “Dividend Policy, Growth, and
the Valuation of Shares.” J. Bus. 34 (October): 411–33.

Modigliani, Franco, and Merton H. Miller. 1958. “The Cost of Capital, Corpora-
tion Finance, and the Theory of Investment.” A.E.R. 48 ( June): 261–97.
Samuelson, Paul. 1965. “Proof That Properly Anticipated Prices Fluctuate Ran-
domly.” Indus. Management Rev. 6 (Spring): 41–49.
Sharpe, William F. 1964. “Capital Asset Prices: A Theory of Market Equilibrium
under Conditions of Risk.” J. Finance 19:425–42.
———. 1991. “The Arithmetic of Active Management.” Financial Analysts J. 47:
7–9.
Behavioral Economics
Richard H. Thaler
Exactly 100 years ago, the JPE was poised to be at the forefront of the field
that would eventually come to be called behavioral economics. John
Maurice Clark, a JPE editor, University of Chicago faculty member, and
son of John Bates Clark, authored the lead article of the January 1918 is-
sue titled “Economics and Modern Psychology: I.” (Part II appeared in
the next issue.) His message was a simple one: “The economist may at-
tempt to ignore psychology, but it is a sheer impossibility for him to ig-
nore human nature. . . . If the economist borrows his conception of
man from the psychologist, his constructive work may have some chance
of remaining purely economic in character. But if he does not he will not
thereby avoid psychology. Rather he will force himself to make his own,
and it will be bad psychology” (4).
A few years later Clark left Chicago to take the position his father had
once held at Columbia, and it seems fair to say that the subsequent edi-
tors of the JPE did not take up his call to arms. Behavioral economics pa-
pers have made only scattered appearances in the journal in the subse-
quent century.1
Thanks to Alex Imas, Emir Kamenica, and Jesse Shapiro for helpful comments.
1
To put a tiny bit of data behind this assertion, I counted the number of papers pub-
lished in a few top journals that are cited in Stefano DellaVigna’s recent survey paper in
the Journal of Economic Literature. The tally is Q JE 32, AER 21, Journal of Finance 16, and
JPE 10. And my informal sense is that the 10 JPE papers contain a greater proportion that
is not behavioral, as compared to those in the Q JE or AER.

As Herbert Simon once said, the term behavioral economics is a bit
strange. “What ‘non-behavioral’ economics can we contrast with it?” he
asked (Simon 1987, 221). One answer to this question is the style of eco-
nomics that the JPE is perhaps best known for: price theory à la Chicago
School led by the intellectual giants Gary Becker, Milton Friedman, and
George Stigler. Becker’s research goal was to apply the standard tools of
maximizing behavior to study a wide variety of topics that were not then
part of the domain of economics including addiction, crime, discrimina-
tion, marriage, divorce, childbearing, and social interactions.
Becker acknowledged that by applying the tools of economics to such
topics he was pushing the envelope. In his Nobel address he discusses this
explicitly (Becker 1993). “I have intentionally chosen certain topics for my
research—such as addiction—to probe the boundaries of rational choice
theory. . . . My work may have sometimes assumed too much rationality,
but I believe it has been an antidote to the extensive research that does
not credit people with enough rationality” (402).
Becker’s last sentence suggests an alternative definition of behavioral
economics: crediting people with just the right amount of rationality and
human foibles. The trick is in figuring out what is just the right amount.
The approach taken by most behavioral economists has been to focus on
a few important ways in which humans diverge from homo economicus.
The basic assumption of standard economic theory is that among all
the affordable consumption bundles, people choose the best one. One
way that assumption might fail is if the utility maximization problem is
too hard to solve; this is the problem of bounded rationality. Another
cause of nonmaximizing behavior is a lack of willpower. The morning af-
ter, many decide that the previous night included at least one drink too
many. Such self-control problems are the subject of my first publication
in the JPE (Thaler and Shefrin 1981), and one of the first behavioral eco-
nomics papers published in the journal since Clark’s.2
Shefrin and I tried to modify the standard approach as little as possi-
ble to accommodate the struggle that people commonly face when choos-
ing between an immediate small pleasure and larger delayed reward. Fol-
lowing Adam Smith’s Theory of Moral Sentiments (1759), our model endows
people with two conflicting sets of preferences, one belonging to a myo-
pic “doer” and the other to a farsighted “planner.” The doer lives just for
one period and cares only about consumption in that period. The planner
seeks to maximize the integral of doer utilities and so sometimes wishes
to constrain or influence the doer’s choices.
2
I cite one earlier paper below. This is a good time to acknowledge that I have likely
missed some important behavioral papers both before and after 1981. My apologies to
the authors of the papers I have missed. Blame it on bounded memory and attention.

One implication of the planner-doer model is that individuals can be
helped by market-supplied commitment strategies. Thaler and Benartzi
(2004) provide evidence to support this prediction. Benartzi and I cre-
ated a strategy to help reluctant savers that we called “Save More Tomor-
row.” Organizations offer their employees an opportunity to sign up for a
program (starting in a few months) in which their pension contribution
rates are increased each year when they get a pay raise. Standard eco-
nomic theory predicts that no one would join such a program (they would
not think they needed it) and that if they did, it would not change their
savings rates (since they were already saving the optimal amount). The pa-
per, published in a special issue of the JPE honoring my advisor Sherwin
Rosen, reported the effects of the program in the first firm to try the idea.
The results were striking: 80 percent of those offered Save More Tomor-
row chose to join, and those who joined more than tripled their savings
rates in just 4 years.
Kaur, Kremer, and Mullainathan (2015) study another type of commit-
ment strategy offered by an employer, this time in the context of increas-
ing output. The article reports on a yearlong experiment in which piece
rate workers were offered a dominated contract on randomly chosen days.
The employees could set a daily goal for themselves with the proviso that
if they meet the goal they are paid normally, but if they fail to meet the
goal they are paid only half the usual rate. Workers chose such contracts
fully 36 percent of the time, and they were wise to do so. For those who
opted in to the dominated contract, output (and thus pay) increased by
6 percent.
One of the most powerful findings of behavioral economics is “loss aver-
sion,” the psychological tendency to feel losses more acutely than gains.
As Adam Smith (1759) put it, “Pain . . . is, in almost all cases, a more pun-
gent sensation than the opposite and correspondent pleasure” (1981, III,
ii, 176–77). Although Daniel Kahneman and Amos Tversky (1979) and I
(1980) had earlier written about this phenomenon, its empirical validity
was still very much in question when Kahneman, Jack Knetsch, and I sub-
mitted an experimental paper on the subject to the JPE, later published in
1990.
In the experiment we randomly assigned half the subjects to receive
some object (often a coffee mug), with the other half getting nothing.
We then conducted a market for the mugs in which both buyers and sell-
ers stated their reservation prices. Since transaction costs were negligible
and the objects were randomly assigned, the Coase theorem predicts that
roughly half the mugs will change hands so that subjects who value mugs
the most end up owning them. Our hypothesis was that fewer than half
the mugs would trade because owners would regard a trade as a loss. This
hypothesis was strongly supported. In a typical experiment, the expected

number of trades was 11 but the empirical average was only 3.4. As pre-
dicted by loss aversion, median reservation prices for selling the mug were
roughly twice the median prices for buying the mug.
The editor handling this paper was George Stigler. He sent us back a
rejection letter based on a highly critical referee report from someone
Stigler described as a “heavyweight in the field.” The referee said that in-
come effects could explain our results since those who received the mugs
had received a windfall relative to those who did not. After taking a few
days to calm myself down (a good self-control strategy) I wrote back on
behalf of my coauthors (who were both away traveling) explaining why
the referee’s comments could not be taken seriously, either theoretically
or empirically. First, the marginal propensity to spend windfalls on uni-
versity insignia coffee mugs must be minuscule. Second, one experiment
explicitly tested and rejected this explanation. Stigler wrote back in his
usual witty style saying that JPE stands for Journal of Perspicacity and Equity,
and he offered to send both my letter and the original referee report to
another referee to adjudicate. That referee said that if forced to choose
between our view and that of the original referee, he would side with us,
which is how the paper came to be accepted.3
Perhaps the subfield of economics in which the behavioral approach
has had the greatest impact is finance, and although the JPE has pub-
lished quite a few influential articles on the subject of financial econom-
ics, not many have been behavioral. One exception is the paper by De
Long et al. (1990), “Noise Trader Risk in Financial Markets,” which takes
on a frequent misconception about the possible role of less than fully ra-
tional investors—“noise traders”—in well-functioning asset markets. De
Long et al. quote the conventional Chicago wisdom (e.g., Friedman
1953; Fama 1965) that noise traders can have little effect on prices and
that any mispricing cannot last long before being wiped out by rational
arbitrageurs.
De Long et al. make the crucial observation that arbitrageurs are likely
to be risk averse and to have short horizons (in part because they are usu-
ally managing other people’s money). Thus when attempting to exploit
mispricing caused by noise traders, arbitrageurs run the risk that what-
ever bias is inducing the noise traders to be excessively optimistic or pes-
simistic about a security might continue or even strengthen before the
arbitrageurs have made their profits. This “noise-trader risk” prevents ar-
3
The self-control paper also involved quite a bit of back and forth with the editor Sam
Peltzman, who somewhat reluctantly agreed to accept it rather than continue to exchange
letters. Both papers were published as the last paper in the issue, which I took as a signal
that they were considered the paper the editors were most ashamed to publish. It is grat-
ifying that both papers were ranked highly on the list of most-cited papers compiled by the
editors for this issue. Perhaps people read the JPE from back to front.

bitrage from eliminating the price effects of noise traders. Indeed, in the
De Long et al. model, noise traders actually make more money than ra-
tional traders because they inadvertently bear more noise trader risk,
which because of the risk aversion of the rational traders pays a positive
risk premium. So in this model noise traders can affect prices and they
do not necessarily go broke—they might even get rich!
It is one thing to demonstrate that noise traders matter in a theoretical
model; showing that noise traders influence actual market prices is an-
other matter. How does one prove that a price is “wrong”? One approach
is to exploit the basic building block of modern finance, the law of one
price: two identical assets must sell for the same price. One counterexam-
ple cited by De Long et al. is the case of closed-end funds in which the
price of a fund’s shares should be equal to the net asset value of the se-
curities the fund owns. But in fact closed-end funds typically sell at a dis-
count relative to net asset value and occasionally sell at a premium.
Owen Lamont and I (2003) published a paper on this theme in the
JPE with the obnoxious title “Can the Stock Market Add and Subtract?”
As you might guess by now, the answer to the question posed by the title
is “no.” Lamont and I study equity carve-outs, focusing on the prominent
example of Palm and 3Com. Here is the story in brief. Palm, a maker of
then-sexy hand-held computers, was owned by 3Com, a profitable tech-
nology company. On March 2, 2000, 3Com sold a small fraction of its stake
in Palm via an initial public offering (IPO). In this carve-out, 3Com re-
tained 95 percent of the shares of Palm but announced that, pending an
expected approval by the Internal Revenue Service, the remaining shares
would be distributed to 3Com shareholders. At that point, 3Com share-
holders would receive about 1.5 shares of Palm for every share of 3Com
that they owned.
The law of one price implies in this case that the price of 3Com must
be at least 1.5 times the price of Palm, since equity prices can never be
negative. However, on the day of the Palm IPO, Palm’s shares traded at
$95.06 a share, but 3Com ended the day trading at $81.81, well short of
the lower bound of $145 implied by the law of one price. Implicitly, the
market was pricing the 3Com “stub” (the company once Palm was gone)
at negative $22 billion!
Though it did not continue to invest much in the topic, the JPE pub-
lished an early and influential paper on nonstandard beliefs, which arise
when people do not use information optimally as traditional economic
theory says they should.4 Camerer, Loewenstein, and Weber (1989) dem-
onstrate that people display a “curse of knowledge,” in the sense that they
4
A recent theoretical paper in the broad theme of biased beliefs by Bordalo, Gennaioli,
and Shleifer (2013) studies cases in which “salient” features of the environment are given
excessive weight.

have a hard time recognizing that others do not know what they know.5 If
Sally has written the code for some app, she is likely to underestimate the
difficulty neophytes will have learning how to use the app. The problem
is that Sally is unable to simulate how she would think in the absence of
her expertise. Camerer et al. demonstrate the curse of knowledge in ex-
perimental markets in which traders with more information make system-
atic errors that affect market prices.
The editors gave us a limited amount of space for these essays, but that
has not proved to be a major problem for the topic to which I was as-
signed. I cannot say for sure whether the small number of behavioral eco-
nomics papers published in the JPE was a shortage of supply or demand,
but there are entire branches of behavioral economics that have not made
(much of) an appearance. Looking over DellaVigna’s (2009) review arti-
cle, one notices many themes that have been largely or entirely absent
from the pages of the JPE, such as framing effects, menu effects (subopti-
mal diversification, effect of defaults, choice overload), peer pressure, and
emotions.
When I came into the profession the JPE had a well-deserved reputa-
tion for having eclectic tastes. This was one reason Shefrin and I submit-
ted our paper on self-control to the JPE. As the field of behavioral econom-
ics continues to grow, it will be a shame if the JPE does not include more
behavioral research in its pages. I suggest the editors all read that paper
by John Maurice Clark. But if the JPE continues to eschew papers on such
topics, one can always quote Stigler and Becker (1977): “De gustibus non
est disputandum.”
References
Becker, Gary S. 1993. “Nobel Lecture: The Economic Way of Looking at Behav-
ior.” J.P.E. 101 ( June): 385–409.
Bordalo, Pedro, Nicola Gennaioli, and Andrei Shleifer. 2013. “Salience and Con-
sumer Choice.” J.P.E. 121 (October): 803–43.
Camerer, Colin, George Loewenstein, and Martin Weber. 1989. “The Curse of
Knowledge in Economic Settings: An Experimental Analysis.” J.P.E. 97 (Octo-
ber): 1232–54.
Clark, John Maurice. 1918. “Economics and Modern Psychology: I.” J.P.E. 26 (1):
1–30.
DellaVigna, Stefano. 2009. “Psychology and Economics: Evidence from the
Field.” J. Econ. Literature 47 (2): 315–72.
De Long, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J.
Waldmann. 1990. “Noise Trader Risk in Financial Markets.” J.P.E. 98 (August):
703–38.
5
A special case of the curse of knowledge is “hindsight bias” (Fischhoff 1975). Once peo-
ple know that something happened, they remember thinking that they knew it all along.

Fama, Eugene F. 1965. “The Behavior of Stock-Market Prices.” J. Bus. 38 ( Janu-
ary): 34 –105.
Fischhoff, Baruch. 1975. “Hindsight ≠ Foresight: The Effect of Outcome Knowl-
edge on Judgment under Uncertainty.” J. Experimental Psychology: Human Percep-
tion and Performance 1 (August): 288–99.
Friedman, Milton. 1953. Essays in Positive Economics. Chicago: Univ. Chicago Press.
Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler. 1990. “Experimental
Tests of the Endowment Effect and the Coase Theorem.” J.P.E. 98 (December):
1325–48.
Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of
Decision under Risk.” Econometrica 47 (March): 263–92.
Kaur, Supreet, Michael Kremer, and Sendhil Mullainathan. 2015. “Self-Control
at Work.” J.P.E. 123 (December): 1227–77.
Lamont, Owen A., and Richard H. Thaler. 2003. “Can the Market Add and Sub-
tract? Mispricing in Tech Stock Carve-outs.” J.P.E. 111 (April): 227–68.
Simon, Herbert A. 1987. “Behavioural Economics.” In The New Palgrave: A Dictio-
nary of Economics, vol. 1, edited by John Eatwell, Murray Milgate, and Peter
Newman. London: Macmillan.
Smith, Adam. [1759] 1981. The Theory of Moral Sentiments. Reprint, edited by
D. D. Raphael and A. L. Macfie. Indianapolis: Liberty Classics
Stigler, George J., and Gary S. Becker. 1977. “De Gustibus Non Est Disputan-
dum.” A.E.R. 67 (March): 76–90.
Thaler, Richard H. 1980. “Toward a Positive Theory of Consumer Choice.” J.
Econ. Behavior and Org. 1 (March): 39–60.
Thaler, Richard H., and Shlomo Benartzi. 2004. “Save More Tomorrow™: Using
Behavioral Economics to Increase Employee Saving.” J.P.E. 112, no. 1, pt. 2
(February): S164–S187.
Thaler, Richard H., and H. M. Shefrin. 1981. “An Economic Theory of Self-
Control.” J.P.E. 89 (April): 392–406.
Corporate Finance
Robert Vishny
Luigi Zingales
The Journal of Political Economy and Chicago economists have played a ma-
jor role in the development of the modern field of corporate finance,

pioneering the agency cost and property rights approach. This brief es-
say is a tribute to those contributions rather than a comprehensive review
of the field. We apologize to those authors excluded by our singular focus.
Over 50 years ago, papers by Modigliani and Miller (1958) and Coase
(1960) provided the foundations by establishing a remarkable set of “ir-
relevance” propositions. Coase established that, under the assumption
of zero transaction costs, the allocation of property rights does not affect
the ability of participants to achieve an efficient outcome. Modigliani and
Miller showed that, with investment policy held constant and with zero
taxes and transaction costs, the mix of securities issued by the firm does
not affect the total value of the firm. These propositions provided a seri-
ous challenge to those who thought institutional arrangements could be
easily explained.
Early Work
One of the fundamental assumptions behind the Modigliani-Miller irrel-
evance proposition is tax neutrality. Modigliani and Miller (1958, 1963)
immediately recognized that in the United States (and most of the world)
debt is tax favored at the corporate level. Thus, firm value increases when
debt replaces equity. Modigliani and Miller’s irrelevance proposition and
its tax implications have been incredibly influential in the practice of fi-
nance. The fundamental valuation techniques (from the weighted aver-
age cost of capital to the adjusted present value approach) are based on
the Modigliani-Miller proposition: they start from the cash flow available
for investors (regardless of whether they are debt or equity holders), and
then they adjust for the effect of taxes and possibly the cost of financial
distress.
Even in a world in which debt is tax advantaged at the corporate level,
capital structure can still be irrelevant for firm value if debt is tax disadvan-
taged from the perspective of personal taxes, as is the case in the United
States (and particularly so before the 1986 tax reform). As Miller (1977)
points out, if there is an interior equilibrium, it will have the characteris-
tic that one dollar of pretax profits paid as interest should deliver inves-
tors the same value as a dollar of pretax profits paid as dividends or capi-
tal gains. Thus, the structure of taxes affects the average leverage in the
economy, but any single company is still indifferent between issuing debt
and equity.
The Agency Cost Approach to Corporate Finance

Starting in the 1950s, the influential managerialist literature (Baumol
1959; Simon 1959; Marris 1964; Williamson 1964) challenged the assump-

tion of value maximization and often abandoned optimizing models al-
together in favor of ad hoc descriptive models. The Chicago response
was to focus on managerial agency problems to explain deviations from
value maximization as well as to understand real-world capital structures.
Alchian and Demsetz (1972) are early adopters of the agency cost per-
spective. They argue that the distribution of cash flow rights is determined
to minimize the expected cost of shirking associated with team produc-
tion. The firm’s owner is a centralized monitor who can measure the pro-
ductivity of team members and reward them accordingly. Ownership of
the residual cash flows provides this centralized monitor with the incen-
tive to be vigilant.
Jensen and Meckling (1976) also adopt an agency cost approach but
try to match the reality of the modern corporation. The corporation is de-
scribed as a “nexus of contracts” among various parties including owner-
managers, employees, suppliers, outside equity holders, and bondhold-
ers instead of a monolith endowed with a single objective such as value
maximization. Jensen and Meckling emphasize the conflicts between the
various parties, especially management versus outside shareholders and
bondholders versus shareholders.
For Jensen and Meckling, the essence of the manager-shareholder con-
flict is the dichotomy between cash flows and firm value on the one hand
and perquisites or nonpecuniary benefits on the other. Their notion of
perquisites has proved flexible enough to encompass various phenomena
including diversion of cash flows, costs of effort, and empire building.
Managers consume too many perquisites because they bear only a frac-
tion of their cost. Outside equity is costly because it drives a wedge be-
tween the benefits of perquisites to the manager and the cash flow cost
of those perquisites, although it allows the firm to pursue valuable invest-
ment opportunities. Borrowing may allow the manager to pursue some of
these investments while still bearing the residual cash flow consequences,
but leverage has its own associated agency costs. Jensen and Meckling
outline how the optimal scale of the firm is determined by the trade-off
between costly external finance and pursuit of positive net present value
investments along with the optimal mix of manager-owned equity, out-
side equity, and debt. Their seminal paper illustrates the sheer explana-
tory power of a simple agency cost framework in corporate finance.
The main weakness of Jensen and Meckling’s model is the omission of
control rights of debt or outside equity as a means of limiting agency costs.
Fama and Jensen (1983) argue that separation of residual risk bearing
from decision management necessitates systems that also separate deci-
sion management from decision control. In large corporations, we typi-
cally have an explicit mechanism such as the board of directors for mon-
itoring and decision ratification. Their paper has stimulated a thriving
empirical literature on the role of boards of directors.

Jensen (1986) argues that corporate cash flow in excess of investment
needs (free cash flow) runs the risk of being wasted by managers through
self-aggrandizing negative net present value projects. The presence of
debt in the capital structure has the benefit of reducing this kind of waste.
This “free cash flow theory” provided a compelling rationale for the 1980s
leveraged buyout wave.
Up to this point, the literature still lacks a good theory of ownership as
distinct from claims to profit shares. Grossman and Hart (1986) fill this
void. Their notion of ownership as the right to make decisions when con-
tracts are incomplete has proved very powerful. It has influenced research
on many topics including corporate governance and ownership patterns
around the world, state-contingent financial contracting, venture capital
contracting, and the role of public enterprises.
There is a large empirical literature on these topics. Early papers in-
clude Demsetz and Lehn (1985), Morck, Shleifer, and Vishny (1988), and
Kaplan (1989).
Demsetz and Lehn (1985) study the variation of ownership concentra-
tion across large US firms and also find that there is no significant rela-
tionship between ownership concentration and profitability. Morck et al.
(1988) find a nonmonotonic relationship between management owner-
ship and market valuation, possibly indicating the interplay of an incen-
tive effect as well as an entrenchment effect that sets in when manage-
ment holds a large block of shares. Kaplan (1989) finds that ownership
and leverage realignments via management buyouts have a positive im-
pact on profitability, consistent with improved incentives.
The Market for Corporate Control

Manne (1965) is the first to point out the role of the market for corpo-
rate control in promoting efficiency and protecting atomistic sharehold-
ers from self-interested managerial behavior. Writing at a time when the
antitrust consensus against horizontal mergers is very strong, he advo-
cates taking into account the efficiency benefits of corporate takeovers.
He explains how a low stock price resulting from inefficient management
provides an opportunity for profit that acts as a stronger force for change
than internal governance mechanisms.
Grossman and Hart (1980) take a similar view on the benefits of the
market for corporate control but point out the limitations of the mech-
anism due to a free-rider problem. Atomistic shareholders have no incen-
tive to tender their shares for anything less than the full value of those
shares after managerial improvements. This makes it impossible for a
more efficient buyer to profit on buying their shares. Shleifer and Vishny
(1986) document the prevalence of large block holdings in US corpora-

tions and analyze the role of large shareholders as activist investors who
can help overcome this free-rider problem.
The Labor Market

The pressure from the corporate control market is not the only remedy
against agency problems. As Fama (1980) points out, an important form
of discipline comes from the labor market, because the current marginal
product of managerial labor contains information about future expected
marginal products. Thus, the wage revision process imposed by the man-
agerial labor market will reward well-performing managers and penalize
poorly performing ones. In some special cases, this ex post settling up
eliminates completely the costs of separation of ownership and control.
For this mechanism to work, however, managerial salaries should vary
significantly on the basis of past performance and so should the proba-
bility of dismissal. Jensen and Murphy (1990) do not observe much evi-
dence for these two predictions, a finding they attribute to unspecified
“political forces” that constrain “the type of contracts that can be written
between management and shareholders” (227).
International Dimension
Corporate finance theory was developed in the United States, inspired
by US stylized facts, and for the first 30 years mostly tested on US data.
Chicago economists have played a significant role in internationalizing
the field. Rajan and Zingales (1995) confronted US-based capital struc-
ture theories with international evidence. They document that corporate
leverage is fairly similar across developed countries. The differences in le-
verage reflect the way bankruptcy is designed rather than the divide be-
tween bank-centered and market-based economies. Where bankruptcy
favors liquidation, firms appear more hesitant to lever up.
In corporate finance, it is often difficult to determine the direction of
causality: does the law drive the behavior or does the behavior drive the
law? La Porta et al. (1998) show that long-standing differences between
legal systems can explain much of the variation in key investor protec-
tion laws across countries. These differences between legal systems can
be traced to their families of origin, which in turn resulted from “a com-
bination of conquest, imperialism, outright borrowing, and more subtle
imitation” (1115). The authors document that the laws of common law
countries (originating in English law) are more protective of outside in-
vestors than those of civil law countries (originating in Roman law) and
that this difference can explain a significant fraction of the variation in
financial development around the world. Their methods of coding inves-

tor protection laws have influenced much research by academics and pol-
icy makers and sparked a lively debate. Subsequent research has shown
that the family of legal origin is highly correlated with a wide range of laws
governing economic activity.
At Chicago, ideas are subjected to intense scrutiny, even among col-
leagues. While recognizing the importance of the legal origins argument,
Rajan and Zingales (2003) thought that a more variable factor is needed
to explain both the time-series variation and the cross-sectional differences
in financial development. Their explanation is the opposition by incum-
bents, who feel threatened by financial markets, because they breed com-
petition. They claim that incumbents’ ability and willingness to resist fi-
nancial development are reduced by free trade and capital flows. Indeed,
financial market development accelerates with open borders and retrenches
in periods of protectionism and restrictions to capital flows.
Given the two Chicago irrelevance propositions mentioned above, it is
important to establish that these observed differences in financing pat-
terns and financial market development are not just a neutral mutation,
but matter for real economic variables. To this end, Rajan and Zingales
(1998) show that differences in financial development can explain a coun-
try’s ability to reallocate resources from sectors in excess to sectors in
need, accelerating the growth of these sectors in need. The difference-
in-difference strategy adopted in their paper has since become a standard
in the literature.
New Directions
Chicago economists have also helped to broaden finance research be-
yond its traditional focus on large corporations. Entrepreneurial finance
and household finance are two important new research areas. The field
of entrepreneurial finance has grown along with the growth of private
equity as an asset class and its increased role in driving innovation. In an
early paper, Kaplan and Strömberg (2003) use the agency cost perspec-
tive to understand the allocation of cash flows and control rights in a
large cross section of venture capital contracts. They find a strong corre-
spondence between the predictions of agency theory and the structure of
real-world venture capital contracts.
Household finance is increasingly recognized as perhaps the key link
between finance and macroeconomics. Amit Seru and his coauthors have
shown how securitization has led to lax lending standards as well as diffi-
culty renegotiating bad loans (Keys et al. 2010; Piskorski, Seru, and Vig
2010). Atif Mian and Amir Sufi have established the close association be-
tween household debt and economic fluctuations using an array of data
sets including detailed zip code–level data in the United States and a

1960–2012 panel data set covering 30 countries (see Mian and Sufi 2014;
Mian, Sufi, and Verner 2017).
References
Alchian, Armen, and Harold Demsetz. 1972. “Production, Information Costs
and Economic Organization.” A.E.R. 62:777–95.
Baumol, William. 1959. Business Behavior, Value and Growth. New York: Macmillan.
Coase, Ronald. 1960. “The Problem of Social Cost.” J. Law and Econ. 3:1– 44.
Demsetz, Harold, and Kenneth Lehn. 1985. “The Structure of Corporate Own-
ership: Causes and Consequences.” J.P.E. 93:1155–77.
Fama, Eugene. 1980. “Agency Problems and the Theory of the Firm.” J.P.E. 88:
288–307.
Fama, Eugene, and Michael Jensen. 1983. “Separation of Ownership and Con-
trol.” J. Law and Econ. 26:301–25.
Grossman, Sanford, and Oliver Hart. 1980. “Takeover Bids, the Free Rider Prob-
lem, and the Theory of the Corporation.” Bell J. Econ. 11:42–64.
———. 1986. “Costs and Benefits of Ownership: A Theory of Vertical and Lateral
Integration.” J.P.E. 94:691–719.
Jensen, Michael. 1986. “Agency Costs of Free Cash Flow, Corporate Finance and
Takeovers.” A.E.R. 76:323–29.
Jensen, Michael, and William Meckling. 1976. “Theory of the Firm: Managerial
Behavior, Agency Costs and Ownership Structure.” J. Financial Econ. 3:305–60.
Jensen, Michael, and Kevin J. Murphy. 1990. “Performance Pay and Top Manage-
ment Incentives.” J.P.E. 98:225–64.
Kaplan, Steven. 1989. “The Effects of Management Buyouts on Operating Per-
formance and Value.” J. Financial Econ. 24:217–54.
Kaplan, Steven, and Per Strömberg. 2003. “Financial Contracting Theory Meets
the Real World: An Empirical Analysis of Venture Capital Contracts.” Rev.
Econ. Studies 70:281–315.
Keys, Benjamin, Tanmoy Mukherjee, Amit Seru, and Vikrant Vig. 2010. “Did Se-
curitization Lead to Lax Screening? Evidence from Subprime Loans.” Q.J.E.
125:307–62.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny.
1998. “Law and Finance.” J.P.E. 106:1113–55.
Manne, Henry. 1965. “Mergers and the Market for Corporate Control.” J.P.E.
73:110–20.
Marris, Robin. 1964. The Economic Theory of “Managerial” Capitalism. Glencoe, IL:
Free Press.
Mian, Atif, and Amir Sufi. 2014. House of Debt: How They (and You) Caused the Great
Recession and How We Can Prevent It from Happening Again. Chicago: Univ. Chi-
cago Press.
Mian, Atif, Amir Sufi, and Emil Verner. 2017. “Household Debt and Business Cy-
cles Worldwide.” Q.J.E. Electronically published May 26.
Miller, Merton. 1977. “Debt and Taxes.” J. Finance 32:261–75.
Modigliani, Franco, and Merton Miller. 1958. “The Cost of Capital, Corporate Fi-
nance and the Theory of Investment.” A.E.R. 48:261–97.
———. 1963. “Corporate Income Taxes and the Cost of Capital: A Correction.”
A.E.R. 53:433 – 43.
Morck, Randall, Andrei Shleifer, and Robert Vishny. 1988. “Management Owner-
ship and Market Valuation: An Empirical Analysis.” J. Financial Econ. 20:293 –
315.

Piskorski, Tomasz, Amit Seru, and Vikrant Vig. 2010. “Securitization and Dis-
tressed Loan Renegotiation: Evidence from the Subprime Mortgage Crisis.”
J. Financial Econ. 97:369–97.
Rajan, Raghuram, and Luigi Zingales. 1995. “What Do We Know about Capital
Structure? Some Evidence from International Data.” J. Finance 50:1421–60.
———. 1998. “Financial Dependence and Growth.” A.E.R. 88:559–86.
———. 2003. “The Great Reversals: The Politics of Financial Development in the
Twentieth Century.” J. Financial Econ. 69:5–50.
Shleifer, Andrei, and Robert Vishny. 1986. “Large Shareholders and Corporate
Control.” J.P.E. 94:461–88.
Simon, Herbert. 1959. “Theories of Decision Making in Economics and Behav-
ioral Science.” A.E.R. 49:253–83.
Williamson, Oliver. 1964. The Economics of Discretionary Behavior: Managerial Objec-
tives in a Theory of the Firm. Englewood Cliffs, NJ: Prentice Hall.
Banking and the Evolving Objectives

of Bank Regulation
Douglas W. Diamond
Anil K. Kashyap
Raghuram G. Rajan
Views on the role played by banks in the economy have evolved greatly
over the last 125 years, as have arguments on the need, as well as the best
way, to regulate them. Some of the key insights in the debate have been
published in the Journal of Political Economy. In what follows, we will out-
line the main contributions to the debate in recent years, with an empha-
sis on work done at the University of Chicago or published in the JPE. We
These views are those of the authors only and are not necessarily shared by the Bank of
England or any other institutions with which we are affiliated.

want to emphasize work that has relevance today, but despite this caveat,
we will probably end up doing injustice to work published long ago.
We begin with a framework for organizing the theories of intermedi-
ation. We then draw out the implications for what the theories say about
regulation and note that in many respects the motivation for regulation
has been only loosely tied to the theory of intermediation. We close with
some open questions for regulators and economists interested in bank-
ing. We do not survey the research that has followed up on work pub-
lished in the JPE, nor will we attempt to provide a detailed overview of
the entire academic literature on banking. For that, we refer the reader
to the excellent work by Gorton and Winton (2003) and Freixas and
Rochet (2008).
Theoretical Overview
We define banks as financial institutions with a substantial fraction of il-
liquid assets financed with demandable liabilities payable at par. Bank-
ing theories typically have focused on one side of the bank’s balance sheet
as critical to its economic role and then argued why the other side of the
bank’s balance sheet has to take the form it does. In general, therefore,
theories tend to emphasize the criticality of both sides of the balance sheet.
Observing that bank-issued certificates of deposit pay market interest rates
and that banks must hold central bank–issued reserves against them that
pay a below-market rate, Fama (1985) argues that banks must provide
some valuable services in order to bear this implicit tax. What might these
services be? Let us start with various forms of liquidity provision.
Liquidity Provision
In environments in which the government does not issue sufficient pay-
ment media, banks could issue short-term demandable paper payable at
par to fill the gap. An argument against privately issued bank money is
that a bank will be tempted to overissue bank notes at par to unsuspect-
ing clients and then default. Such “wildcat banking,” critics argue, justi-
fies a role for government-provided fiat money. Gorton (1996) analyzes
the prices of private bank notes issued in the American Free Banking Era
(1838–63) and finds that the risk of failure was priced into the bank
note; for example, notes of new banks are discounted far more, and the dis-
count declines as banks make payments over time (as predicted by Dia-
mond [1989]; see later). Market discipline was a deterrent against over-
issue, though an open question is whether it sufficiently accounted for
the risks of default.
Krishnamurthy and Vissing-Jorgenson (2012) show that even in mod-
ern times, the private sector supplies more short-term debt when govern-

ment debt issuance contracts, suggesting some degree of “gap filling.”
This has led for calls for the central bank (or government) to “crowd out”
private short-term liability issuance by issuing risk-free excess reserves or
short-term central bank paper (see, e.g., Carlson et al. 2016).
In a related vein to banks issuing payment media, banks may also pro-
vide depositors insurance against liquidity needs (Bryant 1980; Diamond
and Dybvig 1983). Consider, for example, the three-date Diamond-Dybvig
world starting at date 0 with depositors, some of whom may need to con-
sume early (at date 1) and others who may need to consume late (at
date 2). Also, the available investment projects at date 0 return a positive
net return if left to mature at date 2 but return only the capital invested if
liquidated beforehand. If depositors do not know what type they are a
priori and invest directly in the project, the early consumers will have
to settle for less consumption than late consumers. A bank can, however,
act as a risk-sharing mechanism by promising to pay those who have to
withdraw early a little more than the capital invested, while paying those
who withdraw late a little less than the return to maturity of the project
(but more than early withdrawers to incentivize them to stay for the long
run). Unfortunately, though, this can expose banks to runs, because if ev-
eryone decides to withdraw early, the bank is committed to paying more
than the early liquidation value of its assets. In such a situation, because
the bank will not have anything left to pay those who do not demand to
withdraw their money early, it makes sense for everyone to indeed with-
draw their money early.
In sum, Diamond and Dybvig (1983) show that the liquidity risk shar-
ing benefit that they identify is subject to panic-based runs (dominated
Nash equilibria) even when illiquid assets are risk-free. They identify con-
tractual solutions, such as suspension of convertibility of deposits to cash
that can deter runs, but also show that government deposit insurance can
do better in some circumstances because of the taxation authority of the
government. The Diamond and Dybvig model has become the work-
horse model in banking, in part because with a very simple framework,
it rationalizes much of the structure as well as fragility of banks.1
Holmström and Tirole (1998) consider another way intermediaries
can share and alleviate liquidity risk. In their model, the demand for li-
quidity comes from firms that get shocks that require them to infuse new
1
The literature on runs has explored many other possibilities. In Bryant (1980) runs are
based on depositor information and occur when the bank will be insolvent for any level of
withdrawals. These information-based runs (in contrast to the liquidity or panic-based runs
in Diamond and Dybvig [1983]) are also studied in Jacklin and Bhattacharya (1988). Post-
lewaite and Vives (1987) consider runs based on noisy information about when a depositor
and other depositors will need their deposits. Diamond and Rajan (2005) and Goldstein
and Pauzner (2005) examine runs in which depositors consider both information about
solvency and its implications for bank losses due to illiquidity.

funds to protect existing investments. Moral hazard at the firm level en-
sures that there are positive wedges between the amount the firms need
to infuse and what they can actually raise from the markets once the
shock hits. Thus firms need to carry extra value ex ante that they can
dip into if they have liquidity needs, so that positive net present value in-
fusions can be funded.
Holmström and Tirole go on to ask how firms can do this if cash can-
not be stored. One possibility is to hold claims against other firms. The
problem is that this could be an inefficient way to store value since some
firms will not need to infuse much, and will have “excess” liquid assets
while others will have to infuse a lot and have too little. A better solution
is to obtain committed lines of credit from an intermediary who holds
shares in all firms. Only firms that need liquidity will draw down these
lines. The intermediary can thus allocate available scarce liquidity to
those who need it, avoiding trapped pools of liquidity. Holmström and
Tirole go on to ask what would happen if the need for liquidity is aggre-
gate: every firm needs to infuse more at the same time, and the amount
each firm can raise (given the moral hazard wedge between what it earns
and what it can promise to pay out) is too low. In such situations, even an
intermediary cannot help since there is an aggregate shortage of pledge-
able value relative to what needs to be raised. Holmström and Tirole point
to a role for the government, which can get access to the value generated
by the firm (because of the taxation authority of government) that a
private-sector financier cannot access. The government can then lend
more to firms than the private sector. Alternatively, firms can buy and hold
government bonds, as a reliable source of value, to be sold for funds when
the need arises. Holmström and Tirole suggest a liquidity premium for
safe government claims, which is verified by Krishnamurthy and Vissing-
Jorgensen (2012).
Kashyap, Rajan, and Stein (2002) also argue that banks may be useful
in pooling demands for liquidity and that as long as demands to withdraw
deposits are not perfectly positively correlated with drawdowns on lines
of credit, banks may be in the best position to optimally use any given pool
of liquidity. Indeed, they find that banks make more loan commitments
than other intermediaries, and within the banking sector, banks with high
ratios of transaction deposits to total deposits also have high ratios of total
commitments to total loans.
Finally, much of the literature is about using intermediaries to divide
up existing asset liquidity in better ways to meet the needs of investors.
Diamond and Rajan (2001) argue that banks create additional liquidity
because of their special capabilities. In their model, entrepreneurs need
to raise money to fund projects. However, the entrepreneur’s specific
abilities are important to generate value. Because the entrepreneur can-
not commit to stay with the project, he has the ability to hold up lenders,

which creates a wedge between the revenues he can generate and the
amount he can borrow. This is the source of project illiquidity. Bankers,
by learning alternative or second-best uses for the project assets without
the entrepreneur, have a greater ability to extract repayment from him.
This allows them to lend more to him. But what if the banker needs to
raise money himself ? Does the chain of illiquidity reassert itself since
the banker can similarly hold up investors by threatening not to put his
special collection skills to work?
Diamond and Rajan argue that funding through demandable deposits
prevents such banker holdup and tie the banker’s collection skills to the
loans he has made, thus enabling him to borrow against the full value of
the loans. Intuitively, if the banker threatens to pay depositors less than
they are owed, they run on the bank. Importantly, because the banker is
only an intermediary transferring value from the entrepreneur to the de-
positor and does not generate independent value, he can be shut out of
any postrun negotiation. This implies that the run does squeeze out any
intermediary rents and the threat of a run acts as an effective disciplinary
device on the banker. Note that because the entrepreneur adds value to
the project, he cannot tie his human capital to the project by issuing de-
mandable debt directly to investors: demand deposits discipline interme-
diaries, not firms, which distinguishes Diamond and Rajan (2001) from
Calomiris and Kahn (1991), where demand deposits discipline all.
Diamond and Rajan thus argue that demandable debt is a feature of
banks, not a bug, and enables the bank to raise money whenever needed
to fund firms with more than they can borrow from markets. This im-
plies that bank fragility cannot be eliminated without eliminating bank
funding cost advantage. Diamond and Rajan (2000) explore the role of
bank capital in reducing fragility, which has to be traded off against the
enhanced cost of bank funding.
Allen and Gale (1997) suggest a different form of risk sharing in inter-
mediaries: smoothing intergenerational risk. They study a standard over-
lapping generations model with risk-averse investors. The young get an
endowment when born. There is a fixed-supply risky asset that is infi-
nitely lived and pays a nonnegative but independent and identically dis-
tributed dividend, as well as safe assets paying zero dividends that can
serve as a store of value. Because the representative young agent solves
the same problem each period, the equilibrium price of the risky asset is
constant. Given that it pays a nonnegative, and sometimes positive, divi-
dend, its total return dominates that of the safe asset. The safe asset is
therefore never used in the market equilibrium. This means that each gen-
eration bears a substantial amount of risk, since its last-period consump-
tion varies with the entire dividend paid by the risky asset. Allen and Gale
rule out any possibility of market-based intergenerational insurance since

the old know the dividend they will get before the young are born, and
there is no scope for risk sharing after the risk is realized.
The key then to risk sharing is an intermediary, in whom generations
invest their residual endowment (after consumption). The intermediary
holds the risky asset and builds up reserves when the dividend is high,
only to run them down at times of low dividend. Each generation then
gets a smoothed return, which improves on market-based outcomes.
This is probably a better model of intermediary dominated financial sys-
tems (such as what Germany or Japan used to be) than of single intermedi-
aries. More recent work by Dang et al. (2017) develops a related idea in
which banks keep the information about the realized value of their invest-
ments secret so as to facilitate risk sharing. Secrecy provides incentives for
new depositors to deposit to provide liquidity to the previous generation
even when the low value of bank assets, if known, would leave the bank
unable to offer them a rate of return that matches that available in the
market.
Banks as Monitors
We have already reviewed papers that emphasize the role of banks in
monitoring or managing assets to enhance liquidity provision. Other
work emphasizes how monitoring can alter the availability of funding to
borrowers.
Diamond (1984) argues that costly monitoring by banks can resolve
moral hazard or adverse selection problems at firms. But then who mon-
itors the monitor? Does this not simply push the problem one step back:
will investors in the bank not have to engage in costly monitoring to en-
sure the bank monitors? Diamond argues that when the bank is diversi-
fied across a large number of loans, bank asset values will be less sensitive
to the private information in each loan. If investors in the bank hold
debt claims, they will not need to have information about the bank’s
portfolio value to enforce those claims and, if the claims are sufficiently
safe, will not have to individually monitor the bank to see that it is doing
its job. Furthermore, the need to service the debt claims forces the bank
to monitor the loans and to repay the depositors. Banks are special be-
cause diversification reduces the importance of private information at-
tached to each loan the bank makes and makes the bank’s overall bal-
ance sheet more transparent. Banks in Diamond (1984) are thus the
original form of pooling (diversification) and tranching (issuing senior
claims to outside depositors and retaining junior claims inside the bank)
structures that have proliferated in securitization vehicles.
Subsequent work in the JPE contributes to the characterization of the
borrowers who would benefit most from monitoring and be most depen-

dent on banks as a source of finance. The analysis is based on differences
in the severity of borrower moral hazard when borrowers consider the
effect of their current choices on their future access to finance.
Diamond (1989) examines the role of reputation as a way of reducing
borrower moral hazard, and Diamond (1991) extends this to examine
the interaction of monitoring and reputation effects. These models pre-
dict that new borrowers would be subject to severe moral hazard and that
the severity would be reduced over time for borrowers who survive and
acquire a good reputation for repaying investors. The analysis differs
from then-existing models of reputation by predicting that it may take
time to acquire a reputation. The earlier models (Fama 1980; Holmström
1982; Kreps and Wilson 1982; Milgrom and Roberts 1982) focused on the
effects of a prospect of having a reputation in the future rather than on
the costs of losing one’s current reputation.
Diamond (1989) considers a model of borrower moral hazard (dis-
torted incentives for real investment risk choice) in which the project
choice is private information, debt contracts are optimal financial con-
tracts, and the only thing investors observe is the realized payment they
receive. Borrowers who repay debt over time acquire a better reputation
(a better credit rating), and this reputation becomes an asset that they
lose if they subsequently default. Those who repay debt over time consist
of those who always choose safe investments (have no moral hazard) and
those who are subject to moral hazard but whose risky investments had a
realization sufficient to repay investors. These repayments separate the
borrowers from those whose risky investments have realizations insuffi-
cient to repay investors. This learning about survivors improves their rep-
utation, and their moral hazard is reduced sufficiently until they prefer
to avoid risky investments to maintain their reputation. The reputation
of borrowers in this model is measured by their credit rating (their prob-
ability of default for a given level of borrowing).
Diamond (1989) ignores the possibility of borrowing from a lender
who can monitor the investment choices of a borrower. Diamond (1991)
investigates when monitoring will and will not be valuable. A separation
emerges in which new borrowers without a long track record need moni-
toring from banks, while others who always repay such debt for a long
enough time acquire a sufficiently good reputation to borrow directly with-
out monitoring. The second group can issue debt directly to public mar-
kets. Although their record of successful repayments is made while invest-
ment choices have been monitored by banks, it helps future lenders learn
about them: they separate themselves from borrowers who choose risky
investments despite monitoring (those who have only very risky invest-
ments available or, more generally, those with stronger conflicts of inter-
est). This produces a life cycle theory of borrowing: young borrowers
(small and medium-sized businesses) that borrow from banks and ma-

ture ones that acquire a good enough credit rating switch to unmoni-
tored borrowing and no longer depend on bank finance.
A small extension of this model could consider a third, extremely risky
set of borrowers with a limited track record. Suppose that bank monitor-
ing is effective but imperfect: either a competing higher level of moni-
toring exists (from family members) or some initial borrowers have such
strong moral hazard that monitoring cannot improve their incentives.
This implies that these very young and risky borrowers first self-finance
or borrow from family members who can monitor more closely than banks.
Only those who survive this start-up period can borrow from banks. This
identifies the set of borrowers whose access to finance depends on the
banking system and its financial health.
The possibility that banks may have private information about borrow-
ers acquired over time raises the question of costs of bank-firm relation-
ships. Rajan (1992) argues that because firms might be seen as lemons if
they exit such a relationship, banks have holdup power over firms. This
possibility, in turn, can influence firms’ choice of financing between long-
term arm’s-length financing and bank financing and the number of banks
they may want to borrow from. Bolton and Freixas (2000) present another
private information–based model, adding costs of restructuring debt and
of issuing bank equity to provide a theory of the type of firms that issue
bonds, equity, and bank debt.
Aggregate Liquidity Shortages, Fire Sales,

and Contagion
If banks have special skills in evaluating and monitoring loans, then there
is a small group of entities with similar skills that have the ability to pur-
chase such loans. This raises the possibility that when there is an aggre-
gate shortage of financing in the market and a bank has to sell the loans
on its books to repay the short-term debt it has coming due, there will be
a limited pool of buyers with limited resources to buy those loans, and
loans will sell, not for the full value that buyers can collect, but for what
they can pay. The possibility that loans are priced at a fire sale discount
value because the best buyers have limited financing is an important
source of risk in banking. Shleifer and Vishny (1992) provide an early dis-
cussion of this phenomenon and Allen and Gale (1994) discuss the ef-
fects of limited participation in financial markets. To the extent that
banks do not internalize the fact that they will be consuming the limited
common pool of resources available to buy loans when there is a future ag-
gregate shortage, thereby lowering the fire sale price and, importantly, the
ability of other banks to finance themselves, there is a fire sale externality
that causes banks to overissue short-term debt (see Stein 2012).

In Allen and Gale (2000), aggregate liquidity shortages can result in a
contagion of bank failures. Interestingly, these arise from an attempt to
share aggregate liquidity through interbank deposits, as opposed to a
single bank that owns each firm as in Holmström and Tirole (1998). Es-
sentially, as long as there is no aggregate liquidity shortage, a liquidity
shortage in one region can be smoothed over by a bank drawing down
on deposits it has made in banks in liquidity-surplus regions. However,
if there is an aggregate liquidity shortage, banks could fail in a region
that is short, since enough liquidity is not available elsewhere. This will
imply a collapse in deposit values, which will reduce the value of banks
that hold interbank deposits in the failing bank and transmit the failure
to regions connected by interbank deposits. Interestingly, Allen and
Gale argue that partial interconnection can be worse than full intercon-
nection between banks across regions since the latter allows better use of
the available common pool of aggregate liquidity.
In Diamond and Rajan (2005), all banks have access to the common
pool of liquidity, and in contrast to Allen and Gale (2000), there are no
ex ante interbank claims. Nevertheless, even in this structure bank fail-
ures can be contagious; bank insolvencies precipitate runs that cause
them to dump all their assets on the market, thus shrinking the available
pool of liquidity even more, causing other banks to become insolvent and
run. In these views of crises, the failure of banks disrupts lending relation-
ships and causes firms to face credit crunches. Investment and production
then collapse. If banks are special, central bank liquidity provision can
help keep banks solvent in the face of panics (see the evidence in Carlson,
Mitchener, and Richardson [2011]) and avert wider systemic distress.
An alternative view of crises espoused by Friedman and Schwartz
(1963) is that they stem from a shortage of payment media as bank fail-
ures result in a contraction of bank deposits. Once the payment system
collapses, spending declines and falling prices are inevitable. In their
view a series of banking collapses is just one way in which the money sup-
ply could contract, and as in all contractions in the money supply, the
results for the real economy and deflation would be bad.
Financial Regulation
The evolution of financial regulation was dramatically overhauled dur-
ing the Great Depression. The chaos associated with runs and massive
number of bank closures and failures spurred a number of policy propos-
als to prevent that from occurring again. For example, Freidman and
Schwartz (1963) emphasized the critical role of the central bank in not al-
lowing sharp money supply contractions. Other regulations were aimed
more at the financial system itself, imposing constraints on different insti-
tutions or agents and their ability to set interest rates or capital structures.

Many of these important regulatory interventions occurred before the
theories surveyed above had been formally developed, but the intuition
behind them played a role.
One prominent reaction, championed by Henry Simons and others at
the University of Chicago (Simons 1933, 1936), was to call for an end to
fractional reserve banking in what came to be known as the “Chicago
Plan.”2 This idea, which in recent incarnations would be called narrow
banking, is consistent with the view that the value added of the banking
system comes from the payment services it provides. If the liabilities are
what is special, then restricting the asset side of the banks to be boring
and safe is a logical proposal.
While the Chicago Plan attracted many followers and appears to have
received considerable attention at all levels of government, the Banking
Act of 1933 instead created a national deposit insurance scheme. Modern
theories recognize that deposit insurance brings some stabilizing benefits
but also creates distortions by creating incentives to take risks that might
be borne by a deposit insurance fund. Indeed, to mitigate risks, banks
were subject to a plethora of additional regulations including limitations
on interest payments. While these were relatively benign when market in-
terest rates were low, they became problematic as market interest rates
shot up in the 1970s and 1980s.
The next pair of major banking reforms in the United States, the adop-
tion of the first Basel capital standards in 1988 and the FDIC Improve-
ment Act (FDICIA) of 1991, can be viewed as the responses to the long-
standing concerns about deposit insurance distortions. The Basel accords
mandated that banks hold more capital in their liability structure, and
FDICIA forced bank supervisors to close severely undercapitalized banks
without delay so that fewer would operate while insolvent.
The latest wave of regulatory changes came in response to the global
financial crisis of 2007–9. The policy responses appear to reflect differ-
ent interpretations of the root causes of the crisis. Some new regulations
(for instance, the Dodd-Frank Act in the United States) are most naturally
viewed as concluding that some banks were “too big” or “too intercon-
nected” to fail and that reforms to prevent a replay were needed. Others
(e.g., some aspects of the latest Basel reforms) suggest that widespread
runs that were mostly outside the traditional commercial banking system
were the problem.
The lack of a unified diagnosis of what went wrong has led to four ma-
jor regulatory innovations. First, the push for higher capital requirements
was accelerated, especially given the view that financial institutions had
2
Simons describes the plan briefly as an example of how to implement the ideas in his
classic 1936 JPE paper on rules vs. discretion.

gamed existing capital requirements before the crisis. Capital require-
ments for all banks have risen substantially, with special additional rules
for large global entities. While these are understandable given the extent
of leverage before the crisis, it is important to note that banking theories
do suggest that the cost of bank financing can go up as more bank capital
is required. Unlike a pure Modigliani-Miller view of bank capital struc-
ture, this more nuanced view suggests there are trade-offs in setting bank
capital. What the optimal level of bank capital should be is still an open
question.
Second, liquidity regulations are also being imposed for the first time.
These rules force banks with more illiquid assets to use more long-term
funding and also mandate that banks with more runnable funding should
hold more liquid assets that could easily be sold to meet outflows. The the-
oretical emphasis on liquidity outlined earlier and the externality caused
by excessively illiquid assets funded by runnable claims support this fo-
cus. However, calibration of these regulations is proceeding with little the-
oretical guidance.
Third, new regulators have been created, with responsibilities for
looking at the stability of the entire financial system rather than individ-
ual institutions or sectors. The range of tools and authorities for these
“macroprudential” regulators vary greatly. These efforts are in their in-
fancy, so it is too early to tell whether this approach will succeed in deliv-
ering extra stability and, if so, how and why. Once again, though, the
ideas that liquidity and solvency are interlinked both within and across
financial institutions and that the location of excess liquidity as well as
the quantum of aggregate liquidity matters lend support to efforts at mac-
roprudential interventions. More research, both theoretical and empiri-
cal, however, is needed to guide regulations.
Finally, banking regulators everywhere have begun “stress-testing” large
banks. These exercises simulate how banks will fare under different mac-
roeconomic scenarios. Motivated by the same concerns as FDICIA, that
the book value of bank equity is a lagging indicator of bank health, these
tests have become the primary supervisory tool in many jurisdictions.
The postcrisis regulations raise some obvious research questions. For
instance, will macroprudential regulation prove to be a mirage or will it
really change the riskiness of the financial system and its resilience? In
most jurisdictions, activist macroprudential policies have not been pur-
sued, and given the effect they could have on firms’ profitability and busi-
ness practices, some political pressure to avoid acting is likely to be pres-
ent. What will be the most effective way to design a macroprudential
regulator? Which tools are needed to deliver on the mandate? How do
(and should) these policies interact with monetary policy?
This leads to a related set of questions. To what extent does monetary
policy interact with bank lending and bank liquidity? Do central bank
promises of abundant aggregate liquidity if the system is stressed cause

banks to create overly illiquid balance sheets (see, e.g., Diamond and
Rajan 2012; Farhi and Tirole 2012)? To what extent do bank regulation
and monetary policy work at counterpurposes?
Another issue is whether the international regulatory coordination
that has featured so prominently since the crisis will prove to be success-
ful. International monetary and fiscal policy coordination has rarely been
sustained, and many question whether it is desirable except in very ex-
treme circumstances. Is there some reason why financial regulation de-
mands coordination, and is the harmonization that has been pursued
working? Would it make more sense to allow countries to have more reg-
ulatory independence, while improving the frameworks for dealing with
cross-border insolvencies and spillovers?
Finally, what exactly is the role of liquidity regulations? The Modigliani-
Miller propositions (Modigliani and Miller 1958; Miller and Modigliani
1961) serve as a starting point for our thinking on capital regulation,
from which departures have to be justified. There is no equivalent bench-
mark that describes whether the financial system is producing too much
or too little liquidity. Could the new regulations destroy value? Are two li-
quidity requirements necessary? How do the liquidity and capital regula-
tions interact? See Diamond and Kashyap (2016) for discussion of some
of these issues.
Before the recent financial crisis, there was a broad sense that the fi-
nancial system was well understood and relatively stable, especially in de-
veloped countries. The financial crisis has refocused attention on many
of the issues that financial economists were pondering earlier. Despite
the intense research and policy advances made recently, many open ques-
tions remain. Given the importance of the financial system in both bene-
fiting and harming the overall economy, these are areas where more work
will be very valuable.
References
Allen, Franklin, and Douglas Gale. 1994. “Limited Market Participation and Vol-
atility of Asset Prices.” A.E.R. 84 (4): 933–55.
———. 1997. “Intermediaries and Intertemporal Smoothing.” J.P.E. 105 (3):
523–46.
———. 2000. “Financial Contagion.” J.P.E. 108 (1): 1–33.
Bolton, Patrick, and Xavier Freixas. 2000. “Equity, Bonds, and Bank Debt: Cap-
ital Structure and Financial Market Equilibrium under Asymmetric Informa-
tion.” J.P.E. 108 (2): 324–51.
Bryant, John. 1980. “A Model of Reserves, Bank Runs and Deposit Insurance.”
J. Banking and Finance 4 (4): 335–44.
Calomiris, Charles W., and Charles M. Kahn. 1991. “The Role of Demandable
Debt in Structuring Optimal Banking Arrangements.” A.E.R. 81 (3): 497–513.
Carlson, Mark, Burcu Duygan-Bump, Fabio Natalucci, et al. 2016. “The Demand
for Short-Term, Safe Assets and Financial Stability: Some Evidence and Impli-
cations for Central Bank Policies.” Internat. J. Central Banking 12 (4): 307–33.

Carlson, Mark, Kris James Mitchener, and Gary Richardson. 2011. “Arresting
Banking Panics: Federal Reserve Liquidity Provision and the Forgotten Panic
of 1929.” J.P.E. 119 (5): 889–924.
Dang, Tri Vi, Gary Gorton, Bengt Holmström, and Guillermo Ordoñez. 2017.
“Banks as Secret Keepers.” A.E.R. 107 (4): 1005–29.
Diamond, Douglas W. 1984. “Financial Intermediation and Delegated Monitor-
ing.” Rev. Econ. Studies 51 (3): 393–414.
———. 1989. “Reputation Acquisition in Debt Markets.” J.P.E. 97 (4): 828–62.
———. 1991. “Monitoring and Reputation: The Choice between Bank Loans
and Directly Placed Debt.” J.P.E. 99 (4): 689–721.
Diamond, Douglas W., and Philip H. Dybvig. 1983. “Bank Runs, Deposit Insur-
ance, and Liquidity.” J.P.E. 91 (3): 401–19.
Diamond, Douglas W., and Anil K. Kashyap. 2016. “Liquidity Requirements, Li-
quidity Choice, and Financial Stability.” In Handbook of Macroeconomics, vol. 2,
edited by John B. Taylor and Harald Uhlig, 2263–2303. Amsterdam: North-
Holland.
Diamond, Douglas W., and Raghuram G. Rajan. 2000. “A Theory of Bank Capi-
tal.” J. Finance 55 (6): 2431–65.
———. 2001. “Liquidity Risk, Liquidity Creation, and Financial Fragility: A The-
ory of Banking.” J.P.E. 109 (2): 287–327.
———. 2005. “Liquidity Shortages and Banking Crisis.” J. Finance 60 (2): 615–47.
———. 2012. “Illiquid Banks, Financial Stability, and Interest Rate Policy.” J.P.E.
120 (3): 552–91.
Fama, Eugene F. 1980. “Agency Problems and the Theory of the Firm.” J.P.E 88
(2): 288–30.
———. 1985. “What’s Different about Banks?” J. Monetary Econ. 15 (1): 29–39.
Farhi, Emmanuel, and Jean Tirole. 2012. “Collective Morality Hazard, Maturity
Mismatch, and Systemic Bailouts.” A.E.R. 102 (1): 60–93.
Freixas, Xavier, and Jean-Charles Rochet. 2008. Microeconomics of Banking. Cam-
bridge, MA: MIT Press.
Friedman, Milton, and Anna J. Schwartz. 1963. A Monetary History of the United
States, 1867–1960. Princeton, NJ: Princeton Univ. Press.
Goldstein, Itay, and Ady Pauzner. 2005. “Demand-Deposit Contracts and the
Probability of Bank Runs.” J. Finance 60 (3): 1293–1327.
Gorton, Gary. 1996. “Reputation Formation in Early Bank Note Markets.” J.P.E.
104 (2): 346–97.
Gorton, Gary, and Andrew Winton. 2003. “Financial Intermediation.” In Hand-
book of the Economics of Finance, vol. 1A, edited by George M. Constantindes, Mil-
ton Harris, and René M. Stulz. Amsterdam: Elsevier.
Holmström, Bengt. 1982. “Managerial Incentive Problems—a Dynamic Perspec-
tive.” In Essays in Economics and Management in Honor of Lars Wahlbeck. Helsinki:
Swedish School Econ.
Holmström, Bengt, and Jean Tirole. 1998. “Private and Public Supply of Liquid-
ity.” J.P.E. 106 (1): 1–40.
Jacklin, Charles J., and Sudipto Bhattacharya. 1988. “Distinguishing Panics and
Information-Based Bank Runs: Welfare and Policy Implications.” J.P.E. 96 (3):
568–92.
Kashyap, Anil K., Raghuram G. Rajan, and Jeremy Stein. 2002. “Banks as Provid-
ers of Liquidity: An Explanation for the Co-existence of Lending and Deposit
Taking.” J. Finance 57 (1): 33–73.
Kreps, David M., and Robert Wilson. 1982. “Reputation and Imperfect Informa-
tion.” J. Econ. Theory 27 (2): 253–79.
Krishnamurthy, Arvind, and Annette Vissing-Jorgensen. 2012. “The Aggregate
Demand for Treasury Debt.” J.P.E. 120 (2): 233–67.

Milgrom, Paul, and John Roberts. 1982. “Predation, Reputation, and Entry De-
terrence.” J. Econ. Theory 27 (2): 280–312.
Miller, Merton H., and Franco Modigliani. 1961. “Dividend Policy, Growth, and
the Valuation of Shares.” J. Bus. 34 (October): 411–33.
Modigliani, Franco, and Merton H. Miller. 1958. “The Cost of Capital, Corpora-
tion Finance, and the Theory of Investment.” A.E.R. 48 ( June): 261–97.
Postlewaite, Andrew, and Xavier Vives. 1987. “Bank Runs as an Equilibrium Phe-
nomenon.” J.P.E. 95 (3): 485–91.
Rajan, Raghuram G. 1992. “Insiders and Outsiders: The Choice between In-
formed and Arm’s-Length Debt.” J. Finance 47 (4): 1367–1400.
Shleifer, Andrei, and Robert Vishny. 1992. “Liquidation Values and Debt Capac-
ity: A Market Equilibrium Approach.” J. Finance 47 (4): 1343–66.
Simons, Henry C. 1933. “Banking and Currency Reform (with Supplement and
Appendix).” Manuscript, Dept. Econ., Univ. Chicago.
———. 1936. “Rules versus Authorities in Monetary Policy.” J.P.E. 44 (1): 1–30.
Stein, Jeremy C. 2012. “Monetary Policy as Financial-Stability Regulation.” Q. J.E.
127 (1): 57–95.
Monetary Economics
Fernando Alvarez
University of Chicago and National Bureau of Economic Research
Introduction
In this paper I will review contributions to monetary economics, notwith-
standing Lucas’s essay on Chicago developments of monetary theory by
Friedman and Patinkin. I will discuss the most highly cited papers on mon-
etary economics published in the Journal of Political Economy, together with
related work not necessarily published in the JPE to place them in con-
text. I will organize the discussion of these contributions using Jevons’s
functions of money: store of value, medium of exchange, and unit of ac-
count.
Store of Value
I interpret the function of store of value as best captured by the mone-
tary equilibrium of the overlapping generations model. The seminal pa-
per on this topic is the JPE piece by Samuelson (1958), a ground-breaking
contribution. This model, either on the version worked out by Samuel-
son or in subsequent ones, is useful to address several important issues,

such as the desirability of a pay-as-you-go social security or the role of de-
mographic factors and of public debt in capital accumulation; see, for
instance, Blanchard (1985) for an example of such analysis. Besides those
issues, Samuelson’s overlapping generations model is perhaps the clear-
est framework to conceptualize fiat money as a store of value. In the mon-
etary equilibrium, money is an asset backed only by its use as a store of
value—a form of rational bubble. There are multiple uses of the over-
lapping generations models in monetary economics. One such use is the
path-breaking paper “Expectations and the Neutrality of Money” by Lucas
(1972). The popularity of the overlapping generations model comes and
goes as different phenomena fit the role of money as a store of value, such
as the current research on the effects of and causes for the increase in the
demand for safe and/or liquid assets.
Medium of Exchange
There is a long tradition of work on monetary economics to understand
the role of money as a medium of exchange. Earlier seminal contributions
are Baumol (1952) and Tobin (1956) on inventory theoretical models of
cash holdings. During the postwar period there were theoretical and em-
pirical advances on the study of money demand and quantitative theory
led by Friedman and students of his; see Lucas’s paper and the references
therein for details. Later on there were related dynamic versions such as,
for example, those in the work by Sidrauski (1967a, 1967b) with money
in the utility function. Even later, that is, post–rational expectations rev-
olutions, there are new analysis and conceptualizations, such as the cash-
credit model of Lucas and Stokey (1987). All these models give closely
related frameworks to properly define money demand and quantify its ef-
fects. A recent important development is the one in Kiyotaki and Wright
(1989) and follow-up work by Lagos and Wright (2005), both published
in the JPE. These models provide microfoundations for fiat or commod-
ity money as a medium of exchange, based on search and information fric-
tions. Their later versions build a bridge with traditional money demand
theory, as well as provide insights of transactions of other assets beyond
fiat or commodity money.
Standard of Value or Unit of Account

I will consider two related sets of ideas on why monetary policy has an
effect, which I regard as, roughly speaking, related to the standard of value
of unit of account function of money. One is the models that use nominal
rigidities, such as sticky prices, to understand the effect of monetary shocks.
The other set of ideas is the models that use rational confusion between
nominal and real shocks as vehicles for monetary policy to affect output.

There is a long tradition, certainly emphasized by Keynes, on the macro-
economic effects of nominal rigidities and its relationship with the effects
of monetary and fiscal policy. Some form of nominal nonneutrality is also
central to the view of the role of monetary policy portrayed in Friedman’s
(1968) presidential address, which very much inspires several features of
contemporary modeling, as well as the contemporary understanding of
how a monetary policy shock plays it out in the short run (5–7). One pos-
sible reason for the nonneutrality is the effect of monetary policy on expec-
tations, as described in Friedman’s (1977) Nobel lecture, also published in
the JPE. In that lecture, Friedman discusses unexpected nominal changes
as candidate explanations of the traditional Phillips curve type of relation-
ships: “only surprises matter.” Another prominent candidate for such
nonneutrality is nominal wage or nominal price rigidities. I discuss these
candidates separately.
I first turn to contributions published in the JPE on the effect of mon-
etary policy due to rational confusion between real and nominal shocks.
The work by Muth (1961), Lucas and Prescott (1971), and Lucas (1972)
was the cornerstone for the rational expectations revolution. In addition
to the important methodological questions addressed in Lucas (1972),
this paper tackles the classical macroeconomic question of the effect of
monetary disturbances on aggregate prices and output. In his Nobel lec-
ture, Lucas (1996) gives a historical account of the question, describes the
difficulties in assessing it purely on empirical grounds, and sketches the
main ideas behind the nonneutrality of money based on dispersion of in-
formation. One of the features emphasized by the class of models in his
lecture is the differential short-term impact on prices and output of ex-
pected versus unexpected monetary shocks. This differential impact pro-
duces a type of Phillips curve relationship, due to the agents’ rational con-
fusion between nominal and real shocks. These ideas have had a deep
impact on how macroeconomists think about the classical question of
the effect of monetary shock. For instance, there was an early and influ-
ential theoretical work published in the JPE exploring the circumstances
under which systematic monetary policy does have on effect on output.
Sargent and Wallace (1975) is one of the main early rational expectations
contributions emphasizing the generality of the result that the systematic
component of monetary policy does not affect output, as well as the clas-
sical Wicksell indeterminacy of the price level under interest rates rules.
On the one hand, Fischer (1977) is an early example in which, in spite of
rational expectations and as a result of nominal wages set in advance for pe-
riods longer than those for which monetary policy can react, systematic
effects of monetary policy do indeed affect output. On a related matter,
Taylor (1980) studies the effect of setting wages in a staggered fashion on
the propagation of shocks. He argues that in equilibrium this friction im-
plies adjustment to shocks that are even more persistent than the length

1
of their assumed lag. Complementary to these theoretical papers, Barro
(1978) estimates the contribution of the expected and unexpected com-
ponent of monetary changes on output and prices for the United States,
finding evidence consistent with Lucas’s hypothesis. Indeed, the distinc-
tion between expected and unexpected shocks plays an important role
in the method presented in a seminal paper by Sims (1980). This paper
introduced vector autoregressions (VARs), which have become one of the
main conceptual frameworks to measure the importance of the contribu-
tions of different macroeconomic shocks.2
Now I turn to models in which monetary policy nonneutrality is due to
the presence of nominal rigidities on the price (and wage) setting, as op-
posed to informational ones. I also regard these features as examples of
the role of “standard of value of money” or unit of account. The JPE pub-
lished the seminal paper by Rotemberg (1982), constructing a quantita-
tive rational expectations sticky price model. Prices are sticky in the sense
that competitive monopolistic firms face quadratic adjustment costs of
changing them, giving rise to a type of forward-looking Phillips curve kind
of relationship. The author finds a better fit when these costs are positive,
that is, when prices are sticky. An alternative mechanism, whose reduced
form is almost identical, was that proposed by Calvo (1983) and further
developed by Yun (1996). The Calvo price setting results from the aggre-
gation of firms that can change their prices only at idiosyncratic random
times. The models in Rotemberg and Woodford (1997) and in Woodford
(2001) use Calvo pricing to obtain a Phillips curve type relationship and
also clarify and simplify the rest of the environment, producing the core
of the neo-Keynesian model.
As mentioned in the conclusion of Rotemberg (1982), estimating and
testing these types of models using microdata is desirable, since they may
offer more direct evidence of the key pricing friction. The Calvo pricing
assumption goes in this direction, since one can map the exogenous prob-
ability of keeping prices at a fixed value as related to actual durations of
unchanged prices. A key input to this task is to have high-quality measure-
ments of the frequency of price changes. In this regard Bils and Klenow
(2004), also published in the JPE, provided a key input for this measure-
ment for the US economy. While this is not the first paper to present mi-
cro evidence of price rigidities, using the microdata underlying the con-
struction of the consumer price index provided the level of aggregation
proper for macroeconomic analysis and thus initiated a large empirical
literature that uses a variety of data sets and approaches. Surprisingly, Bils
and Klenow found a relatively high frequency of price changes, although
such a conclusion may depend on a more subtle definition of price changes.
1
See Chari, Kehoe, and McGrattan (2000), which revisited the conditions required to
obtain substantial propagation in a modern general equilibrium setup.
2
Regrettably, the JPE missed out on this contribution.

The 2000s has seen an explosion of work on the effect of monetary pol-
icy based on the combination of assuming rational expectations, model-
ing sticky prices and/or wages using Calvo price setting, and using the ba-
sic structure of the neo-Keynesian model as developed in Woodford (2001).
One, if not the, leading quantitative contribution on this area is Chris-
tiano, Eichenbaum, and Evans (2005), published in the JPE. This paper
merges the structure of the real business cycle model with the core neo-
Keynesian model, incorporating additional real frictions that enhanced
the effects of nominal shocks. Additionally, the authors use VAR evidence
from identified monetary shocks to estimate key parameters of the model,
which in turn is used for quantitative evaluation of different policies.3
While Calvo pricing does provide a connection with microdata, these
data strongly suggest the presence of idiosyncratic shocks. Moreover, the
exogenous nature of the timing of price changes in Calvo pricing, while
very tractable, is clearly a shortcut that may have important consequences.
Indeed, Golosov and Lucas (2007) use a quantitative menu cost model, a
model in which firms are subject to idiosyncratic costs and in which they
must pay a fixed menu cost to change nominal product prices. The cali-
brated version of their model shows that while money is not neutral, it is
almost so. Their contribution, together with the availability of rich micro-
data, has also started a literature that aims at using such data to better mea-
sure the impact of the frictions on price setting and to reconcile the small
size of monetary shocks on output with the larger values typically estimated
using identified VARs.
References
Barro, Robert J. 1978. “Unanticipated Money, Output, and the Price Level in the
United States.” J.P.E. 86 (4): 549–80.
Baumol, William J. 1952. “The Transactions Demand for Cash: An Inventory
Theoretic Model.” Q. J.E. 66 (4): 545–56.
Bils, Mark, and Peter J. Klenow. 2004. “Some Evidence on the Importance of
Sticky Prices.” J.P.E. 112 (5): 947–85.
Blanchard, Olivier J. 1985. “Debt, Deficits, and Finite Horizons.” J.P.E. 93 (2):
223–47.
Calvo, Guillermo A. 1983. “Staggered Prices in a Utility-Maximizing Framework.”
J. Monetary Econ. 12 (3): 383–98.
Chari, V. V., Patrick J. Kehoe, and Ellen R. McGrattan. 2000. “Sticky Price Mod-
els of the Business Cycle: Can the Contract Multiplier Solve the Persistence Prob-
lem?” Econometrica 68 (5): 1151–79.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. “Nom-
inal Rigidities and the Dynamic Effects of a Shock to Monetary Policy.” J.P.E.
113 (1): 1–45.
3
Smets and Wouters (2007) is an important complementary, and largely confirmatory,
study of a very closely related model using a different econometric methodology.

Fischer, Stanley. 1977. “Long-Term Contracts, Rational Expectations, and the Opti-
mal Money Supply Rule.” J.P.E. 85 (1): 191–205.
Friedman, Milton. 1968. “The Role of Monetary Policy.” A.E.R. 58 (1): 1–17.
———. 1977. “Nobel Lecture: Inflation and Unemployment.” J.P.E. 85 (3): 451–
72.
Golosov, Mikhail, and Robert E. Lucas Jr. 2007. “Menu Costs and Phillips Curves.”
J.P.E. 115:171–99.
Kiyotaki, Nobuhiro, and Randall Wright. 1989. “On Money as a Medium of Ex-
change.” J.P.E. 97 (4): 927–54.
Lagos, Ricardo, and Randall Wright. 2005. “A Unified Framework for Monetary
Theory and Policy Analysis.” J.P.E. 113 (3): 463–84.
Lucas, Robert E., Jr. 1972. “Expectations and the Neutrality of Money.” J. Econ.
Theory 4 (2): 103–24.
———. 1996. “Nobel Lecture: Monetary Neutrality.” J.P.E. 104 (4): 661–82.
Lucas, Robert E., Jr., and Edward C. Prescott. 1971. “Investment under Uncer-
tainty.” Econometrica 39 (5): 659–81.
Lucas, Robert E., Jr., and Nancy L. Stokey. 1987. “Money and Interest in a Cash-
in-Advance Economy.” Econometrica 55 (3): 491–513.
Muth, John F. 1961. “Rational Expectations and the Theory of Price Movements.”
Rotemberg, Julio J. 1982. “Sticky Prices in the United States.” J.P.E. 90 (6): 1187–
1211.
Rotemberg, Julio J., and Michael Woodford. 1997. “An Optimization-Based Econo-
metric Framework for the Evaluation of Monetary Policy.” In NBER Macroeco-
nomics Annual, vol. 12, edited by Ben Bernanke and Julio J. Rotemberg, 297–
346. Cambridge, MA: MIT Press.
Samuelson, Paul A. 1958. “An Exact Consumption-Loan Model of Interest with
or without the Social Contrivance of Money.” J.P.E. 66 (6): 467–82.
Sargent, Thomas J., and Neil Wallace. 1975. “ ‘Rational’ Expectations, the Opti-
mal Monetary Instrument, and the Optimal Money Supply Rule.” J.P.E. 83 (2):
241–54.
Sidrauski, Miguel. 1967a. “Inflation and Economic Growth.” J.P.E. 75 (6): 796–
810.
———. 1967b. “Rational Choice and Patterns of Growth in a Monetary Economy.”
A.E.R. 57 (2): 534–44.
Sims, Christopher A. 1980. “Macroeconomics and Reality.” Econometrica 48 (1):
1–48.
Smets, Frank, and Rafael Wouters. 2007. “Shocks and Frictions in US Business
Cycles: A Bayesian DSGE Approach.” A.E.R. 97 (3): 586–606.
Taylor, John B. 1980. “Aggregate Dynamics and Staggered Contracts.” J.P.E. 88
(1): 1–23.
Tobin, James. 1956. “The Interest Elasticity of Transactions Demand for Money.”
Rev. Econ. and Statis. 38 (3): 241–47.
Woodford, Michael. 2001. “The Taylor Rule and Optimal Monetary Policy.” A.E.R.
91 (2): 232–37.
Yun, Tack. 1996. “Nominal Price Rigidity, Money Supply Endogeneity, and Busi-
ness Cycles.” J. Monetary Econ. 37:345–70.

Memories of Friedman and Patinkin

Robert E. Lucas Jr.
During the period between the two world wars the University of Chicago
produced an extraordinary group of monetary economists. For these
notes, I will concentrate on two of them: Milton Friedman and Don Pa-
tinkin. I knew them both well, and both were important to my own eco-
nomic growth. Both of them are remembered more for their books than
for their journal articles, but the Journal of Political Economy published
them both, including the interesting exchanges that I will discuss here.
Both Friedman and Patinkin did graduate work at Chicago, Friedman
in the 1930s and Patinkin in the 1940s. Patinkin got his Chicago PhD in
1947, working under Oscar Lange. Friedman got his degree from Co-
lumbia in 1940, supervised by Simon Kuznets. The two did not overlap
at Chicago, but both of them recalled classes on monetary economics
with Henry Simons and Lloyd Mints and, less directly, Jacob Viner and
Frank Knight.
In 1956 Friedman published Studies in the Quantity Theory of Money—a
Restatement, a book consisting of a long introduction by Friedman himself,
followed by the dissertations of four of his students: Phillip Cagan, John
Klein, Eugene Lerner, and Richard Selden. These four dissertations—the
first fruit of the Chicago Money and Banking Workshop—are stunning ex-
amples of economics at its best. I will come back to them, but first I want to
review Friedman’s introduction, which was focused almost entirely on clar-
ifying and reviving a version of the quantity theory of money.
Friedman began with the concern that it “is clear that the general ap-
proach (the quantity theory) fell into disrepute after the crash of 1929
and the subsequent Great Depression and only recently has been slowly
re-emerging into professional respectability” (3). One source of this dis-
repute was “the proponents of the new income-expenditure approach” who
described versions of the quantity theory that were “an atrophied and rigid
charicature.” Friedman argued that Chicago economists—mainly Simons,
Mints, and Knight—had formulated a more sophisticated and useful ver-
sion.1
This was his first attack on the economics of Keynes. Later attacks came
in his 1970 JPE paper “A Theoretical Framework for Monetary Analysis”
1
This Chicago version of the quantity theory has been discussed in much more detail by
many authors. See in particular Laidler (2010), Nelson (2017), and Tavlas (2017).

and again in a 1971 extension, also in the JPE, “A Monetary Theory of
Nominal Income.”
At this point Don Patinkin, who had followed all three of these state-
ments of Friedman’s, lost patience. In 1972 the JPE published Patinkin’s
“Friedman on the Quantity Theory and Keynesian Economics.” Here is
his abstract:
The article is based on textual evidence from the quantity-

theory and Keynesian literature. It shows, first, that the concep-
tual framework of a portfolio demand for money that Friedman
denotes as the “quantity theory” is actually that of Keynesian
economics. Conversely, Friedman detracts from the true quan-
tity theory by stating that its formal short-run analysis assumes
real output constant, while only prices change. Friedman also
incorrectly characterizes Keynesian economics in terms of abso-
lute price rigidity. He does this by overlooking the systematic
analysis by Keynes and the Keynesians of the role of downward
wage flexibility during unemployment, and of the “inflationary
gap” during full employment. Otherwise Friedman’s interpreta-
tion of Keynes is the standard textbook one of an economy in a
“liquidity-trap” unemployment equilibrium. The author restates
his alternative interpretation of Keynesian economics in terms of
unemployment disequilibrium. (1972, 883)
Patinkin went on to develop these assertions in detail. Later, in the text,

he added that “it is obviously no criticism of Friedman—nor does it der-
ogate from his stature as a monetary economist—to say that his analytical
framework is Keynesian. All that is being criticized is Friedman’s persis-
tent refusal to recognize this is so” (886).
In fact this 1972 JPE issue contained, in addition to Patinkin, reactions
to Friedman’s 1971 paper from Karl Brunner and Allan Meltzer, James
Tobin, and Paul Davidson. These were followed by an 85-page counter-
attack from Friedman. (The editors at the time were Robert Gordon and
Harry Johnson.)
Friedman’s 1970 and 1971 papers did not mention Simons, Mints, or
Knight, nor did they refer explicitly to a Chicago tradition. This time Keynes
was discussed at some length. But Friedman continued to refer to “the
Keynesian challenge to the quantity theory” and to cast the two as incompat-
ible contestants. A common reaction from Patinkin, Brunner and Meltzer,
and Tobin was to argue that Keynesian ideas and the quantity theory can
and should be parts of a single model. Reading it now one expects some
kind of unification, and there are times when this seems about to happen.
But Friedman would not have it. The long 1971 debate began with confu-
sion and ended there.

It is surprising to me that Friedman did not connect “the proponents of
the new income-expenditure approach” to the national income account
time series that Simon Kuznets had created in the late 1920s (and that
the US government has maintained ever since). These data opened up
an exciting new world for economists who wanted measurements on the
state of the economy as a whole. Kuznets’s data included “real” series only:
national money supply data were still in the future. The model-building pi-
oneers of quantitative macroeconomics—Jan Tinbergen, Lawrence Klein,
and soon many others—worked with Kuznets’s data because they were the
only good data they had. I think this was all there was to the “new income-
expenditure approach.”
If so, then what was needed to restore the quantity theory was to con-
struct time series on the money supply at the same level of accuracy as
Kuznets’s data on real series. And this is exactly what Friedman’s students
did in the substantive chapters of Studies in the Quantity Theory of Money.
All four dissertations constructed aggregate, economywide time series on
some well-defined measure of a money supply and measures of nominal
prices. Cagan’s study of postwar hyperinflations provides monthly time
series—suitable for his high-frequency data—carefully documented, for
prices, measures of cash balances, and real per capita incomes for seven
countries. There is an explicit theoretical model—set out and calibrated.
Comparisons of theory and time-series data are shown graphically and as-
sessed econometrically.
Cagan’s dissertation was the crown jewel of Studies in the Quantity Theory
of Money, a breathtaking achievement that is still drawn on. The other
three students also produced new monetary time series, shedding light
on interesting situations. John Klein analyzed Germany from 1932 to
1944. Eugene Lerner studied the Southern Confederacy of 1861–65
(see also his 1954 JPE paper). Richard Selden’s “Monetary Velocity in
the United States” covered 1839–1951. In his introduction, Friedman also
provided a valuable discussion of the way these very different situations
all served as natural experiments. What Friedman and his students had
achieved in part, then, was to do for aggregate monetary theory what
Kuznets had done for the aggregate real income and product accounts.
They created a synthesized “money” consisting of many component assets
that can be measured and add up to a whole, just as Kuznets had done with
“consumption” and other real aggregates.
The typology of M0, M1, M2, and so forth was not available in 1956. The
concept M1 was initiated by Homer Jones at the St. Louis Fed in 1960.
Many people were involved in developing it further. Friedman, along with
Allan Meltzer, Karl Brunner, and others, were supporters and users.2 Pa-
2
I thank William Barnett and Stephen Williamson for information on the role of the
St. Louis Fed. It is a fascinating story just touched on here.

tinkin was right, I think, to insist that monetary theory can fit quite nicely
with Keynesian ideas. Friedman and his students showed how to do it with
actual time-series evidence.
References
Brunner, Karl, and Allan H. Meltzer. 1972. “Friedman’s Monetary Theory.” J.P.E.
80 (September/October): 837– 51.
Davidson, Paul. 1972. “A Keynesian View of Friedman’s Theoretical Framework
for Monetary Analysis.” J.P.E. 80 (September/October): 864 –82.
Friedman, Milton, ed. 1956. Studies in the Quantity Theory of Money—a Restatement.
Chicago: Univ. Chicago Press.
———. 1970. “A Theoretical Framework for Monetary Analysis.” J.P.E. 78 (March/
April): 193 –238.
———. 1971. “A Monetary Theory of Nominal Income.” J.P.E. 79 (March/April):
323–37.
———. 1972. “Comments on the Critics.” J.P.E. 80 (September/October): 906–50.
Laidler, David. 2010. “Chicago Monetary Traditions.” In The Elgar Companion to
the Chicago School of Economics, edited by Ross B. Emmett. Northampton, MA:
Elgar.
Lerner, Eugene. 1954. “The Monetary and Fiscal Programs of the Confederate
Government.” J.P.E. 62 (December): 506–22
Nelson, Edward. 2017. “Milton Friedman and Economic Debate in the United
States.” Manuscript, Board Governors, Fed. Reserve System.
Patinkin, Don. 1972. “Friedman on the Quantity Theory and Keynesian Econom-
ics.” J.P.E. 80 (September/October): 883 – 905.
Tavlas, George S. 2017. “The Group: The Making of the Chicago Monetary Tra-
dition, 1927–36.” Manuscript, Bank of Greece.
Tobin, James. 1972. “Friedman’s Theoretical Framework.” J.P.E. 80 (September/
October): 852– 63.
Labor Markets
Robert Shimer
The Journal of Political Economy has been a crucial outlet for pathbreaking
articles on the determination of employment, wages, and inflation. The
JPE published what was arguably the key paper forecasting the demise of

a stable Phillips curve relationship, that is, the end of the trade-off be-
tween employment and inflation. It also published the first theory of in-
tertemporal substitution in labor supply and many subsequent papers
that sought to quantify the elasticity of labor supply. It has been an early
and important source of papers that examined the impact of information
frictions on labor market outcomes. And the JPE has published key pa-
pers that improved our understanding of the sources of fluctuations in
employment over the business cycle. The following paragraphs expand
on those points.
Phelps (1968) sought to understand whether the “Phillips trade-off
[was] real, serious, and not misleading” (681). Phelps recognized that
low unemployment may induce firms to raise wages in an effort to attract
and retain workers. In the most innovative part of his paper, Phelps de-
veloped a notion of a “macroequilibrium” in which actual and expected
wage growth are equal to each other. This was a key component of the ci-
tation for Phelps’s Nobel Prize in 2006. Although this paper does not
have the mathematical sophistication of later studies using rational ex-
pectations, Phelps’s argument predicted that maintaining an unemploy-
ment rate below the steady-state equilibrium value would necessitate spi-
raling increases in inflation. This notion of an expectations-augmented
Phillips curve was of course also emphasized in Friedman’s presidential
address to the American Economic Association as well as in his Nobel lec-
ture (Friedman 1968, 1977).
Given the JPE’s early recognition of the limitations of conventional
Keynesian analysis of the Phillips curve and the economists at the Univer-
sity of Chicago’s tight connection to the rational expectations revolution,
it is equally interesting to note the papers on the employment-inflation
trade-off that the JPE did not publish. The most prominent and impor-
tant of these is Lucas (1972). This omission appears to reflect the reluc-
tance of the journal at the time to engage in the types of mathematical ar-
guments that rational expectations models with strong microeconomic
foundations embraced. Indeed, papers with rational expectations that the
JPE did publish at the time, such as Sargent and Wallace (1975), used ad
hoc models rather than the optimizing framework embraced by Lucas.
Sophisticated models of labor market equilibrium do appear in many
more recent papers published in the journal. And central to any notion
of labor market equilibrium is the interaction between labor supply and
labor demand. While there was historically relatively little controversy
about the role of the marginal product of labor (or the marginal revenue
product of labor) in determining labor demand, the history of models of
labor supply is more interesting. In the 1960s, models of the long run,
like Solow (1956), treated labor supply as fixed, while the Keynesian mod-
els that were used for understanding the short run treated labor supply

as perfectly elastic at some rigid real or nominal wage. In a seminal paper,
Lucas and Rapping (1969) argued that a standard model of the house-
hold provides a strong microeconomic foundation for a dynamic theory
of the elasticity of labor supply. The paper extends a utility-maximizing
theory of the labor-leisure choice to a multiperiod framework and shows
how current and future real wages, as well as interest rates and wealth, af-
fect labor supply. This trade-off is central to all modern models of the la-
bor market.
Subsequent to the largely theoretical contribution of Lucas and Rap-
ping (1969), the JPE has been at the forefront of an empirical literature
measuring how elastic labor supply is. In an important early contribution,
MaCurdy (1981) developed an empirical framework for using variation
in wages and hours worked over the life cycle to measure the Frisch (con-
stant marginal utility of wealth) elasticity of labor supply. Using data on
prime-aged men in the United States, he found that the Frisch elasticity
was positive but small, in the range of 0.10–0.45. This paper spawned
many follow-ups, some using similar sources of wage variation, others us-
ing different sources, such as changes in income tax rates. A prominent
example is Altonji (1986). This paper deals with issues related to missing
and mismeasured wage observations, as well as the components of future
wage growth that are known to the worker but not to the econometrician.
The careful analysis modestly lowers the range of plausible labor supply
elasticities to 0–0.35.
Intertemporal elasticities of labor supply in this range present a chal-
lenge for business cycle models in which flexible wages equate labor sup-
ply and labor demand. These models rely on the assumption that the ob-
served modest variation in wages induces substantial cyclical shifts in
employment because labor supply is elastic (e.g., Kydland and Prescott
1982).
A possible resolution of this is that shifts between market and home pro-
duction effectively raise the relevant elasticity. The modern theory of time
allocation and home production is also closely linked to the JPE and econ-
omists at the University of Chicago. Becker (1965) proposed that time and
market goods are inputs into the production of commodities, which in
turn enter the welfare function. Becker’s theory was both general and
abstract, and so its implications for labor supply were initially unclear.
Gronau (1977) parameterized this framework more tightly, in particular
distinguishing between home production and leisure: “work at home (like
work in the market) is something one would rather have somebody else
do for one (if the cost were low enough), while it would be almost impos-
sible to enjoy leisure through a surrogate” (1104). In turn, Benhabib, Rog-
erson, and Wright (1991) argued that introducing home production into
the Kydland and Prescott (1982) real business cycle model significantly

improved its ability to match the business cycle facts, in large part by effec-
tively raising the intertemporal elasticity of labor supply.1
In a paper that was instrumental to his 1982 Nobel Prize, Stigler (1961)
pioneered the study of information frictions in markets. In Stigler (1962),
he applied these ideas to the labor market, starting a research agenda that
now offers a complementary explanation for the behavior of employment
and the determination of wages. The 1962 paper notes the existence of
pure wage dispersion, meaning that different firms offer the same worker
different wages. He showed how dispersion in wage offers gives workers
an incentive to get multiple offers so as to obtain higher wages on aver-
age. He also examines how the optimal number of offers depends on the
amount of wage dispersion, the cost of search, and the number of years
a job is expected to last. He observed that employers search for workers
at the same time as workers search for employers. Finally, he argues that
large firms pay higher wages because they face a higher cost of monitor-
ing and screening workers. This employer size–wage effect was only care-
fully documented and explored nearly three decades later (Brown and
Medoff 1989).
While Stigler (1962) discussed search both by workers and by firms, he
was not able to articulate a framework in which both happened simulta-
neously. Doing so was the major achievement of Diamond (1982), a pa-
per that features prominently in his 2010 Nobel Prize citation. Although
the model in that paper was stylized and not explicitly cast in terms of the
labor market, it has frequently been reinterpreted in those terms. The
paper argues that thick market effects are a natural implication of search
frictions. There is no point in searching for a job if firms are not looking
to hire workers, and vice versa. The paper also shows that thick market
effects can generate multiple Pareto-rankable steady-state equilibria. In
one equilibrium, search intensity is low because everyone assumes that
it will be difficult to find a trading partner and so engages in minimal
search activity. In another, search intensity is high for the opposite rea-
son. By construction, wages and prices are fixed in Diamond’s model, and
so the model can generate large swings in productive employment with-
out needing any movement in wages and without relying on elastic labor
supply.
Search is not the only important friction in the labor market. Oi (1962)
stressed that training costs are substantial, making it costly to hire and fire
a worker. In his language, this turns labor into a quasi-fixed factor.
1
Other important JPE papers have followed up on other implications of Becker’s and
Gronau’s work. For example, Aguiar and Hurst (2005) showed that time allocation and
home production, rather than myopia, are important for understanding the drop in ex-
penditures on consumption goods at retirement.

A number of papers examine how these two hiring frictions, search and
training costs, affect the determination of employment and wages. Jovan-
ovic (1979) studied an environment in which the quality of a job match
is an experience good. Workers and firms jointly learn about the quality
of the match from the evolution of the cumulative amount of output pro-
duced. Because of hiring frictions, a little bit of bad news will not be
enough to cause the pair to separate. Nevertheless, with enough bad news,
the worker prefers to quit and look for another job. Stochastic match qual-
ity and endogenous separations help break the link between individual
employment histories and wages. This model of symmetric learning has
proved very useful for understanding how turnover varies with job tenure.
Azariadis (1975) and Lazear (1979) examine implicit and explicit long-
term contracts between workers and firms. The starting point for both pa-
pers is the puzzling question of why firms lay off workers rather than cut
wages. For Azariadis, these layoffs happen when the firm contracts. For
Lazear, they occur when a worker retires. Azariadis’s argument is some-
what simpler, since it does not rely on the incentive issues emphasized
by Lazear. Azariadis explained that long-term contracting considerations
can induce firms to lay off workers while maintaining the wage of the
workers they keep. One way to think about this is in terms of optimal con-
tracting with (possibly limited) commitment. To the extent that workers
are less able to insure themselves against idiosyncratic fluctuations in in-
come, firms should compete for workers by promising to smooth wages
as long as the employment relationship lasts. This promise has some chance
of being credible if it is costly for the firm to fire one worker and rehire
an identical one, that is, if there is a hiring friction. But if business con-
ditions become too adverse, the optimal contract may necessitate firing
the worker. Thus Azariadis’s model again predicts smooth wages and large
employment fluctuations, reflecting the fact that the decision to work at
a particular firm is not determined in a spot market with reference only
to the current wage.
A related issue is the structure of pay within organizations. Personnel
managers believe that pay equality is important within an organization, at
least for workers charged with doing similar tasks. A common view is that
this reflects morale concerns within the firm. Lazear (1989) argued that
this explanation is at best incomplete, since highly productive workers
should feel discouraged by a compressed wage scale, and indeed their dis-
couragement may be particularly costly to the firm. He instead empha-
sized the potential destructive role of competition between coworkers.
When one coworker can sabotage another’s output, tournaments may be
counterproductive. This paper offers a distinct explanation for why wages
are smooth within employment relationships.
Finally, Lilien (1982) proposed that sectoral shifts are important for
understanding employment fluctuations. The basic idea is that a market

economy always moves workers from declining sectors into expanding
ones. Owing to the presence of search frictions, this takes time. Indeed,
even though the mean duration of job search may be on the order of 2 or
3 months in the US economy, when a worker moves from a declining sec-
tor to an expanding one, he may cycle through several jobs before find-
ing a stable one, for the reasons that Jovanovic (1979) emphasized. This
does not imply that worker reallocation is socially undesirable. But it does
imply that if the amount of reallocation is unusually high at some point in
time, employment and output will be low; that is, there will be a recession.
Thus search frictions and sectoral shocks naturally give rise to aggregate
fluctuations.
Lilien’s emphasis on sectoral shifts rather than aggregate fluctuations
has proved controversial. Arguably his most lasting insight may lie in how
we think about and model fluctuations in employment. Echoing Phelps
(1968) but armed with more modern terminology and modeling tools,
Lilien wrote, “Given a definition of the natural rate [of unemployment]
based on microeconomic foundations, much of cyclical unemployment is
better described as fluctuations of the natural rate itself” (1982, 778). This
is indeed the current view of equilibrium unemployment.
References
Aguiar, Mark, and Erik Hurst. 2005. “Consumption versus Expenditure.” J.P.E.
113 (5): 919–48.
Altonji, Joseph G. 1986. “Intertemporal Substitution in Labor Supply: Evidence
from Micro Data.” J.P.E. 94, no. 3, pt. 2 ( June): S176–S215.
Azariadis, Costas. 1975. “Implicit Contracts and Underemployment Equilibria.”
J.P.E. 83 (6): 1183–1202.
Becker, Gary S. 1965. “A Theory of the Allocation of Time.” Econ. J. 75 (299):
493–517.
Benhabib, Jess, Richard Rogerson, and Randall Wright. 1991. “Homework in Mac-
roeconomics: Household and Aggregate Fluctuations.” J.P.E. 99 (6): 1166–87.
Brown, Charles, and James Medoff. 1989. “The Employer Size–Wage Effect.”
J.P.E. 97 (5): 1027–59.
Diamond, Peter A. 1982. “Aggregate Demand Management in Search Equilib-
rium.” J.P.E. 90 (5): 881–94.
Friedman, Milton. 1968. “The Role of Monetary Policy.” A.E.R. 58 (1): 1–17.
———. 1977. “Nobel Lecture: Inflation and Unemployment.” J.P.E. 85 (3): 451–
72.
Gronau, Reuben. 1977. “Leisure, Home Production, and Work—the Theory of
the Allocation of Time Revisited.” J.P.E. 85 (6): 1099–1123.
Jovanovic, Boyan. 1979. “Job Matching and the Theory of Turnover.” J.P.E. 87,
no. 5, pt. 1 (October): 972–90.
Kydland, Finn E., and Edward C. Prescott. 1982. “Time to Build and Aggregate
Fluctuations.” Econometrica 50 (6): 1345–70.
Lazear, Edward P. 1979. “Why Is There Mandatory Retirement?” J.P.E. 87 (6):
1261–84.
———. 1989. “Pay Equality and Industrial Politics.” J.P.E. 97 (3): 561–80.

Lilien, David M. 1982. “Sectoral Shifts and Cyclical Unemployment.” J.P.E. 90
(4): 777–93.
Lucas, Robert E., Jr. 1972. “Expectations and the Neutrality of Money.” J. Econ.
Theory 4 (2): 103–24.
Lucas, Robert E., Jr., and Leonard A. Rapping. 1969. “Real Wages, Employment,
and Inflation.” J.P.E. 77 (5): 721–54.
MaCurdy, Thomas E. 1981. “An Empirical Model of Labor Supply in a Life-Cycle
Setting.” J.P.E. 89 (6): 1059–85.
Oi, Walter Y. 1962. “Labor as a Quasi-Fixed Factor.” J.P.E. 70 (6): 538–55.
Phelps, Edmund S. 1968. “Money-Wage Dynamics and Labor-Market Equilib-
rium.” J.P.E. 76, no. 4, pt. 2 (August): 678–711.
Sargent, Thomas J., and Neil Wallace. 1975. “ ‘Rational’ Expectations, the Opti-
mal Monetary Instrument, and the Optimal Money Supply Rule.” J.P.E. 83 (2):
241–54.
Solow, Robert M. 1956. “A Contribution to the Theory of Economic Growth.”
Q. J.E. 70 (1): 65–94.
Stigler, George J. 1961. “The Economics of Information.” J.P.E. 69 (3): 213–25.
———. 1962. “Information in the Labor Market.” J.P.E. 70, no. 5, pt. 2 (Octo-
ber): 94–105.
Chicago Labor Economics

James J. Heckman
Starting in the 1950s, economists influenced by the Chicago approach to

economics revolutionized the field of labor economics and developed
concepts and evidence that now shape the thought of the entire profes-
sion.1 Not all economists working in this tradition were faculty, postdocs,
or students at Chicago, although the dominant figures were.
Chicago labor economics was a natural by-product of Chicago eco-
nomics. The same general principles were relentlessly applied in all fields,
and labor was no exception. Chicago economics emphasized the value of
economic models in interpreting and guiding collection of data and mak-
ing forecasts and constructing policy counterfactuals. The contributions
of labor economists working in the Chicago tradition are so extensive that
no short survey can do justice to them. Instead, I will discuss the core prin-
1
For a comprehensive history of Chicago labor economics, see Kaufman (2010).

ciples of the enterprise that have been relentlessly applied with great the-
oretical and empirical success.
The Chicago approach was in stark contract with the prevalent meth-
ods used in the labor economics of that time: industrial relations. This
was a largely atheoretical institutionalist approach that focused on thick
description, not explanation. It made generalizations to summarize data,
but the summaries were largely disconnected from analytical economics
and often from each other (see, e.g., Kerr et al. 1960). Its hostility to eco-
nomic theory is well exemplified by the paper of Richard Lester (1946),
who interviewed managers of firms, literally asking them if they set wages
to marginal products. Finding they did not report doing so, he pro-
nounced that marginal analysis was useless for the study of labor mar-
kets.2
The Chicago approach emphasized the value of price theory in inter-
preting data. Milton Friedman was a major influence in economics at
Chicago, and his approach to theory and empirical research set the tone.3
In an essay on Wesley Clair Mitchell, he crystallized the Chicago approach
to scientific economics:4
The ultimate goal of science in any field is a theory—an inte-

grated “explanation” of observed phenomena that can be used
to make valid predictions about phenomena not yet observed.
Many kinds of work can contribute to this ultimate goal and
are essential for its attainment: the collection of observations
about the phenomena in question; the organization and arrange-
ment of observations and the extraction of empirical general-
izations from them; the development of improved methods of
measuring or analyzing observations; the formulation of partial
or complete theories to integrate existing evidence. (Friedman
1950, 465)
In a dig at the institutionalists in the same essay, he quoted Marshall:

“‘The most reckless and treacherous of all theorists is he who professes
to let facts and figures speak for themselves.’ [Marshall 1885] And, one
might add ‘The most reckless and treacherous of all empirical workers
2
This paper provoked numerous responses. See, e.g., Machlup (1947) and Friedman
(1953).
3
Friedman made basic contributions to labor economics. Friedman and Kuznets (1945)
pioneered the study of panel data income dynamics and introduced the notion of firm-specific
human capital. Friedman (1951) studied the impact of unionism. Prior to the 1950s, Chi-
cago economist Paul Douglas studied production functions, labor supply, and unionism.
Frank Knight and Henry Simons wrote polemical essays on monopoly unionism.
4
He later expanded on this essay in his well-known essay “The Methodology of Positive
Economics” (Friedman 1953).

is he who formulates theories to explain observations that are the prod-
uct of careless and inaccurate empirical work’” (Friedman 1950, 465–66).
The interplay between data and theory was the hallmark of the Chi-
cago approach. The problems investigated were important for under-
standing the economy and public policy. Careful documentation of these
problems and new empirical regularities were highly valued.
However, economic analysis did not stop there. Interpretation of data—
understanding the problems being studied and the mechanisms generat-
ing them—was a crucial part of policy analysis. All of these activities were
essential for the scientific analysis of counterfactuals that is the basis for
rigorous policy analysis. Both data and theory were taken very seriously.
Economic theory was viewed as an engine for analysis and empirical dis-
covery, and not as an end itself. At the same time, great value was placed
on careful and comprehensive empirical work that produced convincing
evidence. Great value was also placed on rigorous economic theorizing
that made lasting contributions to the understanding of the economy.
Models that were discordant with data were revised and tested on the
same and new data.
Theory was subjected to rigorous testing against the data and was used
to parsimoniously explain phenomena within and across fields in eco-
nomics, including labor economics. Only when the standard tools failed
would the theory be amended. This approach was in stark contrast to that
of the institutionalists who often favored ad hoc generalizations to “let
the facts speak for themselves.” They typically made up new models one
empirical finding at a time and lacked a common core of basic principles
that applied across multiple domains.
Chicago-inspired labor economics applied price theory to the labor
market and emphasized testing models on data and taking great care with
empirical evidence. The first studies used consumer demand theory (the
demand for leisure) to understand the relationship of labor supply with
wages and asset income (Lewis 1956; Mincer 1962) and to estimate in-
come and substitution effects.5 Lewis (1963) investigated counterfactual
union wage impacts.6 A study that addressed the challenging policy ques-
tion of whether there was an undersupply or oversupply of education led
Becker (1962, 1964), a Chicago PhD, to create the edifice of modern hu-
man capital theory.
5
Douglas (1934) and Schoenberg and Douglas (1937) were pioneering empirical papers
on relations of wages (or earnings) to measures of labor supply. See Pencavel (1986) for a
survey of labor supply.
6
His neglected and densely written book defined and estimated economically grounded
counterfactual union “effects” for a variety of market scenarios, including partial and general
equilibrium. Lewis’s (1963) framework is far richer than just the framework he used to mea-
sure union gaps that he estimated in his later work (Lewis 1986).

Becker’s work entailed more than a direct application of consumer de-
mand theory. New analyses were required to explain the relationship of
human capital to earnings and other phenomena. True to Chicago tradi-
tion, human capital theory was parsimonious and had general applica-
tions. It interpreted empirical regularities on earnings, on-the-job train-
ing, life cycle wage growth, the quit and layoff decisions of workers and
firms, and patterns of trade across countries using a core set of basic prin-
ciples consistent with prior theory, but supplementing it when needed. It
is a brilliant operationalization of Friedman’s vision of Chicago econom-
ics.
Becker’s work on fertility (Becker and Lewis 1973; Becker 1991) grew
out of an early failure to explain the time series of fertility and income.7
The systematic study of labor supply and fertility led to novel papers on
the allocation of time and the creation of a new field of household eco-
nomics.8 At each step, theory was tested, and if it failed, it was modified,
but only as needed. The core principles remained.
Becker was, of course, a uniquely creative figure. But the methodolog-
ical principles of the Chicago approach were applied by many labor econ-
omists to explain a variety of phenomena in the labor markets. For exam-
ple, Walter Oi (1962) and Sherwin Rosen (1968) applied and expanded
production theory to analyze the determinants of labor demand using
the Chicago approach.9 Hedonic models of pricing of heterogeneous
goods (and skills) explained variation in quality in both product markets
and labor markets.10 It explained apparent deviations from the law of one
price by invoking quality variations as the source of these differences. In
the same vein, Stigler’s (1961) search theory explained price (and wage
dispersion) not as a failure of competitive markets—as had the institu-
tionalists—but as a consequence of costly search. Rosen’s paper (with Ed
Lazear) on tournaments (Lazear and Rosen 1981) was a basic contribution
to understanding the determinants of compensation of workers.
Research on the economics of the family synthesized and expanded
the Chicago portfolio (see Becker 1991). The three basic ingredients of
Chicago economics were central to the field in his analyses: (a) stable
preferences for agents, (b) agents responding to incentives, and (c) equi-
librium. Studying equilibrium in marriage markets that matched indivisi-
ble agents required new tools not standard in conventional price theory
but present in earlier work by Koopmans and Beckmann (1957). Chiap-
pori (2017) defines the state of the art.
7
See Heckman (2015) for a discussion of the evolution of Becker’s thought.
8
Mincer (1962, 1963) played a seminal role in this activity.
9
See Hamermesh (1993) for a survey.
10
See Lewis (1969), Welch (1969), and Rosen (1974) for early work.

Another example is the work of Becker and Tomes (1979, 1986) analyz-
ing intergenerational mobility. Basic tools of economics were adapted
and applied to the data to investigate the persistence of family influence
across generations. As discussed by Mogstad (2017), this is an active area
of research that integrates theory and empirical work and extends theory.
Many other examples of labor economics in the Chicago tradition could
be given, but space constraints prevent this. The guiding principles in each
of these studies are the same. Theory is used to interpret data. Data are
used to test theory. Understanding the mechanisms producing empiri-
cally estimated “effects” is essential for principled counterfactual analysis
and for explaining phenomena. It was never enough to say an intervention
“worked.” It was required that analysts understand the mechanisms pro-
ducing “the facts,” their generality across multiple empirical domains,
and their relevance for public policy.
References
Becker, Gary S. 1962. “Irrational Behavior and Economic Theory.” J.P.E. 70:1–13.
———. 1964. Human Capital: A Theoretical and Empirical Analysis, with Special Ref-
erence to Education. New York: NBER.
———. 1991. A Treatise on the Family. Enl. ed. Cambridge, MA: Harvard Univ.
Press.
Becker, Gary S., and H. Gregg Lewis. 1973. “On the Interaction between the
Quantity and Quality of Children.” J.P.E. 81, no. 2, pt. 2 (April): S279–S288.
Becker, Gary S., and Nigel Tomes. 1979. “An Equilibrium Theory of the Distribu-
tion of Income and Intergenerational Mobility.” J.P.E. 87 (6): 1153–89.
———. 1986. “Human Capital and the Rise and Fall of Families.” J. Labor Econ. 4,
no. 3, pt. 2:S1–S39.
Chiappori, Pierre-André. 2017. Matching with Transfers: The Economics of Love and
Marriage. Princeton, NJ: Princeton Univ. Press.
Douglas, Paul H. 1934. Theory of Wages. New York: Macmillan.
Friedman, Milton. 1950. “Wesley C. Mitchell as an Economic Theorist.” J.P.E. 58
(6): 465–93.
———. 1951. “Some Comments on the Significance of Labor Unions for Eco-
nomic Policy.” In The Impact of the Union: Eight Economic Theorists Evaluate the
Labor Union, edited by David McCord Wright, 204 –34. New York: Kelley &
Millman.
———. 1953. “The Methodology of Positive Economics.” In Essays in Positive Eco-
nomics, edited by Milton Friedman. Chicago: Univ. Chicago Press.
Friedman, Milton, and Simon Smith Kuznets. 1945. Income from Independent Pro-
fessional Practice. New York: NBER.
Hamermesh, Daniel S. 1993. Labor Demand. Princeton, NJ: Princeton Univ. Press.
Heckman, James J. 2015. “Introduction to a Theory of the Allocation of Time by
Gary Becker.” Econ. J. 125 (583): 403–9.
Kaufman, Bruce. 2010. “Chicago and the Development of Twentieth Century
Labor Economics.” In The Elgar Companion to the Chicago School of Economics, ed-
ited by Ross B. Emmett, 128–51. Northampton, MA: Elgar.

Kerr, Clark, Frederick H. Harbison, John T. Dunlop, and Charles A. Myers. 1960.
Industrialism and Industrial Man: The Problems of Labor and Management in Eco-
nomic Growth. Cambridge, MA: Harvard Univ. Press.
Koopmans, Tjalling C., and Martin Beckmann. 1957. “Assignment Problems and
the Location of Economic Activities.” Econometrica 25 (1): 53–76.
Lazear, Edward P., and Sherwin Rosen. 1981. “Rank-Order Tournaments as Op-
timum Labor Contracts.” J.P.E. 89 (5): 841–64.
Lester, Richard A. 1946. “Shortcomings of Marginal Analysis for Wage-Employment
Problems.” A.E.R. 36 (1): 63–82.
Lewis, H. Gregg. 1956. “Hours of Work and Hours of Leisure.” In Annual Proceed-
ings of the Industrial Relations Research Association, 196–206. Madison, WI: Indus.
Res. Assoc.
———. 1963. Unionism and Relative Wages in the United States: An Empirical Inquiry.
———. 1969. “Employer Interests in Employee Hours of Work.” Cuadernos de
Economia 18:38–54.
———. 1986. Union Relative Wage Effects: A Survey. Chicago: Univ. Chicago Press.
Machlup, Fritz. 1947. “Rejoinder to an Antimarginalist.” A.E.R . 37 (1): 148–54.
Marshall, Alfred. 1885. The Present Position of Economics: An Inaugural Lecture. Lon-
don: Macmillan.
Mincer, Jacob. 1962. “Labor Force Participation of Married Women.” In Aspects of
Labor Economics, edited by H. Gregg Lewis. Princeton, NJ: Princeton Univ.
Press.
———. 1963. “Market Prices, Opportunity Costs, and Income Effects.” Measure-
ment Econ. 48 (1): 67–82.
Mogstad, Magne. 2017. “The Human Capital Approach to Intergenerational Mo-
bility.” J.P.E. 125 (6): 1862–68.
Oi, W. Y. 1962. “Labor as a Quasi-Fixed Factor.” J.P.E. 70 (6): 538–55.
Pencavel, John H. 1986. “Labor Supply of Men: A Survey.” In Handbook of Labor
Economics, vol. 3, edited by Orley Ashenfelter and Richard Layard. Amsterdam:
North-Holland.
Rosen, Sherwin. 1968. “Short-Run Employment Variation on Class-I Railroads in
the U.S., 1947–1963.” Econometrica 36 (3/4): 511–29.
———. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in
Pure Competition.” J.P.E. 82 (1): 34 –55.
Schoenberg, Erika H., and Paul H. Douglas. 1937. “Studies in the Supply Curve
of Labor: The Relation in 1929 between Average Earnings in American Cities
and the Proportions Seeking Employment.” J.P.E. 45 (1): 45–79.
Stigler, George J. 1961. “The Economics of Information.” J.P.E. 69 (3): 213–25.
Welch, Finis. 1969. “Linear Synthesis of Skill Distribution.” J. Human Resources 4
(3): 311–27.

Keeping the ECON in Econometrics:

(Micro-)Econometrics in the Journal
of Political Economy
Stephane Bonhomme
Azeem M. Shaikh
In 1970, John Siegfried wrote an instructive short note in the miscellany

section of the Journal of Political Economy titled “A First Lesson in Econo-
metrics” (Siegfried 1970). The author starts by writing the equation “1 1
1 5 2” but immediately argues that “every budding econometrician must
learn early that it is never in good taste to express the sum of two quan-
tities in [this] form” (1378). He then produces two pages of intricate der-
ivations to arrive at an equivalent but extremely cumbersome expres-
sion.1 From the publication of this note, it is reasonable to infer that
the JPE’s editorial team at the time had some level of distrust in sophis-
ticated econometric analysis.2 Shortly thereafter, however, the journal
began to play a key role in the development of several, novel economet-
ric ideas.
Compared to many of its competitors, the type of econometric re-
search the JPE has published has two distinctive features. The first one
is the promotion of a type of econometric work that is tightly connected
to economic models. In particular, the JPE has been a leading vehicle for
structural econometric modeling. The second main feature is the em-
phasis on empirical applications of the methodology. The JPE seldom
publishes abstract econometric theory. Instead, it promotes econometric
analysis mainly through applications. In agreement with the motto of
While the title echoes Leamer’s (1983) “taking the con out of econometrics,” the expres-
sion “keeping the econ” was previously used by Ehrlich and Liu (1999) in a paper that ap-
peared in the Journal of Law and Economics, also published at the University of Chicago.
1
To add to the irony of that note, Eldridge (2014) points out that Siegfried’s formula is
wrong because of a matrix algebra mistake.
2
For example, the JPE is mentioned only in passing in Christ’s (1994) historical account
of the Cowles Commission between 1939 and 1955, during which it was hosted at the Uni-
versity of Chicago.

the Cowles Commission, the JPE’s style of econometrics is one in which
theory and measurement go hand in hand.
Since trying to review all of the econometrics research in the JPE would
be a daunting task, we will focus on only a handful of contributions, each
of which links economics and econometrics in particularly insightful ways.
Such a choice necessarily means leaving aside a large number of equally
important and influential contributions. In the same spirit, this review will
be limited mainly to microeconometric applications, abstracting from key
contributions to time-series econometrics, macroeconometrics, and fi-
nance that have appeared in the journal.
Econ Meets Metrics: An Econometric

Model of Marriage
Becker’s (1973, 1974) classical theory of marriage appeared in the JPE
and is considered a landmark of the journal. Becker proposed a static
model of the marriage market in which agents of different types, when
matched, share surplus and can transfer utility to each other. Agents rank
potential matches according to their preferences. In equilibrium, all matches
are stable. Viewing marriage as a rational decision leading to an equilib-
rium distribution of matches has strong empirical appeal. For example, the
model could be used to understand the effect of divorce laws or changes
in contraception technology on marriage patterns. Devising an empirical
counterpart to the Becker model, however, remained an unsolved ques-
tion for a long time.
The JPE has been a pioneer in the structural econometric analysis of
marriage markets, and more generally, it has published some of the most
innovative and accomplished work in structural econometrics. The struc-
tural approach tries to build and exploit a tight link between the eco-
nomic model and the empirical econometric model. Its main goal is to es-
timate primitive structural parameters with the hope that such parameters
are invariant to policy and can be used for counterfactual predictions.
Choo and Siow (2006) proposed a structural econometric model of
marriage. They completed the Becker theory to make it an econometric
model that could be taken to the data. In doing so they faced several
challenges: first, how to define an agent’s type empirically and how to
properly account for heterogeneity in preferences; second, how to deal
with the fact that, typically, data on transfers within couples are not ob-
served by the econometrician.
In the Choo and Siow framework, agents (i.e., men and women) have
discrete types defined in terms of covariates such as age, education, eth-
nicity, or geography. Individuals of both genders have preferences for
being married to different types of individuals. The structure of prefer-
ences is key to the tractability of the framework. Specifically, the utility of

a woman of type i (e.g., defined in terms of age and education) for mar-
rying a man of type j is the sum of an (i, j )–specific systematic preference
term, an (i, j )–specific transfer, and an idiosyncratic preference term.
Building on McFadden (1974), the latter is assumed to follow a type I ex-
treme value distribution, independent of the other terms and indepen-
dent across options. For given values of systematic preferences and trans-
fers, women’s choices therefore take the form of logit demand models.
Symmetrically, men’s choices also take a logit form.
In contrast to single-agent choice models, in marriage markets two
types of agents interact with each other and equilibrium constraints must
be met. An innovation of the Choo and Siow framework is that transfers,
which are not observed by the econometrician, are identified as the prices
that clear the market and make women’s and men’s demands equal. As
a result, the overall structure of the model is a two-sided logit demand
model with equilibrium constraints.
A particular implication of the model is that it delivers a closed-form
expression for the utility gains from marriage. A key equation in Choo
and Siow (2006) shows that net gains from marriage for agents of types
(i, j), relative to being single, can be written as a combination of quanti-
ties that are typically easy to estimate: the number of men and women of
types (i, j) who are married, divided by the geometric average of the num-
ber of unmarried women of type i and unmarried men of type j. This trans-
parent expression illustrates the power of a theory that delivers an eco-
nomically interpretable quantity that can be directly estimated from the
data. Taking advantage of this expression, in a way that is typical of many
JPE papers, Choo and Siow illustrate the empirical relevance of their
framework by estimating net gains from marriage in 1970–71 and 1981–
82 by gender, age, and age of the spouse. In addition, they estimate how
gains from marriage evolved after the legalization of abortion in Roe v.
Wade by exploiting variation across states in a difference-in-differences
fashion. This exercise nicely showcases the type of applications that can
be studied with the framework.
Choo and Siow’s (2006) seminal paper has already spurred a long leg-
acy. Important work building on their framework has also appeared in
the JPE (e.g., Chiappori and Oreffice 2008; Chiappori, Oreffice, and
Quintana-Domeque 2012; Dupuy and Galichon 2014; Chiappori, Costa
Dias, and Meghir, forthcoming; Fox, Yang, and Hsu, forthcoming).
Structural Econometric Models of the Labor Market

Among the many studies using the structural approach that have ap-
peared in the JPE, models of education decisions and career choices
have particularly benefited from the development of novel econometric
methods. In such models, individuals choose to select into different ca-

reers depending on the costs they face and their expected returns. The
econometrician must deal with the fact that returns and costs are largely
unobserved. A central challenge is thus how to estimate rates of return to
college or to a type of occupation in the presence of self-selection.
The 1970s and 1980s saw great progress on the understanding of selec-
tion models. Some of the key contributions appeared in the JPE (Gronau
1974; Heckman and Sedlacek 1985). Here, we focus on two contribu-
tions to structural econometrics that have built on this work. We note
that there are several important JPE articles that are closely related to
these two papers, such as Cameron and Heckman (1998, 2001), which
we do not discuss here because of space constraints.
Willis and Rosen (1979) is an early example of a structural econometric
model of education decisions.3 The empirical model builds on the the-
ory of comparative advantage. This work contains a number of strikingly
modern econometric insights that are still relevant to today’s research. A
notable aspect concerns the way the authors specify and analyze the coun-
terfactual—or “potential”—outcomes corresponding to different educa-
tion choices. Their classical discussion of the role of exclusion restrictions,
which are needed for credible identification, includes an exposition of the
distinction between the marginal rate of return to investment and the
marginal cost of funds due to Gary Becker. The role of functional form
assumptions is also carefully discussed.
Another noteworthy aspect of the analysis is the way the economic
model and the econometrics are linked to each other. The model’s pre-
dictions are assessed for two outcomes: initial earnings in the life cycle
and growth rates of earnings. The authors test several of the main struc-
tural restrictions of the model, but they do not interpret the fact that
those restrictions are not violated as definitive success for the structural
model. In the conclusion of the paper, the authors go one step further
and include a small out-of-sample prediction exercise as a validation
check.
Willis and Rosen’s work was extended by Keane and Wolpin (1997),
also published in the JPE, in several dimensions. Keane and Wolpin build
a dynamic life cycle model of human capital investment in which individ-
uals go to school and work in various occupations. Agents, who face un-
certainty in the returns to their choices (i.e., wages), are forward looking
and have rational expectations. Keane and Wolpin work under the con-
straint that the restrictions from the theory must be fully imposed in es-
timation. This structural approach to policy evaluation then allows them
to perform counterfactual policy exercises. Taking such a setup to the
data raises a number of econometric challenges. Setting up a coherent
3
The reader may wonder about the unusual ordering of the authors’ names. The initial
footnote informs us that it was “selected by a random device.”

structure that is rich enough to fit the complex heterogeneity in individ-
ual trajectories, while keeping the model tractable, remains a very diffi-
cult task today.
A central feature of Keane and Wolpin’s econometric model is its dy-
namic nature. Experience is treated as a state variable, and agents form
expectations about streams of income conditional on education and ca-
reer choices. Estimation is based on maximum likelihood. Unlike previ-
ous structural dynamic discrete choice models, however, observed wages
in their model are self-selected since work and experience are choices—
and therefore endogenous, just like schooling. Endogeneity complicates
estimation since it is not possible to proceed sequentially. For computa-
tion, Keane and Wolpin develop an approximate solution to the dynamic
programming problem that allows them to address the computational
curse of dimensionality.
A second key feature of the model is the presence of unobserved types.
Borrowing from Heckman and Sedlacek (1985), Keane and Wolpin al-
low for self-selection in multiple dimensions of skill endowments. They
deal with the presence of multidimensional heterogeneity using a finite
mixture approach, which disciplines the different dimensions of hetero-
geneity.
Since its publication, Keane and Wolpin’s framework has become a
blueprint for structural econometric analysis in labor economics and
elsewhere. Dynamic structural econometric modeling is still a vibrant re-
search area, and some of the best research in this field is appearing in
the JPE, such as two recent contributions by Adda, Dustmann, and Ste-
vens (2017) and Heckman, Humphries, and Veramendi (forthcoming).
Partial Identification Meets Economic Theory

Partial identification is one of the most prominent recent themes in
econometrics. The JPE played an early and important role in promoting
the use of such methods in economics. The defining feature of a partially
identified model is that the parameter of interest is not uniquely deter-
mined by the distribution of the observed data. Instead, it is limited only
to a set of values. As we will see below, one of the main attractions of such
methods is that they permit researchers to avoid making assumptions
that may be deemed unpalatable for one reason or another but that might
have been previously made for tractability.
While not fitting within our theme of microeconometrics, an early and
influential example of partial identification can be found in Hansen and
Jagannathan’s (1991) landmark paper on the implications of security
market data for asset pricing models. In the case of Hansen and Jagan-
nathan, the parameters of interest are the means and standard devia-
tions of the intertemporal marginal rates of substitution and the observed

data consist of security market data. They observe that under weak as-
sumptions one can restrict the set of possible values for the parameters
of interest using the Cauchy-Schwarz inequality. In contrast to previous
approaches, Hansen and Jagannathan need not specify parametric func-
tional forms for the intertemporal marginal rates of substitution. In fact,
their analysis allows them to conclude that certain specifications are in-
consistent with the observed data.
A more recent example of partial identification and one that fits more
closely within our theme is Haile and Tamer’s (2003) analysis of English
or oral ascending auctions. In the case of Haile and Tamer, the parameter
of interest is the distribution of bidders’ (private) valuations and the ob-
served data consist of bids. Instead of relying on a particular model of bid-
ding behavior, such as the “button model” found in Milgrom and Weber
(1982), which they argue may be inconsistent with the observed data,
they instead propose assuming only that (i) bidders do not bid more than
they are willing to pay and (ii) bidders do not allow an opponent to win at
a price they are willing to beat. Using these minimal assumptions on bid-
der behavior and well-known results from the theory of order statistics,
Haile and Tamer derive bounds on the distribution of valuations, which,
in turn, permit them to construct bounds on the optimal reserve price in
such auctions.
A common criticism of partial identification is that weak assumptions
are often accompanied by limited ability to draw meaningful conclusions
from the data. Hansen and Jagannathan and Haile and Tamer both show
that this need not always be the case. In fact, in both settings, weak as-
sumptions lead to remarkably sharp conclusions. In this way, both papers
illustrate clearly the usefulness of approaching empirical work through
the combined lens of economic theory and partial identification and
have provided ample motivation for further applications of partial iden-
tification as well as the development of the accompanying theory for es-
timation and inference. Recent work in this spirit is the estimation of a
structural voting model with deliberation using data from US appellate
courts in Iaryczower, Shi, and Shum (forthcoming).
Conclusion
In this brief and partial review of microeconometrics in the JPE, we have
highlighted the journal’s focus on econometric frameworks that propose
novel ways of taking fundamental economic theories to the data. Influen-
tial examples that we have not discussed include hedonic models (Rosen
1974; Ekeland, Heckman, and Nesheim 2004) and collective models (Chiap-
pori 1992; Browning et al. 1994). In addition to this focus on the interplay
between economic models and empirical analysis, we note that the JPE has
also published several key contributions to traditional areas of economet-

rics such as instrumental variables (Altonji, Elder, and Taber 2005) and
measurement error models (Erickson and Whited 2000).
Despite John Siegfried’s warning against unnecessarily complicated
econometrics, the recent history of the JPE demonstrates the power of
careful econometric thinking in order to blend economic theory and
empirical measurement. We hope that going forward the journal will
continue and reinforce its role as a promoter of pioneering econometric
research.
References
Adda, J., C. Dustmann, and K. Stevens. 2017. “The Career Costs of Children.”
J.P.E. 125 (2): 293–337.
Altonji, J. G., T. E. Elder, and C. R. Taber. 2005. “Selection on Observed and Un-
observed Variables: Assessing the Effectiveness of Catholic Schools.” J.P.E. 113
(1): 151–84.
Becker, G. S. 1973. “A Theory of Marriage: Part I.” J.P.E. 81 (4): 813–46.
———. 1974. “A Theory of Marriage: Part II.” J.P.E. 82, no. 2, pt. 2 (April): S11–
S26.
Browning, M., F. Bourguignon, P.-A. Chiappori, and V. Lechene. 1994. “Income
and Outcomes: A Structural Model of Intrahousehold Allocation.” J.P.E. 102
(6): 1067–96.
Cameron, S. V., and J. J. Heckman. 1998. “Life Cycle Schooling and Dynamic Se-
lection Bias: Models and Evidence for Five Cohorts of American Males.” J.P.E.
106 (2): 262–333.
———. 2001. “The Dynamics of Educational Attainment for Black, Hispanic,
and White Males.” J.P.E. 109 (3): 455–99.
Chiappori, P.-A. 1992. “Collective Labor Supply and Welfare.” J.P.E. 100 (3): 437–
67.
Chiappori, P.-A., M. Costa Dias, and C. Meghir. Forthcoming. “The Marriage
Market, Labor Supply, and Education Choice.” J.P.E.
Chiappori, P.-A., and S. Oreffice. 2008. “Birth Control and Female Empower-
ment: An Equilibrium Analysis.” J.P.E. 116 (1): 113–39.
Chiappori, P.-A., S. Oreffice, and C. Quintana-Domeque. 2012. “Fatter Attrac-
tion: Anthropometric and Socioeconomic Characteristics in the Marriage
Market.” J.P.E. 120:659–95.
Choo, E., and A. Siow. 2006. “Who Marries Whom and Why.” J.P.E. 114 (1): 175–201.
Christ, C. F. 1994. “The Cowles Commission’s Contributions to Econometrics at
Chicago, 1939–1955.” J. Econ. Literature 32 (1): 30–59.
Dupuy, A., and A. Galichon. 2014. “Personality Traits and the Marriage Market.”
J.P.E. 122:1271–1319.
Ehrlich, I., and Z. Liu. 1999. “Sensitivity Analyses of the Deterrence Hypothesis:
Let’s Keep the Econ in Econometrics.” J. Law and Econ. 42 (S1): 455–88.
Ekeland, I., J. J. Heckman, and L. Nesheim. 2004. “Identification and Estimation
of Hedonic Models.” J.P.E. 112 (S1): S60–S109.
Eldridge, D. S. 2014. “A Comment on Siegfried’s First Lesson in Econometrics.”
Econ. Inquiry 52 (1): 503–4.
Erickson, T., and T. M. Whited. 2000. “Measurement Error and the Relationship
between Investment and q.” J.P.E. 108 (5): 1027–57.
Fox, J. T., C. Yang, and D. H. Hsu. Forthcoming. “Unobserved Heterogeneity in
Matching Games.” J.P.E.

Gronau, R. 1974. “Wage Comparisons—a Selectivity Bias.” J.P.E. 82 (6): 1119–43.
Haile, P. A., and E. Tamer. 2003. “Inference with an Incomplete Model of English
Auctions.” J.P.E. 111 (1): 1–51.
Hansen, L. P., and R. Jagannathan. 1991. “Implications of Security Market Data
for Models of Dynamic Economies.” J.P.E. 99 (2): 225–62.
Heckman, J. J., J. E. Humphries, and G. Veramendi. Forthcoming. “Returns to
Education: The Causal Effects of Education on Earnings, Health, and Smok-
ing.” J.P.E.
Heckman, J. J., and G. Sedlacek. 1985. “Heterogeneity, Aggregation, and Market
Wage Functions: An Empirical Model of Self-Selection in the Labor Market.”
J.P.E. 93 (6): 1077–1125.
Iaryczower, M., X. Shi, and M. Shum. Forthcoming. “Can Words Get in the Way?
The Effect of Deliberation in Collective Decision Making.” J.P.E.
Keane, M. P., and K. I. Wolpin. 1997. “The Career Decisions of Young Men.” J.P.E.
105 (3): 473–522.
Leamer, E. E. 1983. “Let’s Take the Con Out of Econometrics.” A.E.R. 73 (1): 31–
43.
McFadden, D. 1974. “Conditional Logit Analysis of Qualitative Choice Behavior.”
In Frontiers in Econometrics, edited by P. Zarembka. New York: Academic Press.
Milgrom, P. R., and R. J. Weber. 1982. “A Theory of Auctions and Competitive
Bidding.” Econometrica 50:1089–1122.
Rosen, S. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in
Pure Competition.” J.P.E. 82 (1): 34–55.
Siegfried, J. J. 1970. “A First Lesson in Econometrics.” J.P.E. 78 (6): 1378–79.
Willis, R. J., and S. Rosen. 1979. “Education and Self-Selection.” J.P.E. 87, no. 5,
pt. 2 (October): S7–S36.
Life Cycle Wage Dynamics and Labor Mobility

Derek Neal
Introduction
The Journal of Political Economy has published a number of seminal papers
on individual investments in human capital and how these investments
vary with ability, preferences, age, and other individual characteristics.1
I thank Stephane Bonhomme, Ronni Pavan, Canice Prendergast, and Christopher Taber
for useful comments.
1
Examples include Mincer (1958), Becker (1962), Ben-Porath (1967), Heckman (1976),
and Rosen (1976).

However, most of this literature has nothing to say about labor mobility.
For the most part, jobs play no role in this literature. All human capital is
general, and the wage earned by a given worker at a particular point in
time is the product of the stock of human capital she possesses and the
rental rate on human capital. There is one market price for the labor ser-
vices of any given worker, and all firms must pay this price in order to re-
ceive her labor services. As a result, these models cannot address data on
labor turnover.
Models of turnover propose mechanisms that explain how both the in-
side and outside options for particular workers evolve over their careers
and therefore produce predictions about the relationships between
wage dynamics and labor mobility that vary with worker skills, tenure,
and labor market experience. While most of this literature has focused
on worker mobility among firms, a more recent literature has explored
links between wage dynamics and patterns of mobility among industries
or careers.
Here I discuss four JPE papers that are intellectual cornerstones of the
literature on individual wage dynamics and mobility decisions and also
comment on how work published in the JPE and elsewhere built on
these seminal papers.
Hiring Costs, Specific Skills, and Other Frictions

Becker (1962) and Oi (1962) are cornerstones of the literature that ex-
plores why the value of a worker to his current firm may diverge from the
worker’s best outside option, even in competitive labor markets. Oi dis-
cusses several fixed costs associated with hiring workers. New workers
need training in order to understand how a given firm organizes work
and communication. Further, they also need to learn how to use capital
that is specific to their new firm. In addition, firms cannot hire workers
or assign workers to tasks without screening them in some way. Screen-
ing activities may include reviewing applications, conducting interviews,
requiring workers to audition during probationary employment periods,
and so forth.
Oi (1962) argues that, in competitive labor markets, workers must pay
for these training and screening costs by accepting wages less than their
marginal products. One can imagine many different payment schedules,
but among workers with some seniority in their current firms, their com-
pensation may well exceed their best option elsewhere, since any new
employer would require them to pay for new hiring costs associated with
training and screening.
Oi (1962) provides some evidence that these hiring costs are greater
among more skilled workers, and he argues that this is the reason that
turnover rates among skilled workers are less cyclical. Assume for a mo-

ment that, in economic downturns, some but not all firms suffer transi-
tory demand shocks that decrease the marginal product of their workers.
Oi claims that those firms hit by negative shocks can ask their most skilled
employees to accept temporary wage cuts without losing them since large
hiring costs create an important wedge between their inside and outside
options.
Becker (1962) stressed the idea that firms and workers may share the
costs and returns associated with firm-specific training. During training pe-
riods, workers earn more than their marginal product but less than their
outside option. After training, workers earn more than their outside op-
tion but less than their full value to the firm. Such sharing rules may pro-
mote efficient separation decisions in settings in which shocks arrive that
change either firm productivity or the outside options of workers.
This framework also suggests that skilled workers, who benefit more
from training and therefore receive more training, should exhibit lower
turnover rates. Further, individual wages should increase with worker
tenure, holding constant wage growth attributable to the accumulation
of general human capital. This insight spawned a large empirical litera-
ture that sought to measure the impact of changes in worker seniority on
wages holding constant years of education and total labor market expe-
rience. I return to this topic below.
Acemoglu and Pischke (1999) note that the types of frictions that
Becker (1962) and Oi (1962) identify may make it profitable for firms
to pay for general training as well. The existence of employment fric-
tions may make it possible for firms and their workers to share in the costs
and returns of training that produces skills that are valuable in many
firms.
Lazear (2009) presents a model in which no skills are truly firm spe-
cific, but each firm employs a different combination of skills and subsi-
dizes investment in the skill mix it desires. This is a two-period model,
and in the second period, a worker remains in her initial firm if her skills
are more valuable to this firm than to the outside firm that makes the
best raiding offer. In this framework, Nash bargaining determines com-
pensation, and separation decisions are efficient, but market thickness
influences skill investment in the first period. Workers are more willing
to invest in skills if their first-period firm employs a mix of skills that is
highly valued by a large number of other potential employers.
These papers sharpen our thinking about how skill specificity affects
both wages and turnover. They offer reasons why skilled workers should
be less mobile than other workers, and they provide reasons why wages
may not equal outside options, even in competitive labor markets. Still,
none of these papers provide a complete characterization of both wage
growth and job mobility over the entire career of a worker. Almost two
decades after Becker and Oi produced their seminal work, another

economist trained at Chicago produced two papers that sparked a sig-
nificant literature on life cycle wage growth and patterns of labor mo-
bility.
Matching, Wage Growth, and Turnover

In 1979, Boyan Jovonavic published two chapters from his PhD thesis in
the JPE. Both are seminal contributions to modern work on individual
mobility decisions and how these decisions affect individual wage growth
over the life cycle. In the late 1960s and throughout the 1970s, research-
ers began analyzing data from the first modern panel surveys of labor
market experiences, and at least three patterns were apparent in these
early panel surveys: (1) wages are higher, on average, among workers with
greater tenure, holding total labor market experience constant; (2) sep-
aration rates decline with wage levels given various sets of controls for
worker characteristics; and (3) separation rates decline with tenure. Both
papers ( Jovanovic 1979a, 1979b) address these empirical regularities.
Jovanovic (1979a) presents a continuous-time model, but to facilitate
exposition, I discuss the discrete-time analogue. When a worker joins a
firm, he draws a match from a distribution, and both worker and firm ob-
serve a signal that provides information about the quality of the match. If
the worker remains with the firm, both worker and firm observe an addi-
tional signal concerning the match quality each period. The competitive
equilibrium in this model is an implicit contract equilibrium that re-
quires all firms to pay workers their expected marginal product based
on the information available about the quality of their current match.
Each period, the worker must decide whether to stay with his current firm
or pay a search cost and start over with a new firm.
Workers live forever in this model, and the distribution of potential
matches is the same in all firms. These features imply that workers follow
a reservation wage rule and that reservation wages rise with seniority.
Once a worker has drawn a match, his current job has value for two rea-
sons. First, he earns a wage equal to the expected value of the match, and
second, his match has option value. If the match turns out to be great, he
can keep it. If it turns out to be terrible, he can leave and try a new firm. In
Jovanovic (1979a), this option value component declines monotonically
with tenure because the posterior variance of the conditional distribu-
tion of match quality shrinks as each new signal arrives. For this reason,
reservation wages rise with seniority.
Average wages are positively correlated with seniority in this model pre-
cisely because reservation wages rise with seniority. Any worker with t pe-
riods of tenure has received signals that produced a sequence of poste-
rior means that exceed more cutoffs and a more stringent set of cutoffs

than another worker with only t 2 k periods of tenure. Thus, on average,
wages must be higher for workers with more tenure.
Further, the probability that the next signal leads to a revision that ends
the match decreases with the expected value of the current match. So, in-
dividual separation rates are negatively related to wage rates.
However, in Jovanovic (1979a), separation rates may rise with seniority
in the early periods of a match. Early on, separation rates may be higher at
tenure t 1 1 than at t because the reservation wage is higher at t 1 1.
Nonetheless, separation rates must eventually decline with seniority be-
cause of the shrinking variance of the conditional distribution of match
quality. At some point, a given worker becomes so confident about the ac-
tual quality of her match that there is almost no chance that future signals
could produce wage decreases that would justify a separation.
In 1979, Jovanovic may have been disappointed that this model did not
produce separation rates that declined monotonically with tenure. How-
ever, he had to be pleased when Farber (1994) reported results from the
first decade of work history information in the National Longitudinal
Survey of Youth—1979 (NLSY79). In contrast to earlier panel studies,
the NLSY79 contained weekly rather than yearly information on separa-
tions, and Farber discovered that individual separation rates actually rose
during the first 6 months of job tenure before falling monotonically, a pat-
tern that one can easily produce using the model in Jovanovic (1979a).
Since a significant fraction of separations occur in the first 6 months of
job spells, Jovanovic managed to predict an important feature of hazard
rates out of employment spells more than a decade before researchers
documented its existence.
Extensions of Jovanovic’s Matching Model

The optimal reservation wage policy in Jovanovic (1979a) is relatively
simple in this model for two reasons. First, Jovanovic assumes that each
worker has a common prior about the distribution of potential matches
that she applies to all potential employers. Second, he assumes that
workers learn about only one match at a time. The signals that one
match produces provide no information about the potential matches
that a worker might have with any other employer. Since workers live for-
ever and the number of potential jobs is infinite, workers never return to
any job they leave. This means that the information that a worker ac-
quires on any given job that eventually ends never affects his next job
choice.
Over the past four decades, economists have tried to generalize this
model. The literature on multiarmed-bandit problems with correlated
arms contains no analytical solutions to the most general formulations
of this matching model. If workers begin their careers with different pri-

ors over each possible job and learn about the distribution of matches in
many potential jobs from observing signals in their current match, the
problem becomes intractable because optimal decision rules depend on
too many state variables and laws of motion. So, subsequent work has ex-
plored different ways to relax the key assumptions in Jovanovic’s frame-
work while maintaining tractability.
Miller (1984) allows workers to have different prior beliefs about the
potential matches associated with different jobs, but he does not allow
the signals from one match to shape beliefs about potential matches in
other jobs. Miller argues that his model explains why young workers are
more likely to enter jobs in which success is rare but particularly lucrative.
In Neal (1999), I assume that jobs involve a firm component and a
career component. I further assume that workers cannot shop effectively
over different careers while working for one firm. This model predicts that
workers should search in two stages. They should focus on finding a career
match first and then search for a firm match. Job histories in the NLSY79
contain several patterns that are consistent with this two-stage search strat-
egy (see Neal 1999; Gathmann and Schoenberg 2010; Pavan 2010).
Gibbons and Waldman (1999) adopt a different approach. They as-
sume that all workers know the mappings between talent and expected
output in every potential job, and they adopt a learning technology such
that the rate of learning about worker talent is the same in all jobs. Here,
the optimal policy is simple. Each period, each worker selects the job in
which she expects to be most productive, given what past signals have
revealed about her talent. This model stands in sharp contrast to that
in Miller (1984) because optimal mobility patterns do not reflect differ-
ences in the value of information produced by different types of work
experience. Nonetheless, this model has enjoyed noteworthy success em-
pirically (see Gibbons and Waldman 1999; Gibbons et al. 2005).
The Role of Experience

Jovanovic (1979a) has been quite influential, but there are features of data
on life cycle wage growth and mobility that it cannot address. Because
workers live forever and learn nothing about themselves in one job that
allows them to search more efficiently for future jobs, wages and separa-
tion rates are not functions of labor market experience among workers
who share the same level of job seniority. One could add general learning
by doing as a quick fix for the implied wage profiles, but each worker
would still make mobility decisions facing an infinite horizon and any in-
formation a worker acquired while working at any one job would not af-
fect her expected match quality in future jobs. So, total market experience
would not affect mobility decisions holding seniority constant.

However, in Jovanovic (1979b), workers have finite work lives, and
among workers with the same seniority, total market experience affects
both wages and separation decisions. Here, workers can allocate time to
three activities: work, investment, and search. Job offers arrive randomly,
and workers decide whether to keep their current match or accept the new
offer. Investment activities build the value of the current match, and search
activities increase the arrival rate of new offers. The key insight of this
model is that good matches last, in part, because of behavioral mecha-
nisms that are not present in Jovanovic (1979a). Agents rationally invest
more in matches that are harder to beat and devote more time to search
when their current match is not that good.
In this model, wages rise with seniority relative to outside options, sep-
aration rates decline with current wages, and separation rates decline
monotonically with tenure. Further, holding tenure constant, separation
rates decline with total labor market experience since workers with more
experience have had more time to receive high offers, and all else equal,
workers devote less effort to search as they near the end of their work
lives because they face a shorter horizon over which to enjoy the returns
from successful search.
Sources of Wage Growth

Both Jovanovic papers highlight a reverse causality concern for the em-
pirical literature on wage determination. Wages are positively correlated
with seniority in both models. However, in Jovanovic (1979a), the value
of matches does not grow with time, and this positive correlation arises
solely because good matches last. And, even in Jovanovic (1979b), a com-
ponent of the relationship between tenure and wage levels is driven by
the fact that good initial matches last longer, in part, because they in-
duce low search intensity. The selection effects highlighted by these
models have frustrated many efforts to pin down the causal effect of se-
niority on wages. However, the ideas in these two papers have shaped
decades-long debates concerning the best ways to measure the contribu-
tion of growth in firm-specific, industry-specific, or career-specific skills
to life cycle wage growth.
This literature is too large to review here (see Altonji and Shakotko
1987; Topel 1991; Kambourov and Manovskii 2009). Yet Pavan (2011) de-
serves our attention. Pavan estimates a structural model of search and wage
growth. In the model, workers search for both firm and career matches.
Further, wages grow over the life cycle for four reasons: workers accumu-
late general human capital through schooling and experience, matches
with careers evolve over time, matches with firms evolve over time, and
workers change both firms and careers optimally. In Jovanovic (1979b),

good matches last, in part, because workers invest in them. In Pavan
(2011), matches last if they grow fast enough.2
Pavan’s estimates imply that both firm-specific and career-specific com-
ponents of wages grow as workers gain more seniority in a firm and more
experience in a career.3 In the final section of the paper, Pavan simulates
data from his model and then estimates an instrumental variable (IV) re-
gression quite similar to one found in Kambourov and Manovskii (2009).
The results of this regression imply that the firm-specific component of
wages does not grow with seniority, which parallels the results reported
in Kambourov and Manovskii’s paper. This finding casts serious doubt
on the value of such IV methods, since Pavan ran his IV regressions on
data simulated from a model in which the firm-specific component of
wages does grow substantially with tenure.
Pavan (2011) shows that researchers who seek to pin down the sources of
life cycle wage growth must employ empirical models that explicitly treat
wage growth and mobility decisions as joint outcomes. Jovanovic (1979a,
1979b) pointed this literature in the right direction, and Pavan (2011)
demonstrates the value of this approach. However, more work remains. Pa-
van needed several restrictive assumptions on the processes that deter-
mine wage growth in order to make his model computationally tractable.
Conclusion
Models of life cycle investment in general human capital are powerful
tools for explaining how the average earnings of particular types of work-
ers evolve, rather smoothly, over the life cycle. However, individual wage
and earnings histories are quite jagged. They often contain large increases
or decreases, and these changes are usually coincident with promotions,
changes of employer, or career changes.
Thus, researchers cannot explain individual wage histories without
formulating models that treat both individual wage growth and individ-
ual mobility decisions as endogenous life cycle outcomes. Even in a set-
ting in which demands for various types of workers are stationary, a worker
must discover what type she is and what forms of employment allow her
type to be most productive. In addition, a worker must constantly weigh
the returns from investing in her current match against the possible gains
from searching for better.
These search and investment activities often create ex post rents, that
is, wedges between current productivity and outside options, but the im-
2
Keane and Wolpin (1997) structurally estimate a dynamic model that allows young
men to choose dynamically working in either the blue-collar or white-collar sector. This
model includes sector-specific skill growth and shocks to individual skill stocks, but firms
play no role. All firms pay the same prices for both skills.
3
Pavan (2010) defines careers using industry and occupation codes. Neal (1995) and
Poletaev and Robinson (2008) produce similar results using data on wage changes among
displaced workers.

plicit contract equilibrium in Jovanovic (1979a) suggests that the exis-
tence of these rents does not necessarily imply that any jobs are rationed,
that workers do not invest efficiently in skills, or that mobility decisions
are inefficient.4
Further, if workers sort among jobs on the basis of information about
their skills that econometricians cannot measure but employers observe,
researchers may measure firm, industry, or career differences in residual
wage levels and turnover rates that are not evidence of ex ante labor mar-
ket rents but rather evidence that workers sort among firms and careers
on valuable traits that are in relatively fixed supply (see Neal 1998).
The next century of JPE papers will likely contain important efforts to
position explicit life cycle models of wage growth within models of labor
market equilibrium that contain shocks to firms or sectors. Here, mobil-
ity and wage growth may reflect shocks to what a worker knows about
herself or shocks to demand and technology that change the productiv-
ity of her type. This hybrid approach may help economists better under-
stand the causes and consequences of the ex post rents that experienced
workers enjoy in their current firms and careers.
References
Acemoglu, Daron, and Jörn-Steffen Pischke. 1999. “The Structure of Wages and
Investment in General Training.” J.P.E. 107 (3): 539 –72.
Altonji, Joseph G., and Robert A. Shakotko. 1987. “Do Wages Rise with Job Se-
niority?” Rev. Econ. Studies 54 (3): 437–59.
Becker, Gary S. 1962. “Investment in Human Capital: A Theoretical Analysis.”
J.P.E. 70, no. 5, pt. 2 (October): 9 –49.
Ben-Porath, Yoram. 1967. “The Production of Human Capital and the Life Cycle
of Earnings.” J.P.E. 75 (4): 352–65.
Farber, Henry S. 1994. “The Analysis of Interfirm Worker Mobility.” J. Labor Econ.
12 (4): 554 – 93.
Gathmann, Christina, and Uta Schoenberg. 2010. “How General Is Human Cap-
ital? A Task-Based Approach.” J. Labor Econ. 28 (1): 1– 49.
Gibbons, Robert, Lawrence F. Katz, Thomas Lemieux, and Daniel Parent. 2005.
“Comparative Advantage, Learning, and Sectoral Wage Determination.” J. La-
bor Econ. 23 (4): 681–724.
Gibbons, Robert, and Michael Waldman. 1999. “A Theory of Wage and Promo-
tion Dynamics Inside Firms.” Q. J.E. 114 (4): 1321– 58.
Heckman, James. 1976. “A Life-Cycle Model of Earnings, Learning, and Con-
sumption.” J.P.E. 84, no. 4, pt. 2 (August): S11– S44.
Jovanovic, Boyan. 1979a. “Job Matching and the Theory of Turnover.” J.P.E. 87
(5): 972– 90.
———. 1979b. “Firm-Specific Capital and Turnover.” J.P.E. 87 (6): 1246 – 60.
Kambourov, Gueorgui, and Iourii Manovskii. 2009. “Occupational Specificity of
Human Capital.” Internat. Econ. Rev. 50 (1): 63 –115.
Keane, Michael, and Kenneth Wolpin. 1997. “The Career Decisions of Young
Men.” J.P.E. 105 (4): 473–522.
4
However, these ex post rents may create inefficiency. See Sanders and Taber (2012).

Lazear, Edward P. 2009. “Firm-Specific Human Capital: A Skill-Weights Ap-
proach.” J.P.E. 117 (5): 914 – 40.
Miller, Robert A. 1984. “Job Matching and Occupational Choice.” J.P.E. 92 (6):
1086–1120.
Mincer, Jacob. 1958. “Investment in Human Capital and Personal Income Distri-
bution.” J.P.E. 66 (4): 281–302.
Neal, Derek. 1995. “Industry-Specific Human Capital: Evidence from Displaced
Workers.” J. Labor Econ. 13 (4): 653 –77.
———. 1998. “The Link between Ability and Specialization: An Explanation for
Observed Correlations between Wages and Mobility Rates.” J. Human Resources
33 (1): 173–200.
———. 1999. “The Complexity of Job Mobility among Young Men.” J. Labor
Econ. 17 (2): 237–61.
Oi, Walter Y. 1962. “Labor as a Quasi-Fixed Factor.” J.P.E. 70 (6): 538 – 55.
Pavan, Ronni. 2010. “The Role of Career Choice in Understanding Job Mobility.”
Labour 24 (2): 107–27.
———. 2011. “Career Choice and Wage Growth.” J. Labor Econ. 29 (3): 549 – 87.
Poletaev, Maxim, and Chris Robinson. 2008. “Human Capital Specificity: Evi-
dence from the Dictionary of Occupational Titles and Displaced Worker Sur-
veys, 1984–2000.” J. Labor Econ. 26 (3): 387–420.
Rosen, Sherwin. 1976. “A Theory of Life Earnings.” J.P.E. 84, no. 4, pt. 2 (August):
S45–S67.
Sanders, Carl, and Christopher Taber. 2012. “Life-Cycle Wage Growth and Het-
erogeneous Human Capital.” Ann. Rev. Econ. 4:399 – 425.
Topel, Robert. 1991. “Specific Capital, Mobility, and Wages: Wages Rise with Job
Seniority.” J.P.E. 99 (1): 145–76.
The Human Capital Approach to

Intergenerational Mobility
Magne Mogstad
Introduction
In two closely related papers, Becker and Tomes (1979, 1986) develop a
model of the transmission of earnings, assets, and consumption from par-
ents to children. The model is based on utility maximization of parents
concerned about the income or welfare of their children, in contrast to
Thanks to Jorge Luis García, Jim Heckman, John Eric Humphries, Jack Mountjoy, and
Alessandra Voena for helpful comments and suggestions.

contemporaneous empirical and statistical work that was not explicitly
based on a model of maximizing behavior.1
Three decades later, the Becker-Tomes model remains the main build-
ing block of economic research on intergenerational mobility. Not only
did the model help clarify key economic mechanisms that may be pro-
ducing or preventing intergenerational mobility, but it also started a pro-
cess in which new data, empirical evidence, and theoretical models were
brought forward and scrutinized.2 By now, this process has generated a
large and growing literature on intergenerational mobility.
Different strands of the literature have different goals. Some analyses
advance knowledge by addressing measurement challenges or uncover-
ing new facts. Some seek to identify causal impacts of specific interven-
tions or policy changes. Other analyses try to understand the mechanisms
producing or preventing intergenerational mobility. Some studies even
do all three. To give a full picture of the literature on intergenerational
mobility is a daunting task and is not within the scope of this short article.3
Instead, I outline a stripped-down Becker-Tomes model, discuss a few of
its insights, and review some exciting recent advancements.
The Becker-Tomes Model

Becker and Tomes (1979, 1986) present multigenerational models with
one period of childhood, one period of adulthood, one child per family
(no fertility decisions), and a single parent. Parents begin with income
Yt, a combination of earnings and financial transfers they received from
their parents. Parents spend on three items: their own consumption Ct,
investments in the human capital of their child It11 , and financial trans-
fers to their child Xt11 . Parents exogenously transmit ability or endow-
ment At11 to their children through a stochastic linear autoregressive
process. After observing the child’s ability, the human capital of the child
is determined by parental investment in human capital. In adulthood,
labor is supplied inelastically.
Parents care about their own consumption and the income available
for consumption and investment for their children. The optimization
problem of the parent is
max U ðCt , Yt11 Þ (1)

Ct ,It11
1
Goldberger (1989) famously critiqued the Becker-Tomes model for having little added
value relative to statistical, “nonoptimizing” models of intergenerational transmission. Becker
(1989) replies to the criticism, while Mulligan (1999) and Solon (2004) clarify and discuss the
predictions of the Becker-Tomes model.
2
An early example is Behrman, Pollak, and Taubman (1982). While contemporaneous
work on intergenerational mobility appealed to stylized facts in an attempt to rationalize or
test the theories, Behrman et al. estimate a general preference model for analyzing paren-
tal allocations of resources among their children.
3
For a recent review of the literature, see Heckman and Mosso (2014).

subject to
Yt 5 Ct 1 Xt11 1 It11 ,
(2)
Yt11 5 wt11 f ðIt11 , At11 Þ 1 ð1 1 rt11 ÞXt11 1 Ut11 ,
and, possibly, the borrowing constraint
Xt11 ≥ 0, (3)
where wt11 is the return to human capital in period t 1 1, f() is the hu-
man capital production function, r is the return on financial assets, and
Ut11 is the idiosyncratic component of children’s income (market luck).
There are two distinct versions of the Becker-Tomes model. In both ver-
sions, the parent can affect the consumption allocation of the family by
investing in children’s human capital and by leaving bequests. A key dif-
ference between the models is the possibility that credit constraints may
influence parental investment decisions and thereby alter the nature of
intergenerational mobility as compared to a situation without credit con-
straints.
In the 1979 version of the model, the constraint (3) is not imposed, and
intergenerational mobility is driven by persistence in ability and the vari-
ance of labor market shocks. Because this version of the model places
no restrictions on the ability of parents to borrow against the earnings po-
tential of their children, there is no role for parental income (or the mag-
nitude of parental altruism) in determining the optimal level of invest-
ment. No matter their income, parents can borrow freely in the market
to finance the optimal investment level. All parents will therefore choose
to invest the privately efficient amount in children’s human capital so that
the marginal return is driven down to the opportunity cost of investments,
which is the forgone interest on financial investments. As a consequence,
equally able children will receive equal investments independent of their
parent’s income or human capital, and parental influence on intergener-
ational mobility is limited to the heritability of ability. This does not, how-
ever, mean that all children receive the same level of investments. In the
Becker-Tomes model, the specification of f() assumes that the marginal
return from investment in a child’s human capital is positively related to
the endowment he inherits. This assumption implies that children with
greater endowments will receive larger investments, contributing to earn-
ings inequality across families within a generation.
In the 1986 version of the model, the constraint (3) is invoked, restrict-
ing parents from borrowing against the earnings potential of their chil-
dren. This constraint captures the idea that children cannot credibly
commit to repay the loans parents would take on their behalf. If a parent
is credit constrained, the child acquires less human capital, and so the re-

turn to such investments is higher than that on financial assets. In this set-
ting, earnings persist across generations both because ability persists (as
in the 1979 version) and because credit constraints limit human capital
investments. Building on the insights of Becker and Tomes, researchers
have tried to test for and quantify the distortions caused by credit con-
straints. As discussed by Heckman and Mosso (2014), the early literature
did not make a compelling case for the importance of credit constraints
for investments in human capital and intergenerational mobility. How-
ever, there is a now a growing body of evidence—based on more recent
data—supporting the empirical importance of credit constraints in af-
fecting educational attainment (see, e.g., Lochner and Monge-Naranjo
2012, 2015; Hai and Heckman 2017). In particular, the constrained seem
to fall into two groups: those who are permanently poor over their life-
times and a group of well-endowed individuals with rising high levels of
acquired skills who are constrained early in their life cycles.
Beyond the Becker-Tomes Model

Since Becker and Tomes (1979, 1986), analyses of intergenerational
mobility have made important progress by using multiple models and
sources of data in a back and forth in which both models and measure-
ments are augmented as learning evolves. Recently, particular attention
has been devoted to three assumptions of the Becker-Tomes model: (i)
investments at any stage of childhood are equally effective, (ii) earnings
depend on a single skill in the form of human capital, and (iii) parental
engagement with the child is in the form of investment in educational
goods, analogous to firm investments in capital equipment. An active
body of research suggests that these assumptions are at odds with the
data and that the Becker-Tomes model misses key implications of richer
models of intergenerational mobility. Work by Cunha and Heckman
(2007), Heckman and Mosso (2014), and Lee and Seshadri (forthcom-
ing) highlights three ingredients of particular significance for measuring
and understanding intergenerational mobility: multiple periods of child-
hood and adulthood, multiplicity of skills, and several forms of investments.
Multiple Periods
Given the assumption that investments at any stage of childhood are
equally effective, Becker and Tomes can model parental choices of in-
vestment in children through a simple lifetime budget. Therefore, what
matters for parental investments is the lifetime or permanent income,
and not the timing of receipt (or uncertainty) of income over the life cy-
cle. In models with multiple periods of childhood and adulthood, how-

ever, the timing of income can be important as it interacts with restric-
tions on credit markets and the technology of skill formation. Parents
may not be only restricted from borrowing against the earnings potential
of their children (i.e., an intergenerational credit constraint, as in [3]),
but also prevented from borrowing fully against their own future earn-
ings (i.e., an intragenerational credit constraint). The intragenerational
credit constraint induces a suboptimal level of investment (and con-
sumption) in each period in which it binds. How harmful this constraint
will be depends on the technology of skill formation and the life cycle
profile of parental earnings. Cunha, Heckman, and Schennach (2010)
estimate the elasticity of substitution parameters for inputs at different
periods that govern the trade-off of investment between the early years
and the later years. They present evidence of dynamic complementari-
ties in the production of human capital, implying that early investment
in children’s skill development will have large returns because they raise
the payoffs to future investments. As a consequence, if the intragenera-
tional credit constraint is binding during the early periods, late invest-
ments will be lower, even if the parent is not constrained in later periods.
Multiple Skills
An active body of research measures various types of traits or skills and ex-
amines how well they predict or explain various socioeconomic outcomes
(see Borghans et al. [2008] and the references therein). Recently, impor-
tant progress has been made in accounting for measurement error and in
trying to establish causal rather than merely predictive effects. The evi-
dence points to the importance of including sufficiently broad and nu-
anced measures of skills in studies of intergenerational mobility. Both cog-
nitive and noncognitive skills predict adult outcomes, but they have
different relative importance in explaining different outcomes. For exam-
ple, schooling seems to depend more strongly on cognitive skills, whereas
earnings are equally predicted by noncognitive abilities. Importantly for
models of intergenerational mobility, both cognitive and noncognitive
skills can be affected by parental investment (Cunha and Heckman
2007). However, sensitive periods in which investments have particularly
high returns appear to come earlier for cognitive as compared to noncog-
nitive skills.
Many Forms of Investments

The Becker-Tomes model (and many of the extensions of the model) con-
siders only a single child investment good. Recent evidence, however,
points to the importance of allowing for multiple forms of investments
and letting the returns to these investments vary over the life cycle of the

child. For example, Del Boca, Flinn, and Wiswall (2014) develop and esti-
mate a model of intergenerational mobility in which parents make a num-
ber of specific input choices, ranging from various time inputs to child
good expenditures, each with a child age-specific productivity. Their em-
pirical results indicate that both parents’ time inputs are important for
the cognitive development of their children, particularly when the child
is young. In contrast, the productivity of monetary investments in children
has limited impacts on child quality no matter what the stage of develop-
ment.
Concluding Remarks
The human capital approach considers how the productivity of people
in market and nonmarket situations is changed by investments in educa-
tion, skills, and knowledge. The approach was pioneered by scholars as-
sociated with the University of Chicago in the late 1950s and early 1960s,
and many of the seminal contributions were published in the JPE. In
1962, for example, the JPE published a special issue on human capital
with several landmark papers. Nearly two decades later, Becker and
Tomes (1979, 1986) developed the human capital approach into a gen-
eral theory for income inequality, both across families within a genera-
tion and between different generations of the same family. Much of
the progress since, however, has focused on improving measurements,
uncovering new facts, or identifying causal impacts of various determi-
nants of mobility. With some notable exceptions, the JPE and the Univer-
sity of Chicago played a smaller role in this endeavor, possibly because of
preferences for empirical work with tighter links to theory. Recently,
however, important progress has been made by combining theory and
empirics. In particular, Heckman and coauthors have adjusted the theo-
ries of intergenerational mobility in light of new evidence and then taken
those theories to new data sets, getting us closer to fulfilling the goal of
Becker and Tomes (1986, 3) of an “analysis that is adequate to cope with
the many aspects of the rise and fall of families.”
References
Becker, Gary S. 1989. “On the Economics of the Family: Reply to a Skeptic.”
A.E.R. 79 ( June): 514 –18.
Becker, Gary S., and Nigel Tomes. 1979. “An Equilibrium Theory of the Distribu-
tion of Income and Intergenerational Mobility.” J.P.E. 87 (December): 1153 –
89.
———. 1986. “Human Capital and the Rise and Fall of Families.” J. Labor Econ.
4 ( July): 1–39.
Behrman, Jere R., Robert A. Pollak, and Paul Taubman. 1982. “Parental Prefer-
ences and Provision for Progeny.” J.P.E. 90 (February): 52–73.

Borghans, Lex, Angela Lee Duckworth, James J. Heckman, and Bas ter Weel.
2008. “The Economics and Psychology of Personality Traits.” J. Human Re-
sources 43 (4): 972–1059.
Cunha, Flavio, and James J. Heckman. 2007. “The Technology of Skill Forma-
tion.” A.E.R. 97 (2): 31– 47.
Cunha, Flavio, James J. Heckman, and Susanne M. Schennach. 2010. “Estimating
the Technology of Cognitive and Noncognitive Skill Formation.” Econometrica
78 (3): 883 –931.
Del Boca, Daniela, Christopher Flinn, and Matthew Wiswall. 2014. “Household
Choices and Child Development.” Rev. Econ. Studies 81 ( January): 137–85.
Goldberger, Arthur S. 1989. “Economic and Mechanical Models of Intergenera-
tional Transmission.” A.E.R. 79 ( June): 504 –13.
Hai, Rong, and James J. Heckman. 2017. “Inequality in Human Capital and En-
dogenous Credit Constraints.” Rev. Econ. Dynamics 25 (April): 4 –36.
Heckman, James J., and Stefano Mosso. 2014. “The Economics of Human Devel-
opment and Social Mobility.” Ann. Rev. Econ. 6:689 –733.
Lee, Sang Yoon, and Ananth Seshadri. Forthcoming. “On the Intergenerational
Transmission of Economic Status.” J.P.E.
Lochner, L. J., and A. Monge-Naranjo. 2012. “Credit Constraints in Education.”
Ann. Rev. Econ. 4:225–56.
———. 2015. “Student Loans and Repayment: Theory, Evidence and Policy.”
Working Paper no. 20849, NBER, Cambridge, MA.
Mulligan, Casey B. 1999. “Galton versus the Human Capital Approach to Inher-
itance.” J.P.E. 107, no. 6, pt. 2 (December): S184 – S224.
Solon, Gary. 2004. “A Model of Intergenerational Mobility Variation over Time
and Place.” In Generational Income Mobility in North America and Europe, edited
by Miles Corak. Cambridge: Cambridge Univ. Press.
Health Economics: A Selective Historical

Review for the 125th Anniversary of the
Journal of Political Economy
Robert H. Topel
Introduction
Only a small fraction of my own work can be categorized as health eco-
nomics (HE), so the reader might consider my views on the field as those
of a modestly informed outsider who participates on occasion. My taxon-

omy of HE contributions also reflects my background and “Chicago”
tastes. Simplifying more than a bit, I group important HE papers appear-
ing in the JPE into two main areas. The first area treats health as a form of
human capital, with all that entails. Given the place of the University of
Chicago in developing the human capital model, one might expect that
a larger share of HE papers in the JPE would fall in this category than at
other journals, and that is indeed the case.1 One might also expect that I
would find this category the most interesting and important, and that is
also the case. The second main area is the economics of health insur-
ance, including the design of public health insurance schemes and the
effects of insurance on incentives, the demand for medical care, and the
prices and allocation of health care services. I confine most of my discus-
sion to these two areas.2
Because this is the 125th anniversary issue of the JPE, I shall engage in
a bit of intellectual history, following the trail of HE back to the early
1900s, and spend some ink on articles unlikely to appear on modern grad-
uate reading lists. Yet the data indicate that health economics is a fairly
young field. A rough but fairly reliable method of dating its emergence
is to find the first mentions of the word “health” in the title of research
papers published in leading general-interest journals. The JPE began pub-
lication in 1892, but a paper with “health” in its title did not appear for
70 years, with the publication of Selma Mushkin’s “Health as an Invest-
ment” (1962), a landmark (in my view, at least) contribution about which
I will say more below. The same method of carbon dating yields similar re-
sults in other leading journals: the first appearances of “health” in the
Quarterly Journal of Economics and the American Economic Review occurred
in 1952 and 1954, respectively. A broader search for health-related terms
(such as longevity, mortality, morbidity, disease, sickness, medical, medi-
cine, and life span) does not alter my general conclusion that HE was
not an especially active area of scholarship before the 1950s, though, as
discussed below, important work on the desirability and design of univer-
sal health insurance—a very active topic in HE since 1970—appeared in
the JPE as early as 1904.
Of necessity this is a very selective review, with a historical perspective.
I emphasize original contributions in areas that I find interesting and im-
portant, which means that later and perhaps even better contributions are
neglected. I apologize if yours is one of those.
1
An unscientific review of PhD reading lists in health economics indicates that other
general-interest journals also publish more HE papers, broadly defined. This suggests that
the JPE ’s human capital “slant” is greater than simple shares would suggest.
2
Back in graduate school one of my professors explained that there are only two fields
in economics: labor and industrial organization. My taxonomy might be an application of
his more general principle.

Health as Human Capital
Investing in Health Capital
Mushkin’s “Health as an Investment” (1962) appears in a special issue
of the JPE titled “Investment in Human Beings,” edited by T. W. Schultz
and including now-classic papers by Becker, Mincer, Stigler, and Sjaastad.
Becker (2007) describes it as “not particularly insightful,” and he even at-
tributes the slow growth of the health-as-human-capital field to this paper.
I disagree with Gary. Mushkin’s paper is a tour de force that places human
health within the then-nascent theory of human capital.3 Unlike later,
more specialized, contributions in this area, Mushkin’s paper is practi-
cally a survey of what health economists studying the production, main-
tenance, and value of health will work on in later decades, and most of
her admittedly preliminary and nontechnical analysis is spot-on and (I
think) remarkably insightful. Among other topics, she analyzes (1) com-
plementarities between health and other forms of human capital, partic-
ularly education, at both the private and social levels;4 (2) the difficulties
of measuring health-related changes in the quality of life; (3) the contri-
butions of health capital and increasing longevity to economic growth;
(4) the public-good nature of health research and treatment, which mo-
tivates government support of both medical research and the supply of
health care; (5) inherent difficulties in identifying the social returns to
health research as an input to policy evaluation; and (6) methods of as-
signing a monetary value to health improvements.
Mushkin’s discussion of methods for valuing health improvements is a
bit primitive and even contradictory. Like many others before and after,
she attempts to value improved health in terms of the increased labor
supply and production or earnings that come with living longer or health-
ier, whereas the economically correct approach would measure the value
in terms of individuals’ willingness to pay for health improvements—
which embeds any value derived from productivity and the ability to work.
For example, her approach assigns zero value to reductions in mortality
or morbidity that affect only the retired population, yet these individuals
would nevertheless enjoy the benefits of living longer and better. That is
valuable. In fact, this is the margin where health gains in advanced econ-
omies have been concentrated since about 1960: older individuals are liv-
ing longer and better (Murphy and Topel 2006). Mushkin appears to
3
She cites Schultz (1961), but JPE papers by Schultz (1960) and Mincer (1958) are
equally relevant.
4
“A lengthening of life expectancy through improved health reduces the rate of depre-
ciation of investment in education and increases the return to it. . . . Improved education,
on the other hand, increases the return on a lifesaving investment in health. . . . Educational
levels determine to a large extent the seeking out of health services and the selection of ap-
propriate kinds of services” (131).

have recognized the issue, however, and in another part of her paper she
offers this prescient econometric strategy for valuing debilitating health
risks and lost lives: “If workmen’s compensation charges . . . can be used
to measure the risk of death and disability [by occupation or industry],
it may be possible to use [wage] premiums paid for extra occupational
risks—or ‘hazard pay’—as an index of the market evaluation of the risk
of continuing debility. . . . Injury rates become one of the many factors
considered in wage negotiation even when separate hazard premiums
are not paid” (135). This is a concise description of methods that would
be applied many years later, and are still widely used today, to estimate the
“value of a statistical life,” which became a key tool for government cost-
benefit analyses of policies or programs that save lives.5
Human capital research accelerated in the 1960s, but beyond noting
that health could be viewed as a form of human capital (e.g., Becker 1964),
a full decade passed before an important paper followed Mushkin’s early
lead. Grossman (1972) is the first formal treatment of health as a form
of capital and an object of an individual’s life cycle choices, so that both
health and longevity are endogenous. It is the most-cited health econom-
ics paper to have appeared in the JPE, so it is worth a somewhat closer
look.
There is no uncertainty in Grossman’s setup: he assumes perfect fore-
sight. The stock of human capital at each age, Ha, produces an output of
“healthy time” that enters life cycle utility. The stock follows a standard
law of motion for a durable good, Ha11 2 Ha 5 Ia 2 da Ha , where Ia and
da are gross investment in health and depreciation, respectively. Longev-
ity is endogenous because death occurs when the stock of human capital
falls below some exogenous level, H . Gross investment is produced with
purchased medical care and the stocks of health and other human capi-
tal, so the demand for medical care is not a form of consumption but is
derived from the more primitive demand for health itself. With assumed
constant returns in goods and time inputs, and thus constant marginal
cost of investment, death can occur only if depreciation rises with age,
which Grossman assumes. This yields additional predictions for the life
cycle patterns of purchased medical care and gross investment, as people
attempt to offset rising depreciation. Gross investment and the demand
for medical care also rise with age if the elasticity of the marginal efficiency
of capital is smaller than unity. Production of health capital and consump-
tion of health rise with the wage because health time is more valuable,
making health a normal good. Grossman also builds in complementarity
between education, or other forms of human capital, and the production
of health, so more educated individuals are also endogenously healthier.
5
The literature is reviewed by Viscusi and Aldy (2003).

Grossman’s (1972) paper stimulated a substantial literature treating
health as a form of human capital, much of which appeared in the JPE.6
But until recently research in this area remained a tiny fraction of the
much larger human capital literature, which has mainly focused on edu-
cation and training. Becker (2007) is a sweeping attempt to rectify this
and also a useful summary of developments in the area.
Valuing Medical Advances and the Measurement

of Health Capital
In 1900, 18 percent of newborn males in the United States died before
their first birthday. By the end of the century it took until age 63 to reach
the same level of cohort mortality. Early in the twentieth century reduced
mortality rates were concentrated among the young, as infant mortality
plummeted and a variety of infectious diseases affecting children were
controlled or eradicated. Later gains were mainly concentrated at older
ages as death rates from age-related diseases—especially cardiovascular
diseases—declined after 1970.7 Overall morbidity also declined, so that
not only were people living longer, they were living better.
What drove these health improvements, and what are they worth? What
would be the social value of further gains, and what might those gains
cost? These questions are important for several reasons. First, improved
health is an obvious contributor to human welfare, yet health improve-
ments are not measured in traditional growth and welfare metrics such
as per capita income. The uncounted gains can be large, so traditional mea-
sures can be very misleading. Second, basic medical research is a public
good: a new idea or treatment can be extended across the current popu-
lation and to future generations as well. This is one of the most powerful
arguments for public support of medical research. In the United States,
federal government support for medical research currently totals about
$35 billion annually, almost all of which flows through the National In-
stitutes of Health. As a matter of public policy, are the medical advances
and improved health flowing from this research worth their cost? The an-
swer requires that we connect medical research to treatments and health
outcomes and that we properly value any changes in mortality and mor-
bidity.
Weisbrod (1971) is the first attempt of which I am aware to measure
the costs and benefits of medical research and to perform a cost-benefit
6
See, e.g., Ben-Porath (1976), Cropper (1977), Rosenzweig and Schultz (1983), Wolfe
(1985), Ehrlich and Chuma (1990), Ehrlich and Lui (1991), Kenkel (1991), Sah (1991),
Philipson and Becker (1998), and Acemoglu and Johnson (2007).
7
See Murphy and Topel (2006) for details.

analysis. Like many papers that followed, this is a case study of a partic-
ular disease (polio) for which (1) reasonable historical data on the flow of
research support are available; (2) a preventative treatment (vaccination)
was successfully discovered and widely applied, effectively eradicating the
disease in the United States; and (3) once discovered and perfected, the
marginal cost of treatment is fairly low.8 Thus the connection between re-
search, treatment, and outcomes is fairly clear. Weisbrod measures bene-
fits in terms of increased productivity and reduced costs of care that would
have been otherwise necessary. He recognizes that these measures are in-
adequate but at least conservative. He also recognizes that the benefit of
a polio cure affected not just the current generation but all future ones
as well, so future generations entered his calculations using various dis-
count rates. Applying his model to the United States alone (other popula-
tions also gained because of the public-good character of new knowledge),
he estimates an internal rate of return on polio research of 11 percent. It
turned out to be a good investment, at least ex post.
Case studies such as this are important, and Weisbrod makes a con-
vincing argument that his estimates are conservative, conditional on the
successful research outcome that occurred and his assumption that re-
search support caused it.9 But even a retrospective assessment of the over-
all value of medical research requires that we integrate over successes and
failures and that we have a more convincing method of valuing benefits
and costs than Weisbrod was able to apply, including the widely varying
costs of treating heterogeneous diseases. My paper with Kevin Murphy
(Murphy and Topel 2006) makes some initial progress on these questions,
offering evidence that the past and potential future gains from medical
research may be quite large.
An important contribution of our paper is to recognize that the pri-
vate economic value of health is determined by willingness to pay. With
diminishing marginal utility of full consumption, which includes the value
of leisure, life extension is valued because average utility exceeds mar-
ginal utility—consumption yields a flow of consumer surpluses over the
life cycle—so the ability to spread consumption over more life years is val-
ued (see also Usher 1973; Rosen 1988). This yields a formal expression
for the value of a statistical life, which other research put at about $6 mil-
lion at the time we wrote. We allocate that total to create a life cycle pat-
8
Cutler and Kadiyala (2003) and Cutler, Rosen, and Vijan (2006) provide overviews of
later literature. Chandra and Staiger (2007) study productivity in the treatment of heart at-
tacks, explaining geographic differences in the use of intensive medical care.
9
Even here, the attribution of costs and benefits to a particular research program is
problematic. Weisbrod points out that Watson and Crick’s work on DNA was partially sup-
ported by grants to study polio, so spillovers are important. And who knows what other
grants polio researchers received.

tern of values for life years that declines at older ages. We then use these
values to estimate the gains from past and potential future reductions in
age-specific mortality rates from various diseases and overall.
The numbers are huge (see also Murphy and Topel 2003; Nordhaus
2003). Net of medical expenditures, the value of increased life expectancy
between 1970 and 2000 in the United States totaled roughly $60 trillion,
or a flow of about $2 trillion per year. Compare this to an average annual
value of GDP of about $6 trillion over this period and the importance of
improvements in health capital as a contributor to welfare is obvious.
Becker (2007) extends these calculations to all OECD countries and es-
timates benefits that are triple the values for the United States (see also
Becker, Philipson, and Soares 2005). Murphy and Topel (2006) also ex-
tend the analysis to measure the prospective gain (net of treatment costs
and other possible distortions) to future generations from, say, a 10 per-
cent reduction in cancer mortality. We find that a 10 percent reduction
in cancer mortality would be worth about $500 billion to current and fu-
ture US generations. Of course there are a number of caveats to these es-
timates, which we attempt to cover in some detail in the paper. Some are
noted in the last section of this piece.
Health Insurance and Markets for Medical Care

On the basis of the flow of articles published, I claimed earlier that HE
was not an active field of economic research until at least the 1950s. The
extreme outlier is a series of papers on various types of social insurance
by Isaac Rubinow that appeared in the JPE beginning in 1904. Rubinow
(1875–1936) might lay claim to being the original health economist, and
the JPE was his favorite outlet; he had 11 papers in the journal between
1904 and 1930.10 A Russian immigrant, he earned his doctor of medicine
degree from New York University and began his professional life treating
poor immigrants in New York City. This experience alerted him to the
complex interaction of poverty and health, and he concluded that he
could have greater impact as an economist (it’s true) than as a physician.
So he earned his degree at Columbia in 1903 and began a long econom-
ics career in government and nonprofit organizations. His “Labor Insur-
ance” (1904) was “the first publication by an American to call unequivo-
cally for social insurance” (Kreader 1988, 56) and the first that I know
of that explicitly argued for compulsory participation in government-
subsidized health insurance; the intellectual debate over the “individual
10
Health and social insurance were not his only topics. His JPE contributions also cov-
ered unemployment (1917) and even retail coupons and discounting (1905).

mandate” (as we call it today) was heated even then, and the issues were
precisely the same. Foreshadowing the reasoning of the US Supreme Court
(2012) concerning the individual mandate in the Affordable Care Act,
Rubinow argued that compulsory participation is not forced consump-
tion; it is simply a form of “tax” that the government is empowered to levy.
He later greatly expanded these ideas in “Standards of Sickness Insur-
ance” (1915). This paper is a nontechnical exercise in policy design, using
the experiences of European health care programs to advocate for uni-
versal mandated coverage in the United States. Agree with him or not,
Rubinow’s work was way ahead of his time and his economic reasoning
was not bad.
I am sure that Rubinow would have been very surprised at how things
played out over the next 100 years.
Health insurance and the demand for health care did not materially
reemerge in the literature until Arrow’s influential 1963 paper in the
AER. Publications from this strand did not appear in the JPE until the
1970s, and one is struck by the changes in style and methodology of eco-
nomic research. The best of the lot is a classic paper by Feldstein (1973),
who models and estimates the distortions and welfare implications of ex-
isting health insurance plans.11 In modern parlance the paper might be
described as “structural industrial organization,” except that it is penetra-
ble. The key idea is this. Virtually by definition, insurance plans include
coinsurance rates such that the marginal price of medical care for an in-
sured person is lower than the marginal cost of supplying that care. Then
insurance increases the demand for both the quantity and quality of
health care, which further increases the demand for insurance. With a ris-
ing supply price of medical services, this (stable) feedback process raises
prices and allows suppliers to earn rents, while buyers of health insurance
are overinsured. Feldstein formulates and estimates a structural model
of health care demand and this dynamic process and then calibrates the
welfare gains of raising the average coinsurance rate from 0.33 to 0.5 or
higher. Because raising the coinsurance rate reduces the demand for
medical care, equilibrium prices fall, which provides an offsetting benefit
to consumers. The net welfare gain turns out to be positive and large:
Feldstein estimates the gain from such reforms to be about $4 billion
per year in early 1970s dollars. It is a very nice piece that foreshadows
many others; the most recent example in the JPE is a fine piece by Clem-
ens and Gottlieb (2017).
11
Much of the health insurance literature seems to ebb and flow with public policy in-
terests in creating national health insurance, perhaps driven by the supply of research
funds. Several papers in the 1970s came out of the RAND health insurance project; see
Mitchell and Phelps (1976).

What Can We Learn?
Of necessity this has been a highly selective review, in part because of my
emphasis on the historical development of HE in one general journal and
in part because of my own interests. I close with some comments on what I
believe to be a very important area of research that is connected to both
of the areas I discussed above: health as human capital and the role of pub-
lic and private insurance schemes in allocating health care resources.
My work with Murphy suggests that the potential value of medical in-
novations is vastly larger than the flow of government-supported medical
research, so the public-good case for greater research funding is some-
what compelling. But an important caveat is that the costs of research
might be very small compared to the costs of treatments that are discov-
ered, and these treatment costs could offset even the large potential sur-
plus we calculate. Part of the reason has to do with insurance providers as
downstream allocators of health care resources and technologies. In a
nutshell, health care allocation via third-party payers, combined with con-
sumers’ desire for whatever care might help, creates an “invent it and they
will use it” waste in the allocation of health care resources. Back upstream
at the level of research, this distortion creates incentives for inefficiently
costly health care advances. Going further, even technologies that would
yield positive social surplus in a world of efficient downstream allocation
can be a net social waste because the new technology is vastly overused.
In these circumstances the major risk in medical research might not be
the risk of failure to develop a new treatment after large up-front invest-
ment, but instead the risk of unaffordable success. This means that health
care markets and the value of research are complements: greater efficiency
in downstream allocation raises the value of medical research. Future re-
search improving downstream allocations can therefore yield a sort of
double dividend, so that is one area where I think substantial effort is war-
ranted. Similarly, it is important to know the extent of the upstream distor-
tion in research incentives, which I think is an even more challenging
problem.
References
Acemoglu, Daron, and Simon Johnson. 2007. “Disease and Development: The
Effect of Life Expectancy on Economic Growth.” J.P.E. 115 (6): 925–85.
Arrow, Kenneth. 1963. “Uncertainty and the Welfare Economics of Medical
Care.” A.E.R . 53 (5): 941–73.
Becker, Gary S. 1964. Human Capital: A Theoretical and Empirical Analysis, with Spe-
cial Reference to Education. New York: NBER.
———. 2007. “Health as Human Capital: Synthesis and Extensions.” Oxford Econ.
Papers 59:379–410.
Becker, Gary S., Tomas J. Philipson, and Rodrigo R. Soares. 2005. “The Quan-
tity and Quality of Life and the Evolution of World Inequality.” A.E.R . 95:
277–91.

Ben-Porath, Yoram. 1976. “Fertility Response to Child Mortality: Micro Data from
Israel.” J.P.E. 84, no. 4, pt. 2 (August): S163 –S178.
Chandra, Amitabh, and Douglas O. Staiger. 2007. “Productivity Spillovers in
Health Care: Evidence from the Treatment of Heart Attacks.” J.P.E. 115 (1):
103 – 40.
Clemens, Jeffrey, and Joshua Gottlieb. 2017. “In the Shadow of a Giant: Medi-
care’s Influence on Private Physician Payments.” J.P.E. 125 (1): 1–39.
Cropper, M. L. 1977. “Health, Investment in Health, and Occupational Choice.”
J.P.E. 85 (December): 1273–94.
Cutler, David M., and S. Kadiyala. 2003. “The Return to Biomedical Research:
Treatment and Behavioral Effects.” In Measuring the Gains from Medical Re-
search: An Economic Approach, edited by Kevin M. Murphy and Robert H. Topel.
Cutler, David M., Allison B. Rosen, and Sandeep Vijan. 2006. “Value of Medical
Innovation in the United States: 1960–2000.” New England J. Medicine 355 (9):
920–27.
Ehrlich, Isaac, and Hiroyuki Chuma. 1990. “A Model of the Demand for Longev-
ity and the Value of Life Extension.” J.P.E. 98 (4): 761–82.
Ehrlich, Isaac, and Francis T. Lui. 1991. “Intergenerational Trade, Longevity, and
Economic Growth.” J.P.E. 99 (5): 1029 –59.
Feldstein, Martin. 1973. “The Welfare Loss of Excess Health Insurance.” J.P.E. 81,
no. 2, pt. 1 (March/April): 251–80.
Grossman, Michael. 1972. “On the Concept of Health Capital and the Demand
for Health.” J.P.E. 80 (March/April): 223 –55.
Kenkel, Donald S. 1991. “Health Behavior, Health Knowledge, and Schooling.”
J.P.E. 99 (2): 287–305.
Kreader, J. L. 1988. “America’s Prophet for Social Security: A Biography of Isaac
Max Rubinow.” PhD diss., Univ. Chicago.
Mincer, Jacob. 1958. “Investment in Human Capital and the Personal Distribu-
tion of Income.” J.P.E. 66 (August): 281–302.
Mitchell, Bridger M., and Charles E. Phelps. 1976. “National Health Insurance:
Some Costs and Effects of Mandated Employee Coverage.” J.P.E. 84 ( June):
553–72.
Murphy, Kevin M., and Robert H. Topel. 2003. “The Economic Value of Medical
Research.” In Measuring the Gains from Medical Research: An Economic Approach,
edited by Kevin M. Murphy and Robert H. Topel. Chicago: Univ. Chicago Press.
———. 2006. “The Value of Health and Longevity.” J.P.E. 114 (5): 871–904.
Mushkin, Selma J. 1962. “Health as an Investment.” J.P.E. 70, no. 5, pt. 2 (Octo-
ber): 129 –57.
Nordhaus, William. 2003. “The Health of Nations: The Contribution of Im-
proved Health to Living Standards.” In Measuring the Gains from Medical Re-
search: An Economic Approach, edited by Kevin M. Murphy and Robert H. Topel.
Philipson, Tomas J., and Gary S. Becker. 1998. “Old-Age Longevity and Mortality-
Contingent Claims.” J.P.E. 106 (3): 551–73.
Rosen, Sherwin. 1988. “The Value of Changes in Life Expectancy.” J. Risk and Un-
certainty 1:285 –304.
Rosenzweig, Mark R., and T. Paul Schultz. 1983. “Estimating a Household Pro-
duction Function: Heterogeneity, the Demand for Health Inputs, and Their
Effects on Birth Weight.” J.P.E. 91 (5): 723 – 46.
Rubinow, I. M. 1904. “Labor Insurance.” J.P.E. 12 ( June): 362–81.
———. 1905. “Premiums in Retail Trade.” J.P.E. 13 (September): 574–86.
———. 1915. “Standards of Sickness Insurance: I.” J.P.E. 23 (March): 221–51.

———. 1917. “Medical Benefits under Workmen’s Compensation: I.” J.P.E. 25
(6): 580 – 620.
Sah, Raaj K. 1991. “The Effects of Child Mortality Changes on Fertility Choice
and Parental Welfare.” J.P.E. 99 (3): 582–606.
Schultz, T. W. 1960. “Capital Formation by Education.” J.P.E. 68 (December):
571–83.
———. 1961. “Investment in Human Capital.” A.E.R. 51 (March): 1–17.
Usher, D. 1973. “An Imputation of the Measure of Economic Growth for Changes
in Life Expectancy.” In The Measurement of Economic and Social Performance,
edited by Milton Moss, 193–225. Conference on Research in Income and
Wealth, Studies in Income and Wealth, vol. 38. New York: Columbia Univ. Press
(for NBER).
Viscusi, W. K., and J. E. Aldy. 2003. “The Value of a Statistical Life: A Critical Re-
view of Market Estimates throughout the World.” J. Risk and Uncertainty 27:5–
76.
Weisbrod, Burton A. 1971. “Costs and Benefits of Medical Research: A Case
Study of Poliomyelitis.” J.P.E. 79 (3): 527– 44.
Wolfe, John R. 1985. “A Model of Declining Health and Retirement.” J.P.E. 93
(6): 1258–67.
Agency Issues
Canice Prendergast
Modern agency theory begins with Mirrlees (1976) and Holmstrom (1979)
with a general prescription for how compensation can be used to alleviate
agency issues. It proposes that an agent’s pay should vary whenever there
is information about her effort and that all information on performance
should be used. Furthermore, the ability to resolve agency problems is
limited only by risk-sharing considerations associated with random varia-
tion in performance measures. With some parametric restrictions, this also
has the implication that greater randomness (uncertainty) reduces the in-
tensity between pay and performance measures.
This work has been appropriately feted as the root from which modern
agency theory derives. It does, however, suffer from one important prob-
lem: most people do not get paid this way. Instead, the current pay of
most workers is insensitive to pretty much anything, relevant information
is consciously not used, and the relationship between uncertainty and the

intensity of pay for performance if anything goes the wrong way. Because
of this, the literature published in the journal has instead originated a se-
ries of avenues that—while respecting the logic of these contributions—
at times better accord with observed practices. These avenues largely fit
into two categories: (i) there are other ways to skin the cat of motivating
workers, and (ii) in many settings, some of the assumptions of the canon-
ical model do not hold. I consider each in turn.
Other Ways of Motivating

Workers are often embedded in firms, in markets, and in long-lasting re-
lationships. Each of these issues has been explored to offer alternatives
to the pay for performance logic of the canonical model.
First, most of us work in organizations that are hierarchical, with work-
ers sorted into positions on the basis of performance and ability. Given
this, it should not be surprising that an alternative source of motivation
for workers is the possibility of promotion. The earliest and most impor-
tant contribution here is Lazear and Rosen (1980), which argues for the
role of tournaments in providing incentives. Tournaments are settings in
which a group of agents compete for a set of fixed prizes rather than face
a pay schedule that varies explicitly with their actions. Agents exert effort
to change the probability of getting a better prize. In contrast to an indi-
vidual piece rate setting, here what matters is relative performance in a
particularly stark way, as only rank order performance determines pay.
In simple settings, Lazear and Rosen show how tournaments can induce
the first-best outcome without explicit pay for performance and offer
some results on when tournaments dominate piece rates based on only
individual performance. This logic has subsequently been extended by
Green and Stokey (1983) and Malcomson (1984).
One difficulty that agency theory has faced has been the absence of clean
empirical testing. Tournament theory is a welcome exception. There are
two natural tests. First, do bigger prizes cause agents to work harder? In a
number of (mostly sports) settings, this has been shown to be so. The sec-
ond test—do perceived marginal probabilities of winning affect effort?—
is both more subtle and more directly related to the notion that agents
compete in probabilities of winning. The strongest work in that vein is
Brown (2011), which showed that the presence of Tiger Woods in golfing
tournaments acted as a disincentive for other golfers. They perform worse
as they perceived the marginal value of effort to be lower.
Second, workers are embedded not only in firms but also in markets.
Another theme that agency theory has explored successfully over the last
three decades is how labor markets can act as an alternative to contracts.
This logic derives from Fama (1980), which described how external audi-
ences can constrain opportunities for moral hazard through the process

by which outside opportunities evolve. A simple example would be a pro-
fessional baseball player. Despite the plethora of performance measures
on his performance, pay for performance is rare. The reason, of course, is
that there is an external market that sets the market pay of baseball play-
ers and a player who shirks can expect his reputation and future oppor-
tunities to worsen.
Fama does not claim that markets always offer enough discipline to
solve agency problems. This issue has been elegantly formalized by Holm-
strom (1999), and these two papers have now helped to develop a sub-
field of agency theory known as “career concerns.” A more general char-
acteristic of the orientation of the JPE is a focus on theoretically informed
empirical testing. A particularly notable example of this in this field is the
paper by Gibbons and Murphy (1992). They study the interaction be-
tween career concerns and formal pay for performance in the market for
CEOs. Using a tightly specified model of career concerns interacting with
formal pay for performance, they show how greater career concerns in-
centives reduce the need for formal pay for performance, as arises in
their empirical results.
Finally, workers are located not only in hierarchies and markets but of-
ten in long-term employment relationships. This allows temporal consid-
erations to enter optimal contracting, where long-term employment re-
lations allow the possible use of deferred compensation. This is where an
agent may not be rewarded for current performance today, but rather
sometime in the future. There are by now many dynamic contracting pa-
pers whose optimal outcome is to hold back some component of pay un-
til late in the agent’s career. The logic is usually simple: by deferring a
performance-related bonus until a worker is older, incentives for older
workers improve while maintaining incentives for the younger worker
(because by working hard now, she can be in line for that bonus later).
One of the earliest formalizations of this logic is the study by Lazear
(1979). His interest in that work is the need for mandatory retirement
in optimal employment relationships. However, the reason why manda-
tory retirement is needed is that older workers are overpaid relative to
their contemporaneous marginal productivity, which itself derives from
the desire to defer compensation for the reason above.
The Assumptions Do Not Hold

Another series of extensions to the canonical model have arisen from
the realization that many of its assumptions do not hold in many impor-
tant settings. Here I provide a number of important examples.
The canonical model assumes that more pay for performance changes
behavior only in ways that benefit the principal. However, by now it is
well known that incentive pay can also induce the kind of dysfunctional

behavioral responses that cause pay for performance to backfire. Cheva-
lier and Ellison (1997) offer a nice example. This line of research has
become known as multitasking and typically mutes the use of perfor-
mance pay. An elegant and tractable example of this is Baker (1992).
Other early contributions in the JPE on dysfunctional responses have
focused on another characteristic of organizations, namely, the use of
rules to allocate resources over allowing discretion. A series of papers have
related this to dysfunctional behavioral responses. Consider the motivat-
ing example in Milgrom’s (1988) important contribution on what he terms
“influence activities.” American Airlines needs to staff routes, and flight
attendants have preferences over which routes they are given. Rather
than allowing supervisors to assign “shifts” on the basis of the idiosyn-
cratic preferences of employees, it uses a much simpler rule: routes
are assigned by seniority, where workers with the most seniority pick first.
(As a more substantive example, many firms use “last in, first out” rules
for layoffs, despite the fact that some more senior workers could be
less productive for the firm than their junior counterparts.) This arises
in Milgrom’s work as a way of deterring dysfunctional lobbying behavior
by workers, where time is spent influencing superiors for resources for
themselves rather than spending time on more productive activities. Said
another way, while such bureaucratic rules may be inefficient at the point
at which the decision is made, it may save sufficient resources at an earlier
point to be worthwhile. Similar logic underlies Prendergast and Topel
(1996) in a setting in which favoritism arises.
Public agencies are often accused of being unaccountable to their con-
stituents. Foremost among these is the behavior of police forces. Another
example of how ex post inefficient rules can be part of optimal oversight
arises in Prendergast (2003). In that setting, legitimate consumer com-
plaints are ignored. This is done because if public agents believe that
their behavior is likely to be investigated on the basis of such complaints,
they will simply capitulate to those consumers in settings in which they
should not. (In the police example, they become more resistant to arrest-
ing suspects.) Once again, the response to dysfunctional behavioral ac-
tions is to ignore valuable information.
Probably the most important parallel exploration to agency contract-
ing has been to understand governance in settings in which outcomes
are noncontractible. There have been two lines of inquiry. The first has
derived from the seminal work of Grossman and Hart (1986), where own-
ership of assets (or, in later work, the control of assets) can be used to mit-
igate agency concerns. This has led to the modern theory of the firm.
This contribution and its extensions are described elsewhere in this issue
by Rob Vishny and Luigi Zingales. The second area dealing with non-
contractibility has addressed relational contracts, where repeated inter-
actions can potentially resolve noncontractibility issues.

Work on noncontractible environments has also led to consideration
of other instruments than compensation to align incentives. Contribu-
tions to the JPE have played a central role in this. An important example
is Aghion and Tirole’s (1997) work on real and formal authority. Con-
sider a setting in which a worker cares not just about current pay and effort
but also about the kinds of activities that they engage in. Furthermore, by
exerting effort, she can identify which activity she prefers. Yet she will exert
effort only if that preferred outcome is likely to be implemented. The prob-
lem for her is that her boss may not agree with her and instead overturns
her recommendation. The principal will overturn, though, only if he is suf-
ficiently sure of the right action. In settings in which he is not sure, the
agent has real authority even though the agent may be subject to the for-
mal authority of her boss. However, if the fear of being overturned is suffi-
ciently salient to the agent, she will not exert any effort. In these settings,
the firm may delegate formal authority to the agent. This work has been
influential not just for its elegant and tractable modeling but also as the
picture it paints—of organizations characterized by conflicts, with individ-
uals vying for control—resonates with reality.
A feature of many institutions is conflict. Recent work in noncon-
tractible settings has focused on the value of using workers who do not
share the beliefs and preferences of their superiors. An early example
is Che and Kartik (2009). The authors consider a setting in which an ex-
pert collects information to determine the right course of action, but she
may be biased for or against that action (compared to the beliefs of her
principal). Suppose that the principal could choose the bias of the agent:
should she share the preferences of the principal? The intuitive sugges-
tion that their beliefs should be aligned turns out not to be right. The rea-
son is that an agent who shares the principal’s belief realizes that if she
exerts no effort, the principal is likely to do what the agent already thought
was the right answer. Instead, the optimal strategy is to introduce disagree-
ment between agent and principal, because an agent who believes that the
principal’s prior is wrong is more likely to work hard to dissuade him. Yet
this is not costless, as she is less likely to reveal her information clearly.
Many workers, arguably most, are not rewarded on output measures.
Instead their inputs are monitored, where they follow instructions pro-
vided by their superiors. Showing up on time and doing what is asked
of them is the reality for most workers. Imagining input monitoring as
an alternative to pay for performance has also helped to make progress
on understanding one of the empirical difficulties faced by the litera-
ture. Specifically, there is little empirical support for what has become
known as the trade-off between risk and incentives, where more uncer-
tain environments would result in less pay for performance. Instead, the
evidence seems more supportive of greater uncertainty leading to more

pay for performance. Prendergast (2002) addresses this by noting that
for many workers, the alternative to pay for performance is a situation
in which a superior tells a worker what to do. Now consider a setting with
more uncertainty. In the canonical model, this only adds measurement
noise to the ability to infer agent effort. This alone would attenuate pay
for performance. However, in more uncertain settings, a superior may
now additionally be less able to tell the agent what to do (as he knows less
about what is going on). If so, the principal may need the agent to decide
the right course of action in these uncertain settings, which likely leads to
more pay for performance. Said another way, uncertainty may indeed
render pay for performance costly; it may render the alternative even
worse.
When a CEO increases the earnings of a company or a sales agent sells
more, we can be pretty sure that this is a good thing. A final area of explo-
ration in the field has been to consider settings in which it is not clear
what output means. For example, when an auto mechanic tells you to
have your car repaired at some expense, it is unclear if this reflects good
or bad performance by the mechanic. Taylor (1995) studies this problem.
A beautiful example of this inability to interpret performance measures
is the work of Dewatripont and Tirole (1999) on advocates. Once again,
consider a setting in which information needs to be collected on whether
to carry out an action. The canonical agency model relies on a monotone
likelihood ratio property (MLRP), where if it holds, pay will strictly in-
crease in output (or profits). The difficulty here is that information
can be positive or negative and can be offsetting. Specifically, an agent
who has worked hard and found one piece of positive information and
one piece of negative information finds herself in the same position as
one who collects no information. Formally, this means that the MLRP
of the canonical model fails, and it can render it impossible to have
one person collect all the information. The alternative is advocacy: where
one person collects only information that is positive and the other col-
lects only the negative information, and they are rewarded if the outcome
reflects the kind of information they collect. As such advocacy is so per-
vasive inside and outside organizations, this paper offers an insight far
from the canonical model, but does so with a minimum of additional as-
sumptions.
To conclude, it should be clear from this short essay that the literature
on agency has come far from its original focus on the shape of compen-
sation functions, and the central role played by the JPE in that develop-
ment. It is perhaps worthwhile to conclude by noting the paucity of em-
pirical work among the contributions above. Fields can thrive only when
numbers are added to the Greek alphabet, and it is hoped that the JPE
can play a significant role in promoting such work going forward.

References
Aghion, Philippe, and Jean Tirole. 1997. “Formal and Real Authority in Organi-
zations.” J.P.E. 105:1–29.
Baker, George. 1992. “Incentive Contracts and Performance Measurement.”
J.P.E. 100:598–614.
Brown, Jennifer. 2011. “Quitters Never Win: The (Adverse) Incentive Effects of
Competing with Superstars.” J.P.E. 119 (5): 982–1013.
Che, Yeon-Koo, and Navin Kartik. 2009. “Opinions as Incentives.” J.P.E. 117 (5):
815–60.
Chevalier, Judith, and Glen Ellison. 1997. “Risk Taking by Mutual Funds as a Re-
sponse to Incentives.” J.P.E. 105:1167–1201.
Dewatripont, Matthias, and Jean Tirole. 1999. “Advocates.” J.P.E. 107 (1): 1–39.
Fama, Eugene. 1980. “Agency Problems and the Theory of the Firm.” J.P.E.
88:288–307.
Gibbons, Robert, and Kevin J. Murphy. 1992. “Optimal Incentive Contracts in
the Presence of Career Concerns.” J.P.E. 100:468–506.
Green, Jerry, and Nancy Stokey. 1983. “A Comparison of Tournaments and Con-
tracts.” J.P.E. 91:349 – 64.
Grossman, Sanford, and Oliver Hart. 1986. “The Costs and Benefits of Owner-
ship: A Theory of Vertical and Lateral Organization.” J.P.E. 94:691–719.
Holmstrom, Bengt. 1979. “Moral Hazard and Observability.” Bell J. Econ. 10:74–91.
———. 1999. “Managerial Incentive Problems: A Dynamic Perspective.” Rev.
Econ. Studies 66 (1): 169– 82.
Lazear, Edward. 1979. “Why Is There Mandatory Retirement?” J.P.E. 87 (6): 1261–
84.
Lazear, Edward, and Sherwin Rosen. 1980. “Rank Order Tournaments as Opti-
mal Labor Contracts.” J.P.E. 88 (5): 841–64.
Malcomson, James. 1984. “Work Incentives, Hierarchy, and Internal Labor Mar-
kets.” J.P.E. 92 (3): 486–507.
Milgrom, Paul. 1988. “Employment Contracts, Influence Activity, and Efficient
Organization.” J.P.E. 96:42– 60.
Mirrlees, James. 1976. “The Optimal Structure of Incentives and Authority within
an Organization.” Bell J. Econ. 7 (1): 105–31.
Prendergast, Canice. 2002. “The Tenuous Trade-off between Risk and Incen-
tives.” J.P.E. 110 (5): 1071–1102.
———. 2003. “The Limits of Bureaucratic Efficiency.” J.P.E. 111 (5): 929 – 59.
Prendergast, Canice, and Robert Topel. 1996. “Favoritism in Organizations.”
J.P.E. 104 (5): 958 –78.
Taylor, Curtis. 1995. “The Economics of Breakdowns, Checkups, and Cures.”
J.P.E. 103:53–74.

Information Economics
Emir Kamenica
When George Stigler published “The Economics of Information” in this

journal a little over 50 years ago (1961), he was justified in his complaints
about the absence of research on the topic. Yet, this complaint was soon
to become obsolete. Figure 1 reports the incidence of the phrases “infor-
mation economics” and “economics of information” normalized by the
use of the word “economics” in books published from 1900 to 2008. The
figure clearly shows that Stigler’s article was a harbinger of a new field
of inquiry.1 The field grew quickly for the next 25 years, reached a plateau
that lasted from the mid-1980s until the late 1990s, and then experienced
another growth spurt. During the last decade, 13 percent of the articles
published in the J PE had the word “information” in their abstract.
I. Information Acquisition
One early strand of the literature developed search models in which individ-
uals incur costs to acquire information for private use. Stigler’s (1961)
model, in which agents decide ex ante on the number of alternatives to
sample, was replaced by a sequential search formulation (McCall 1970).
This formulation provided sharp and lasting intuitions about individual
motives for acquiring additional information given the distribution of op-
tions but was not well suited to explain how such distributions arise in the
first place (Diamond 1971). This line of research eventually grew into the
matching-and-bargaining and directed-search models that are now widely
used in labor macroeconomics but have little direct contact with the rest
of information economics.
In other parts of macroeconomics, agents’ information is now often en-
dogenized through rational inattention models (Sims 2003) that postulate
that the cost of information acquisition is proportional to the reduction
in Shannon entropy.2 While the rational inattention approach may seem
1
Of course, one can almost always push intellectual origins of any field further into the
past. Hayek (1945) is an important early reference in information economics. It is likely
responsible for the 1-year blip in use of the phrases in 1946 visible in fig. 1.
2
Matejka and McKay (2015) show that rational inattention can also be used to provide a
microfoundation for the multinomial logit model of choice often used in empirical indus-
trial organization.

F IG . 1.—Information economics over time. Data from Google Books Ngram Viewer (Michel et
al. 2011). The y -axis depicts the share of “information economics” among all 2-grams plus
the share of “economics of information” among all 3-grams divided by the share of “eco-
nomics” among all 1-grams (all phrases case insensitive) in books published that year.
The dots indicate raw data by year while the solid line depicts a 5-year moving average.
like a return to the ex ante formulation of information-gathering costs,

Hebert and Woodford (2016) and Morris and Strack (2017) develop se-
quential sampling models that generate the static cost functions employed
in the rational inattention literature.
The aforementioned papers focus on information acquisition by isolated
decision makers. Grossman and Stiglitz (1980) show that in certain mar-
ket settings, the value of acquiring information is lower when a higher
fraction of other agents are informed; then full-information equilibrium
outcomes are not possible. Information externalities have also been used
to understand herd behavior, fashions, and customs. Banerjee (1992) and
Bikhchandani, Hirshleifer, and Welch (1992) point out that if a large
number of individuals all share the same state-dependent preferences,
observe discrete signals about the state, and take their actions sequentially
(after seeing what those before them have done), information cascades
will arise: individuals start to ignore their own information and simply
mimic the behavior of others.3 Following Bikhchandani et al. in the pages
3
Smith and Sorensen (2000) and Eyster and Rabin (2014) discuss some generalizations
and limitations of this result.

of this journal, Bernheim (1994) proposed an alternative theory of con-
formity, one built on signaling considerations. These papers were a part
of a broader shift whereby information joined the center stage, alongside
preferences and technology, as a key factor for understanding socioeco-
nomic phenomena.4
II. Asymmetric Information

Parts of the literature that focused on endogenizing information through
information acquisition tended to consider environments with limited
strategic interaction. At the same time, a largely separate literature on
Bayesian games (with exogenously specified information) was coming of
age. The early work by Harsanyi and others focused on “pure” game the-
ory,5 but then in the 1970s, a series of applied theory papers—Akerlof’s
(1970) market for lemons, Spence’s (1973) job signaling, and Stiglitz
and coauthors’ analysis of screening (Stiglitz 1975; Rothschild and Stiglitz
1976; Stiglitz and Weiss 1981)—identified the crucial importance of asym-
metric information. Remarkably, Rothschild and Stiglitz’s (1976) analysis
of insurance markets opens with the sentence, “Economic theorists tradi-
tionally banish discussions of information to footnotes” (629). Yet, with
hindsight, it is clear that by the mid-1970s, information economics was
flourishing.
While Akerlof, Spence, and Stiglitz played key roles in the “asymmetric
information revolution,” related ideas were also being explored by other
scholars around the same time. For example, writing in the J PE, Nelson
(1970, 1974) proposed, in informal terms, a Spence-like channel through
which advertising could be understood as a costly signal of a firm’s quality.
Milgrom and Roberts (1986) subsequently formalized this view (and filled
an important gap in Nelson’s original argument). The core ideas about
asymmetric information developed in the 1970s continue to play a key
role in our analysis of health care, banking, education, and many other
markets.
III. Communication
The fact that asymmetric information can have stark consequences—and
is often detrimental to social welfare—poses the question of whether eco-
nomic agents will share their private information so as to eliminate the
informational asymmetry. The 1980s saw the development of two widely
used frameworks for studying information exchange. Crawford and Sobel
4
As an aside, one might argue that all technology is a form of information, but information-
theoretic concepts have so far not been widely applied to the study of technological change.
5
One important exception is Vickrey’s (1961) analysis of auctions.

(1982) consider a cheap talk environment in which an informed sender
can costlessly convey any message to an uninformed receiver, who then
takes an action that affects the welfare of both parties. Grossman (1981)
and Milgrom (1981) introduced verifiable message models in which the
sender can choose how much of his private information to disclose but
cannot tell outright lies. In cheap talk settings, the sender’s problem tends
to be the receiver’s inertia:6 when the player’s interests are insufficiently
aligned, it is not possible for the sender to convey any information even
if both parties would be better off if he were able to do so. In contrast,
in verifiable message models, the sender’s problem is often the receiver’s
reaction to silence: when the sender’s preferences are monotone, full in-
formation is always conveyed in equilibrium even if the sender wishes this
were not so.7 This result has played an important role in discussions of gov-
ernment policies regarding disclosure mandates.
Economists’ analysis of strategic obstacles to information exchange com-
plements the computer science and engineering literatures that focus
on the technological constraints on information transmission (Cover and
Thomas 2006). One exciting and underexplored area is the interaction
between the two types of constraints. For example, Blume, Board, and
Kawamura (2007) point out that the presence of technological limita-
tions can alleviate the impact of the strategic obstacles and thus improve
communication.
IV. Recent Developments

The currently most active area of research in information economics
is probably information design, a confluence of work on Bayes correlated
equilibria (Bergemann and Morris 2013) and Bayesian persuasion (Kame-
nica and Gentzkow 2011). Bayes correlated equilibria take as given play-
ers’ state-dependent preferences and describe the set of all possible out-
comes that could arise regardless of what each player knows about the
state and about what others know. Bayesian persuasion models seek to
identify the best outcome from this set given some objective function.
Thus, research on information design seeks to identify the optimal infor-
mational environment (who should know what and when) taking as given
the preferences of the players and some objective function over the play-
ers’ actions. Information design can be seen as a parallel to mechanism
design: in the latter, the designer can choose the game but has no control
over the information structure, whereas in the former the designer can
6
It is possible to construct a cheap talk game with an equilibrium that gives the sender a
lower payoff than he would get under no communication, but such examples tend to be
somewhat contrived.
7
This result requires that the receiver is certain of what information the sender has (Dye
1985).

choose the information structure but has no control over the game (cf. Ber-
gemann and Morris 2016; Taneva 2016). In just the few years since its in-
ception, information design has been used to address issues in banking
regulation, internet advertising, censorship, entertainment, price discrim-
ination, traffic congestion, and so forth.8
In closing, it may be worthwhile to note that, over the last century, in-
formation has come to play an important role in other disciplines besides
economics. In biology, we discovered that all complexity of life is encoded
as information about sequences of nucleic acids. In physics, not only are
informational constraints (expressed as the uncertainty principle) a cen-
tral feature of quantum mechanics, a provocative “it from bit” doctrine pro-
poses that “all things physical are information-theoretic in origin” (Wheeler
1990, 311). Perhaps in some distant future, information-theoretic ap-
proaches may reveal structures shared by biological, physical, and eco-
nomic systems.
References
Akerlof, George. 1970. “The Market for ‘Lemons’: Quality Uncertainty and the
Market Mechanism.” Q. J.E. 84 (3): 488–500.
Banerjee, Abhijit. 1992. “A Simple Model of Herd Behavior.” Q. J.E. 107 (3): 797–
817.
Bergemann, Dirk, and Stephen Morris. 2013. “Robust Predictions in Games with
Incomplete Information.” Econometrica 81 (4): 1251–1308.
———. 2016. “Information Design, Bayesian Persuasion, and Bayes Correlated
Equilibrium.” A.E.R. Papers and Proc. 106 (5): 586–91.
Bernheim, B. Douglas. 1994. “A Theory of Conformity.” J.P.E. 102 (5): 841–77.
Bikhchandani, Sushil, David Hirshleifer, and Ivo Welch. 1992. “A Theory of Fads,
Fashion, Custom, and Cultural Change as Informational Cascades.” J.P.E. 100
(5): 992–1026.
Blume, Andreas, Oliver Board, and Kohei Kawamura. 2007. “Noisy Talk.” Theoret-
ical Econ. 2 (4): 395– 440.
Cover, Thomas, and Joy Thomas. 2006. Elements of Information. 2nd ed. Hoboken,
NJ: Wiley.
Crawford, Vincent, and Joel Sobel. 1982. “Strategic Information Transmission.”
Diamond, Peter. 1971. “A Model of Price Adjustment.” J. Econ. Theory 3 (2): 156–
68.
Dye, Ronald. 1985. “Disclosure of Nonproprietary Information.” J. Accounting
Res. 23 (1): 123–45.
Ely, Jeffrey, Alexander Frankel, and Emir Kamenica. 2015. “Suspense and Sur-
prise.” J.P.E. 123 (1): 215–60.
Eyster, Erik, and Matthew Rabin. 2014. “Extensive Imitation Is Irrational and
Harmful.” Q. J.E. 129 (4): 1861–98.
Grossman, Sanford. 1981. “The Informational Role of Warranties and Private
Disclosure about Product Quality.” J. Law and Econ. 24 (3): 461–83.
8
Contributions to information design published in the J PE include Rayo and Segal
(2010), Kremer, Mansour, and Perry (2014), and Ely, Frankel, and Kamenica (2015).

Grossman, Sanford, and Joseph Stiglitz. 1980. “On the Impossibility of Informa-
tionally Efficient Markets.” A.E.R. 70 (3): 393–408.
Hayek, F. A. 1945. “The Use of Knowledge in Society.” A.E.R. 35 (4): 519–30.
Hebert, Benjamin, and Michael Woodford. 2016. “Rational Inattention with Se-
quential Information Sampling.” Working paper, Stanford Univ.
Kamenica, Emir, and Matthew Gentzkow. 2011. “Bayesian Persuasion.” A.E.R.
101 (6): 2590–2615.
Kremer, Ilan, Yishay Mansour, and Motty Perry. 2014. “Implementing the ‘Wis-
dom of the Crowd.’” J.P.E. 122 (5): 988–1012.
Matejka, Filip, and Alisdair McKay. 2015. “Rational Inattention to Discrete
Choices: A New Foundation for the Multinomial Logit Model.” A.E.R. 105
(1): 272–98.
McCall, J. J. 1970. “Economics of Information and Job Search.” Q. J.E. 84 (1):
113–26.
Michel, Jean-Baptiste, Yuan Kui Shen, Aviva Presser Aiden, et al. 2011. “Quanti-
tative Analysis of Culture Using Millions of Digitized Books.” Science 331 (6014):
176–82.
Milgrom, Paul. 1981. “Good News and Bad News: Representation Theorems and
Applications.” Bell J. Econ. 12 (2): 380–91.
Milgrom, Paul, and John Roberts. 1986. “Price and Advertising Signals of Prod-
uct Quality.” J.P.E. 94 (4): 796–821.
Morris, Stephen, and Philipp Strack. 2017. “The Wald Problem and the Equiva-
lence of Sequential Sampling and Static Information Costs.” Working paper,
Princeton Univ.
Nelson, Phillip. 1970. “Information and Consumer Behavior.” J.P.E. 78 (2): 311–
29.
———. 1974. “Advertising as Information.” J.P.E. 82 (4): 729–54.
Rayo, Luis, and Ilya Segal. 2010. “Optimal Information Disclosure.” J.P.E. 118
(5): 949–87.
Rothschild, Michael, and Joseph Stiglitz. 1976. “Equilibrium in Competitive In-
surance Markets: An Essay on the Economics of Imperfect Information.”
Q. J.E. 90 (4): 629–49.
Sims, Christopher. 2003. “Implications of Rational Inattention.” J. Monetary Econ.
50 (3): 665–90.
Smith, Lones, and Peter Sorensen. 2000. “Pathological Outcomes of Observa-
tional Learning.” Econometrica 68 (2): 371–98.
Spence, Michael. 1973. “Job Market Signaling.” Q. J.E. 87 (3): 355–74.
Stigler, George. 1961. “The Economics of Information.” J.P.E. 69 (3): 213–25.
Stiglitz, Joseph. 1975. “The Theory of ‘Screening,’ Education, and the Distribu-
tion of Income.” A.E.R . 65 (3): 283–300.
Stiglitz, Joseph, and Andrew Weiss. 1981. “Credit Rationing in Markets with Im-
perfect Information.” A.E.R . 71 (3): 393– 410.
Taneva, Ina. 2016. “Information Design.” Working paper, Univ. Edinburgh.
Vickrey, William. 1961. “Counterspeculation, Auctions, and Competitive Sealed
Tenders.” J. Finance 16 (1): 8–37.
Wheeler, John A. 1990. “Information, Physics, Quantum: The Search for Links.”
In Complexity, Entropy, and the Physics of Information, edited by W. Zurek, 309–36.
Redwood City, CA: Addison-Wesley.

The Continuing Impact of Sherwin Rosen’s

“Hedonic Prices and Implicit Markets: Product
Differentiation in Pure Competition”
Michael Greenstone
Sherwin Rosen’s landmark paper “Hedonic Prices and Implicit Markets:

Product Differentiation in Pure Competition” (1974) fundamentally al-
tered understanding of several fields of economics, including environ-
mental, labor, public, and urban economics. At its most general level,
the paper outlines how the market solves the problem of matching buy-
ers and sellers of multidimensional goods. Since virtually all goods have
multiple characteristics, the paper’s framework has proven to be broadly
applicable across a range of economic topics. It is therefore no surprise
that the paper is the sixth most cited in the Journal of Political Economy’s
history. To note just a few examples of its application, the paper has served
as the foundation for inferring households’ valuations of air quality (Chay
and Greenstone 2005), understanding “equalizing differences” in the la-
bor market (e.g., Brown 1980), estimating the incidence of government
policies (e.g., Gruber 1994), and describing the equilibrium allocation
of individuals and firms across locations (e.g., Roback 1982; Greenstone,
Hornbeck, and Moretti 2010). Part of the enduring appeal of the paper is
that it outlines a method for estimating relationships of extraordinary im-
portance for the determination of optimal policy, particularly individuals’
willingness to pay for goods and services for which there are not explicit
markets. Canonical examples include environmental quality, school qual-
ity, crime, other amenities, and mandated government benefits.
The paper’s starting point is that virtually all goods are heterogeneous
(e.g., houses, jobs, and cities) and that while we can observe their overall
price, this alone does not shed much light on the demand for and supply
of their characteristics. To make progress on these more fundamental
economic questions, Rosen’s paper outlines an approach that considers
goods to be a vector of their “utility-bearing attributes or characteristics.”
The paper’s central contribution is to model how consumers’ and sup-
I thank Lucas Davis for sage advice and criticisms and Michael Galperin for outstanding
research assistance.

pliers’ optimizing behavior governs the data-generating process that de-
livers the potentially observable equilibrium relationship between char-
acteristics and their prices.
This paper briefly reviews the model that Rosen outlined, discusses
the success of efforts to apply it empirically to gain understanding of sev-
eral key relationships, and outlines areas for future research.
I. A Brief Review of Rosen’s Hedonic Model

In Rosen’s (1974) formulation, a differentiated good is described by a
vector of its characteristics, C 5 ðc1 , c2 ,:::, cn Þ. In the case of a house,
these characteristics may include structural attributes (e.g., number of
bedrooms), neighborhood public services (e.g., local school quality),
and local environmental amenities (e.g., air quality). Thus, the market
price of the ith house can be written as
Pi 5 P ðci1 , ci2 ,:::, cin Þ: (1)

The partial derivative of P() with respect to the jth characteristic, ∂P =∂cj ,
is referred to as the marginal implicit price. It is the marginal price of
the j th characteristic, holding constant all other characteristics, and is
implicit in the overall price of the house.
In the hedonic model, the locus between housing prices and a given
characteristic, called the hedonic price schedule (HPS), is generated by
the equilibrium interactions of consumers and producers. It is assumed
that markets are competitive and that all consumers rent one house at
the market price. Consumers’ utility depends on consumption of the
numeraire X (with price equal to one) and the vector of house charac-
teristics:
u 5 u ðX , CÞ: (2)
The budget constraint is expressed as I 2 P 2 X 5 0, where I is income.
Maximization of (2) with respect to the budget constraint reveals that in-
dividuals choose levels of each of the characteristics to satisfy ð∂U =∂cj Þ=
ð∂U =∂xÞ 5 ∂P =∂cj . Thus, the marginal willingness to pay for cj (e.g., air
quality) must equal the marginal cost of an extra unit of cj in the market.
It is convenient to substitute the budget constraint into (2), which
gives u 5 uðI 2 P , c1 , c2 ,:::, cn Þ. Inverting this equation and holding all
characteristics but j constant results in an expression for willingness to
pay for cj :
Bj 5 Bj ðI 2 P , cj ; C*2j , u * Þ: (3)
Here, u* is the highest level of utility attainable given the budget con-
straint and C*2j is the vector of the optimal quantities of other character-

F IG . 1.—Bid curves, offer curves, and the equilibrium HPS in a hedonic market for air
quality.
istics. This is referred to as a bid (or indifference) curve, because it re-

veals the maximum amount that an individual would pay for different
values of cj, holding utility constant.
Heterogeneity in individuals’ bid functions due to differences in pref-
erences and/or incomes leads to differences in the chosen quantities of
a characteristic. This is depicted in figure 1, which plots the HPS and bid
curves for cj of two consumer types. The consumers are denoted as types #1
and #2; there is potentially an unlimited number of consumer types, each
of which has a bid curve that is tangent to the HPS. Each bid function re-
veals the standard declining marginal rate of substitution between cj and
X (because X 5 I 2 P ). The two types choose houses in locations where
their marginal willingness to pay for cj is equal to the market-determined
marginal implicit price, which occurs at cj1 and cj2 , respectively. Given mar-
ket prices, these consumers’ utilities would be lower at sites with any other
level of local environmental quality.
The other side of the market comprises suppliers of housing services.
It is assumed that suppliers are heterogeneous because of differences in
their cost functions. This heterogeneity may result from differences in
the land they own. For example, it may be very expensive to provide a
high level of air quality on a plot of land located near a steel factory.
By inverting a supplier’s profit function, we can derive the supplier’s of-
fer curve for the characteristic c j :
Oj 5 Oj ðcj ; C*2j , p* Þ, (4)

where p* is the maximum available profit given the supplier’s cost func-
tion and the HPS. Figure 1 depicts offer curves for two types of suppliers.
With this setup, individuals who live in a house that they own would be
both consumers and suppliers, and their supplier self would rent to their
consumer self.
The HPS is formed by tangencies between consumers’ bid and sup-
pliers’ offer functions. At each point on the HPS, the marginal price
of a housing characteristic is equal to an individual consumer’s marginal
willingness to pay (MWTP) for that characteristic and an individual
supplier’s marginal cost of producing it. From the consumer’s perspec-
tive, the gradient of the HPS with respect to air quality gives the equilib-
rium differential that compensates consumers for accepting the disutil-
ity (e.g., increased health risk, aesthetic disamenities) associated with
poorer air quality. Put another way, areas with high levels of air pollution
must have lower housing prices to attract potential homeowners, and the
HPS reveals the price that allocates consumers across locations (and pol-
lution levels). From the suppliers’ perspective, the gradient of the HPS
reveals the equilibrium marginal cost of supplying a cleaner local envi-
ronment.
The HPS itself is useful for a limited range of welfare analysis. The gra-
dient at each point along the HPS reveals the MWTP for the set of con-
sumers that have sorted themselves to the pair of prices and quantities of
the relevant characteristic. Thus, it is possible to infer the welfare gain
associated with a marginal change for different segments of the popula-
tion. The overall HPS can be used to determine the average MWTP in
the relevant population.
Figure 1 illustrates that knowledge of the HPS is not sufficient to con-
duct welfare analysis for nonmarginal changes, at least in partial equilib-
rium. Consider an improvement in air quality from cj1 to cj2 . Consumer 1’s
valuation or willingness to pay for this change is equal to the difference
between p12 and p1; yet the HPS would suggest that the gain is much larger,
equal to the difference between p2 and p1. The difficulty is that we observe
only one point on each person’s bid function (i.e., the pair of prices and
quantities that they choose). Other points are observed only for other in-
dividuals who presumably have different tastes or income levels. Thus,
the HPS cannot be used to determine the welfare consequences of non-
marginal changes in key characteristics.
Rosen’s solution to this problem is a two-step econometric procedure
that, in principle, delivers the HPS and consumer’s bid functions.1 Con-
sider again the case of houses and air quality. The procedure’s first step
is to regress house prices against all housing characteristics, including air
quality, allowing for their effects to be nonlinear:
1
An analogous procedure delivers suppliers’ offer functions.

p 5 a 1 f ðc1 , c2 ,:::, cn Þ 1 ε: (5)

The marginal implicit price of air quality is the derivative of housing
prices with respect to air quality. This quantity is used in the following
second-stage equation:

∂p ðCÞ=∂cair quality 5 a0 1 g cair quality 1 ε0 , (6)
where ε0 includes all demand shifters, such as tastes and income, that in
principle are observable and can be included as covariates. The Rosen
method then calls for evaluating g(cair quality) at different values of cair quality
to trace out the bid function.
Successful implementation of this two-step procedure would have tre-
mendous practical value, because it would reveal consumers’ demand
primitives. Thus, it would be possible to obtain measures of the welfare
effects of nonmarginal changes in the characteristics of goods. This helps
to explain the enduring influence of Rosen’s model.
II. The Rosen Model in Action

Rosen developed this method with the aim of improving understanding
of the world. Indeed, he wrote, “We anticipate that the basic conceptual
framework outlined above will have a variety of applications to many prac-
tical problems” (Rosen 1974, 54). However, at the time of Rosen’s paper,
the economics profession did not adequately appreciate the difficulty in
obtaining causal estimates of empirical relationships, particularly cross-
sectional ones. Thus, the paper paid little attention to the challenges of
consistent estimation of the HPS. Furthermore, the paper refers to the
second-stage estimation as a “garden variety identification problem,” which,
with the benefit of more than 40 years of continued research, also ap-
pears quite optimistic.
For the first 30 years after its publication, the hedonic approach was
largely unsuccessful empirically, and indeed, my judgment is that its
practical value was in question. However, the last 10–15 years have seen
great advances in the empirical credibility focused on the estimation of
the first stage, or HPS. As Rosen originally conceived it, there are few, if
any, instances of credible estimation of the second stage.
This section outlines the challenges with estimation, where there have
been successes, and where more work is needed.
A. Estimation of the HPS

The consistent estimation of equation (1) is the foundation on which
any welfare calculation rests. The reason is that the welfare effects of a

marginal change in a characteristic are obtained directly from the he-
donic price schedule. Furthermore, inconsistent estimation of the HPS
will result in an inconsistent MWTP function, invalidating any welfare
analysis of nonmarginal changes regardless of the method used to recover
preference or technology parameters. These challenges were apparent
to at least one researcher just 1 year after the publication of the Rosen pa-
per: “I have entirely avoided . . . the important question of whether the
empirical difficulties, especially correlation between pollution and un-
measured neighborhood characteristics, are so overwhelming as to ren-
der the entire method useless. I hope that . . . future work can proceed
to solving these practical problems. . . . The degree of attention devoted
to this [problem] . . . is what will really determine whether the method
stands or falls” (Small 1975, 107).
For roughly 30 years, these words proved prophetic as researchers found
consistent estimation of the HPS as in equation (1) to be extraordinarily
challenging. For example, the cross-sectional estimation of the HPS exhib-
ited signs of misspecification (e.g., great sensitivity to the exact set of con-
trols and frequent findings of perversely signed estimates) in a number of
settings, including the relationships between land and/or house prices and
air quality (Smith and Huang 1995), school quality (Black 1999), proximity
to hazardous waste sites (Greenstone and Gallagher 2008), and climate var-
iables (Deschenes and Greenstone 2007). Similar problems characterized
the estimation of compensating wage differentials for job characteristics,
such as the risk of injury or death: in a wide range of studies, the regression-
adjusted association between wages and many job amenities was found
to be weak and often had a counterintuitive sign (e.g., Smith 1979; Brown
1980; Black and Kneisner 2003).
One attempted solution to the challenges of estimating cross-sectional
hedonic equations was to move to a panel setting. An especially impor-
tant example of this method is Brown’s (1980) paper on equalizing dif-
ferences in the labor market, which included person-specific fixed ef-
fects and year fixed effects. However, the results were disappointing in
that they remained sensitive to small changes in specification and often
had the counterintuitive sign. Indeed, Brown concluded, “The impacts
of the intercepts on the coefficients of job characteristics vary consider-
ably, and there is no marked improvement in the correspondence be-
tween these coefficients and a priori predictions” (130).
Table 1, taken from Chay and Greenstone (2005), clearly illustrates
the problems the literature encountered with the cross-sectional and
panel data approaches. It presents “conventional” estimates of the capi-
talization of total suspended particulates (TSPs) air pollution into housing
values based on fitting regressions with county-level census data. Across dif-
ferent panels, the model is estimated using cross-sectional data from 1970

TABLE 1
Estimates of the Effect of TSPs Pollution on Log Housing Values
(1) (2) (3) (4)
A. 1970 Cross Section
Mean TSPs (1/100) .032 2.062 2.040 2.024
(.038) (.018) (.017) (.017)
R2 .00 .79 .84 .85
988 987 987 987
Sample size
B. 1980 Cross Section
Mean TSPs (1/100) .093 .096 .076 .027
(.066) (.031) (.030) (.028)
R2 .00 .82 .89 .89
988 984 984 984
Sample size
C. 1970–80 (First Differences)
Mean TSPs (1/100) .102 .024 .004 2.006
(.032) (.020) (.016) (.014)
R2 .02 .55 .65 .73
Sample size 988 983 983 983
County Data Book covariates no yes yes yes
Flexible form of county covariates no no yes yes
Region fixed effects no no no yes
Note.—The housing and overall consumer price index series are used to deflate all
housing values to 1982–84 dollars. The TSPs data are derived from the Environmental Pro-
tection Agency’s network of pollution monitors. The 1970 (1980) mean TSPs concentra-
tion is the average across all counties’ mean TSPs concentration from 1969 to 1972
(1977 to 1980). Each county’s annual mean TSPs concentration is calculated as the weighted
average of the geometric mean concentrations of each monitor in the county, using the num-
ber of observations per monitor as weights. The county-level mean across multiple years (e.g.,
1969–72) is the average of the annual means. The flexible functional form includes qua-
dratics, cubics, and interactions of the variables as controls. The mean of the natural log of
1970 housing prices is 10.55. The means of the dependent variables in panels B and C are
10.82 and 0.27, respectively. Standard errors (in parentheses) are estimated using the Eicker-
White formula to correct for heteroskedasticity. This table appears as table 3 in Chay and Green-
stone (2005, 408).
(panel A) and 1980 (panel B) and by first-differencing the 1970 and 1980
data (panel C) to remove the influence of time-invariant unobservables.
Across columns 1–4, an increasing number of covariates are used to adjust
the effect of TSPs on housing values.
The instability of the estimates across specifications and within a spec-
ification across panels is striking and suggests that the conventional ap-
proach to estimating the HPS is prone to misspecification. The column 2
results illustrate this point powerfully, because they use a specification typ-
ical of the three decades of research following the publication of Rosen’s
paper. With this specification, the 1970 data suggest that a one-unit de-
cline in TSPs is associated with a 0.06 percent increase in housing values,
the 1980 data suggest that it is associated with a 0.10 percent decrease, and

the first-difference data find no meaningful relationship. The point is that
even when holding the specification constant, it is possible to find virtually
any effect that one desires; not even the sign is constant across the dif-
ferent data sets. Unfortunately, this pattern of results is not specific to
the relationship between housing prices and TSPs; for example, Black
and Kneisner’s (2003) paper on the value of a statistical life convincingly
demonstrates a similar variability when estimating the relationship be-
tween wages and on-the-job fatality risks. The result of this uncertainty
was that despite the publication of perhaps hundreds of empirical papers
based on cross-sectional and panel data estimation of the hedonic model,
the prevailing sense was that the resulting estimates were plagued by omit-
ted variable bias. The hedonic method’s practical relevance was in doubt.
The late 1990s and early 2000s marked a turning point in the field as
the “credibility revolution” sparked renewed interest in estimation of
Rosen-style hedonic equations. Specifically, researchers began to iden-
tify quasi-experimental variation in the variables of interest that was plau-
sibly unrelated to unobserved determinants of the studied outcome
(e.g., housing prices, wages). In a quasi-experiment, variation in the var-
iable of interest is determined by nature, politics, an accident, or some
other action beyond the researcher’s control. The identifying assump-
tion is that this variation is exogenous, and in high-quality quasi exper-
iments, researchers understand well the source of this variation and
characterize it clearly for the reader.
This turning point manifested itself with great advances in under-
standing of a wide range of topics. As part of my dissertation, ultimately
published in the JPE, Kenneth Chay and I exploited spatial and temporal
variation in the introduction of the Clean Air Act, along with knowledge
of the exact rule that determined that variation, to study the relationship
between TSPs concentration and housing values. We found that the elas-
ticity of housing values with respect to TSPs concentrations ranges from
20.20 to 20.35 (Chay and Greenstone 2005). In a compelling study, Da-
vis (2004) found that an increase in the incidence of pediatric leukemia
in a Nevada county reduced housing prices by about 15 percent. Build-
ing on the important Black (1999) paper, Bayer, Ferreira, and McMillan
(2007) estimate households’ willingness to pay for school quality using
variation across boundaries for catchment zones.2 There are many com-
pelling examples, and any effort to be comprehensive will surely fall short;
but a few other quasi-experimental findings include a negative MWTP for
proximity to a convicted sex offender (Linden and Rockoff 2008), willing-
2
This paper’s contribution goes beyond credible estimation of the HPS. It provides
great insight into the relationship between estimates of willingness to pay from hedonic
price regressions vs. those from random utility model discrete choice approaches to de-
mand estimation.

ness to pay for Superfund cleanups that is smaller than their costs (Green-
stone and Gallagher 2008), and an 11 percent decline in housing values
within 0.5 mile of newly opened industrial plants that emit toxic air pollut-
ants (Currie et al. 2015).
Overall, the application of the quasi-experimental approach has breathed
new life into the Rosen model. The proliferation of papers in recent years
has generated tremendous insights across a wide range of areas where
there was previously little credible empirical work that could guide peo-
ple’s understanding. There is currently tremendous interest in random-
ized control trial experiments in economics, but I am not aware of any
field experiment applications of Rosen’s hedonic model to date (although
they would be an incredible addition both substantively and methodolog-
ically).
B. What about the Bid Functions?

The greatest promise of the Rosen model is not just to consistently esti-
mate the HPS but to recover individuals’ bid functions (and suppliers’
offer functions). Estimation of these primitives allows for assessing the
welfare consequences of nonmarginal changes and counterfactual poli-
cies; this is the difference between estimating what has been and what
might be. The great challenge, of course, is that estimating individuals’
bid functions requires the quite difficult task of observing the same indi-
vidual or taste type facing two sets of prices.
While the estimation of the HPS faces a formidable enemy (i.e., omit-
ted variables), the estimation of bid functions has found itself facing
seemingly even more formidable foes. Following Rosen’s paper, there
were some initial efforts to recover these bid functions (e.g., Palmquist
1984). However, Brown and Rosen (1982) cautioned researchers by dem-
onstrating the hedonic method’s reliance on potentially strong func-
tional form assumptions. A pair of later papers showed that efforts to in-
fer consumers’ bid functions from the HPS are further undermined by
taste-based sorting and that the difficulty of addressing this problem us-
ing standard exclusion restrictions means that quite strong assumptions
are necessary (Bartik 1987; Epple 1987). Thus by the late 1980s, there was
a sense that credible estimation of bid functions with the Rosen approach
was not possible, and efforts to even attempt their estimation began to
disappear.
In response to this decline, several papers tried to revive Rosen’s meth-
od’s aim of recovering demand primitives through structure or by invok-
ing alternative assumptions. Epple and Sieg (1999) outlined a “locational
equilibrium” model that can be used to develop estimates of the demand
primitives. Returning to the Rosen model’s roots, Ekeland, Heckman,
and Nesheim (2004) outline the assumptions necessary to identify the de-

mand (and supply) functions in an additive version of the hedonic model
with data from a single market.3 Bajari and Benkard (2005) similarly re-
turn to the Rosen model and, through alternative approaches to the first
and second stages, outline an approach to construct bounds on individ-
uals’ utility parameters and other economic objects. Each of these ap-
proaches is promising, but they all require potentially strong assumptions
and to date their influence on applied work has not been tremendous (al-
though researchers continue to experiment with them).
III. Conclusions
Sherwin Rosen’s hedonic method is a great achievement of economic
theory. Taking as its starting point an observable relationship that by it-
self does not shed any light on the economic behavior underlying it, the
paper outlines a model of buyer and seller optimizing behavior to ex-
plain the process that generates what is observed in the data. In outlin-
ing this framework, Rosen fundamentally altered how we understand the
world.
On the applied side, the application of quasi-experimental techniques
to the estimation of the HPS has reinstated the hedonic method as a work-
horse in environmental, labor, public, urban, and other parts of econom-
ics. Although consistent estimation of the HPS cannot be used for counter-
factual policy analysis of nonmarginal changes, the decades of empirical
research that have been guided by Rosen’s paper demonstrate that there
are many marginal changes worth analyzing, and the last 10–15 years illus-
trate that it is possible to produce credible evidence on their welfare con-
sequences.
With respect to nonmarginal changes, the picture is not quite as bright
when it comes to using the Rosen method. However, there are already
some promising approaches that merit greater investigation, application,
and exploration. If history is any guide, the coming years will see the
development of new methods that build on Rosen’s method to recover
bid functions as the questions that can be answered remain vital and
urgent.
Although consistent estimation of the HPS does not recover the un-
derlying bid functions, it can be used to estimate the welfare impacts of
nonmarginal changes of past changes in amenities. Specifically, Green-
stone and Gallagher (2008) demonstrate that knowledge of the HPS can
be used to infer the historical welfare consequences of a nonmarginal
3
Heckman, Matzkin, and Nesheim (2005, 2010) examine identification and estimation
of nonadditive hedonic models and the performance of estimation techniques for additive
and nonadditive models.

amenity change for landowners in a particular location (e.g., a neighbor-
hood, city, or county) before the change occurs.4 In their empirical appli-
cation, Greenstone and Gallagher use the HPS to learn about the welfare
consequences of Superfund clean-ups of industrial sites, but the historical
record is filled with countless other consequential nonmarginal amenity
changes. Use of the HPS in this manner has potentially tremendous prac-
tical value, because the benefits of previous policies are an important
guide about the benefits of future or alternative policies; this is especially
so in the all too frequent case in which reliable estimates of the underlying
demand primitives or bid functions are unavailable.
Perhaps most notably, more than four decades after its publication,
Rosen’s model is being used to generate insights about the world, and
researchers are still actively engaged with its methods. It is evident that
Rosen’s hedonic method has fundamentally altered our understanding
of the world and in so doing has passed the acid test of time.
References
Bajari, P., and C. Benkard. 2005. “Demand Estimation with Heterogeneous Con-
sumers and Unobserved Product Characteristics: A Hedonic Approach.” J.P.E.
113 (6): 1239–76.
Bartik, Alexander W., Janet Currie, Michael Greenstone, and Christopher R.
Knittel. 2018. “The Local Economic and Welfare Consequences of Hydrau-
lic Fracturing.” Manuscript, Soc. Sci. Res. Network. https://ssrn.com/abstract
52692197.
Bartik, T. J. 1987. “The Estimation of Demand Parameters in Hedonic Price Mod-
els.” J.P.E. 95 (1): 81–88.
Bayer, P., F. Ferreira, and R. McMillan. 2007. “A Unified Framework for Measur-
ing Preferences for Schools and Neighborhoods.” J.P.E. 115 (4): 588–638.
Black, D. A., and T. J. Kniesner. 2003. “On the Measurement of Job Risk in He-
donic Wage Models.” J. Risk and Uncertainty 27 (3): 205–20.
Black, S. E. 1999. “Do Better Schools Matter? Parental Valuation of Elementary
Education.” Q. J.E. 114 (2): 577– 99.
Brown, C. 1980. “Equalizing Differences in the Labor Market.” Q. J.E. 94 (1):
113 –34.
Brown, J. N., and H. S. Rosen. 1982. “On the Estimation of Structural Hedonic
Price Models.” Econometrica 50 (3): 765–68.
Chay, K., and M. Greenstone. 2005. “Does Air Quality Matter? Evidence from the
Housing Market.” J.P.E. 113 (2): 376 – 424.
Currie, J., L. Davis, M. Greenstone, and R. Walker. 2015. “Environmental Health
Risks and Housing Values: Evidence from 1,600 Toxic Plant Openings and
Closings.” A.E.R. 105 (2): 678 –709.
4
This is a key population in their own right. Further, the welfare impacts on them are
equal to the social welfare impacts with the admittedly strong assumption of zero moving
costs, as well as no change in the overall price level. See Roback (1982) and more recent
papers (e.g., Moretti 2011; Bartik et al. 2018) that consider the cases of nonzero moving
costs and elastic housing supply.

Davis, L. W. 2004. “The Effect of Health Risk on Housing Values: Evidence from a
Cancer Cluster.” A.E.R. 94 (5): 1693 –1704.
Deschenes, O., and M. Greenstone. 2007. “The Economic Impacts of Climate
Change: Evidence from Agricultural Output and Random Fluctuations in
Weather.” A.E.R. 97 (1): 354 – 85.
Ekeland, I., J. Heckman, and L. Nesheim. 2004. “Identification and Estimation
of Hedonic Models.” J.P.E. 112 (S1): S60 – S109.
Epple, D. 1987. “Hedonic Prices and Implicit Markets: Estimating Demand and
Supply Functions for Differentiated Products.” J.P.E. 95 (1): 59 – 80.
Epple, D., and H. Sieg. 1999. “Estimating Equilibrium Models of Local Jurisdic-
tions.” J.P.E. 107 (4): 645 – 81.
Greenstone, M., and J. Gallagher. 2008. “Does Hazardous Waste Matter? Evi-
dence from the Housing Market and the Superfund Program.” Q.J.E. 123 (3):
951–1003.
Greenstone, M., R. Hornbeck, and E. Moretti. 2010. “Identifying Agglomeration
Spillovers: Evidence from Winners and Losers of Large Plant Openings.” J.P.E.
118 (3): 536–98.
Gruber, J. 1994. “The Incidence of Mandated Maternity Benefits.” A.E.R. 84 (3):
622–41.
Heckman, J. J., R. L. Matzkin, and L. P. Nesheim. 2005. “Simulation and Estima-
tion of Hedonic Models.” In Frontiers in Applied General Equilibrium Modeling,
edited by T. J. Kehoe, T. N. Srinivasan, and J. Whalley. New York: Cambridge
Univ. Press.
———. 2010. “Nonparametric Identification and Estimation of Nonadditive He-
donic Models.” Econometrica 78 (5): 1569 –91.
Linden, L., and J. E. Rockoff. 2008. “Estimates of the Impact of Crime Risk on
Property Values from Megan’s Laws.” A.E.R. 98 (3): 1103 –27.
Moretti, Enrico. 2011. “Local Labor Markets.” In Handbook of Labor Economics,
vol. 4b, edited by Orley Ashenfelter and David Card, 1237–1312. Amsterdam:
Elsevier.
Palmquist, R. B. 1984. “Estimating the Demand for the Characteristics of Hous-
ing.” Rev. Econ. and Statis. 66 (3): 394 – 404.
Roback, J. 1982. “Wages, Rents, and the Quality of Life.” J.P.E. 90 (6): 1257–78.
Rosen, S. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in
Pure Competition.” J.P.E. 82 (1): 34 –55.
Small, K. A. 1975. “Air Pollution and Property Values: Further Comment.” Rev.
Econ. and Statis. 57 (1): 105 –7.
Smith, R. S. 1979. “Compensating Wage Differentials and Public Policy: A Re-
view.” Indus. and Labor Relations Rev. 32 (3): 339–52.
Smith, V. K., and J. Huang. 1995. “Can Markets Value Air Quality? A Meta-
Analysis of Hedonic Property Value Models.” J.P.E. 103 (1): 209 –27.

Assignment Problems
Philip J. Reny
I. Introduction
An assignment problem is one in which a number of goods, each in some
fixed quantity, must be assigned to a number of individuals. The class of
assignment problems that will concern us here are those in which no mon-
etary transfers are possible.1 Assigning committee positions to members of
Congress or dormitories to students are but two of many such examples.
When the individuals’ tastes are known, it is not difficult in principle
to achieve an assignment of goods to individuals that is Pareto efficient.
But this becomes considerably more difficult when preferences are pri-
vate information because one must then ensure that no individual has
any incentive to misreport his or her preferences.
In a seminal J PE paper, Hylland and Zeckhauser (1979) consider as-
signment problems in which each individual can receive at most one good
and at most one unit of it (as in the two examples above). They showed
that if individuals are endowed with fiat money and participate in a mar-
ket that sets nominal prices for the probabilities with which goods can
be obtained, then competitive equilibrium prices (for probabilities) exist
and yield ex ante efficient lotteries that can be resolved to produce ex
post efficient outcomes.2 Consequently, when there are sufficiently many
individuals so that no single individual has any significant impact on
prices, each individual would be willing to report his preferences truth-
fully in a mechanism that computes and implements the competitive equi-
librium outcome for those preferences. Such is the mechanism proposed
by Hylland and Zeckhauser.
An even more challenging class of assignment problems are the so-
called combinatorial assignment problems. In such a problem, there are
I thank Eric Budish for helpful comments and the National Science Foundation (SES-
1227506 and SES-1724747) for financial support.
1
In contrast, e.g., to Koopmans and Beckman (1957).
2
That the difficulties created by indivisibilities might be circumvented by introducing
probabilities was first recognized by von Neumann (1953), whose work influenced Koop-
mans and Beckman (1957), who interpret probabilities as fractions of perfectly divisible sur-
rogate goods. Birkhoff ’s (1946) theorem, that doubly stochastic matrices are convex com-
binations of permutation matrices, is the central mathematical result that is at the heart of
this approach.

again many units of many goods to be allocated, but there are no a priori
restrictions on the bundles of goods that individuals can receive. Espe-
cially challenging are cases in which individual preferences over bundles
of goods exhibit complementarities.
Recently, an important combinatorial assignment problem has been
considered by Budish (2011). He considers the challenging problem of
assigning classes to students, a problem in which complementarities arise
naturally from course scheduling constraints even if student preferences
over classes, without those constraints, are additively separable.3
As in the Hylland-Zeckhauser model, the goods in Budish’s (2011)
model, namely classes, are indivisible. Hylland and Zeckhauser (1979)
circumvent the indivisibility problem by creating a market for probabil-
ities. Unfortunately, as Budish observes, in the presence of complemen-
tarities there may be no prices for the probabilities with which individual
classes can be obtained that lead students to choose lotteries over bun-
dles of classes that efficiently exhaust the total available probability and
that are feasible to carry out. In short, the combinatorial assignment prob-
lem cannot, in general, be solved by using the lottery technique of Hyl-
land and Zeckhauser. One must deal with the indivisibilities and comple-
mentarities head on.4
Like Hylland and Zeckhauser (1979), Budish (2011) uses a market
mechanism with fiat money to attack the problem. Students are given “in-
come” in the form of fiat money that can be used to purchase classes at
prices that are determined in market equilibrium. Importantly, Budish
allows the students’ preferences to be almost completely general and does
not assume that the fiat money has any intrinsic value to them (unlike
other mechanisms in use for this problem).
Because issues such as “fairness” are particularly important in this and
other assignment contexts, it would be natural for each student to re-
ceive the same amount of fiat money. However, as Budish (2011) shows,
this can lead to the nonexistence of the type of market equilibrium that
he considers. Budish’s striking result is that, with arbitrarily small depar-
tures from equal relative incomes, existence is restored, and several attrac-
tive efficiency and fairness criteria can be obtained. Furthermore, with
large numbers of individuals, Budish’s market mechanism is approxi-
mately incentive compatible.
The objective in this short note marking the 125th anniversary of
the Journal of Political Economy is modest. It is shown here that Budish’s
(2011) result can be generalized to allow arbitrary preferences and both
3
I am grateful to Eric Budish for pointing this out.
4
But see Budish et al. (2013) for particular conditions under which the lottery tech-
nique can be made to work.

divisible and indivisible goods, as would exist, for example, in the com-
mittee assignment problem when the workload on some committees can
be divided up in any way among committee members.5 Thus, while the ab-
sence of divisible goods is important in Budish’s proof, it is inessential for
his results.
On the technical side, the proof offered here is rather simple. Indivis-
ibilities create discontinuities in demand as prices vary, because strictly pre-
ferred bundles can suddenly become affordable. These discontinuities
are the source of most of the complications that arise in Budish’s (2011)
clever proof. The main technical contribution here is to note that one
can avoid discontinuities altogether by considering a surrogate economy
in which agents, instead of maximizing utility subject to their budget con-
straint, maximize a Lagrangian in which violations of their budget con-
straint yield a suitably high utility cost per unit of overexpenditure. Equilib-
ria of this surrogate economy are shown to yield equilibria in the sense of
Budish’s paper. It is entirely possible that this surrogate economy, because
it is continuous, might lead to more efficient and/or stable algorithms for
computing the requisite equilibria. But these computational issues have
not been explored here in any detail whatsoever.6
II. Assignment Problems

An assignment problem, (I, L, X, u, q), consists of the following items:
1. I and L are positive integers, where I is the number of agents and

L is the number of commodities;
2. X 5 Ii51 X i , where each consumption set X i is a compact set of non-
negative vectors in RL with 0 ∈ X i ;
3. q is an aggregate endowment vector in RL such that ql > 0 for every
l 5 1, … , L;
4. u 5 ðu1 , … , uI Þ, where each utility function ui : X i → ½0, 1 is contin-
uous.7
5
Budish (2011) allows arbitrary preferences except for the assumption that no agent is
indifferent between any two bundles in his finite consumption set. Because divisible goods
are allowed here, my consumption sets can be uncountably infinite. Therefore, to accom-
modate continuous preferences, indifference must be permitted, and this is done whether con-
sumption sets are finite or infinite.
6
The proof technique employed here can also provide a generalization of the results of
Dierker (1971) to include both indivisible and divisible goods, while at the same time sim-
plifying his proof.
7
All the results can be derived under the slightly more general assumption that for each
consumer i there is a reflexive and transitive (but possibly incomplete) binary relation, ≿i,
on Xi that is continuous; i.e., for every xi ∈ Xi , the sets fyi ∈ Xi : yi ≿i xi g and fyi ∈ Xi : xi i yi g
are closed.

Remark 1. Consumer i’s consumption set Xi need not be convex or
even connected. In particular, an assignment problem here can accom-
modate the simultaneous presence of divisible and indivisible goods.
For any p ∈ ½0, ∞ÞL , for any ci > 0, and for any agent i, let Di ð p, ci Þ ⊆ X i
be the set of solutions to the maximization problem,
max ui ðxi Þ subject to px i ≤ ci :
xi ∈Xi
Let kk denote the Euclidean norm. For any ε > 0 and for any c1 , … , cI > 0,
let c 5 ðc1 , … , cI Þ and define
dε ðp, c Þ ≔ sup k xi 2 yi k,
where the supremum is over all agents i ∈ f1, … , I g and all xi , yi ∈
[ε ∈½0,ε Di ðp, ci 1 ε0 Þ.
0
Throughout the paper, for any l ∈ f1, …, Lg, el 5 ð0, … , 0, 1, 0, … , 0Þ

denotes the lth unit vector in RL .
The main result below replicates the main result in Budish (2011), but
does not require the sets Xi to be finite, requiring instead only that they be
compact. Also included is the minor modification that the positive in-
comes c1 , … , cI of the agents can be arbitrary, whereas Budish focuses
on the most central case in which c1 5 ⋯ 5 cI 5 1.8
Theorem 1. For any assignment problem (I, L, X, u, q), for any ε > 0
and for any positive real numbers c1 , … , cI , there exist p * ∈ ½0, ∞ÞL and
x * ∈ X such that
a. for every agent i,

i. p * xi* ≤ ci 1 ε, and
ii. xi* solves maxy ∈X ui ðyi Þ subject to p * yi ≤ maxðp * xi* , ci Þ;
i i
pffiffiffiffi
b. k z * k ≤ dε ðp * , cÞ L =2, where c 5 ðc1 , … , cI Þ and where for each
l 5 1, … , L,
8
<o
>
> xil* 2 ql , if pl* > 0
i
zl* ≔
>
>
: max oxil* 2 ql , 0 if pl* 5 0;
i
and
c. p * q ≤ oi ðci 1 εÞ.
8
The present result is not an exact replication of the result in Budish (2011)
pffiffiffiffi because
the bound on the market-clearing error here (specifically, the coefficient of L =2 in part b
of theorem 1) can be smaller or larger than the bound that he describes in his n. 15. How-
ever, the important feature of both bounds is that they are of the same order of magnitude
and, most importantly, that they are independent of the number of agents, I.

Remark 2. To obtain the result in Budish (2011), restrict each Xi to be
a finite set of nonnegative vectors with integer coordinates, set c1 5 ⋯ 5
cI 5 1 above, and define his bi* ≔ maxðp * xi* , 1Þ for every i 5 1, … , I .
Note that theorem 1 cannot be obtained from Budish’s result by dis-
cretizing the compact X i with a finite grid and then taking the limit as
the grid size shrinks to zero. Such a limiting argument can ensure in part
a(ii) only that xi* solves maxy ∈X ui ðyi Þ subject to p * yi < maxðp * xi* , ci Þ, be-
i i
cause strictly better bundles that are unaffordable along the sequence
might become exactly affordable at the limit.
Remark 3. Endowing the agents with even slightly different amounts
of fiat money, instead of with real goods, and allowing the vector of goods
prices to be any nonnegative vector, including the zero vector, means that
differences in real incomes can be arbitrarily large and will be determined
by the goods prices in equilibrium.9 This can be advantageous since effi-
ciency might sometimes require such real income differences when pref-
erences are not strictly monotone (e.g., as in the problem of assigning
classes to students).
Budish (2011) introduces two fairness-related concepts that can be use-
fully applied to problems that include indivisible goods.10 The first of these
is an agent’s “maxmin utility.”11 Before defining this, first define, for
any positive integer n, agent i’s n-maxmin utility to be the utility number
max minðui ðy1 Þ, … , ui ðyn ÞÞ, where the maximum is over all y1 , … , yn ∈ Xi
such that onj51 yj ≤ q. Then, define agent i’s maxmin utility to be his I-
maxmin utility. Maxmin utility is one way to generalize the “I cut you
choose” method of fair division to many agents. Budish’s second fairness
concept presumes that X i contains only vectors with integer coordinates
and is as follows. An allocation x ∈ X is envy-free up to a single unit of a
single good iff for every pair of agents i and j, if ui ðxj Þ > ui ðxi Þ, that is, if i
envies j, then there is a commodity l such that either ui ðxi Þ ≥ ui ðxj 2 el Þ
or xj 2 el ∉ Xi .
Budish (2011) shows that the allocations that he obtains are envy-free
up to a single unit of a single good and that, while they might not yield
each agent his maxmin utility, they do yield each agent his (I 1 1)-maxmin
utility. Budish also shows that his allocations cannot be Pareto improved
on if the agents are allowed to trade among themselves after the assign-
ment is made. Budish does not rule out the possibility that if there is ex-
9
Relative incomes can be made arbitrarily similar. With c1 5 ⋯ 5 cI 5 1, the ratio of
any pair of the incomes b1* , … , bI* defined in the previous remark is between 1 and 1 1 ε.
10
Indivisibilities can lead to nonexistence of efficient and envy-free allocations. See
Budish (2011).
11
Budish (2011) uses the term maxmin share since he focuses on the bundle that
achieves the maxmin utility. I find it more convenient to define the utility level, even though
this number obviously depends on the particular utility representation.

cess supply, then that excess supply could be used to achieve a Pareto im-
provement.
The remarks to follow, in part, describe the sense in which Budish’s
important fairness and efficiency results can be maintained and some-
times slightly improved on. In these remarks, p* and x* are as in theorem 1.
Remark 4 (h-envy-free). If c1 5 ⋯ 5 cI 5 1, then for any h > 0, there
is ε > 0 sufficiently small so that the allocation x* is “h-envy-free” in the
following sense. For any agents i and j, if xj* ∈ Xi and ui ðxj* Þ > ui ðxi* Þ,
that is, if i envies j, then there is a commodity l such that ui ðxi* Þ ≥ ui ðyi Þ
for every yi ∈ Xi such that yi ≤ xj* 2 hel .12 In particular, if X i 5 f0,
1, … , kgL as in Budish (2011), then we can choose yi 5 xj* 2 el and x* is
1-envy-free, that is, envy-free up to a single unit of a single good.
Remark 5 ((I 1 1)–maxmin utility). If c1 5 ⋯ 5 cI 5 1 and ε < 1=I ,
then ui ðxi* Þ ≥ maxðminðui ðy1 Þ, ui ðy2 Þ, … , ui ðyI 11 ÞÞ for each agent i, where
11
the maximum is over all y1 , … , yI 11 ∈ Xi such that oIj51 yj ≤ q. That is, xi*
is at least as good for i as his (I 1 1)–maxmin share (Budish 2011).13
Remark 6 (Weak stability and efficiency). There do not exist S ⊆
f1, … , I g and x^ ∈ X such that xî ≠ xi* for at least one i ∈ S, oi∈S xîl ≤
oi∈S xil* for every l with pl* > 0, and ui ð^
xi Þ > ui ðxi* Þ for every i ∈ S such that
xî ≠ xi . In particular, setting S 5 f1, … , I g shows that x* is weakly Pa-
* 14
reto efficient in this economy for any aggregate endowment vector q*

that satisfies q*l 5 oi xil* if pl* > 0, and q*l ≥ oi xil* if pl* 5 0. If, as in Budish
(2011), all Xi are finite and preferences exhibit no indifference, then
weak stability and efficiency are equivalent to standard stability and effi-
ciency, wherein blocking requires only one individual in the coalition to
be made strictly better off.
The next two remarks indicate that the efficiency and maxmin results
described in the previous remarks can be improved on with an arbitrarily
small degradation of the bound in part b of theorem pffiffiffiffi 1. Fix any arbitrarily
small h > 0. Choose ε > 0 small enough so that εI L < ðoLl51 pl Þh for every
12
Given h > 0, choose ε > 0 so that maxl pl > ε=h for every price vector p ∈ ½0, ∞ÞL such
that px j ≥ 1 for some consumer j and some xj ∈ X j . Such an ε exists by the compactness of
the consumption sets. With this choice of ε, suppose that ui ðxj* Þ > ui ðxi* Þ. Then p * xj* > 1,
and so there is l such that pl* h > ε. Hence, because also p * xj* ≤ 1 1 ε by part a(i), we have
p * ðxj* 2 hel Þ < 1. So any yi ≤ xj* 2 hel that is in Xi satisfies p * yi < 1 5 ci , and so ui ðxi* Þ ≥
ui ðyi Þ by part a(ii).
13
Otherwise, if y1 , … , yI 11 is a solution to the maxmin problem, then ui ðyj Þ > ui ðxi* Þ for
11
every j. But then by part a(ii), p * yj > 1 for every j, and so p * oIj51 yj > I 1 1 > ð1 1 εÞI ≥ p * q,
where the final inequality is by part c. But the outer strict inequality contradicts the feasi-
bility of y1 , … , yI 11 for the maxmin problem.
14
If such an x^ were to exist, then part a(ii) implies that p * xî > p * xi* for every i ∈ S such
that xî ≠ xi* . Since there is at least one i ∈ S such that xî ≠ xi* , p * oi∈S xî > p * oi∈S xi* . Hence,
there is an l such that pl* > 0 and oi∈S xîl > oi∈S xil* , contradicting the feasibility of x^ for the
coalition S.

price vector p ∈ ½0, ∞Þ satisfying maxx∈X poi xi ≥ I . Such an ε > 0 exists
L
by the compactness of X.
Remark 7 (Maxmin utility). If c1 5 ⋯ 5 cI 5 1,pifffiffiffiffiε < ε, and if the
bound in part b is weakened to k z * k ≤ h 1 dε ðp * , cÞ L =2, then we can
ensure that ui ðxi* Þ ≥ maxðminðui ðy1 Þ, ui ðy2 Þ, … , ui ðyI ÞÞ, where the maxi-
mum is over all y1 , … , yI ∈ Xi such that oIj51 yj ≤ q. That is, xi* yields agent
i at least his maxmin utility.15
Remark 8 (Weak Pareto efficiency). pffiffiffiffi If the bound in part b is weak-
ened to k z* k ≤ h 1 dε ðp * , cÞ L =2, then we can ensure that there does
not exist x^ ∈ X distinct from x* such that oIi51 xîl ≤ ql for every l, and
ui ð^xi Þ > ui ðxi* Þ for every i such that xî ≠ xi* .16
Remark 9. The relevance of the efficiency and fairness results in the
previous remarks is called into question by the possibility that x* might
not be feasible. This difficulty can be mitigated as follows. Define dεp ðcÞ
ffiffiffiffi 5
sup dε ðp, cÞ, where the supremum is over all p ∈ ½0, ∞ÞL .17 If ql ≥ dε ðcÞ L =2
for every l, then an application pffiffiffiffi of theorem 1 using the endowment
vector q ~ 5 q 2 ðdε ðcÞ L =2Þ1 instead of q yields p* and x* satisfying
part a of ptheorem
ffiffiffiffi 1 and, by part b, satisfying oi xil* ≤ ql for every l and
ql 2 dε ðcÞ L p ≤ffiffiffio
ffi i xil ≤ ql for every l with pl > 0, and, by part c, satisfying
* *
p * ðq 2 ðdε ðcÞ L =2Þ1Þ ≤ oi ðci 1 εÞ. All of the efficiency and fairness re-
sults in the remarks above then go through, but with respect to the econ-
omy with aggregate endowment q ~ instead of q. In particular, x* is feasible,
h-envy-free, and weakly stable (because the h-envy-free and weak stabil-
ity properties do not depend on the aggregate endowment). However,
while x* is weakly Pareto efficient using only the aggregate endowment
~ , x* need not be Pareto efficient for the actual economy with aggregate
q
endowment q. A second way to handle infeasibility is to note that, except
for at most L agents, every agent i can actually receive his bundle xi* .18
The L exceptional agents can then be assigned any bundles that are fea-
pffiffiffiffi
15
Let y1 , … , yI 11 be a solution to the maxmin problem. Replace q with q 1 ðh= L Þ1 in
the statement of theorem 1. Now suppose, by way of contradiction, that ui ðyj Þ p >ffiffiffiuffi i ðxi* Þ for
every j. Then by part a(ii), p * yj > 1 for every j. Therefore, p * oIj51 yj ≥ p * ðq 1 ðh= L Þ1Þ 2 εI ,
ffiffiffiffi final inequality is by part c. But p * oj51 yj > I implies by our choice of ε that
I
where p the
p * ððh= L Þ1Þ 2 εI > 0 and so p * oIj51 yj > p * q, contradicting the feasibility of y1 , … , yI for
the maxmin problem. pffiffiffiffi
16
Suppose that such an x^ exists. Replace q with q 1 ðh= L Þ1 in the statement of
theorem 1. Then for every i, either ui ð^ xi Þ > ui ðxi Þ or xî 5 xi , and so part a(ii) implies
* *
that p * xî ≥ 1 for every i with p ffiffiffi
ffi strict inequality whenever xî ≠ xi* . Therefore, since x^ ≠ x * ,
p * oIi51 xî > I ≥ p * ðq 1 ðh= L Þ1Þ pffiffiffi2
ffi εI , where the second inequality is by part c. But our
choice of ε implies that p * ððh= L Þ1Þ 2 εI > 0 and so p * oIi51 xî > p * q, contradicting the fea-
sibility of x^.
17
The function dε ðcÞ exists and is finite because dε(p, c) is bounded above by max k xi 2 yi k,
where the maximum is over all i and all xi , yi ∈ Xi .
Observe that j ≤ L in the proof below because oi51 ð#Si 2 1Þ ≤ L and, in the summand,
18 j
each ð#Si 2 1Þ ≥ 1. Hence, q ≥ y* 5 y1* 1 ⋯ 1 yj* 1 xj11 * 1 ⋯ 1 x * for some y * in the con-
I i
vex hull of Si, i 5 1, … , j.

sible given the remaining goods. It then follows as in remarks 6 and 7
that all but the L exceptional agents receive at least their maxmin utility
and that no coalition involving only agents outside the exceptional set
can trade among themselves following the assignment so that each is bet-
ter off.
Remark 10. For any closed subset S of RL we may follow Starr (1969)
and define
rðS Þ 5 sup inf sup ∥ a 2 a 0 ∥:
y∈coS A ⊆ S;y∈coAa,a 0 ∈A
The function r is a measure of the convexity of the set S, with rðSÞ 5 0 if

and only if S is convex. Defining Si ðp, ci , εÞ ≔ [ε ∈½0,ε Di ðp, ci 1 ε0 Þ, thepproof
0
ffiffiffiffi
shows that the bound in part b can be reduced to supi rðSi ðp * , ci , εÞÞ L =2.
Remark 11. If there is a continuum of each type of agent i 5 1,
2, … , I , then no single agent would have an incentive to misreport his util-
ity function because he could not affect the price or the bundle that he re-
ceives. Furthermore, the per capita market-clearing error, being bounded
above by the bound in part b divided by the (infinite) number of agents,
is zero. Hence, as in Hylland and Zeckhauser (1979) and Budish (2011),
the solution here can be the basis for an ε-incentive-compatible mecha-
nism with small per capita market-clearing error when aggregate endow-
ments are proportional to the large but finite number of agents.
III. Proof of Theorem 1

Fix any ε > 0. Because every coordinate of q is strictly positive, we may
choose a positive real number B such that
I
B> oðc
i51
i 1 εÞ=ql for every l 5 1, … , L: (1)
For any a ∈ R, let a1 5 maxða, 0Þ.

Define an I 1 1–person game between players i 5 0, 1, … , I as fol-
lows. Player i 5 0’s set of pure strategies is [0, B]L. For i ∈ f1, … , I g,
player i’s set of pure strategies is Xi. For any ðp, xÞ ∈ ½0, BL X , the
players’ payoffs are as follows:
Ui ðp, x Þ ≔ εui ðxi Þ 2 ðpx i 2 ci Þ1 for i 5 1, … , I

and

U0 ðp, xÞ ≔ p oxi 2 q :
i
All of the payoff functions Ui : ½0, B X → R are continuous.

L

Let us allow each player i > 0 to use a mixed strategy, that is, a proba-
bility measure, mi on Xi. For i > 0, let Mi denote player i’s space of mixed
strategies, henceforth simply strategies, and let M 5 Ii51 Mi . Each Mi is
endowed with the weak* topology. Like [0, B]L, each Mi is nonempty, com-
pact, and convex.
For any m 5 ðm1 , … , mI Þ ∈ M , let m denote the product measure
m1 ⋯ mI .
Extend the players’ payoff functions to ½0, BL M , by an expected
utility calculation. Then, all of the payoff functions Ui : ½0, BL M → R
are continuous. Since each player’s payoff function is quasi-concave (lin-
ear in fact) in his own strategy for any fixed strategies of the others, the
players’ best reply correspondences satisfy all of the hypotheses of Glicks-
berg’s (1952) fixed-point theorem. Hence, this game possesses a Nash
equilibrium (p*, m*).
For each player i > 0, equilibrium implies that the support of mi* con-
tains only elements xi of Xi that maximize Ui(p*, xi).19 Hence, if xi is in the
support of mi* , then xi solves
maxðεui ðyi Þ 2 ðp * yi 2 ci Þ1 Þ: (2)

yi ∈Xi
Since ui is nonnegative and 0 ∈ Xi , the maximum value in (2) is non-

negative. Consequently, since ui is bounded above by 1, it must be the
case that
p * xi ≤ ci 1 ε for every xi in the support of mi* : (3)

Furthermore, any solution xi to the maximization problem (2), and hence
any xi in the support of mi* , must solve
max ui ðyi Þ subject to p * yi ≤ maxðp * xi , ci Þ: (4)

yi ∈Xi
To see this claim, let xi solve (2). Then for every yi ∈ X i ,
εui ðxi Þ ≥ εui ðyi Þ 2 ðp * yi 2 ci Þ1 1 ðp * xi 2 ci Þ1

≥ εui ðyi Þ if p * yi ≤ maxðp * xi , ci Þ,
as claimed.
Define
ð
y* ≔
X
ox m * ðdx Þ:
i
i
19
Recall that the support of a probability measure in a separable metric space is the
smallest closed subset having probability one.

Then player 0’s equilibrium payoff, U0(p*, m*), satisfies
U0 ðp * , m * Þ 5 p * ðy* 2 qÞ ≥ pðy* 2 qÞ 8p ∈ ½0, B L :

Consequently, for any l ∈ f1, … , Lg, yl* < ql ⇒ pl* 5 0 and yl* > ql ⇒
pl*5 B. But then yl* > ql is impossible since pl* 5 B implies, by (3), that
il ≤ ðci 1 εÞ=B for every agent i and so yl ≤ oi ðci 1 εÞ=B < ql by (1). We
y* *
may conclude that
yl* ≤ ql for every l 5 1, … , L (5)

and that pl* 5 0 for any l for which the inequality is strict.
For each i 5 1, … , I , let ei denote the ith unit vector (0, . . . , 0, 1, 0, . . . ,
0). Then
ð I
1 1
oðei , xi Þ m * ðdx Þ 5 I ð1, … , 1, y* Þ ∈ DI RL ,
X I i51
(6)
where (ei, xi) denotes the concatenation of ei and xi, and DI denotes the I 2
1–dimensional unit simplex.
The equality in (6) says that ð1, : : :, 1, y * Þ=I is in the convex hull of the
closed subset C of DI RL that consists of all points of the form (ei, xi),
where xi is in the support of mi* for each i. By Caratheodory’s theorem
(Rockafellar 1970), ð1, … , 1, y* Þ=I can therefore be written as a convex
combination of I 1 L or fewer points belonging to C. Thus, for some pos-
itive integer K we may write
1 I K
I
ð1, … , 1, y * Þ 5 ool
i51 k51
ik ðei , xik Þ, (7)
where the lik’s are nonnegative and sum to one, and at most I 1 L of the
lik are strictly positive and lik > 0 implies that xik is in the support of mi* .
For each i 5 1, … , I , let Si 5 fxik : lik > 0g. Since the first I coordinates
of the vector on the left-hand side of (7) are positive, each Si contains at
least one element. Reindexing if necessary, let S1 , … , Sj denote those Si
that contain two or more elements. So Sj11 , … , SI are singletons, and
since at most I 1 L of the lik are strictly positive, the union of S1 , … , Sj
contains no more than L 1 j elements.
For every i 5 1, … , j, every xi in Si is in the support of mi* and there-
fore satisfies (3) and solves (4). In particular, for any xi ∈ Si letting ε0 5
ðp * xi 2 ci Þ1 , we have ε0 ∈ ½0, ε and xi ∈ Di ðp * , ci 1 ε0 Þ. Consequently,
the distance between any two points in Si is no greater than dε ðp * , cÞ,
where c 5 ðc1 , … , cI Þ. Therefore, the distance between any point in Si and
the simple average of all of the points in Si is no greater than dε ðp * , cÞð#Si 2
1Þ=ð#Si Þ.

The equality in (7) for the first I coordinates implies that o I lik 5 1 K
k51
for each i, and the equality for the last L coordinates then implies that y*
is contained in the sum of the convex hulls of the sets S1 , … , SI . Hence,
y* is contained in the convex hull of S1 1 ⋯ 1 SI .20 Consequently, by the
Shapley-Folkman theorem (see Starr 1969) we may select xi* ∈ Si for each
i 5 1, … , I , so that

ð#Si 2 1Þdε ðp * , cÞ 2
k k
I 2 j
y * 2 oxi* ≤ o :
i51 i51 #Si
Since #Si ≥ 2 for every i 5 1, … , j, we have ½ð#Si 2 1Þ=ð#Si Þ2 ≤
ð#Si 2 1Þ=4 for every i 5 1, … , j. Hence,
ð#Si 2 1Þd2ε ðp * , c Þ
J
k y * 2 ox * k
I 2
i51
i ≤ o
i51 4
≤ d2ε ðp * , c ÞL=4,
where the second inequality follows because the union of the sets
S1 , … , Sj contains no more than L 1 j elements, and so oi51 ð#Si 2 1Þ ≤
j
L. Hence, we may conclude that

pffiffiffiffi
k y* 2 ox * k
I
i ≤ dε ðp * , c Þ L =2: (8)
i51
For every l, either oi xil* ≤ ql or yl* ≤ ql < oi xil* , by (5). Consequently, for
every l 5 1, … , L,

*
*
max oxil 2 ql , 0 ≤ yl 2 oxil : *
i i
Hence, by (8) and the fact that p*

l > 0 implies that yl* 5 ql (by [5]), we
have
pffiffiffiffi
k z* k ≤ dε ðp * , c Þ L =2,
where for each l 5 1, … , L,
8
<o
>
> xil* 2 ql if pl* > 0
i
zl* ≔
>
>
: max oxil* 2 ql , 0 if pl* 5 0,
i
which establishes part b of theorem 1.
20
Because the sum of the convex hulls of any finite number of sets is equal to the convex
hull of their sum.

Part a is established by noting that for each i > 0, xi* ∈ Si implies that
x* *
i is in the support of mi , and so parts i and ii follow by (3) and (4).
Finally, part c is established by noting that p * q 5 p * y * ≤ oi ðci 1 εÞ,
where the equality follows from (5) and the inequality follows from
(3). QED
Remark 12. If one were to use the game in the proof of theorem 1 as
the basis for an algorithm to compute the prices and allocations p*, x*,
then one would need to ask agents to report their ordinal preferences
(e.g., their indifference maps), not their utility functions. The mechanism
would then use some canonical procedure to generate a utility represen-
tation. The reason is that if the algorithm maximizes εui ðyi Þ 2 ðpy i 2 ci Þ1
for each agent i, then, for any fixed ε, any agent who cannot affect the
price would have an incentive to scale up his reported utility numbers
ui(yi).
References
Birkhoff, Garrett. 1946. “Three Observations on Linear Algebra.” Univ. Nacional
Tucumán Revista A 5:147–51.
Budish, Eric. 2011. “The Combinatorial Assignment Problem: Approximate Com-
petitive Equilibrium from Equal Incomes.” J.P.E. 119 (6): 1061–1103.
Budish, Eric, Yeon-Koo Che, Fuhito Kojima, and Paul Milgrom. 2013. “Designing
Random Allocation Mechanisms: Theory and Applications.” A.E.R. 103 (2):
585–623.
Dierker, Egbert. 1971. “Equilibrium Analysis of Exchange Economies with Indi-
visible Commodities.” Econometrica 39 (6): 997–1008.
Glicksberg, Irving L. 1952. “A Further Generalization of the Kakutani Fixed
Point Theorem, with Applications to Nash Equilibrium Points.” Proc. American
Math. Soc. 3 (1): 170–74.
Hylland, Aanund, and Richard Zeckhauser. 1979. “The Efficient Allocation of In-
dividuals to Positions.” J.P.E. 87 (2): 293–314.
Koopmans, Tjalling, and Martin Beckmann. 1957. “Assignment Problems and
the Location of Economic Activities.” Econometrica 25 (1): 53–76.
Rockafellar, R. Tyrrell. 1970. Convex Analysis. Princeton, NJ: Princeton Univ.
Press.
Starr, Ross M. 1969. “Quasi-Equilibria in Markets with Non-convex Preferences.”
von Neumann, John. 1953. “A Certain Zero-Sum Two-Person Game Equivalent
to the Optimal Assignment Problem.” In Contributions to the Theory of Games,
vol. 2, edited by H. W. Kuhn and A. W. Tucker, 5–12. Princeton, NJ: Princeton
Univ. Press.

Auctions in the Journal of Political Economy,

1894–2017
Ali Hortaçsu
A quick JSTOR query for the keyword “auction” in the JPE returns 215
articles. The very first article in the list is “The Assignats: A Study in the
Finances of the French Revolution,” by Emile Levasseur, published in
1894 in the second issue of the second volume of the journal. In this en-
gaging account of the monetary trials and tribulation of the French Rev-
olution, there is a single reference to an auction: assignats were claims to
public lands (mostly confiscated from the clergy) that were also decreed
to become the currency of the revolutionary government, and auctions
were used to sell the public lands to interested parties. Levasseur asserts
that very few assignat holders actually used them to claim land, and as a
paper money, the assignat’s credibility was very much shaken by the lack
of tax revenue for the new government and the many new additional as-
signat issues. Indeed, by the time the assignats were taken out of circula-
tion in 1794 (in quite spectacular fashion: “at nine o’clock in the morning,
with a great crowd of people looking on, all the tools which had been used
in printing assignats were brought to the Place Vendome; the plates and
stamps were broken, and reams of paper and 1167 millions of assignats
burned” [191]), they had depreciated by 98 percent.
The topic of assignats in the French Revolution is taken up a century
later by Thomas Sargent and François Velde, in their “Macroeconomic
Features of the French Revolution” (1995), which appears in the 103rd vol-
ume of the JPE. Sargent and Velde reexamine the data on the price level
and government expenditures during the French Revolution through
the lens of modern macroeconomic theory. Their reading of the data is
more favorable toward the assignat than Levasseur’s: they write that “the
tax-backed money scheme functioned adequately until a war broke out
in 1792, which initially went badly for France. The government wanted
more resources, so it divorced note issues from the land sales. The tax-
backed money plan devolved into a fiat money scheme, causing real bal-
ances to drop and prices to rise quickly in early 1793 and threatening
I would like to thank John Asker, Alex Wolitzky, and Robert Porter, who generously com-
mented on an early version of this note.

the base of the tax (the inflation tax) that was the government’s lifeline”
(475).
Perhaps more interestingly from an auction’s perspective, Sargent and
Velde’s appendix provides a detailed description of the auction format
used to sell clerical lands. The format is that of a two-stage English auc-
tion. In the first auction, a highest bidder is declared, but the highest bid
is advertised as the starting value for a second auction. We should note
an interesting rule used in the second auction that we will revisit below:
“At the second auction, one-minute candles were lit in sequence and bid-
ding started at the highest bid established during the first auction. If two
candles expired without any new bids, the estate was awarded to the
highest bidder from the first auction” (1995, 514).
The use of auctions in the sale or lease of government lands is also men-
tioned in G. O. Virtue’s “Public Ownership of Mineral Lands in the United
States,” published in 1895 in the third volume of the journal. What we
learn from this article is that, in its early history, the US government re-
frained from selling land with valuable mineral deposits, especially lead
and salt. “No doubt the chief motive which induced Congress to reserve
the salines from sale was the fear that they might become a monopoly”
(187), says Virtue. Only shorter-term leases were allowed on reserved gov-
ernment mineral land. However, in 1847, the policy was reversed, and the
government began to sell saline lands at public auctions.
The sale of the public’s assets remains an important problem to this
day, and auctions remain a very popular method of sale. Fast-forwarding
a century from Virtue, we find several very interesting papers regarding
the sale of government property through auctions. The first, “How Does
Privatization Work? Evidence from the Russian Shops,” by Barberis et al.
(1996), is an empirical study of 452 shops in Russia, of which 413 were
privatized, through either auctions or another method of privatization.
This is a small but empirically insightful cross section of one of the most
important privatization episodes in history, that of Soviet Russia, in which
auctions played an important role.
Two other very important privatization efforts in recent memory in
which auctions played an important role were in the auctioning of the
wireless spectrum, which essentially led to the birth of the wireless com-
munications sector, and the privatization of the electricity generation sec-
tor, which, once again, was an important application of auctions. The JPE
is very fortunate to have featured two seminal papers on both topics: “Put-
ting Auction Theory to Work: The Simultaneous Ascending Auction” by
Paul Milgrom, published in 2000, and, in 1992, “Competition in the Brit-
ish Electricity Spot Market,” by Richard Green and David Newbery.
Paul Milgrom is one of the co-inventors (along with Robert Wilson and
Preston McAfee) of the simultaneous ascending auction (SAA) mecha-

nism used to auction off wireless spectrum licenses in many countries
around the world. This is a package auction, where a large number of re-
gional licenses are sold simultaneously, and where bidders may wish to
purchase multiple licenses. The SAA proceeds in multiple rounds: in each
round, the auctioneer sets the minimum bid as a price increment above
the latest high bid on a license. Bidders on each round have to decide
whether to not bid or meet/exceed the minimum bid requirement to
keep the auction going. If no new bids are submitted, the auction ends.
Milgrom first analyzes an essentially nonstrategic version of the auction,
where bidders evaluate, at each round, what their preferred bundle of li-
censes are given the current minimum bid prices and their standing high
bids. In the case in which the licenses are substitutes, Milgrom shows that
straightforward bidding, in which bidders keep bidding for their pre-
ferred bundle at the current prices, leads to an efficient, competitive allo-
cation.
What is key to this result is the fact that bidders never regret having
become the standing high bidder on a license when the substitutes con-
dition is met. In the case in which some licenses are complements, how-
ever, an “exposure” problem emerges, where some bidders may indeed
regret having bid so high for a particular license, whose complement
may have become unaffordable. Milgrom provides the example of a spec-
trum auction in the Netherlands in which this exposure problem became
an important issue.
An important design feature that Milgrom discusses is the need for
“activity rules” that will prevent bidders from playing “wait-and-see” tac-
tics that can slow down progress in the auction. Note that in the clerical
land auctions used in the French Revolution, the 1-minute candles played
a similar role—to speed up the auction. Milgrom gives the example of an
auction in which a laxer activity rule was utilized, which led to a much
slower progression of bids and convergence to the final allocation.
In their 1992 piece studying electricity deregulation in England and
Wales, Green and Newbery discuss the design for the spot market for
electricity. What is used is a uniform price multi-unit auction in which gen-
erators submit supply schedules comprising multiple price-quantity bids.
Green and Newbery assume, for the sake of modeling, that the supply
schedules are continuous and thus invoke the “supply function equilib-
rium” model of Klemperer and Meyer (1989). There are multiple poten-
tial equilibria in this game, though Green and Newbery argue that capac-
ity constraints and/or the threat of entry will reduce the range of
potential equilibria. They then utilize calibrations of the supply function
equilibrium model using data on generation costs to demonstrate the
potential for considerable market power exercise and efficiency loss in
the duopoly market structure of England and Wales. Beyond its specific

findings, the supply function equilibrium approach to modeling elec-
tricity markets has been very influential in the literature on privatized
electricity generation markets.
Another market that, like the electricity market, operates through a
multi-unit auction mechanism is the Treasury securities auction market.
Treasury securities are typically sold by one of two mechanisms. The first
is the discriminatory or “pay-as-bid” auction in which bidders submit
multiple price-quantity bids and are charged the area under their revealed
“demand” curve (which, of course, is not their true demand curve). The
other mechanism is the “uniform price” auction, in which the aggregate
“demand” curve of the bidders is intersected with the Treasury’s supply
to determine a single market-clearing price at which all inframarginal
bids are fulfilled.
Which method is better? Discussed in Henry Goldstein’s 1962 piece
“The Friedman Proposal for Auctioning Treasury Bills,” Milton Friedman
had strong objections to the pay-as-bid mechanism in favor of the uniform
price auction. Friedman argued that the uniform price auction would
lead to competitive outcomes if no bidder had the power to affect the
market-clearing price, which would lead to straightforward bidding (in
the Milgrom sense), where each bidder would reveal his true willingness
to pay. Friedman further argued that not needing to strategize in the auc-
tion would give incentive to small bidders to participate in the auction and
thus lead to higher participation and potentially better rates for the Trea-
sury. Friedman also thought that the pay-as-bid system would create more
incentive for communication and, potentially, collusion among the bid-
ders, as it required guessing where the market-clearing price would be.
Goldstein’s article provides some argument as to why collusion in the
Treasury auction market may be difficult to sustain, a point that is also ech-
oed by Michael Rieber in the 1964 piece “Collusion in the Auction Market
for Treasury Bills.” In his “Comment on ‘Collusion in the Auction Market
for Treasury Bills’” (1964), Friedman concedes that he may have overstated
his case for the presence of potential collusion in the Treasury market but
reiterates that his other points regarding the desirability of the uniform
price auction remain unassailed.
Interestingly, the Green and Newbery article on England and Wales’s
uniform price auctions for electricity points out a case in which Fried-
man’s assumption of “no market power” does not hold true and in which
bidders can strategically withhold supply (or demand) to affect the market-
clearing price, with important allocational and revenue/cost implications.
Whether bidders have significant market power in (multi-unit) auctions is
an empirical question that depends on the elasticity of the residual supply
(or demand) function that each bidder faces. Empirically estimating the
distribution of residual supply/demand functions that a bidder may ex-

pect to face in auction, and whether a given bidder is able to strategically
“shade” her bid in response, is the topic of Hortaçsu and McAdams (2010).
The empirical strategy also allows us to compute counterfactual revenues
in the uniform price auction, under the assumption of equilibrium bid-
ding. We find, in the context of Turkish Treasury auctions, that switching
to the uniform price auction from a discriminatory format would not have
significantly increased revenues. That said, our analysis does not take into
account the differential participation incentives that Friedman hypothe-
sized, which, I am sure, will be a topic for further discussion in the next de-
cades to come.
On the topic of collusion in auctions, it is difficult to eschew a mention
to George Stigler’s classic “A Theory of Oligopoly” (1964), which appeared
in volume 72 of the journal. Aside from laying out a theory of collusive be-
havior that is embodied in the vast literature on repeated games, Stigler
had strong opinions regarding the effectiveness of public procurement
auctions. In particular, Stigler thought that publishing bids after the close
of each auction was an invitation to collusive behavior. The idea that price
transparency facilitates collusion is a proposition that has been taken for
granted for many decades, though a forthcoming paper by Takuo Sugaya
and Alex Wolitzky points out interesting counterexamples to Stigler’s in-
tuition.
The topic of empirically detecting collusion in auctions, a topic of dis-
cussion between Friedman and his respondents in the context of Trea-
sury auctions, is well represented in the JPE through two pioneering pa-
pers: Porter and Zona (1993) and Baldwin, Marshall, and Richard (1997).
Porter and Zona, especially, argue how difficult it is to formulate rigorous
empirical strategies to tell collusion apart from competition. To this date,
much of the literature on collusion in auctions has followed an ex post
strategy, where one analyzes documented bid rigging cases to gain insight
into what the ring members were doing.1
Because of lack of space, with sincere apologies, I have omitted men-
tion of many other well-known papers on auctions published by the JPE.
Still, I hope that this brief survey has managed to convey its simple point:
a lot has been written about auctions in the last 125 years, but it appears
that many questions, some more than a century old, lie open for further
inquiry. I hope that future volumes of the JPE will continue to capture
this lively discussion.
1
They could be implementing very sophisticated mechanisms to allocate rents. See,
e.g., Asker (2010).

References
Asker, John. 2010. “A Study of the Internal Organization of a Bidding Cartel.”
A.E.R. 100 (3): 724 –62.
Baldwin, Laura H., Robert C. Marshall, and Jean-Francois Richard. 1997. “Bidder
Collusion at Forest Service Timber Sales.” J.P.E. 105 (4): 657–99.
Barberis, Nicholas, Maxim Boycko, Andrei Shleifer, and Natalia Tsukanova. 1996.
“How Does Privatization Work? Evidence from the Russian Shops.” J.P.E. 104
(4): 764–90.
Friedman, Milton. 1964. “Comment on ‘Collusion in the Auction Market for
Treasury Bills.’” J.P.E. 72 (5): 513–14.
Goldstein, Henry. 1962. “The Friedman Proposal for Auctioning Treasury Bills.”
J.P.E. 70 (4): 386–92.
Green, Richard J., and David M. Newbery. 1992. “Competition in the British Elec-
tricity Spot Market.” J.P.E. 100 (5): 929–53.
Hortaçsu, Ali, and David McAdams. 2010. “Mechanism Choice and Strategic Bid-
ding in Divisible Good Auctions: An Empirical Analysis of the Turkish Treasury
Auction Market.” J.P.E. 118 (5): 833–65.
Klemperer, Paul D., and Margaret A. Meyer. 1989. “Supply Function Equilibria in
Oligopoly under Uncertainty.” Econometrica 57:1243–77.
Levasseur, Emile. 1894. “The Assignats: A Study in the Finances of the French
Revolution.” J.P.E. 2 (2): 179–202.
Milgrom, Paul. 2000. “Putting Auction Theory to Work: The Simultaneous As-
cending Auction.” J.P.E. 108 (2): 245–72.
Porter, Robert H., and J. Douglas Zona. 1993. “Detection of Bid Rigging in Pro-
curement Auctions.” J.P.E. 101 (3): 518–38.
Rieber, Michael. 1964. “Collusion in the Auction Market for Treasury Bills.” J.P.E.
72 (5): 509–12.
Sargent, Thomas J., and François R. Velde. 1995. “Macroeconomic Features of
the French Revolution.” J.P.E. 103 (3): 474 –18.
Stigler, George J. 1964. “A Theory of Oligopoly.” J.P.E. 72 (1): 44 –61.
Sugaya, Takuo, and Alexander Wolitzky. Forthcoming. “Maintaining Privacy in
Cartels.” J.P.E.
Virtue, G. O. 1895. “Public Ownership of Mineral Lands in the United States.”
J.P.E. 3 (2): 185–202.
The Economics of Crime

Steven D. Levitt
The Journal of Political Economy has played an absolutely central role in

the birth and the evolution of the literature devoted to the economics

of crime. By my calculations, eight of the 10 most highly cited works,
including Becker’s (1968) seminal article, either were published in the
journal or were coauthored by an editor of the journal. It could be cred-
ibly argued that no single journal has had such a profound influence on
a particular field of economics as the J P E has had on the economics of
crime.
The modern literature on the economics of crime traces its roots to
Becker’s remarkable publication in this journal in 1968.1 Filling 48 jour-
nal pages in a time when the typical article was fewer than 20 pages in
length, Becker’s article was a tour de force. At the heart of the paper
was a relatively straightforward price theory model of the supply of and
demand for offenses—a model of deterrence that has come to be known
as “the economic model of crime.” Interestingly, many of the predictions
of Becker’s model turn out to be quite at odds with what we observe in
the real world. For instance, in his framework, fines are a more efficient
means of punishment than imprisonment, which should make fines pre-
ferred to incarceration as a punishment for crime, but is not what we ob-
serve in practice. His model also argues that the combination of a low
probability of punishment accompanied by an extremely severe penalty
when a criminal is caught is the most efficient way to deter crime (because
detection is costly but extracting wealth once a criminal is caught is
“cheap”). Again, this is not how modern criminal justice systems func-
tion. It is not, however, the particulars of the predictions that make
Becker’s paper so remarkable, but rather the amazing breadth of facts,
insights into human nature, and theoretical conjectures contained in the
article. The basic idea underlying almost every theoretical paper in the
voluminous literature that subsequently emerged can be traced back to
Becker (1968).2 This paper would become Becker’s fourth most cited work,
with roughly 10 times as many cites as any other article in the economics
of crime literature.
One of the few theoretical insights that Becker missed in his 1968 pa-
per was the idea of “marginal deterrence,” a point developed by Stigler
(1970) in an influential piece in this journal. In Becker’s paper, all crimes
would be punished with extremely severe sanctions. As Stigler notes, how-
ever, that leads to distorted incentives on the margin. If, for example, the
punishment for robbery is the same as for murder, then a cornered rob-
ber might be willing to kill the police officers who apprehend him. More
generally, to the extent that crimes are substitutes for one another, there
1
Of course, as Becker (1968) notes, a number of great economic thinkers had written
about some of the same issues more than 100 years earlier.
2
I have often joked that Becker did a severe disservice to the economics of crime field by
writing a paper that was too good and left too little for those who followed him to improve
on.

is a role for distorting expected punishments in a manner that shifts crimes
toward those that are least socially costly.
Interestingly, although a great deal of academic effort was devoted gen-
erally to exploring the economic model of crime in the 1970s and 1980s,
these contributions were largely incremental, and the JPE published very
little work on crime in those decades.3 It was not until the influential work
of Sah (1991) that the journal once again began to put its imprimatur on
the field. Sah’s paper extends Becker (1968) by endogenizing individual
perceptions of punishment and adding dynamics to the economic model
of crime. Because individual perceptions depend critically on local condi-
tions, large differences in criminal propensities can emerge and persist
across groups or communities that are observably similar in Sah’s study.
This work was important not only within the economics of crime but also
more generally as an early example of endogenizing individual beliefs.
The late 1990s marked a turning point in the field. Up until then, vir-
tually all of the most-cited works in the area were theory papers; since
that time, highly cited papers have almost exclusively had a large empir-
ical component.4 An early example of data-driven crime studies in J P E is
Levitt (1998), which exploits the natural experiment associated with the
sharp discontinuity in expected punishment as an offender transitions
from the juvenile justice system to the adult system. The juvenile justice
system emerged in the 1800s as a response to the perception that mixing
juveniles and adults in a prison system constituted cruel and unusual
punishment for juveniles and had a corrupting influence. To this date,
all states handle crimes committed by juveniles and adults very differently.
Conveniently for the economist wishing to study the impact of punish-
ment on crime, there is state-level variation in the age of majority (the
age at which jurisdiction switches from the juvenile to the adult system)
and in the relative severity of the juvenile and adult justice systems. Levitt
finds strong evidence that criminal behavior is highly responsive to changes
in expected punishment that arise with the transition to the adult sys-
tem. In a field bedeviled by reverse causality (Fisher and Nagin 1978), this
study was one of the first to provide empirical support for the economic
model of crime using plausibly exogenous variation.
Duggan’s (2001) article “More Guns, More Crime” proved influential
in shaping an important public policy debate surrounding the impact
of gun availability on crime. A few years earlier, in a highly controversial
and frequently challenged paper published in the Journal of Legal Studies,
Lott and Mustard (1997) argued empirically that laws making it easier for
3
This is particularly notable given that George Stigler edited the journal for almost the
entirety of those two decades.
4
Becker, Murphy, and Grossman (2006) is an exception to this trend. This paper is an
“old-school” application of price theory to the question of illegal drugs. While disarmingly
simple, this paper sheds a plethora of important and nonintuitive policy-relevant insights.

people to carry concealed weapons sharply reduced crime. A number of
papers provided specific critiques of Lott and Mustard’s paper; see, for
instance, Ludwig (1998). Duggan (2001) took a very different approach
to the problem. Rather than focusing narrowly on concealed weapons,
he tackled the issue of guns and crime in a more holistic manner. An
enormous obstacle to such a study was the absence of data; somewhat
amazingly, there are no reliable data about gun ownership at the state
or local level, making it extremely difficult to address the issue with stan-
dard micro-empirical techniques. Duggan’s first insight was to recognize
that circulation counts of magazines devoted to gun issues—which were
available at the local level—could be used as a credible proxy for gun
ownership.5 Duggan’s most important conclusion is that increases in
gun ownership (as proxied by gun magazine sales) strongly predict gun
homicides but are essentially unrelated to nongun homicides and other
crimes.
Fisman and Miguel (2007) provide another example of creative empir-
ical work. They address the question of corruption and, in particular, the
extent to which observed differences in corruption across countries can
be tied to either cultural norms or enforcement of law. Corruption is a
notoriously difficult question to address empirically, because it typically
requires cross-country comparisons in a setting in which data are rare
and of poor quality. Fisman and Miguel exploit a unique natural experi-
ment tied to diplomatic immunity with respect to parking violations for
United Nations diplomats. Prior to 2002, there is complete diplomatic
immunity. Interestingly, Fisman and Miguel find a strong correlation be-
tween widely used measures of corruption in a country (based on surveys
of people conducting business in that country) and the number of park-
ing violations generated by the country’s UN diplomats. This suggests
that the sort of cultural norms that lead to corruption spill over into other
behaviors, even among diplomats stationed thousands of miles away. Af-
ter 2002, New York police were given the power to punish parking viola-
tions. In one of the clearest examples ever of the power of deterrence, vi-
olations fall by 98 percent when legal enforcement becomes possible.
Another influential empirical contribution published in JPE is Drago,
Galbiati, and Vertove (2009), which exploits a remarkable natural experi-
ment in Italy. At Pope John Paul II’s urging, Italian lawmakers passed a law
that led to the release on August 1, 2006, of any inmate with less than 3 years
remaining on his or her sentence. Forty percent of all Italian prisoners
5
One might think it would be an impossible task to convince the reader of the validity of
a proxy when the reason a proxy is needed in the first place is the absence of direct data on
guns. Duggan, however, manages to do this methodically and artfully by demonstrating a
high degree of correlation between his proxy, which is available both at a geographically
disaggregated level and with variation over time, and other gun-related outcomes that
are available only at high levels of geographic aggregation or in the cross section.

were released that day! Notably, if a released prisoner was convicted of a
new crime, an amount of time equal to the commuted sentence was added
to the new sentence. Therefore, some criminals (i.e., those who had 3 years
left to serve when they received early release) faced the specter of having
an extra 3-year penalty if they got caught doing another crime. Those
who were only a month away from completing their sentences on August 1,
2006, faced only an extra month. This policy thus induced plausibly exog-
enous variation in the severity of punishment, allowing Drago et al. to es-
timate how responsive former prisoners were to deterrence. They find that
the Italian ex-convicts are extremely responsive to these variations in ex-
pected punishment, providing some of the most compelling evidence there
is in support of deterrence.
After nearly 50 years of theoretical and empirical work, much is now
understood about the economics of crime. Nonetheless, huge questions
remain unanswered. It remains a mystery, for instance, why crime rose so
much in the 1960s. We continue to have relatively little insight into the
question of why some individuals become criminals and others do not. It
is also difficult to explain why crime varies so much both spatially and
temporally (although for a start on this question, see Glaeser, Sacerdote,
and Scheinkman [1996] and Glaeser and Sacerdote [1999]). In some
sense, however, public policies to reduce crime (many of them informed
by economic thinking) have proven too successful from the perspective
of the academic interested in studying crime. With the crime rate at less
than half the level it was two decades ago in the United States and lower
almost everywhere else in the world as well, the demand for crime re-
search has no doubt also been diminished.
References
Becker, Gary S. 1968. “Crime and Punishment: An Economic Approach.” J.P.E.
76 (2): 169 –217.
Becker, Gary S., Kevin Murphy, and Michael Grossman. 2006. “The Market for
Illegal Goods: The Case of Drugs.” J.P.E. 114 (1): 38 – 60.
Drago, Francesco, Roberto Galbiati, and Pietro Vertova. 2009. “The Deterrent Ef-
fects of Prison: Evidence from a Natural Experiment.” J.P.E. 117 (2): 257– 80.
Duggan, Mark. 2001. “More Guns, More Crime.” J.P.E. 109 (5): 1086 –1114.
Fisher, Franklin M., and Daniel S. Nagin. 1978. “On the Feasibility of Identifying
the Crime Function in a Simultaneous Model of Crime Rates and Sanction
Levels.” In Deterrence and Incapacitation: Estimating the Effects of Criminal Sanc-
tions on Crime Rates. Washington, DC: Nat. Acad. Sci.
Fisman, Raymond, and Edward Miguel. 2007. “Corruption, Norms, and Legal
Enforcement: Evidence from Diplomatic Parking Tickets.” J.P.E. 115 (6): 1020–
48.
Glaeser, Edward, and Bruce Sacerdote. 1999. “Why Is There More Crime in Cit-
ies?” J.P.E. 107, no. 6, pt. 2 (December): S225 – S258.
Glaeser, Edward, Bruce Sacerdote, and Jose Scheinkman. 1996. “Crime and So-
cial Interactions.” Q. J.E. 111 (2): 507– 48.

Levitt, Steven D. 1998. “Juvenile Crime and Punishment.” J.P.E. 106 (6): 1156–
85.
Lott, John, and David Mustard. 1997. “Crime, Deterrence, and Right-to-Carry
Concealed Handguns.” J. Legal Studies 26 (1): 1– 68.
Ludwig, Jens. 1998. “Concealed-Gun-Carrying Laws and Violent Crime: Evi-
dence from State Panel Data.” Internat. Rev. Law and Econ. 18:239–54.
Sah, Raj. 1991. “Social Osmosis and Patterns of Crime.” J.P.E. 99 (6): 1272–95.
Stigler, George J. 1970. “The Optimum Enforcement of Laws.” J.P.E. 78 (3): 526–
36.
Experimental Economics in the

Journal of Political Economy
John List
Similar to the spirit in which astronomy draws on the data from mechan-
ics and physics to make deeper insights, experiments can help to provide
the necessary behavioral principles to permit sharper inference from nat-
urally occurring data in economics. Indeed, experiments have helped to
uncover the key causes and underlying conditions necessary to produce
data patterns observed in the field. At the same time, the experimental
method has gone beyond merely a complementary role, as today experi-
ments generate data that provide crisper tests of economic theory than
previously achieved.
In contrast to other sciences, the experimental approach has not pro-
gressed to the point of being the cornerstone of the scientific method in
economics just yet, but it has progressed sufficiently to find itself in the
center of key debates and is well represented in every major economics
journal. This was not always the case. Indeed, the Journal of Political Econ-
omy has played a central role in the general acceptance of the experi-
mental approach.
To showcase this fact, I focus narrowly on two areas of experimental in-
quiry: market institutions and individual choice. Given that two standard
assumptions that underlie standard economic theory are that (i) markets
are cleared via the institution of Walrasian tâtonnement and (ii) agents
aim to maximize utility, it is fitting that the J P E has contributed to both
experimental research agendas.

I. Market Institutions
Empirical attempts to explore the effects of institutions have been plen-
tiful, but traditional tools have had limited success. In what is broadly
believed to be the first market experiment, Chamberlin (1948) used Har-
vard students participating in decentralized one-shot bargaining mar-
kets to explore whether price and quantity converged to the intersection
of supply and demand. Chamberlin observed that volume was typically
higher and prices typically lower than predicted by competitive models
of equilibrium. Efficiency was also frustrated in these bilateral negotiat-
ing markets.
Vernon Smith, a Harvard student at the time and one of Chamberlin’s
experimental subjects, later refined Chamberlin’s work (Smith 1962) by
varying two key aspects of the experimental design: (i) centrally occur-
ring open outcry of bids and offers (commonly termed “double auction
markets”) and (ii) multiple market periods (allowing agents to learn).
Empirical results from Smith’s experiments were staggering—quantity
and price levels were very near competitive levels—and served to present
the first evidence that Walrasian tâtonnement, conducted by a central
auctioneer, was not necessary for market outcomes to approach neoclas-
sical expectations. It is fair to say that this general result remains one of
the most robust findings in experimental economics today.
Smith complemented this early work with another JP E study in 1965
that presented the “excess rent” model, which is defined as the total rent
that would be obtained if all agents who want to trade at the prevailing
price were to trade minus the total rent at the competitive equilibrium.
In his experimentum crucis, Smith (1965) reports that predictions of the
excess rent model are more consonant with his data than the Walrasian
model using a series of double auction markets.
Several research areas have arisen from these seminal contributions of
Chamberlin and Smith. List (2004) represents a field experiment that
moves the Chamberlin bilateral bargaining institution from the lab to
the natural setting in which the actors actually undertake decisions. List’s
field experiment therefore represents an early empirical test in an actual
marketplace in which agents engage in face-to-face continuous bilateral
bargaining in a multilateral market context. Much as in Smith’s (1962,
1965) setup, the market mechanics in these bilateral bargaining markets
are not Walrasian. In contrast to Smith’s work, however, List’s design shifts
the task of adaptation from the auctioneer to the agents; in doing so, the
market structure transforms the problem of stability of equilibria as a ques-
tion about the behavior of actual people as a psychological question—as
opposed to a question about an impersonal market. Two key results of
List’s study are the high efficiency obtained in his markets and the strong

tendency for exchange prices to approach the neoclassical competitive
model predictions, especially in symmetric markets.
In a dramatic extension of Smith’s (1962, 1965) double auction mar-
ket, Gode and Sunder (1993) design markets that compare results from
human traders with those from “zero-intelligence traders.” These traders
are irrational in the sense that they are not maximizing; rather they are
submitting random bids and offers. The sole constraint is that buyers can-
not pay more than their willingness to pay and sellers cannot sell for less
than their marginal cost. Their amazing results suggest that even with such
traders, the market approaches 100 percent efficiency. Their conclusion
is that the efficiency of the double auction market arises from the institu-
tion itself, not from market experience, learning, or the like (this is not to
imply that there is no role for human traders to significantly improve
their ability to gain surplus when learning about the state of the world
in double auction markets; see Plott and Sunder 1982).
Of course, there are hundreds of other excellent examples of notewor-
thy experiments in this area that could fill more than five tomes, but I
need to move on to discuss individual choice, however unjust and risky
that might appear.
II. Individual Choice

A second important strand of experimental work that the JPE has con-
tributed to advancing is in the area of individual choice. An early choice
experiment that set into motion several branches of subsequent work
was due to Thurstone (1931). In a series of hypothetical choice exercises,
Thurstone attempted to measure individuals’ indifference curves by ask-
ing individuals to make choices between bundles of hats and coats, hats
and shoes, and shoes and coats. In an exciting early test of ordinal prefer-
ence theory, Thurstone concluded that indifference curves could ade-
quately represent his choice data.
Thurstone’s work drew the ire of Wallis and Friedman (1942), who in a
provocative study argued that the choice set was ill-specified and that the
choices themselves were hypothetical. For example, they argued that “it is
questionable whether a subject in so artificial an experimental situation
could know what choices he would make in an economic situation; not
knowing, it is almost inevitable that he would, in entire good faith, system-
atize his answers in such a way as to produce plausible but spurious re-
sults” (179–80).
The Wallis and Friedman (1942) critique led immediately to a follow-
up study due to Rousseas and Hart (1951), who constructed (arguably) a
more realistic choice scenario whereby experimental subjects in the lab

made choices over breakfast items: number of eggs and strips of bacon,
for example. In a key innovation, they had individuals make a single
choice repeatedly. This avoided certain complications, but the aggrega-
tion of choices across individuals to test the relevant theory became an
issue. Related to this inquiry was the early lab experimental work of Mos-
teller and Nogee (1951), which explored utility over additional money
income. This overarching line of research even induced a mini debate
about the merits, and what had been accomplished, within the area of
experimental economics that was published in the JPE (Castro and Wein-
garten 1970; Naylor 1972).
This early work anticipated several lines of interesting research. For
example, within the area of hypothetical choice, Cummings et al. (1997)
explored whether referenda for nonmarketed goods and services were in-
fluenced by hypothetical bias. They found that considerably more people
voted in favor of the referendum when it was hypothetical versus when it
was real. In an early field experiment, List and Shogren (1998) executed
hypothetical and real auctions and reported similar results: the average
bid in the hypothetical auction was roughly three times higher than in
the real auction.
An area of individual choice following in the spirit of Wallis and Fried-
man (1942) is on the “demand side” of charitable fund-raising. An early
field experiment on the demand side is summarized by List and Lucking-
Reiley (2002), who raised money for the Center for Environmental Policy
Analysis at the University of Central Florida in a field experiment. They
found key success for the theoretical prediction of Andreoni (1998): that
seed money increases the amount of public-good provision in a charita-
ble fund-raiser, from zero to some positive equilibrium level G * (greater
than or equal to the threshold level). This research led to many related
field experiments, with the recent work of Andreoni, Rao, and Tracht-
man (2017) and Perez-Truglia and Cruces (2017) representing two ex-
cellent examples.
Relatedly, to explore the importance of social preferences, List (2006)
carries out various field experiments analyzing gift exchange. The games
have buyers making price offers to sellers, and in return sellers select the
quality level of the good provided to the buyer. The artefactual field exper-
imental (lab-like) results mirror the typical lab findings with other subject
pools: strong evidence for social preferences was observed through a pos-
itive price and quality relationship. Yet, when the environment is moved
to the marketplace via a natural field experiment, where dealers are un-
aware that their behavior is being recorded as part of an experiment, little
statistical relationship between price and quality emerges. Similar in-
sights on the effect of the situation in an entirely different setting and
game form can be found in List (2007).

Much as in the brief summary of market institutions above, there are
several other examples of noteworthy individual choice experiments. Cam-
erer (1998) is one neat example that explores betting choices at a race-
track. Cason and Plott (2014) explore recent challenges of revealed pref-
erence theory by using lab experiments with interesting framings. They
conclude that “mistakes in choices obscured by a possible error at the
foundation of the theory of framing can masquerade as having been pro-
duced by nonstandard preferences” (1235). Kahneman, Knetsch, and
Thaler (1990), in seminal work, test how well predictions of reference-
dependent theory explain trading decisions of lab subjects. They find ev-
idence consonant with reference dependence, and this work has touched
off a line of research that continues to thrive today.
References
Andreoni, James. 1998. “Toward a Theory of Charitable Fund-Raising.” J.P.E. 106
(6): 1186–1213.
Andreoni, James, Justin M. Rao, and Hannah Trachtman. 2017. “Avoiding the
Ask: A Field Experiment on Altruism, Empathy, and Charitable Giving.” J.P.E.
125 (3): 625–53.
Camerer, Colin F. 1998. “Can Asset Markets Be Manipulated? A Field Experiment
with Racetrack Betting.” J.P.E. 106 (3): 457–82.
Cason, Timothy N., and Charles R. Plott. 2014. “Misconceptions and Game Form
Recognition: Challenges to Theories of Revealed Preference and Framing.”
J.P.E. 122 (6): 1235–70.
Castro, Barry, and Kenneth Weingarten. 1970. “Toward Experimental Econom-
ics.” J.P.E. 78 (3): 598–607.
Chamberlin, Edward. 1948. “An Experimental Imperfect Market.” J.P.E. 56 (2):
95–108.
Cummings, Ronald G., Steven Elliott, Glenn W. Harrison, and James Murphy.
1997. “Are Hypothetical Referenda Incentive Compatible?” J.P.E. 105 (3):
609–21.
Gode, Dhananjay K., and Shyam Sunder. 1993. “Allocative Efficiency of Markets
with Zero-Intelligence Traders: Market as a Partial Substitute for Individual Ra-
tionality.” J.P.E. 101 (1): 119–37.
Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler. 1990. “Experimental
Tests of the Endowment Effect and the Coase Theorem.” J.P.E. 98 (6): 1325–
48.
List, John A. 2004. “Testing Neoclassical Competitive Theory in Multilateral De-
centralized Markets.” J.P.E. 112 (5): 1131–56.
———. 2006. “The Behavioralist Meets the Market: Measuring Social Prefer-
ences and Reputation Effects in Actual Transactions.” J.P.E. 114 (1): 1–37.
———. 2007. “On the Interpretation of Giving in Dictator Games.” J.P.E. 115 (3):
482–93.
List, John A., and David Lucking-Reiley. 2002. “The Effects of Seed Money and
Refunds on Charitable Giving: Experimental Evidence from a University Cap-
ital Campaign.” J.P.E. 110 (1): 215–33.

List, John A., and Jason F. Shogren. 1998. “Calibration of the Difference between
Actual and Hypothetical Valuations in a Field Experiment.” J. Econ. Behavior
and Org. 37 (2): 193–205.
Mosteller, Frederick, and Philip Nogee. 1951. “An Experimental Measurement
of Utility.” J.P.E. 59 (5): 371– 404.
Naylor, Thomas H. 1972. “Experimental Economics Revisited.” J.P.E. 80 (2): 347–
52.
Perez-Truglia, Ricardo, and Guillermo Cruces. 2017. “Partisan Interactions: Evi-
dence from a Field Experiment in the United States.” J.P.E. 125 (4): 1208– 43.
Plott, Charles R., and Shyam Sunder. 1982. “Efficiency of Experimental Security
Markets with Insider Information: An Application of Rational-Expectations
Models.” J.P.E. 90 (4): 663 –98.
Rousseas, Stephen W., and Albert G. Hart. 1951. “Experimental Verification of a
Composite Indifference Map.” J.P.E. 59 (4): 288 –318.
Smith, Vernon L. 1962. “An Experimental Study of Competitive Market Behav-
ior.” J.P.E. 70 (2): 111–37.
———. 1965. “Experimental Auction Markets and the Walrasian Hypothesis.”
J.P.E. 73 (4): 387–93.
Thurstone, Louis L. 1931. “The Indifference Function.” J. Soc. Psychology 2 (2):
139 –67.
Wallis, W. Allen, and Milton Friedman. 1942. “The Empirical Derivation of Indif-
ference Functions.” In Studies in Mathematical Economics and Econometrics, in
Memory of Henry Schultz, edited by Oscar Lange, Francis McIntyre, and Theo-
dore O. Yntema, 175–89. Chicago: Univ. Chicago Press.


The Past, Present, and Future of Economics: A Celebration of The 125-Year Anniversary of The JPE and of Chicago Economics

Uploaded by

Copyright:

Available Formats

The Past, Present, and Future of Economics: A Celebration of The 125-Year Anniversary of The JPE and of Chicago Economics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Past, Present, and Future of Economics: A Celebration of The 125-Year Anniversary of The JPE and of Chicago Economics

Uploaded by

Copyright:

Available Formats

The Past, Present, and Future of Economics:

A Celebration of the 125-Year Anniversary

The Journal of Political Economy is celebrating its 125th anniversary this

[ Journal of Political Economy, 2017, vol. 125, no. 6]

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Chicago and the Origins of Modern

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

is continuous, nonempty, and convex-valued, and if each Ui is continu-

opi 5 1g be the set of normalized prices. Each consumer has a real-

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Economic Growth: The Past, the Present, and the

Is there some action a government of India could take that

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Physical Capital Accumulation, Spillovers, and Growth

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Human Capital, Fertility, and Growth

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Political Economics in the Journal of Political

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Aggregative Fiscal Policy

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

II. Debt Finance

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

III. The Role of Commitment

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

IV. The Size of Government

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

Business Cycles and International Trade

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

This content downloaded from 202.003.077.210 on January 30, 2018 01:26:17 AM

International Trade and Exchange Rates