Beruflich Dokumente
Kultur Dokumente
Economics"
Author(s): Wolfgang Pesendorfer
Source: Journal of Economic Literature , Sep., 2006, Vol. 44, No. 3 (Sep., 2006), pp. 712-
721
Published by: American Economic Association
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic
Literature
WOLFGANG PESENDORFER*
712
The Endowment Effect: In standard models that allow players' utility to depend
consumer theory, demand is a function on of
their own and their opponent's monetary
wealth and prices but does not depend on
payoffs. Closest to standard models is the
work
the composition of the endowment. Thaler, of Ernst Fehr and Klaus M. Schmidt
(chapter 9) in which the utility of a player
in his 1980 paper, coined the term "endow-
ment effect" to describe the experimental
depends on the monetary payoff of all play-
finding that subjects value a good moreers.
if There
it is additional evidence that players
not
is part of their endowment than if it is only care about the material outcomes of
not.
The endowment effect is addressedtheir by opponents but also about their oppo-
assuming that agents treat additions tonent's
theircharacter. As a result, players may care
about what other players reveal during the
endowments differently from subtractions.
Hence, the endowment point is treatedcourse
as a of the interaction about their charac-
reference point and agents are assumed
ter. John Geanakoplos, David Pearce, and
have a kink in their valuations around the
Ennio Stacchetti (1989) develop a frame-
endowment point. work that allows for such interdependence
Hyperbolic Discounting: Standard dy- and Matthew Rabin (chapter 10) applies that
namic decision theory assumes that framework.
intertemporal choices do not depend on the As these examples illustrate, the focus of
decision date. Whether the agent chooses research in behavioral economics is on indi-
consumption in the initial period or sequen- vidual choice and the motives underlying
tially has no effect on the choice if the budg- that choice. Starting with an experimental
et constraint is the same in both cases. finding that shows violations of standard eco-
Strotz (1955/56) develops a model of deci-
nomic assumptions, research in behavioral
economics
sion making that relaxes this assumption. As proceeds by introducing new
variables that are used to "parameterize"
David Laibson (1997) points out, this model
can be used to address experimental deviations
evi- from standard models. In many
cases, the new variable is used to describe a
dence of an "immediacy effect" in behavior:
subjects have a tendency to choose earlier,
"bias" in decision making, i.e., some form of
smaller rewards over later, larger rewards
irrationality or systematic mistake.
when the earlier reward offers immediate This essay contains three observations on
consumption but reverse this preferencethe direction and the focus of behavioral
when both rewards are delayed. economics. The next section previews the
Social Preferences: Standard incentivethree observations and sections 3-5 discuss
theory assumes that the choices of an agent in more detail. Section 6 concludes.
them
depend only on his own monetary payoff. 2. Three Observations
This assumption has been challenged by a
variety of experiments, perhaps most (1) Much of behavioral economics builds
famously the experiments on ultimatum bar-on experimental evidence in which a new
gaining. Introduced by Werner Guth, Rolf variable that is ignored in standard econom-
Schmittberger, and Bernd Scbwarze (1982), ic models is shown to "matter." While the
this experiment pairs subjects to play a sim-
new variable may be observable in experi-
ple bargaining game. The proposer makesmental
an settings, it is often unobservable
offer and the responder accepts or rejects. when
A the researcher must deal with eco-
nomic
rejection leaves both players with a zero pay- (field) data. A prominent example of
off. Responders routinely reject small offers
this is the reference point in prospect theo-
and therefore do not maximize their selfish ry. The reference point can be manipulated
monetary payoff. To address this evidence, in experimental settings but is essentially
behavioral economists have introduced unobservable outside the laboratory. This
which the agent previously bought an asset, value of asset when it was purchased and
or the expected earnings at the end of a does not adjust over time. Moreover, it is
working day. Colin F. Camerer (chapter 5) assumed that assets are treated in isolation so
illustrates how prospect theory is applied to that the agent cannot use a different asset
a variety of economic settings. with superior return to recover losses.
Shlomo Benartzi and Thaler (1997) sug- Camerer, Linda Babcock, George Loew-
gest an explanation of the equity premium enstein, and Thaler (chapter 19) suggest that
based on prospect theory. Investors' utility New York cab drivers use a daily income tar-
depends on the annual return of their port- get as a reference point and stop working
folio. In particular, investors are loss averse after the target is achieved. Income targeting
and compute gains and losses with respect to is an extreme form of loss aversion with the
the value of their portfolio at the beginning target income level as a reference point. As
of the planning horizon (approximately one Henry S. Farber (2005) points out, this
year). In this model, the reference point is implies that cab drivers quit early on days
the value of the asset in the portfolio at the where it is easy to make money and quit late
beginning of the year (the beginning of the on days when it is hard to make money.
"planning horizon"). If investors are suffi- Farber reexamines the evidence presented
ciently risk averse around the reference by Camerer et al. and finds that there is no
point, this formulation will imply a large evidence of a target-income behavior among
equity premium. To match the observed New York cab drivers. Farber concludes that
equity premium, it is essential that gains and the primary determinant of stopping work is
losses are computed over relatively short hours worked and that cumulative income is
periods (annually rather than over five or ten at most weakly related to stopping work.
years) and that the reference point adjusts Farber (2004) estimates a more general util-
over time. If the reference point stays con- ity function that allows intertemporal substi-
stant (for example at the price the asset was tution but also includes loss aversion around
purchased), then over time the effect of loss a reference point. For that model, he finds
aversion fades. evidence that a reference level may effect
Studies on stock trading (Terrance Odean the labor supply decision on a given day. But,
1998) and housing transactions (David even for a given driver, the estimated refer-
Genesove and Christopher Meyer 2001) find ence level varies substantially from day to
that agents are less likely to sell assets thatday. Farber concludes that "This [variation]
have incurred losses than assets that have seriously limits the predictive power of the
incurred gains. Prospect theory explains thisreference point, and undermines the useful-
behavior by assuming that investors treat ness of the construct of the reference income
each asset as a separate decision and use the level as a determinant of labor supply"
purchase price as a reference point. Investors (Farber 2004, p. 4).
are assumed to tolerate more risk when they In all these applications, we cannot
try to recover a loss than when they try toobserve variations in the reference point in
increase their gains. As a result, investors the same way that experimenters can fix
may hold on to assets that have made lossesand manipulate the reference point.
even if those assets yield a lower expected Therefore, the reference point becomes a
return than some other asset in their portfo- parameter that is calibrated to match the
lio.1 In this case, the reference point is theobserved data. But unlike risk aversion or
the discount factor, the reference point
1 Nicholas Barberis and Wei Xiong (2006) show that
need not be consistent across applications
this argument is flawed. A formal model corresponding to
the intuitive description by Camerer (chapter 5) may notor even consistent across periods for the
be consistent with the evidence in Odean. same application. Essentially, it captures a
large purchase. As a consequence, it is diffi- Rather, the calibration has to be done sepa-
cult to use economic data to calibrate utility rately for every context. As a result, the exper-
functions that depend on those variables. imental evidence offers little quantitative
guidance for economists.
3.3 Calibrating from Experiments
Since novel behavioral variables such as 4. Humans as Stubborn Operators of
Broken Machines
the reference point can be manipulated in
experiments, it seems tempting to use exper- Behavioral economics grew out of a cri-
imental data to estimate the parameters tique
of of standard economic assumptions.
This tradition sometimes leads to a view of
utility functions and then apply these esti-
behavioral economics as a collection of
mates to economic applications. The experi-
mental evidence on discounting offers"biases,"
a that is, violations of standard
good illustration of the difficulty of this assumptions in economics.
approach. Behavioral models often take as a starting
The impatience displayed by subjects in point a standard economic model and re-
experimental settings is striking. Thaler interpret the model as a description how the
(1981) finds annual discount rates over 300 person thinks and feels. Next, an (often
percent for time-delays of a month. A recent compelling) case is made that many of the
experiment by Benhabib, Bisin, and assumptions are unrealistic because humans
Schotter (2005) finds annual discount rates cannot perform the difficult mental tasks
of 472 percent. These findings are striking embodied in the formalism. The mistake or
because there are market prices for these bias is typically modeled by assuming that
intertemporal trade-offs. Real interest rates some aspect of the optimization procedure
are typically in the low single digits-orders in the decision model is done incorrectly.
of magnitude off the experimental findings. Typically, behavioral economists take great
Moreover, economic agents tend to hold care in motivating the particular mistake and
assets that offer those low rates of return.2 provide detailed evidence that humans
A second striking feature of the experi- indeed have trouble performing the task
ments is the large cross-study differences in required by the model. Finally, the psycho-
estimated discount rates. Frederick, logical and sometimes also the economic
Loewenstein, and O'Donoghue write "theconsequences of the model with the bias are
spectacular cross-study differences in dis- explored. However, rarely do these theories
count rates also reflect the diversity of con-ask whether--once the mistake is taken for
siderations that are relevant in intertemporal granted-the original model makes sense.
choices and that legitimately affect different In other words, why would humans go to the
types of intertemporal choices differently"trouble of maximizing objective fiunctions,
(p. 211). The argument by Frederick, formulating complicated beliefs only to get
Loewenstein, and O'Donoghue implies that things systematically wrong?
experimental evidence cannot be used to cal- For example, consider the O'Donoghue-
ibrate utility functions that then are applied Rabin model of naive, dynamically inconsis-
in (very different) economic contexts. tent decisionmakers. These decisionmakers
maximize a standard utility function but
have incorrect beliefs about their own future
2 Of course, there is substantial diversity in savings
behavior. As Laibson et al. (2005) point out, many U.S. behavior. The assumption that humans have
households hold credit card debt at high interest rates wrong, even systematically biased, beliefs
which suggests a great deal of impatience. On the other
hand, savings behavior in Germany has puzzled
seems plausible and strikes a cord with most
researchers (see Axel Boersch-Supan et al. 2001) because readers. At the same time, it seems highly
it suggests a great deal of patience. implausible that individuals would go to the
motives are behind the assumed behavior. objective function or process probabilities
By contrast, economic psychology ex- incorrectly.
plores the psychological mechanisms behind Behavioral economics emphasizes the
economic behavior. An implication of this context-dependence of decision making. A
shift in focus is that the interesting observa- corollary of this observation is that it is diffi-
cult to extrapolate from experimental set-
tions of behavioral theories are often psycho-
logical, that is, about the mental processes tings to field data or, more generally,
behind behavior. The objective of a broadly economic settings. Moreover, not all vari-
applicable economic framework becomes ables that are shown to matter in some
less important, is left for future work or is experiment are useful or relevant in eco-
viewed as a psychologically unrealistic goal. nomic applications. The question whether a
6. Conclusions particular variable is useful or even observ-
able for economics rarely comes up in
Behavioral economics has reached the sta- behavioral models, yet the success or failure
tus of an established discipline. The central of modeling innovations often depends on its
issues and theories discussed in Advances answer.
Barberis, Nicholas,
At the same time, behavioral economics and Wei Xiong. 2005. "What Driv
the Disposition Effect? An Analysis of a Lon
remains a discipline that is organized around
Standing Preference-Based Explanation." Princeto
the failures of standard economics. University.
The typ-Mimeo.
Benartzi, Shlomo, and Richard H. Thaler. 1995.
ical contribution starts with a demonstration
of a failure of some common economic "Myopic Loss Aversion and the Equity Premium
Puzzle." Quarterly Journal of Economics, 110(1):
73-92.
assumption (usually in some experiment)
B6rsch-Supan,
and proceeds to provide a psychological Axel, Anette Reil-Held, Ralf Rodepeter,
Reinhold Schnabel, and Joachim Winter. 2001. "The
explanation for that failure. This symbiotic
German Savings Puzzle." Universitat Mannheim.
Mimeo.
relationship with standard economics works
Camerer, Colin F. 2000. "Prospect Theory in the Wild:
well as long as small changes to standard
Evidence from the Field." In Choices, Values, and
assumptions are made. In that case, the Frames, ed. D. Kahneman and A. Tversky.
behavioral evidence can be the impetus for
Cambridge; New York and Melbourne: Cambridge
University Press; New York: Russell Sage
small changes of standard models that leave
Foundation, 288-300.
the basic structure of the theory intact.
Camerer, Colin F., Linda Babcock, George
Loewenstein, and Richard H. Thaler. 1997. "Labor
Examples of this are theories that allow
Allais-type behavior and loss aversion, or theSupply of New York City Cabdrivers: One Day at a
Time." Quarterly Journal of Economics, 112(2):
models that allow a preference for commit-
407-41.
ment (such as Strotz model of consistent
Epstein, Larry G., and Stanley E. Zin. 1989.
"Substitution, Risk Aversion, and the Temporal
planning).
Behavior of Consumption and Asset Returns: A
With the success of behavioral economics, Theoretical Framework." Econometrica, 57(4):
937-69.
more radical departures are being consid-
ered. In that case, the traditional blue- Farber, Henry S. 2004. "Reference-Dependent
Preferences and Labor Supply: The Case of New
print-evidence plus small modification-is York City Taxi Drivers."
less compelling. For example, the economic Farber, Henry S. 2005. "Is Tomorrow Another Day?
model of individual behavior as the result of The Labor Supply of New York City Cabdrivers."
Journal of Political Economy, 113(1): 46-82.
constrained optimization is not well suited to Fehr, Ernst, and Klaus M. Schmidt. 1999. "A Theory of
describe wrong or biased behavior. There is Fairness, Competition, and Cooperation." Quarterly
no "small" modification of the standard Journal of Economics, 114(3): 817-68.
Frederick, Shane, George Loewenstein, and Ted
model that can deal convincingly with the
O'Donoghue. 2002. "Time Discounting and Time
hypothesis that people are wrong about theirPreference: A Critical Review." Journal of Economic