Sie sind auf Seite 1von 58

Journal of Economic Literature 2012, 50:2, 368–425

http:www.aeaweb.org/articles.php?doi=10.1257/jel.50.2.368

Economic Incentives and Social


Preferences: Substitutes
or Complements?
Samuel Bowles and Sandra Polanía-Reyes*

Explicit economic incentives designed to increase contributions to public goods and to


promote other pro-social behavior sometimes are counterproductive or less effective
than would be predicted among entirely self-interested individuals. This may occur
when incentives adversely affect individuals’ altruism, ethical norms, intrinsic motives
to serve the public, and other social preferences. In the fifty experimental studies that we
survey, these effects are common, so that incentives and social preferences may be either
substitutes (crowding out) or complements. We provide evidence for four mechanisms
that may account for these incentive effects on preferences, based on the fact that incen-
tives may (i) provide information about the person who implemented the incentive, (ii)
frame the decision situation so as to suggest appropriate behavior, (iii) compromise a
control averse individual’s sense of autonomy, and (iv) affect the process by which peo-
ple learn new preferences. An implication of the fact that incentives affect preferences
is that the evaluation of public policy must be restricted to allocations that are support-
able as Nash equilibria when account is taken of these crowding effects. We show that
well designed fines, subsidies, and the like minimize crowding out and may even do the
opposite, making incentives and social preferences complements rather than substitutes.
( JEL D02, D03, D04, D83, E61, H41, Z13)

1.  Introduction the President. “People worked long hours,”


he remembered in a recent communication

T homas Schelling recalls his “exciting and


stimulating times” in the early 1950s as
a young staffer in the Executive Office of
to one of us, “and felt compensated by the
sense of accomplishment and . . . personal
importance. Regularly a Friday afternoon
* Bowles: Santa Fe Institute and University of Siena. Irlenbusch, Magnus Johannesson, Esther Kessler, Nicola
Polanía-Reyes: University of Siena and University Col- Lacetera, Maria Claudia López, Thomas Schelling, Carlos
lege London. For their contributions to this paper, we Sickert-Rodriguez, David Skuse, Joel Sobel, John Stranlund,
thank Mahzarin Banaji, Abigail Barr, Iris Bohnet, Stephen Elisabeth Wood, and Erte Xiao, as well as three anonymous
Burks, Juan Camilo Cárdenas, Jeff Carpenter, David Ech- referees and the editor. Financial support by the Behavioral
everry, Ernst Fehr, Urs Fischbacher, Christina Fong, Simon Sciences Program of the Santa Fe Institute, the U.S. National
Gächter, Roberto Galbiati, Joshua Greene, Sung-Ha Hwang, Science Foundation, the University of Siena, and the
Jonathan Haidt, Daniel Houser, Steffen Huck, Bernd ­European Science Foundation is gratefully acknowledged.

368

02_Bowles_502.indd 368 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 369

meeting at the White House would go on c­ounterproductive incentives are com-


until 8 or 9, when the chairman would sug- mon. On December 1, 2001, the Boston
gest resuming Saturday morning. Nobody Fire Department terminated its policy of
demurred. We all knew it was important, unlimited paid sick days, replacing it with
and we were important. . . . What happened a fifteen-day sick day limit; pay would be
when the President issued an order that any- docked for firemen exceeding the limit. The
one who worked on Saturday was to receive firemen responded to the new incentives:
overtime pay…? Saturday meetings virtually those calling in sick on Christmas and New
disappeared.” Year’s Day increased tenfold over the previ-
Was Schelling’s experience atypical? ous year. The Fire Commissioner retaliated
Incentives work, often affecting the tar- by canceling their holiday bonus checks
geted behavior almost exactly as conven- (Belkin 2002). The firemen were unim-
tional economic theory predicts: textbook pressed: the year following they claimed
examples include the work response of 13,431 sick days; up from 6,432 the previ-
Tunisian sharecroppers and American ous year (Greenberger 2003). Many of the
windshield installers, and the reduced firemen, apparently angered by the new
criminal activities of those former Italian system, abused it or abandoned their previ-
convicts who could expect more severe ous ethic of serving the public even when
sentences if convicted again (Laffont injured or not feeling well.
and Matoussi 1995; Lazear 2000; Drago, Not surprisingly, then, since Richard
Gal-biati, and Vertova 2009). Titmuss’s The Gift Relationship: From Blood
But explicit economic incentives some- Donations to Social Policy, economists have
times have surprisingly limited effects and been intrigued by the claim that policies
may even be counterproductive. Substantial based on explicit economic incentives may
rewards for high school matriculation in a be counterproductive when they induce
randomized experiment in Israel had no ­people to adopt what Titmuss called a “mar-
impact on boys and little effect on girls ket mentality” or in some other way com-
except among those already quite likely to promise preexisting values to act in socially
matriculate (Angrist and Lavy 2009). Large beneficial ways. But few were persuaded
and, in most cases, immediate cash pay- (Solow 1971; Arrow 1972; Bliss 1972).
ment in return for tested scholastic achieve- At the time of its publication, there
ment in 250 urban schools in the United were two strong reasons to doubt Titmuss’s
States were almost entirely ineffective, claim: there was little hard evidence that
while incentives for student inputs (reading social preferences are important influences
a book, for example) had the intended, if on individual behavior; and there was even
modest effects (Fryer 2011). In an unusual less evidence (in the Titmuss 1971 book or
natural experiment, the imposition of elsewhere) that social preferences would
fines designed to reduce hospital stays in be undermined by explicit economic incen-
Norway had the opposite effect (Holmås tives (which we will call simply “incentives”
et al. 2010), while in England hospital stays without the adjectives, meaning interven-
were greatly reduced by a policy designed tions to influence behavior by altering the
to evoke shame and pride in hospital man- economic costs or benefits of some targeted
agers rather than the calculus of profit and activity).
loss (Besley, Bevan, and Burchardi 2009). Theoretical and empirical advances over
Anecdotal accounts of what appear the intervening years provide the basis for a
to be even more dramatic cases of reconsideration of these issues (Kreps 1997;

02_Bowles_502.indd 369 5/25/12 11:42 AM


370 Journal of Economic Literature, Vol. L (June 2012)

Elster 1998; Rabin 1998; Loewenstein 2000; 2.  Overview: Incentives and Social
Sobel 2002; surveyed in Meier 2007  b). Preferences as Substitutes or Complements
First, evidence from both the behavioral
experimental laboratory and the field is
­ We use the term “social preferences” to
consistent with the view that social prefer- refer to motives such as altruism, reciprocity,
ences are important influences on economic intrinsic pleasure in helping others, inequity
behavior (Fehr, Gächter, and Kirchsteiger aversion, ethical commitments, and other
1997; Bewley 1999; Fehr and Schmidt 1999; motives that induce people to help others
Fehr and Gächter 2000; Young and Burke more than would an own-material-payoff
2001; Bandiera, Barankay, and Rasul 2005; maximizing individual. Our use of the term is
Falk, Fehr, and Fischbacher 2005; Fehr, thus not restricted to cases in which the actor
Klein, and Schmidt 2007; DellaVigna 2009; assigns some value to the payoffs received by
Leider et al. 2009; Sloof and Sonnemans another person, as in the utility functions of
2011). Fehr and Schmidt (1999), Rabin (1993), and
Second, the importance of incomplete Levine (1998). While these functions pro-
contracts has been widely recognized in vide a convenient way to model some of the
theoretical works and studied empirically motivations for pro-social behavior, we use
(Stiglitz 1987; Laffont and Matoussi 1995; the broader definition because moral, intrin-
Tirole 1999). Partly as a result, the terms sic, or other reasons unrelated to a concern
trust, reciprocity, fairness, gift exchange, for another’s payoffs often motivate people to
and social capital now appear in the model- help others, adhere to social norms, and act in
ing and empirical study of principal–agent other pro-social ways even when it is person-
relationships, the provision of public goods, ally costly to do so. A person, for example, may
and other standard economic applications, adhere to a social norm not because of the
often referring to the social norms and iden- harm that a transgression would do to others,
tities that underwrite mutually beneficial but because of the kind of person she would
exchange in the absence of complete con- like to be; helping the people who are home-
tracts (Arrow 1971; Becker 1976; Akerlof less may be motivated by Andreoni’s “warm
1984; Helsley and Strange 2000; Benabou glow” of giving rather than a concern with the
and Tirole 2006; MacLeod 2007; Sliwka wellbeing of the poor (Andreoni 1990).
2007; Ellingsen and Johannesson 2008; The standard (if generally implicit)
Akerlof and Kranton 2010). assumption in economics is that the behav-
Third, advances in the theory of public ioral functions relevant for mechanism
policy have addressed cases in which incen- design, public economics, and related fields
tives affect both beliefs and preferences and are separable in social preferences (should
may thus have unintended effects (Lucas they exist) and incentives. This means, for
1976; Taylor 1987; Bowles 1989; Aaron example, that the citizen’s response to varia-
1994; Sunstein 1996; Frey 1997; Cooter tions in a subsidy for contributions to a pub-
2000; Bowles 2004; Bar-Gill and Fershtman lic good is independent of her preexisting
2005; Sobel 2005; Cervellati, Esteban, and level of social preferences. It also means
Kranich 2010). that the effect of variations in her preexist-
Finally, a substantial literature (surveyed ing ­noneconomic motivations on the citizen’s
in Frey and Jegen 2001, Bowles 2008, and level of contributions does not depend on
Gneezy, Meier, and Rey-Biel 2011) emerged the presence or magnitude of incentives.
showing that explicit incentives are some- We call this the separability assumption. (A
times counterproductive. recent example is Levitt and List 2007.) It is

02_Bowles_502.indd 370 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 371

Own-material
(+) payoffs
(+)

Incentive Action
s a

(–/+) (+)
Social
preferences

Figure 1. Crowding Effects of Incentives


Notes: The direct and indirect effects of incentives on contribution to a public good (a). The effect of an
incentive (s) on social preferences may be either to reduce their behavioral salience for the action (social
preferences are state-dependent) or to affect the manner in which preferences are updated, thereby altering
the individual’s social preferences (endogenous social preferences). Crowding out occurs when the effect of
an incentive on social preferences is negative (assuming that the effect of social preferences on the action is
positive, as shown). Crowding in (the opposite) also occurs.

a useful simplification for problems in which s­ ubstitutes: the effect of each on the targeted
social preferences can safely be ignored, as activity declines, the greater is the level of
in the choice of apples versus oranges and the other. Where the effect on social pref-
other standard textbook examples. But it erences is positive, crowding in occurs and
can also be misleading as it implies that social preferences and incentives are com-
taxes, subsidies, and other incentives affect plements, the level of each enhancing the
behavior only by altering the economic costs effect of the other.
and benefits of the targeted activities. We The possibility that incentives designed
will present evidence that the separabil- for material payoff-maximizers might have
ity assumption often does not hold, and as adverse effects is a familiar theme in political
a result social preferences may be either science (Taylor 1987; Grant 2011), psychol-
heightened by incentives appealing to self- ogy (Deci 1975), sociology (Healy 2006), and
interest or, the more commonly observed the other social sciences; but it has found
case, affected adversely by incentives. few adherents in economics. The reason is
This is illustrated in figure 1 where, due that we have adopted a simplifying strategy
to the effect of incentives on preferences, that goes back at least to John Stuart Mill
the total—direct and indirect—effect of (1867[1848], 97): “[Political economy] does
the incentive may fall short of that which not treat of the whole of man’s nature . . .
works directly on the costs and benefits of it is concerned with him solely as a being
the targeted activity. In this case, we say that who desires to possess wealth, . . . it predicts
incentives crowd out social preferences and only such . . . phenomena . . . as take place
that incentives and social preferences are in consequence of the pursuit of wealth.

02_Bowles_502.indd 371 5/25/12 11:42 AM


372 Journal of Economic Literature, Vol. L (June 2012)

It makes entire abstraction of every other decision situation—interacting with a domi-


human passion or motive.” In other words, neering supervisor, shopping, or ­relating to
we ignore the two lower arrows in figure 1. one’s neighbors, for example. To see how this
But recent experimental and other evidence works, think about gifts. Economists know
has prompted many economists to recon- that money is the perfect gift—it replaces
sider Mill’s simplification. To further this the giver’s less well-informed choice of a
reconsideration we here provide a taxonomy present by the recipient’s own choice. But
of incentive effects on preferences based when the holidays come around few econo-
on two distinctions: their nature and their mists give money to their friends, family and
causes. colleagues. This is because we also know that
Concerning the first, people often react money cannot convey thoughtfulness, con-
to the mere presence of incentives rather cern, whimsy, or any of the other messages
than their extent: giving to charity when tax that nonmonetary gifts sometimes express.
breaks are involved (whatever their magni- A gift, we know, is more than a transfer of
tude) may feel different or send a different resources; it is a signal about the giver and
signal than would be the case in the absence her relationship to the recipient, and money
of these incentives. But the extent of the changes the signal.
incentive may also matter (Gneezy 2003). Can the same be said of incentives? A
Thus the effects of incentives on social pref- long tradition in psychology has concluded
erences may be either marginal (depending that it can: “The multiple meanings of . . .
continuously on the level of the incentive) tangible rewards are reflected in our every-
or categorical (the presence of incentives day distinction among bribes and bonuses,
affecting social preferences independently incentives and salaries. . . . they carry differ-
of their level) or a combination of the two. ent connotations concerning, for example, (i)
We also distinguish between two causes the likely conditions under which the reward
of incentive effects on preferences. First, was offered, (ii) the presumed motives of the
behavior is acutely sensitive to the nature person administering the reward, and (iii)
of the decision situation (Ross and Nisbett the relationship between the agent and the
1991; Tversky and Kahneman 1981); and, recipient of the reward” (Lepper et al. 1982,
as we will see, the presence or extent of numbers added).
incentives provides information about the As Lepper and his coauthors (1982) say,
situation. A psychologist might say that pref- incentives may affect preferences for a rea-
erences are “situation-dependent” and that son familiar to economists, that is because
incentives provide situational clues. We say they indicate “the presumed motives of the
that the preferences are state-dependent, person administering the reward.” By imple-
with differing incentives constituting dif- menting an incentive, a principal reveals
ferent states. In the next section we offer a information about his or her intentions (own
model of incentive-state-dependent pref- payoff maximizing versus fair-minded, for
erences and provide data indicating that example) as well as beliefs about the target
both categorical and marginal crowding out of the incentives (hardworking or not) and
occurs. the targeted behavior (how onerous it is,
State dependence arises because actions for example.) This information, in turn, may
are motivated by a heterogeneous repertoire then affect the target’s noneconomic motiva-
of preferences—from spiteful to payoff- tion to undertake the task at hand. In section
maximizing to generous, for example—the 5, we present experimental evidence that
salience of which depends on the nature of the the information provided about the principal

02_Bowles_502.indd 372 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 373

can sometimes attenuate or even reverse the who wore (nonprescription) dark glasses
intended effect of the incentive. Of course were much less generous to their partner in
when an ­ incentive ­ provides good news a Dictator Game than were those outfitted
about the principal’s intentions or type— with clear glasses (Zhong, Bohns, and Gino
when rewards are offered, for example, 2010). The dark glasses and darkened room
rather than fines—it may recruit the target’s gave the subjects a sense of anonymity, the
social preferences to work synergistically researchers found. But it was entirely illu-
with the direct effect of the incentive. In sory: it is difficult to imagine that a subject
this case, incentives and social preferences could really think that his own wearing dark
become complements rather than substi- glasses would make him less observable,
tutes. We will see (in section 9 and 11) that especially given that the experiment was
this crowding in phenomenon is sometimes conducted at computer terminals in closed
observed in experiments, for example, when cubicles.
the principals implementing incentives are The degree of anonymity differs dramati-
peers in a Public Goods Game who pay to cally as we move between family, workplace,
fine free riders in order to support coopera- marketplace and other domains of social
tive norms. interaction. Fiske (1992) provides a taxonomy
But there are other reasons, less familiar of four psychological models correspond-
to economists, for state-dependence: rea- ing to distinct kinds of social relationships:
sons that do not concern information about authoritarian, communal, egalitarian and
the principal, and that may be at work even market, each with culturally prescribed pat-
in nonstrategic settings. A second mecha- terns of appropriate behavior. Depending on
nism is that incentives provide cues about the information they convey, incentives may
(as Lepper and his coauthors put it) “the signal that the situation corresponds to any
likely conditions under which the reward one of these four types, and therefore evoke
was offered”: by framing a decision situation, distinctive responses.
economic incentives may provide cues for We will see that a plausible explanation
appropriate behavior. This second mecha- of some of the framing effects of incentives
nism is distinguished from the first in the observed in experiments is that it occurs
experimental evidence by the fact that in the because market-like incentives trigger
former the incentives are implemented by a what psychologists term “moral disengage-
principal who is a player in the game; while ment” (Bandura 1991), a process that occurs
in the latter the targets of the incentive are because “people can switch their ethicality
not playing against the incentive designer; on and off  ” (Shu, Gino, and Bazerman 2011,
rather the incentives are introduced by the 31). In section 6, we review experiments in
experimenter. which crowding out appears to have been the
Situational cues may be very subtle, and result of moral disengagement. Depending
our responses to them unwitting. When on the information they convey, incen-
experimental subjects had the opportunity tives may also trigger the opposite—moral
to cheat on a test and as a result to gain engagement—and, as we will see in section
higher monetary rewards, less than quarter 9, experiments provide a few examples of
did so when the room was brightly lit, but this form of crowding in, illustrating the pos-
more than half cheated when the room was sible synergy or complementarity between
slightly less well lit (the variations in light- social preferences and incentives.
ing had no effect on the observability of The third mechanism that makes social
cheating). In another experiment, subjects preferences state-dependent is the crowding

02_Bowles_502.indd 373 5/25/12 11:42 AM


374 Journal of Economic Literature, Vol. L (June 2012)

out of intrinsic motives by incentives (or con- incentive designer, but instead altered the
straints) that compromise a subject’s sense meaning of the activity itself from one that
of autonomy (Deci and Ryan 1985; Deci, expressed autonomy to one that expressed
Koestner, and Ryan 1999). These effects may compliance. The interpretation that self-
occur in strategic situations where the bad determination is involved in the negative
news that incentives convey concerns the response to incentives is consistent with the
desire of a principal to control the agent. But fact that close supervision or arbitrary tem-
most of the experimental evidence for this poral deadlines for completion of an oth-
third crowding out mechanism comes from erwise enjoyable activity have effects very
nonstrategic settings (the experimenters, similar to financial rewards (Lepper et al.
not a principal implements the incentive). 1982). In section 7, we survey experimental
The underlying psychological mechanism evidence for this “control aversion” mecha-
appears to be a fundamental desire for “feel- nism for state-dependent preferences.
ings of competence and self-determination” We have just described three (partially over-
that are associated with intrinsically moti- lapping) reasons why the s­tate-dependent
vated behavior (Deci 1975). nature of preferences might lead to crowding
According to this interpretation, where out. For ease of reference, we will call them
people derive pleasure from an action per se “bad news,” “moral disengagement,” and
in the absence of other rewards, the intro- “control aversion.” But, in addition to incen-
duction of incentives may “overjustify” the tives altering the preferences that motivate
activity and reduce the individual’s sense of an individual’s action by altering the subject’s
autonomy. This self-determination mecha- sense of the situation, there is a second and
nism differs from the previous two mecha- quite different way that incentives may affect
nisms—bad news about a principal and moral preferences. The type and extent of a soci-
disengagement—because it arises from the ety’s use of economic incentives also may
target’s desire for autonomy and does not affect the process of preference-updating
depend on the target inferring negative by which individuals acquire new tastes or
information about a principal or clues about social norms that will persist over long peri-
appropriate behavior. This is particularly evi- ods. Models from biology, anthropology, and
dent in some early “overjustification” experi- economics allow us to formalize this learning
ments in which when a financial reward was process (Cavalli-Sforza and Feldman 1981;
offered by the experimenter, children often Boyd and Richerson 1985; Güth and Kliemt
forsook previously uncompensated activities 1994; Bowles 1998; Bisin and Verdier 2011;
in which they had enthusiastically engaged, Bowles and Gintis 2011).
like painting. More recent experiments show The key difference between endogenous
the same negative effects of incentives on and state-dependent preferences is that,
altruistic behavior (Warneken and Tomasello in the former case, the effect of the incen-
2008): in the absence of rewards, children tive on preferences persists in the long run
less than two years old avidly helped an adult because the updating process on which cul-
retrieve an out of reach object; but after tural transmission is based typically occurs
being rewarded with a toy for their help- during youth and its effect endures over
ing behavior the helping rate fell by forty decades if not entire lifetimes. We say that
percent. incentives affect preferences in both the
The fact that the incentive was a reward state-dependent and endogenous preference
rather than a penalty suggests that it did case, but the mechanism of the effect is dif-
not convey negative information about the ferent: in the former case the incentive is a

02_Bowles_502.indd 374 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 375

reversible signal about the principal or the Our empirical strategy (based on experi-
situation, in the latter the incentive alters the mental results) is to observe the total effect
preference-updating process. of incentives on behavior and to note
An example unrelated to incentives may whether this differs from the predicted
clarify the difference between endogenous direct effect (the top arrows in figure 1) in
and state-dependent preferences. As Italian order to infer the effects of incentives on
residents, your authors now eat a lot more (unobserved) social preferences and thereby
pasta than we did in our countries of ori- on actions (the bottom two arrows). Our data
gin. Abstracting from possible international set includes all the economic experiments
price differences, this could be another case we have been able to locate that allow a test
of “when in Rome, do as the Romans.” Or it of the separability assumption. Our tables
might be that we have newly come to enjoy include more than a hundred different sub-
the taste of pasta, perhaps through extensive ject pools, over twenty-six thousand subjects
exposure to it while in Italy. Which case it from thirty-six countries, playing Dictator,
is—state-dependent or endogenous prefer- Trust, Ultimatum, Public Goods, Third Party
ences—would be revealed by what we will Punishment, Common Pool Resource, Gift
eat back in Bogotá or Santa Fe. If we go Exchange, and other principal–agent games.
back to arepas or potatoes, then our taste These are all settings in which one’s actions
for pasta was state-dependent. If we remain affect the payoffs of others so that social pref-
­pastaphiles, then our preferences have erences may affect a subject’s experimental
endogenously changed. behavior. We find evidence of nonseparabil-
Preferences may be endogenous in this ity in all of these games. Because nonsepa-
sense because the extent to which a soci- rability, as we will see, arises from the social
ety relies on economic incentives—as relationships among those imposing incen-
opposed to other kinds of motivations and tives and their targets and the nature of the
controls—may affect how people learn new incentive, and because game structures dif-
preferences (evidence for the endogene- fer in this respect, it would be surprising if
ity of preferences is surveyed in Bowles the nature and degree of non separability did
(1998, 2004) and Bowles and Gintis (2011). not differ across these games. However, lack-
The learning on which preference endo- ing a metric for nonseparability that is com-
geneity is based is of course a long-term parable across games, we have not explored
process unlikely to be observed in a brief this possibility.
experiment. Nonetheless, experiments may Few experiments have thus far been
provide clues that learning is affected: we designed to address the causes of nonsepara-
take as evidence consistent with preference bility, so the inferences that we draw must be
endogeneity those cases in which crowd- provisional. The experimental methods that
ing effects of incentives persist after the have become standard in economics include
removal of the incentive. (Other explana- playing for real stakes, excluding deception,
tions not involving endogenous prefer- and making explicit use of game theoretic
ences are generally also possible in these concepts to clarify the role of incentives.
situations.) As experimental methods differ consider-
In sections 3 and 4, we make explicit the ably across disciplines, and for reasons of
underlying causal mechanisms through the space, we limit the entries in the tables to
use of models of state-dependent and endog- experiments done by economists. We refer
enous preference formation. Table 1 sum- to a number of important experiments done
marizes the differences. using other methods in the text. All reported

02_Bowles_502.indd 375 5/25/12 11:42 AM


376 Journal of Economic Literature, Vol. L (June 2012)

Table 1
Economic Incentives and Social Preferences:
Endogenous and State-Dependent Effects and Mechanisms

Source and Characterization Mechanisms Description (§: section with empirical evidence
(modeled in section §) relevant to this mechanism)
State-dependent preferences
Incentives affect the Information Incentive signals the designer’s type or beliefs about
behavioral salience of an “bad news” the target or the nature of the targeted task, and
individual’s social preferences. §3 may convey illegitimate pursuit of self-interest by
principal. §5

Framing Incentive signals the type of situation and hence


“moral appropriate behavior for the target, and may a­ ctivate
disengagement” own payoff-maximizing modes of thought. §6
Self-determination Incentive affects target’s sense of autonomy, and
“control aversion” may signal unacceptable control and motivate
resistance. §7
Endogenous preferences
Incentives affect the Conformist Incentives reduce the perceived population fraction
environment in which preferences preference-updating of social preference types. The extent to which a
are learned and therefore the society relies on economic incentives—as opposed
stationary distribution of preference to other kinds of motivations and controls—will
types in the population (i.e. the affect how people learn new preferences that may
fraction of population with social persist over long periods. §8
preferences). §4

Notes: As a result of the mechanisms listed incentives and social preferences may be either complements (crowd-
ing in) or substitutes (crowding out). In the conclusion we consider cases in which the degree of endogenous or
state-dependent nonseparability is subject to public policy because the crowding parameters introduced below
(​λm
​ ​   , ​λc​​, ​Λm
​ ​  , and/or Λ
​ c​​) may themselves be affected by incentives. Additional mechanisms for endogenous crowding
out are provided in Bowles (2004).

results are statistically significant at conven- of ­others, with ­ possibly counterintended


tional levels unless noted. effects. This is often thought to be the case
Incentives may have counterintuitive and when tax authorities announce stiff penalties
counterproductive effects for reasons other for underpayment, unwittingly letting the
than nonseparability (Seabright 2009). public know that cheating is common and
Strong monetary incentives, for example, thereby promoting rather than deterring it.
may overmotivate an agent leading to We do not consider these and other cases
greater than the optimal level of arousal. of counterproductive incentives where the
This appears to be the mechanism under- mechanisms are unrelated to the nonsepa-
lying the negative effects of high incen- rability of incentives and social preferences,
tives found in three experiments by Ariely which is the focus of this paper.
et al. (2009). In other cases, incentives Some of the experimental results pre-
alter ­individuals’ beliefs about the actions sented below may be explained by more than

02_Bowles_502.indd 376 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 377

Substitutes

Information
8% 29%
2%
Complements
4% 31%

14% Self
2%
determination
2% 25%
10%

10%
6%
8%

Learning 4% Framing
18% 49%

Figure 2. Summary of Experimental Evidence on the Four Crowding Out Mechanisms and Crowding In
Notes: In the figure on the left, the mechanisms accounting for crowding out are shown. The intersections
show cases in which more than one mechanism may be involved. For example 14 percent of the experiments
are consistent with both the framing and information about the incentive designer mechanisms. The circle
in the upper right refers to crowding in (we have separated out the mechanisms in this case.) The numbers
indicate the percentage of the total of fifty studies that exhibit the mechanisms indicated. There are no studies
in the intersections that are blank.

one of our four mechanisms that account for that exhibit each of the four crowding out
non-separability, either because the mecha- mechanisms in question, and the intersec-
nisms are not mutually exclusive so that mul- tions giving the cases where multiple mecha-
tiple mechanisms are at work, or because the nisms may be involved.
experiment does not provide sufficient infor- There are two reasons why, despite the
mation to say which mechanism accounts considerable number of experiments in
for the evidence of nonseparability. In these which incentives appear to affect social pref-
ambiguous cases, we classified the experi- erences, it is difficult to estimate how preva-
ment as an illustration of the mechanism lent these effects are in real economies. First,
which we thought best accounted for the the experimental games involved are about
crowding result that we report (We indicate social dilemmas or sharing with others, that
in each table where such ambiguities occur is, settings in which social preferences are
and which other mechanisms may have been likely to be important and therefore there
at work.) As a preview, figure 2 presents is something to be crowded out or in. While
a summary of our findings, the size of the the experimental evidence suggests that
ellipses indicating the percentage of ­studies crowding out may affect blood donations or

02_Bowles_502.indd 377 5/25/12 11:42 AM


378 Journal of Economic Literature, Vol. L (June 2012)

participation in community service projects, We make explicit the sources of nonsepara-


it does not have much to say about the effect bility by the value function:
of incentives on shopping behavior or clean-
ing hotel rooms. Second, while section 10 (2)  v(s;​λ​0​,​λc​​,​λm
​ ​) = ​λ​0​(1 + 1{s > 0}​λ​c​+ s​λ​m​),
presents evidence that experimental play in
these social dilemmas predicts behavior in
some nonexperimental situations, isolating where the indicator 1{s > 0} = 1 if s > 0 and
social preferences from other influences on zero otherwise. In equation (2), ​λ​0​  ≥ 0 mea-
behavior in natural settings is difficult. We sures the citizen’s baseline social preferences
conclude in sections 11 and 12 with policy namely the citizens values in the absence of
implications. a subsidy, λ​ ​c​  (which may be of either sign)
measures the categorical effect of the pres-
ence of an incentive, and λ ​ ​(which also may
​m
3.  Incentives as Signals: A Model of
be of either sign) measures the marginal
State-Dependent Preferences
effect of variations in s on values for s > 0.
In this section (following Hwang and The crowding effects represented by ​λc​​ and​
Bowles 2011a and Bowles and Hwang 2008), λ​m​ in (2) may arise because of any of the
we model incentive effects on state-depen- three mechanisms by which state-dependent
dent preferences and clarify the distinction preferences arise: bad news, moral disen-
between categorical and marginal incentive gagement, or control aversion.
effects by means of an empirical illustration. The individual’s utility is thus
We consider an individual who may bear a
cost to take an action that confers benefits (3)  u = φA − g(a) 
on others, which may be encouraged by a
subsidy implemented by a social planner. +  a(s + ​λ0​ ​(1  + 1{s > 0}​λc​​ + s​λm
​ ​))
Citizens also have values that may motivate
such pro-social actions even in the absence and the individual’s utility maximizing con-
of the subsidy. We study a single member of tribution (​a​*​) equates the marginal cost of
a community of identical citizens who may contributing to the marginal benefits (the
contribute to a public project by taking an returns from the public good plus the sub-
action a at a cost g(a) that is increasing and sidy plus the effect on the individual’s val-
convex in its argument, and that may be ues), or:
offset partially by a subsidy s, that is pro-
portional to the individual’s level of contribu- (4)  g′ (​a*​ ​) = φ + s 
tion. The output of the project is available in
equal measure to all, and it varies positively + ​λ0​ ​(1  + 1{s > 0}​λc​​  + s​λm
​ ​).
and linearly with A, the sum of the n mem-
bers’ contributions, according to φA where φ We assume that in the absence of a subsidy
is a positive constant. the contributions of the citizens to the public
We express the individual’s social prefer- good given by (4) are inefficient in the sense
ences as  v, the effect of an increase in the that there exists a mutual increase in contri-
contribution level on the individual’s utility butions that would make all citizens better
that is unrelated to material payoffs. Thus we off. The causal structure of the model is illus-
have the individual’s utility trated in figure 3.
The introduction of a subsidy increases
(1) u  =  φA  −  g(a)  +  as  +  av. contributions by raising the m ­ arginal

02_Bowles_502.indd 378 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 379

Material cost
net of incentive
g′ (a) ̵̶ s

Incentive Action
s a(s; λ 0, λ c, λ m)

Values
v(s; λ 0, λ c, λ m)
Baseline
social
preferences
λ0

Figure 3. Crowding Effects of Incentives for an Individual with State-Dependent Preferences


Notes: Baseline social preferences are the individual’s nonmaterial motivations to contribute to the public good
in the absence of an incentive. Incentives reduce the net cost of contributing to the public good; but unless
​λ​c​ = 0 = ​λm
​ ​ (separability) or λ
​ ​0​ = 0 (no social preferences to crowd out) they also affect the m
­ otivational
salience of the individual’s social preferences.

­enefits of contributing, that is, the right


b values of the individual (​λ​0​). We likewise see
hand side of (4) which we denote, θ. that
Considering the case in which there initially
is no incentive, the effect of an incentive on Δθ 
(6) ​ _  ​  = 1 +  1{s > 0}​λc​​  + s​λm
​ ​,
the net benefits of contributing (expressed Δ​λ0​ ​
in discrete terms so as to be able to account
for the discontinuity in the value function at which, in the case of crowding out, is declin-
s = 0) is ing in s.
Equations (5) and (6) make it clear that

( )
when ​λc​​and λ
​ ​m​are negative, incentives and
Δθ ​  ​λ​​
(5) ​ _ λ​0​ _
  = 1 + ​ ​  c  ​ 
  +  ​λm
​​ baseline values are substitutes: the effect
Δs Δs of each on the marginal benefits of con-
tributing varies inversely with the level of
and is composed (as expected) of a direct the other. The fact (from equation 5) that
effect (that is, 1, the top arrows in fig- the crowding effect is larger for those with
ure 1),and the indirect state-dependent greater baseline social preferences makes
effect that will be negative in the case of sense and is consistent with experiments
crowding out (​λ​c​or ​λm
​ ​negative), and larger that have identified the strength of indi-
in absolute value the greater are the baseline viduals’ social ­ preferences independently

02_Bowles_502.indd 379 5/25/12 11:42 AM


380 Journal of Economic Literature, Vol. L (June 2012)

Action Separability c Categorical crowding out


(a) λc = λm = 0
a* = φ + λ 0 + s λ c < 0, λ m = 0
a* = φ + s + λ 0 (1 + λ c )
b Marginal crowding out
λ m < 0, λ c = 0
a a* = φ + s + λ 0 (1 + sλ m)

Self-interested contribution
Social preferences λ0 = λc = λm = 0
(with no incentive) d a* = φ + s
φ + λ0

Social preferences
(with incentive) Strong (marginal) crowding out
φ + λ 0 (1 + λ c) λ 0λ m < – 1, λ c = 0
a* = φ + λ 0 + s (1 + λ 0λ m)
Own-payoff
maximizing preferences
(with no incentive)
φ Incentive
0 s– s+ (s)

Figure 4. The Sophisticated Planner’s Implementation Technology: Citizen’s Contribution to the Public
Good with State-Dependent Social Preferences
Notes: Under separability (top dashed line), incentives and social preferences are additive. Under strong
­marginal crowding out, the use of the incentive is counterproductive (i.e., reduces contributions). Under
­categorical crowding out (dot-dashed line), incentives are also counterproductive for sufficiently small,
namely, s < s–.

of ­incentives and found that crowding out What we term strong crowding out holds if
effects are larger for those with greater Δθ/Δs < 0, which from the right-hand side
baseline values (Bohnet and Baytelman of (5), we see may occur if categorical crowd-
2007; Kessler 2008; Carpenter and Myers ing out is large relative to the size of the sub-
2010). This substitutability between incen- sidy, or its marginal effect (​λ​​Ox​ ​​ ​λ​m​) is small,
tives and baseline values will be important or if the marginal effect is negative.
when we address questions of public policy The two forms of nonseparability are illus-
in the penultimate section. trated by crowding out in figure 4. Crowding
Using (5), we say that a particular change in, which we do not show, would either
in incentives Δs has crowded out social pref- shift the “separability” function upwards—
erences if Δθ/Δs < 1, that is, if the total categorical crowding in—or increase its
effect of the incentive is less than the direct slope—marginal crowding in. Because the
effect, and conversely for the case of crowd- functions in figure 4 represent the citizens’
ing in. Crowding will not occur if ​λ​c​and ​λm​​ best responses to the planner’s choice of an
or ​λ​0​are zero (that is, if social preferences incentive and thus constitute one of the con-
are not state-dependent, or they are absent). straints making up the planner’s optimizing

02_Bowles_502.indd 380 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 381

problem, we call these functions the plan- i­ncentive and the predicted contributions
ner’s implementation technology. had an arbitrarily small incentive been in
An experiment allows an estimate of effect (the vertical intercept of the observed
both categorical and marginal crowding line in figure 5) or 34.55. The incentive thus
out. Irlenbusch and Ruchala implemented categorically crowded out 21 percent of the
a public goods experiment in which the effect of social preferences (measured by
192 German student subjects faced three the excess in contribution levels above Nash
conditions: no incentives to contribute and equilibrium for self-interested subjects,
a bonus, given to the highest contribut- 12.04).
ing individual, that was either high or low Categorical crowding out is also evident in
(Irlenbusch and Ruchala 2008, details are in other experiments. In one, reported willing-
table 3 in section 5 and results are shown in ness to help a stranger load a sofa into a van
figure 5). Payoffs were such that even with was much lower under a small money incen-
no incentive individuals would maximize tive than with no incentive at all, yet a moder-
their payoffs by contributing 25 units. In ate incentive increased the willingness to help
the no-incentive case, contributions aver- over the no incentive condition (Heyman and
aged 37 units, or 48 percent above what Ariely 2004). Using these data as we did in
would have occurred if the participants the Irlenbusch and Ruchala study, we esti-
had been motivated only by the material mate that the mere presence of the incentive
rewards of the game. Contributions in the reduced the willingness to help by 27 percent
low-bonus case were not significantly dif- (compared to the no incentive condition).
ferent from the no-bonus treatment. In the Another experiment that allows us to dis-
high-bonus case, significantly higher contri- tinguish categorical and marginal crowding
butions occurred, but the amount contrib- was implemented by Cardenas (2004), but
uted (53 units) barely (and insignificantly) here (as in some other experiments) we
exceeded that predicted for self-interested observe categorical crowding in. Cardenas
subjects (50 units). implemented an experimental Common
In figure 5, we use the observed behavior Pool Resource Game very similar in struc-
in the high and low bonus case along with the ture to the kind of real world commons prob-
assumption that marginal crowding affects lem faced by his subjects—rural Colombian
the slope of the citizens’ best response func- eco-system users. In the absence of any
tion by a given amount (so that the function explicit incentives, the villagers on average
remains linear as in figure 4) to estimate the extracted 44 percent less of the experimental
marginal effect of the bonus. We find that a “resource” than would have maximized their
unit increase in the bonus is associated with a individual payoffs, providing evidence of a
0.31 increase in contributions. This contrasts significant willingness to sacrifice individual
with the marginal effect of 0.42 that would gain so as to protect the resource and raise
have occurred under separability, that is, had group-average payoffs. When they were lia-
subjects without social preferences simply ble to pay a small fine (imposed by the exper-
best responded to the incentive. Crowding imenter) if they overextracted the resource,
out thus affected a 26 percent reduction in as expected, they extracted even less than
the marginal effect of the incentive. The esti- without the fine, showing that the fine had
mated response to the incentive also gives the intended effect.
us the level of categorical crowding out, The fact that the average extraction under
namely the difference between the observed the small fine treatment was 55 percent less
contributions (37.04) in the absence of any than the Nash equilibrium for self-interested

02_Bowles_502.indd 381 5/25/12 11:42 AM


382 Journal of Economic Literature, Vol. L (June 2012)

Contribution Low bonus High bonus


(a)
62.04

53.12
52 —

50
42.04
Social preferences
(with no incentive)
37.04 38.27
34.55
Social preferences Predicted contributions for purely self-interested subjects
–predicted (with 30 Predicted contributions under separability
incentive close to 0) Observed data-average contribution in each scheme
25


Own-payoff 0 12 60
maximizing
Bonus level
preferences (with
(s)
no incentive)

Figure 5. Categorical and Marginal Crowding Out


Source: Calculated from Irlenbusch and Ruchala (2008). See text. The experimental design is an adapted
Public Goods Game comparing two team-based compensation schemes without and with a relative bonus.

subjects (when account is taken of the fine) well-designed policies can make incentives
suggests that the fine had increased the and social preferences complements rather
salience of the villagers’ social preferences than substitutes.
(by 25 percent, if the 44 percent deviation Unfortunately, unlike the Irlenbusch
from the self-interested Nash behavior is and Ruchala and Cardenas studies, many
taken as the measure of social preferences). experiments do not establish the response
Interestingly, raising the fine from a low to a to incentives that would be observed under
high level had virtually no effect. Variations separability, so it is impossible to determine
in the fine thus did not work as an incentive, if incentives are “underperforming.” A com-
but rather (in Cardenas’s view) the very pres- mon misinterpretation of experimental
ence of the fine (high or low) was a signal, results is to infer from the observation that
one that alerted subjects to the public good an incentive has an effect in the intended
nature of the interaction. We will present direction that crowding out has not occurred
other examples of fines as signals (section (Rigdon 2009). But observing a positive
3) and crowding in (section 9). These cases incentive effect in an experiment does not
hold important lessons for why incentives preclude crowding out. It is clear from fig-
sometimes are counterproductive and how ure 4 (or equation 5 and the definition of

02_Bowles_502.indd 382 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 383

c­ rowding out) that a positive incentive effect two empirical regularities. The first is the
may occur in the presence of marginal crowd- powerful effect of mere exposure on pref-
ing out (as long as it is not “strong”) and in erences, documented by the social psycholo-
the presence of categorical crowding out (as gist Zajonc (1968) and in subsequent works
long as the incentive is sufficiently large). For (Birch and Marlin 1982; Murphy and Zajonc
example, consider some substantial incentive 1993; Murphy, Monahan, and Zajonc 1995).
indicated by s+ in figure 4. Under both mar- The exposure effect is one of the reasons
ginal and categorical crowding out, the action that cultural transmission may have what is
taken (points a and b respectively) is greater termed a conformist component that favors
than in the absence of the incentive (d), so the numerous over the rare, independently
the incentive “worked”: it affected the action of their economic success. (See Boyd and
in the intended direction. But the diagnostic Richerson 1985, 223ff; Ross and Nisbett
for the presence of crowding is a compari- 1991, 30ff; Bowles 1998 and the works cited
son of these two action levels with the level there.) Following Boyd and Richerson,
that would have occurred under separability, Hwang and Bowles assume a degree of
namely point c, and this comparison makes it conformist cultural transmission, so that
clear that crowding out occurred. the likelihood that an individual will adopt
a particular preference varies not only with
relative payoffs associated with the behaviors
4.  Incentives Alter Cultural Learning:
motivated by the preference but also with
A Model of Endogenous Preferences
the prevalence of individuals with that pref-
A quite different mechanism by which erence in the population.
crowding might occur has also been studied, The second empirical regularity captured
one in which preferences are endogenous in their model of individual updating is that
so that one or more of the parameters of the presence and extent of incentives to con-
the individual’s value function—​λ​0​,  ​λc​​ and​ tribute to a public project (or to engage in
λ​m​—are altered by incentives (Bar-Gill and similar activities that benefit others) make
Fershtman 2005; Hwang and Bowles 2011b). the action (contribution) a less convincing
Hwang and Bowles present a model of cul- signal of an individual’s social preferences,
tural evolution in which the presence or level resulting in observers interpreting some
of incentives affects the process by which generous acts as merely self-interested.
preferences are acquired or abandoned, so This is the key mechanism in the model of
that a population’s equilibrium distribution Benabou and Tirole (2006) showing how
of preferences depends on incentives. By incentives may crowd out prosocial behav-
equilibrium preferences they mean a con- ior. Similarly, in his “Generous actors, selfish
figuration of incentives and preferences such actions” paper and in his subsequent work
that the latter are stationary given the pro- with Dufwenberg et al. (2011), Sobel (2009)
cess of preference-updating. and his coauthors provide not very restric-
In the Hwang and Bowles model, prefer- tive conditions on individual utility functions
ences are endogenous because (i) schools, such that “agents who care directly about the
families, religious organizations, and other welfare and opportunities of others cannot
societal institutions seek to promote civic be distinguished from selfish agents in mar-
minded values and (ii) individuals periodi- ket settings” (19). The reason is that for a
cally alter their preferences in response to class of utility functions admitting such other
their own recent experiences. Their model regarding preferences as inequality aversion,
of endogenous preferences is based on (paraphrasing the main theorem in their

02_Bowles_502.indd 383 5/25/12 11:42 AM


384 Journal of Economic Literature, Vol. L (June 2012)

2011 paper, 6) the “set of Walrasian equilib- Taken together, these two assumptions
ria of an economy [with other regarding pref- imply that the extensive use of incentives may
erences] coincides with the set of Walrasian reduce the perceived population frequency
equilibria of its corresponding . . . economy of individuals with social preferences, leading
[in which] agents care only about their own (via the conformist learning effect) to an evo-
direct consumption.” Thus the use of mar- lutionary disadvantage of generosity over self-
ket-like incentives may make it impossible interest in the preference-updating process.
to infer generous or fair-minded behaviors To show this, Hwang and Bowles (2011b)
from the observed actions of one’s fellow adapt the model of endogenous preferences in
citizens. Bowles (1998) and (2004) to study the effects
There are two reasons why the presence of incentives on the p ­reference-updating
of an incentive may lead people to mistake process. In terms of the state-dependent
­
a generous act—helping another at a cost to model of the previous section, they study the
oneself—for a self-interested one. The first effect of incentives on the equilibrium frac-
is that the incentive provides a competing tion of the population for whom ​λ​0​is posi-
explanation of the generous act: “he did it tive and sufficient to motivate contribution to
for the money.” The second is that incentives a public good. The causal structure of their
often induce individuals to shift from an model is shown in figure 6.
ethical to a payoff maximizing frame (even Suppose there are two types: a Civic gives
relocating the neural activity to different to the public good at a personal cost equal to
regions of the brain); and knowing this, the g that may be partially offset by a subsidy s,
presence of an incentive for an individual to while Homo economicus does not contribute
help another may suggest to an observer that and receives no subsidy. Both types update
the action was self-interested (Gneezy and their traits by myopic best response, observ-
Rustichini 2000b; Heyman and Ariely 2004; ing the material payoffs and public goods
Irlenbusch and Sliwka 2005; Li et al. 2009). contribution of a sample of the population
The first “he did it for the money” reason (they do not observe the utility of others) and
depends on the magnitude of the incentive a signal ​ p​  ˜  (possibly inaccurate when the plan-
because in order to provide a convincing ner implements a subsidy, s > 0) of p, the true
self-interested interpretation for the help- frequency of the Civics in the population, so
ing act the subsidy would have to exceed ​  ˜  = ​ p​

p, p​   ˜ (s, p), which is decreasing in s. To
the cost of helping. The second reason— capture the fact that the effect of the incentive
“when ­ incentives are in force, everyone on citizen’s perception of the fraction of their
maximizes their payoffs”—is categorical; it fellow citizens who are Civics may depend on
is simply the presence of the incentive that the mere presence of the incentive or on its
matters. Of course, an observer could make extent (or both), let
the opposite mistake, inferring that the gen-
erous act that was motivated entirely by an  ˜  =  p(1  +  1{s > 0}​Λ​​  + s​Λ​ ​),
(7) ​ p​  c m
incentive, was done for ethical rather than
payoff maximizing reasons. In the model where as before the indicator 1{s > 0} = 1
that follows the incentive is assumed on if s > 0 and zero otherwise and ​ Λc​​  ≤ 0
balance to degrade helping as the signal ­measures the categorical effect of the pres-
of a generous individual’s type rather than ence of an incentive on one’s inference about
motivating self-interested individuals to act another individual’s type based on observing
in ways that are mistakenly taken as signals his or her contribution to the public good
of a generous type. and ​Λ​m​  ≤ 0 measures the marginal effect of

02_Bowles_502.indd 384 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 385

Payoff differences
π C(s) ̵̶ π F(s)

Incentive Equilibrium fraction C’s


s p*

Perceived fraction C’s


p̃ = p̃ (s, p, Λc, Λm)

Socialization
γ

Figure 6. Crowding Effects of Incentives for a Population with Endogenous Preferences


Notes: Incentives raise the relative payoff of those contributing to the public good, supporting a larger frac-
tion of civic-minded citizens; but unless Λc = 0 = Λm (separability) or ​p​
_  = 0 in which case p  = 0 (no civic

minded citizens in the absence of a subsidy) a subsidy also alters the preference updating process by reducing
the perceived fraction of civic-minded citizens p̃.

the level of an incentive on one’s inference. that each individual bearing the trait will
Note that, when p = 0 or s =  0, p​ ​  ˜  = p so in
  leave in the subsequent period. (If person k
the absence of the subsidy or when Civics switches to j’s type and j does not switch then
are absent, the citizen’s perception of the k has left no replica and j has two replicas.)
fraction of the population who are Civics is To capture the effect of socialization institu-
accurate. tions on the evolution of preferences in this
The incentive has two offsetting effects population, the authors suppose that in any
on the distribution of types in the popu- period some fraction γ of the H-types will be
lation, one intended and the other not: it converted to a C-type. (Because it plays little
raises the relative payoffs of the Civics, but role in what follows, Hwang and Bowles do
it also reduces their apparent prevalence not model the manner in which socialization
in the population. To see how this affects institutions accomplish this, other than to
the equilibrium distribution of types in the assume that the process is not affected by the
population, suppose that individuals live for- level of incentives). They define α ∈ (0, 1]
ever but they periodically may switch their as the relative weight of conformism rather
type. Denote the cultural fitness of trait i as​ than payoffs in the updating process, β as
r​  i​(i = C, H for Civic and Homo economicus) the weight of payoff differences relative to
defined as the expected number of replicas the socialization effect and ​ π​C​, ​π​H​ as the

02_Bowles_502.indd 385 5/25/12 11:42 AM


386 Journal of Economic Literature, Vol. L (June 2012)

expected payoffs of the two types. Then, tak- which requires that the conformist effect
ing account of the common influences on the favoring the more common trait (the left-
number of replicas, r − 0, the cultural fitness hand side) offset the net effect of that
of the two traits can be written: trait’s payoff disadvantages and the societal
level socialization effects (the right-hand

( )
side). Values of p satisfying (11) are termed
​  ˜  − ​ _
(8)  r​ ​C​  = ​r​0​  +  α p​
  1  ​  
2
the population’s equilibrium preferences

[ ]
and denoted as p ​ *​ ​(s).
+ (1 − α) β(​πC​ ​  − ​πH​ ​)  + γ ​  _
1 − p Figure 7 illustrates the cultural equilib-
p ​  and
rium condition (11). The solid lines show the
two sides of equation (11)—the conform-
r​0​  +  α _
(9)  r​ ​H​  = ​ (​  )
2 ​  − ​ p​   
1  ˜ ist effect and the payoff plus socialization
effects—and their intersection, satisfying
equation (11) when s = 0  and giving ​p*​ ​(0)
+ (1 − α)[β(​π​H​  − ​πC​ ​)  − γ]. that is, the equilibrium distribution of pref-
erences in the absence of incentives. The
The second term in both equations is the dotted lines show the effect of the imple-
conformism effect, and it favors the Civics mentation of a subsidy. The intended effect
if it is perceived that they constitute more is to reduce the payoff advantage of the H
than half of the population. The third term types (they do not receive the subsidy) shift-
is the net effect of socialization and payoff ing downward the payoff cum socialization
based updating. The socialization effect in function. A naive social planner, unaware of
equation (8) (the second term in the square the conformist effect would thus expect the
brackets) is derived as follows: noting that introduction of the incentive to increase the
poulation size is normalized to unity, each fraction of Cs in the population to p ​ N​ ​(s).
of the 1 − p H types in the population has But the unintended effect of the subsidy is
a γ probability of converting to C (shown in to reduce the perceived fraction of the popu-
(9)) and thus appearing as γ (1 − p) replicas lation who are Cs and thereby to diminish the
assigned to the p Cs in the population. The conformist advantage of the Cs. The down-
final expression in (8) is thus the per C share ward shift in the conformist effect function
of these socialized former Hs. thus partially offsets the payoff effect, with
From these cultural fitness equations, one the resulting stationary distribution equal to​
readily derives the familiar replicator equa- p​*​(s). In the case of strong crowding out (not
tion for the movement of p over time: shown), the second effect would more than
offset the first, resulting in a ​p*​ ​(s) < ​p*​ ​(0).
(10) dp/dt  =  p(1 − p)(​r​C​  − ​r​  H​). The source of the nonseparability between
socialization and incentives is clear if we
Taking account of the costs of contribut- return to equation (11) and consider the
ing to the public good and the subsidy, and effect of an increase in s on the cultural fit-
noting that the payoff difference between ness of the C types relative to the H types,
the types π ​ ​H​  − ​πC​ ​is just g − s, the resulting evaluated at the status quo distribution of
stationary condition for an interior value of p types in the population. In the initial of any
(namely r​ C​ ​  − ​r​  H​ = 0, so that dp/dt = 0) is subsidy, is just the vertical distance at p ​ *​ ​(0)
between the two functions that have been
(11)  (p​ ​  ˜  − ​ _12 ​)    ​_
  α   ​ 
1 − α
γ
  =  β(g − s) − ​ _  ​ 
2p
,
displaced by the introduction of the incen-
tive (the dashed lines). Because this effect is

02_Bowles_502.indd 386 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 387

Cultural fitness
(number of replicas)

Payoff and socialization effect


γ
β (g − s) −
2p

Conformist effect
1 α
p~−
2 1 −α
½
p
0 1
β
κ

p*(0) p*(s) p N(s)

Figure 7. Incentives and Equilibrium Preferences


Notes: The figure (solid lines) shows the determination of fraction of citizens with social preferences in a cul-
tural equilibrium under the influence of payoff-based and conformist updating, namely p* when s = 0. The
subsidy (dotted lines) reduces the payoff difference between Homo economicus and Civics, and in the absence
of the effect on the perceived frequency of conformism in the population, the fraction of Cs in the popula-
tion would increase from p*(0) to pN(s). However, the associated reduction in the conformism effect partially
offsets this. The resulting equilibrium outcome is p*(s). The distance indicated by β is the effect anticipated
by the naive planner who assumes separability, while κ is the true effect taking nonseparability into account.
Source: Hwang and Bowles (2011b).

the cultural fitness advantage of the C-types Δκ ​ ​ 


(12) ​ _  
| ​​p*​ ​​  =  β 
Δs

[ ]
following the introduction of the incentive
when p = p*(0), Hwang and Bowles term α   ​ ​p​ ​Λc​​
+ ​ _ ​ ​  + ​ _
 *​(γ, s) ​Λm   ​ ,
it the evolutionary impact of the incentive, 1 − α Δs
denoted by  κ (analogous to θ in the state-
dependent model of the previous section, where the left-hand side means the change in
κ is the difference between the two sides the equilibrium condition associated with the
of (11)). The direct effect of incentives on κ change in s, for the given level of p, namely​
is just β, but as is clear from the following p​*​. (As in the model of state-dependent
expression, there also is an indirect effect: preferences, we consider discrete changes

02_Bowles_502.indd 387 5/25/12 11:42 AM


388 Journal of Economic Literature, Vol. L (June 2012)

here rather than simply differentiating (11), A summary of the two sources of nonsepa-
in this case because of the d ­iscontinuity rability—state-dependence and endogene-
 ˜  at s = 0 in the presence of categorical
of ​ p​ ity of preferences—and the mechanisms
crowding.) involved is provided in table 2.
The indirect effect will be negative in the The design of effective incentives in cases
case of crowding out, so the total effect of the where separability may not hold requires
incentive is less than the direct effect. The a better understanding of the cognitive or
absolute size of the indirect effect (the sec- affective effects of incentives that explain the
ond term on the right of (12)) is (as expected) categorical and marginal crowding out effects
increasing in the extent of conformism in observed in experiments. We turn in the
updating and in the (absolute magnitude next three sections to the mechanisms that
of the) crowding parameters. Importantly, make preferences incentive-state-dependent,
the negative indirect crowding out effect resulting in crowding out effects before con-
will be larger in absolute value, the greater sidering (in section 8) the evidence for the
​ *​ ​. Because p
is p ​ ​*​ varies positively with the adverse effects of incentives on preference-
socialization effect (γ), the total effect of the updating. (We consider crowding in—the
incentive is less, the more effective is a soci- case where incentives and social preferences
ety’s socialization institutions. The crowding are complements—in section 9.)
effect will absent (separability will hold) if​
Λ​m​ = ​Λc​​ = 0 in which case ​ p​  ˜  = p, so there
are no misperceptions of the fraction of Cs in 5.  Bad News: Incentives Provide
the population, or γ = 0 in which case there Information about the Principal
are no Cs to misperceive as self-interested, or
α = 0 in which case there is no conformism Incentives are implemented for a purpose,
in updating so the misperceptions induced and because the purpose is often evident to
by the incentives have no effect. the target of the incentives, the target may
We also have that the evolutionary impact also infer information about the person who
of socialization institutions is designed the incentive, about his or her
beliefs concerning the target, and the nature
Δκ ​  
(13) ​ _  | ​p​*​  = ​ _ 1   ​  ,
of the task to be done (Benabou and Tirole
Δγ 2​p*​ ​ (γ, s) 2003; Fehr and Rockenbach 2003).
We will illustrate this incentives-as-infor-
which diminishes with greater use of incen- mation-about-the-incentive designer effect
tives because (in the absence of strong by the negative response to fines imposed by
crowding out) incentives raise p​ *​ ​. experimental “investors” and “trustees” in the
Thus where crowding out occurs incen- Trust Game, a principal–agent experiment
tives and socialization institutions are implemented by Fehr and Rockenbach.
­substitutes in the sense that the marginal German students in the role of “inves-
effect of one on the evolutionary advan- tor” were given the opportunity to transfer
tages of the civic minded types diminishes some amount to the other player, called the
with the level of the other. We will return to “trustee.” This amount was then tripled by
the property of incentives and socialization the experimenter. The trustee, knowing the
as substitutes and the possibility of making investor’s choice, could in turn “back-transfer”
them complements when we consider the some (or all, or none) of this tripled amount,
policy implications of these models and the returning a benefit to the investor (Fehr
data to follow. and Rockenbach 2003). When the investor

02_Bowles_502.indd 388 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 389

Table 2
Separability and Crowding when Social Preferences are State-Dependent or Endogenous

Preferences State dependent Endogenous

Exogenous determinant Individual baseline values Population level socialization effect


of social preferences ​λ​0​ γ

Crowding mechanism Salience of values Perceived fraction C’s


v = ​λ0​ ​(1  + 1{s > 0}​λc​​ + s​λm
​ ​)  ˜  = p(1 + 1{s > 0} ​Λ​​  + s​Λ​ ​)
​ p​  c m

Intended target Individual best response Fraction of Cs in population


of the incentive ​a*​ ​(s, g(a), ​λ0​ ​   , ​λ​c​  , ​λ​m​) ​p*​ ​(s, g, γ, ​Λ ​c ​ ​Λm
​ ​)

Separability ( ​λ​c​
​λ​0​ ​ _
Δs )
  ​  +  s​λ​m​   = 0 ​  α   ​ ​ 
_
1 − α ( ​ ​ + ​ _
p​*​(γ, s) ​Λm
Δs )
​Λc​​
  ​  = 0

Sufficient Conditions: ​λ​0​  = 0 γ = 0, α = 0


separability or ​λc​​ = ​λ​m​ = 0 or ​Λ​c​ = ​Λm
​ ​= 0

Necessary Conditions: ​λ0​ ​  > 0, γ > 0, α > 0


crowding out (in) ​λc​​ or ​λ​m​ < (>)0 ​Λc​​ or ​Λm
​ ​ < (>)0

Notes: In both models, the citizens may bear a cost (g(a) or g) in order to contribute to a public good where a subsidy,
s, may partially offset the cost. In the endogenous preference model, those who contribute are Cs. Additional
notation: ​λ​m​, ​λ​c​ and ​Λm
​ ​, ​Λ​c​ are the marginal and categorical crowding parameters (in the state-dependent and
endogenous cases, respectively) and α is the relative importance of conformism in the endogenous preferences
model.

t­ransferred money to the trustee, he or she transfer, while renouncing the use of the fine
also specified a desired level of the back- when it was available to the investor increased
transfer. The experimenters implemented an back-transfers. Only one-third of the inves-
incentive condition in which the investor had tors renounced the fine when it was available;
the option of declaring that he would impose their payoffs were 50 percent greater than
a fine if the trustee’s ­back-transfer were less the investors who threatened use of the fines.
than the desired amount. The investor could The proximate causes of the nega-
also decline the use of the fine, the choice tive impact of incentives in this case are
of using or declining the fine option being suggested by evidence on the neural
known to the trustee and taken prior to the responses of the trustees in another Trust
trustee’s decision. There was also a “trust” Game experiment (Li et al. 2009). As in
condition in which no such incentives were the Fehr and Rockenbach experiment, the
available to the investor. investor’s threat of sanctions negatively
Trustees reciprocated generous initial affected back-transfers by trustees. To iden-
transfers by investors with greater back- tify the proximate causes of this result, Li
transfers. But the use of the fine reduced and his coauthors used functional magnetic
return transfers conditional on the investor’s resonance imaging (f  MRI) to compare the

02_Bowles_502.indd 389 5/25/12 11:42 AM


390 Journal of Economic Literature, Vol. L (June 2012)

activation of distinct brain regions of trust- now turn [16, 17, 19, 24, 28]. Please note that
ees when faced with an investor who had the numbers in square brackets throughout
threatened to sanction the trustee for insuf- the paper refer to the experiments as num-
ficient back-transfers and an investor who bered in the tables). Crowding out as the
had not threatened a sanction. Sanction result of the “bad news” mechanism may
threats deactivated the Ventromedial be prevalent in principal–agent settings and
Prefrontal Cortex (VMPFC), a brain area can be averted where the principal has a
whose activation was greater in trustees means of signaling trust or fairness (experi-
who made larger back-transfers, as well as ments [1–3]). Not surprisingly crowding out
other brain areas thought to be involved affects individuals who are intrinsically moti-
in the processing of social rewards. The vated or fair-minded [5–6]; for own payoff
threat activated the parietal cortex, an area maximizers, it appears there is nothing to
thought to be associated with cost–benefit crowd out.
analysis and other self-interested optimiz-
ing processes. The interpretation by Li and
his coauthors is that the sanctions induced 6.  Moral Disengagement: Incentives May
a “perception shift” favoring a more self- Suggest Permissible Behavior
interested response.
The signaling interpretation of counter- In most situations, people look for clues
productive incentives in the Trust Game of appropriate behavior, and incentives often
suggested by Fehr and Rockenbach is that provide them. In table 4, we survey experi-
in the trust condition, or when the fine was ments in which this framing effect appears
renounced by the investor, a large initial to have been at work. These experiments dif-
transfer signaled that the investor trusted fer from those in table 3, in which incentives
the trustee. The positive response to the were deployed by experimental subjects in
investor’s renunciation of the fine option is the role of a principal interacting with an
a categorical effect, analogous to the nega- agent. Here incentives are implemented
tive categorical effect of the use of incentives exogenously, that is by the experimenter, so
in the Irlenbusch and Ruchala experiment that they provide no information about the
described above. The threat of the fine, intentions or beliefs of other experimen-
however, conveyed a different message and tal subjects. As can be seen from the table,
diminished the trustee’s reciprocity. incentives appear to affect moral disengage-
Similar cases of crowding out due to the ment not only among students but also (as we
“bad news” conveyed by the incentive are at have seen) among poor Colombian villagers
work in experiments among student subject [13, 14] and top U.S. CEOs [16]. Moral dis-
pools in Switzerland, the United States, Italy, engagement was evident in the Ultimatum
France and Costa Rica (as well as Germany) Game and the Common Pool Resource
and in a diverse set of games including Gift Game [11–14; 20] as well as in the games for
Exchange, Public Goods, and a charity giv- which the bad news mechanism was at work
ing setting similar to a Dictator Game. [1, 3, 6, 7, 9, 10]. In addition, this mecha-
Costa Rican businessmen also responded nism may be clearly recognized in settings
negatively to the bad news that incentives of 1-player games (i.e., Dictator Game or a
conveyed. Table 3 summarizes experiments performance Task) [30, 32].
in which this incentives-as-signals effect Hoffman and her coauthors (1994) illus-
appears to have been at work (in some cases trated the framing power of names: generos-
along with other mechanisms, to which we ity and fair-minded behavior were diminished

02_Bowles_502.indd 390 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 391

Table 3
Bad News: Incentives Provide Information about the Person who Implements the Incentive (I)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[01] Fehr and German Trust Game •  Optional punishment Trustee’s back-transfers Explicit incentives
Rockenbach students as an incentive are lower when investors undermine altruistic
(2003) (238) contract (i.e. a fine if impose fines. Not using cooperation and
less than the desired the punishment option reciprocity; forgoing the
back-transfer amount when it is available punishment option is a
is returned). The level results in larger back- signal of good will and
of the fine is fixed by transfers and a larger trust. See Fehr and List
the experimenter and joint surplus. (2004). Negative effects
the only choice of the of use of the punishment
investor is whether to option are greater when
impose the fine or not the investor demands a
larger share of the joint
surplus. Categorical
crowding out when the
investor chooses the fine. F
[02] Fehr and Costa Rican Trust Game •  Optional punishment CEO principals trust Key to performance: “the
List (2004) CEOs (126) as an incentive more and are more psychological message…
and students contract (i.e., a fine if trustworthy than students conveyed by incentives –
(76) less than the desired and as a result they whether . . . kind or
back-transfer amount achieve allocations closer hostile . . .” (p. 745). See
is returned) to the maximum surplus Fehr and Rockenbach
that could be generated (2003).
by the two parties. Joint
surplus is highest when
the punishment option
is available and not
used and lowest if the
punishment option is
used.
[03] Borges and German Buyer–Seller •  Three rights of When sellers voluntarily “Buyers are more inclined
Irlenbusch students Game withdrawal: none, offer a withdrawal right, to behave fairly toward the
(2007) (179) voluntary offer of a buyers make order sellers if they have . . .
right of withdrawal decisions that are less granted the withdrawal
(with a return cost for harmful for the seller right voluntarily than if it is
the seller) and imposed than if the withdrawal constituted by law”. (p. 17)
•  The right of right is imposed on [because it is] “perceived
withdrawal when sellers exogenously. . . . as a generous act and
imposed has a return they might feel inclined
cost for the buyer to reciprocate by not
or not exploiting the
seller . . .” (p. 12). F
[04] Fehr and German Gift- •  Two internal forms Most principals do not “Explicit and implicit
Schmidt students Exchange of enforcement: The use the fine. The joint incentives are substitutes
(2007) (70) Game principal (employer) surplus under the pure rather than complements”
can choose to rely on bonus contract is 20 (p. 3). Agents perceive that
- an announced percent greater than principals who are less fair
unenforceable bonus under the combined are more likely to choose a
contract contract. Wages are combined contract and less
- a combination of the 54 percent higher likely to pay the announced
bonus contract with in the pure bonus bonus. The effect of effort
a fine contract. Profits are not on the bonus paid is twice
significantly different in as great in the pure bonus
the two contracts. case.

(continued)

02_Bowles_502.indd 391 5/25/12 11:42 AM


392 Journal of Economic Literature, Vol. L (June 2012)

Table 3
Bad News: Incentives Provide Information about the Person who Implements the Incentive (I)
(continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)
[05] Fehr et al. German Gift- •  Three internal forms Under the “Bonus contracts
(2007) students Exchange of enforcement: The unenforceable that offer a voluntary
(130) Game principal can choose bonus contract and unenforceable
to rely on subjects contribute bonus for satisfactory
- a trust (pure fixed more than the performance provide
payment) contract, payoff maximizing powerful incentives
or a price deduction Nash equilibrium, and are superior to
(i.e., fine) contract - outperforming explicit incentive
a trust, a fine, or an the enforceable contracts when there
unenforceable bonus incentive contract are some fair-minded
contract (fine). The results players” (121).
•  Different frames: are the same
employer–employee independently of the
or buyer–seller framing.

[06] Dickenson French Gift- •  Stranger or Partner In the partner While intrinsic
and Villeval students Exchange with communication treatment, when motivation is
(2008) (182) Game with •  Employer payoffs employer payoffs evident in subject
a computer dependent on depend on employee behaviors, in the
task employee effort effort less monitoring Partner relationship
(variable) or not induces substantially the effect of more
higher performance. monitoring appears to
Consistent with Frey be a reciprocity based
(1993). negative response to
the principal’s lack
of trust or intent to
benefit at the agent’s
expense. F, S

[07] Irlenbusch German Public •  An external form High (but not Both categorical and
and Ruchala students Goods of enforcement low) bonuses marginal crowding out
(2008) (192) Game (i.e., experimenter increase average occur. The tournament
imposed): Team- effort, and joint structure reduces
based compensation surplus increases voluntary cooperation.
with and without significantly only if F (See text, pp. 14–15)
a reward for the the bonus is high,
highest contributor but decreases over
in the team time. Only with the
•  The reward is a low purely team-based
or a high bonus. compensation
•  Pure individual (no individual
bonus without team- incentives) do agents
based compensation contribute more
than self-interest
would motivate.
Pure tournament
incentives induce
effort levels below
the self-interested
Nash equilibrium
prediction.
(continued)

02_Bowles_502.indd 392 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 393

Table 3
Bad News: Incentives Provide Information about the Person who Implements the Incentive (I)
(continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[08] Ariely, U.S. students Charity •  An external form of In the public The signaling value of
Bracha, (161) giving enforcement: treatment subjects giving is compromised
and Meier based on With monetary exert more effort by incentives. “Image
(2009) task compensation or for a good cause and motivation is crowded out
performance without effort is substantially by monetary incentives
•  Donation choices are lower in the incentive [that are] more likely to
public or private treatment. Monetary be counterproductive for
•  Different frames: incentives increase public pro-social activities
“good” and “bad” effort in the private than for private ones.”
charitable causes treatment. (p. 1) Categorical crowding
out. See Tenbrunsel and
Messick (1999), Mulder,
van Dijk, De Cremer, et al.
(2006).

[09] Stanca, Italian Gift- •  In the first move, Second movers’ Reciprocity is stronger in
Bruni, students Exchange Information (player amounts returned response to actions that
and (96) Game 1 knows there is a are more correlated are perceived as driven by
Corazzini second move) or No with the first mover’s intrinsic motivation, than
(2009) Information (player 1 amounts sent in the to be in response to actions
does not know there No Information that are perceived as
is a second move treatment. extrinsically motivated. F
and hence thinks the
game is a Dictator
Game)
[10] Fehr and Swiss Gift- •  Three external forms Incentives reduce Effects of incentives are
Gächter students Exchange of enforcement: A agent’s effort. If the due to the perceived
(2002b)* (182) Game Trust (pure fixed incentive is framed fairness, kindness and
wage) contract, a as a price deduction hostility of the principal’s
deduction (i.e., fine) the effort reduction action. F, S
contract, and bonus is greater than where
incentive contract the incentive is framed
as a bonus. Incentives
reduce total surplus,
increase principal’s
profits.

Notes: The bold entries in the comments column—I, F, S, E, and C—indicate that the experiment in question
could also have been included in tables 3 (Information about the principal), 4 (Framing), 5 (Self-determination),
6 (Endogenous preferences), or 7 (Complementary relations between incentives and social preferences). All the
papers but those marked with an * are published or forthcoming in a publication. The entries for each table are
organized as follows: First, those studies that are published in a journal, ordered by year and first author. Second,
working papers, ordered by year and first author. In a multiperiod experiment, the “stranger” treatment is one in
which the pairing of subjects randomly changes from period to period. The “partner” treatment is one in which the
pairing of subjects continues from period to period.

02_Bowles_502.indd 393 5/25/12 11:42 AM


394 Journal of Economic Literature, Vol. L (June 2012)

Table 4
Moral Disengagement: Incentives may Suggest Permissible Behavior (F)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[11] Hoffman U.S. Ultimatum •  Roles are assigned Offers are lower Institutional cues affect
et al. students Game; by contest (the right and fewer low behavior: with property
(1994) (270) Dictator to be the Proposer is offers are rejected rights (i.e. legitimate
Game ‘earned’ or randomly in the “Exchange” ‘earning’ right to be
assigned). frame or when the proposer), a market
•  Different frame: proposer earns the framing or total anonymity
“Exchange” game right to his role. proposers and responders
(between a “seller” and Proposers accurately are more self-interested. S
a “buyer”) or no frame gauge willingness of
•  Anonymity: Double responders to accept
blind or not lower offers. Dictators
send lower amounts in
double blind.
[12] Schotter U.S. Ultimatum •  Survival treatment Competitive threats The context affects
et al. students Game (two-stage): subjects to survival induce behavior: ‘earning’ right
(1996) (247) with higher payoffs lower offers, and fewer to be the first mover or
“survive” to proceed to rejections of low offers. threat to survival induces
stage 2. proposers to behave in
•  Non survival treatment a more self-interested
(one stage): the manner.
proposer is randomly
assigned
•  Contextual framing:
a simultaneous move
normal or a sequential
extensive form game
[13] Cardenas, Colombian Common •  External enforcement Fines induce more Weakly (exogenously)
Stranlund, forest area Pool device with a low self-interested behavior enforced fines diminish
and Willis dwellers Resource probability inspection and are ineffective at socially motivated behavior.
(2000) (112) Game and a fine reducing common pool Fine appears to have
•  Communication overexploitation in the induced a shift from moral
longer run. Socially to self-interested frame.
optimal individual See Tenbrunsel and
deviations from the Messick (1999).
self-interested Nash
equilibrium behavior
(and the implied
foregone payoffs by
subjects) are least under
the fines.
[14] Cardenas Colombian Common •  Different levels of Deviation from self- Marginal Crowding
(2004) users of Pool external enforcement interested behavior is Out. (See text and also
rural Resource (low and high fines) much greater under table 7; where Categorical
ecosystems Game with announcement communication (no fine) Crowding In also occurs).
(265) of socially optimal than under either high
extraction level or low fines without
and without communication. The
communication behavioral effect of high
•  Communication (compared to low) fines
without fines and is less than 6 percent
announcement. of the predicted effect
assuming self-interested
preferences.

(continued)

02_Bowles_502.indd 394 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 395

Table 4
Moral Disengagement: Incentives may Suggest Permissible Behavior (F) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[15] Heyman 240 U.S. A computer •  Different forms of In both the cash and The level and form of
and students task and a compensation (cash, the candy conditions, compensation affect
Ariely (150+90) puzzle task candy or a cash- effort increases when performance. “Monetary
(2004) denominated amount the compensation compensation may act as
of candy) level increases from a strong signal invoking
•  Different levels low to medium. In norms of money markets
of monetary the no-compensation instead of social-market
compensation (none, treatment, effort is relations” (p. 6)
low, medium) higher than the low- Monetary incentives
compensation condition influence the ways in
for both the cash and which tasks are framed
the cash in terms of and the motivation to
candy conditions and engage in them. The
is not different from type of market in which
low-compensation in the the exchange takes place
candy condition. influences the relationship
between reward and
motivation.
[16] Bohnet and Senior Trust Game •  No communication, Repetition and “The availability of
Baytelman executives and a face-to-face pre-play communication increase punishment destroys
(2007) in U.S. Dictator communication amount transferred and intrinsic trust and lowers
(353) Game or post-play back-transferred; the people’s willingness
communication option of punishment to reward trust” (p. 1) I
•  An external form for low back transfers
of enforcement reduces back-transfers of
(Post-play monetary other regarding trustees
punishment or not) (those who send more in
•  Stranger and Partner the Dictator Game).
[17] Houser U.S. Gift- •  A form of When back-transfer “Subjects interpret
et al. students Exchange enforcement requests are high in punishment as the
(2008) (532) Game (Punishment as an relation to the sanction’s price for self-interested
incentive contract size, regardless of behavior and the price,
(i.e., a fine)) whether the request is regardless of whether
•  Intention treatment: fair and regardless of it was intentionally
Punishment is whether punishment is imposed, is an excuse for
assigned exogenously intentional, punishment selfishness” (p. 15)
or imposed by incentives have Categorical crowding out
investors detrimental effects on when the investor chooses
the amount returned. the fine. See Fehr and
Rockenbach (2003) and
Mulder et al. (2006) I
[18] Mellstrom Swedish Subjects are •  With and without The incentive reduces The monetary incentive
and students offered the a monetary the supply of prospective may make it more
Johannesson (262) opportunity compensation for blood donors from 52% difficult to signal social
(2008) to take a becoming blood to 30% among women. preferences, diminishing
health donors No effect among men. the signaling value of
exam to •  To choose between Allowing individuals to contributing. Charity
become a monetary donate the payment to option facilitates signaling.
blood compensation and charity eliminates the Overjustification appears
donors donating the same negative effect of the also to be involved. S
amount to charity monetary compensation.

(continued)

02_Bowles_502.indd 395 5/25/12 11:42 AM


396 Journal of Economic Literature, Vol. L (June 2012)

Table 4
Moral Disengagement: Incentives may Suggest Permissible Behavior (F) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[19] Li et al. U.S. citizens Trust Game •  Optional punishment Trustees reciprocate Monetary sanctions
(2009) (104) as an incentive relatively less when “encourage activity within
contract (i.e., a facing sanction threats, neural networks associated
monetary sanction if and the presence of with self-interested
less than the desired sanctions significantly economic decision making
backtransfer amount reduces trustee’s brain while simultaneously
is returned) activities involved in mitigating activity in
social reward valuation networks implicated in
(VMPFC, LOFC, social reward evaluation
and amygdala), while and processing” (p. 3) I
significantly increasing
activities in parietal
cortex previously
implicated in economic
decision making.
[20] Henrich 15 societies Dictator •  Differences between In many populations in The presence of the fine
et al. Including Game, societies the TPG the incentives in the TPG appears to
(2010) and U.S. Ultimatum •  Subjects played in the provided by the fine do have reduced the salience
Barr et al. students, Game, and following sequence not induce higher offers, of moral reasoning
(2009)* and African Third-Party keeping their role but rather have the (derived from the
personal workers, Punishment (active or passive): opposite effect; factors teachings of the world
communica- Amazonian, Game (TPG) first DG, then the UG that may influence religions) and enhanced
tion from Arctic, and and finally the TPG self-interest calculations subjects concerns with
Barr and African (an explicit incentive, (i.e. wealth, income their own economic needs.
Henrich Hunter- i.e., fine) and household size) are (See text, pp. 29 –30.)
(March gatherers. significant predictors of
2009) (428) allocations in the TPG
(but not in the DG).
Membership in a “world
religion” positively
associated with offers in
the DG but not in the
TPG.

Notes: The bold entries in the comments column—I, F, S, E, and C—indicate that the experiment in question
could also have been included in tables 3 (Information about the principal), 4 (Framing), 5 (Self-determination),
6 (Endogenous preferences), or 7 (Complementary relations between incentives and social preferences). All the
papers but those marked with an * are published or forthcoming in a publication. The entries for each table are
organized as follows: First, those studies that are published in a journal, ordered by year and first author. Second,
working papers, ordered by year and first author. In a multiperiod experiment, the “stranger” treatment is one in
which the pairing of subjects randomly changes from period to period. The “partner” treatment is one in which the
pairing of subjects continues from period to period.

by simply relabeling an Ultimatum Game the Loewenstein, and Murnighan 2007) but in
“Exchange Game” and relabeling proposers some cases (Ellingsen et al. 2011) the fram-
and responders “sellers” and “buyers.” The ing effect appears to have altered subjects
power of names has been confirmed in many beliefs about the actions of others rather
(but not all) experiments since then (Zhong, than their preferences.

02_Bowles_502.indd 396 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 397

But literally naming the game is not neces- to allocate some all or none of it to a p­ assive
sary for framing effects to occur. Incentives ­recipient. The Third Party Punishment Game
alone may provide powerful frames for the is a Dictator Game with an active onlooker
decisionmaker. A year before the first reality (the third party) who observes the dictator’s
TV Survivor show, Schotter and his coauthors allocation. If the third party deems the dicta-
found that market-like competition for “sur- tor’s allocation worthy of punishment he or
vival” among subjects reduced their concern she may then pay to impose a monetary fine
for fairness in an Ultimatum Game experi- on the dictator. One would expect that in the
ment (Schotter, Weiss, and Zapater 1996). presence of a third party, the dictators would
In this game, player 1 is given an endow- adjust their allocations upwards (compared
ment and asked to propose a division of it to the two party standard Dictator Game)
with player 2. Player 2, knowing the size of and thus avoid being fined. But this was not
the endowment, decides whether to accept the case; fining was common; it occurred
or reject the division. If player 2 accepts, in 30 percent of the interactions across the
then the proposed division is implemented. study sites.
If player 2 rejects, both players receive zero. Surprisingly, in only two of the fifteen
As is commonly observed in the Ultimatum populations were the offers significantly
Game, player 1 made quite generous offers higher in the Third Party Punishment Game
and low offers were frequently rejected. But than in the Dictator Game, and in four of the
when the experimenters told the subjects that populations the allocations were significantly
those with lower earnings would be excluded (and in some cases substantially) lower. In
from a second round of the game, player 1 Accra, for example, where 41 percent of the
subjects offered less generous amounts to dictator’s allocations resulted in fines by the
player 2, and player 2 accepted lower offers. third party, the allocations were 30 percent
The authors’ interpretation was that: “ . . . the lower in the Third Party Punishment Game
competition inherent in ­markets . . . offers than in the Dictator Game. The incentives
justifications for actions that, in isolation, provided by the fine did not induce higher
would be unjustifiable” (38). allocations, but rather had the opposite
While plausible, direct evidence for effect. (For two groups, there was a signifi-
this “moral disengagement” explanation is cant positive effect of the fine option indi-
­lacking because the social preferences that cating that, in these two cases, the incentive
apparently accounted for fair behavior in had some effect; but as we have seen does
the nonsurvival condition of the experiment not preclude crowding out.)
were not measured. There are cases, how- Crowding out of ethical motives is sug-
ever, in which the reduction in the salience gested by the fact that the dictator’s adher-
of ethical reasoning induced by the presence ence to one of the world’s religions (Islam or
of incentives can be detected. Christianity, including Russian Orthodoxy)
A large team of anthropologists and econo- raised allocations in the Dictator Game by
mists implemented both Dictator and Third 23 percent (compared to those unaffiliated
Party Punishment Games in fifteen societies with a world religion). But in the Third Party
ranging from Amazonian, Arctic and African Punishment Game, the estimated “religion
hunter gatherers to manufacturing workers effect” was reduced to just 7 percent of its
in Accra, Ghana, and U.S. undergraduates value in the Dictator Game and it was not
(Barr et al. 2009; Henrich et al. 2010). In significantly different from zero. The pres-
the Dictator Game, an experimental sub- ence of the incentive based on the fine
ject is assigned a sum of money and asked appears to have defined the setting as one in

02_Bowles_502.indd 397 5/25/12 11:42 AM


398 Journal of Economic Literature, Vol. L (June 2012)

which the moral teachings of these r­ eligions the “bad news” about the principal rubric
were not relevant. Consistent with a crowd- presented in table 3 [6,10]. Moreover, fram-
ing out interpretation of these results, the ing effects may result in moral disengage-
negative effect on the dictator’s allocations ment in some of these experiments [24, 28].
of his or her economic need (number of Crowding out effects of intrinsic motivation
children, conditional on a given level of may be recognized in Ultimatum games
income and wealth) was substantial (and [11, 12] and games where the experimenter
statistically significant) in the Third Party is the principal [18, 28, 31]. We think it is
Punishment Game, but in the Dictator likely that in these and other cases more than
Game this “economic need effect” was an one mechanism is at work.
order of ­magnitude smaller and not signifi- Falk and Kosfeld used a principal–agent
cantly different from zero. game to explore the idea that “control aver-
In the Accra sample (Barr 2004), the dic- sion” based on the self-determination motive
tator’s allocation covaried significantly with may be a reason why incentives sometimes
the frequency of attendance at church or degrade performance (Falk and Kosfeld
mosque in the standard two party Dictator 2006). Experimental agents in a role similar
Game; but this large “religion effect” van- to an employee chose a level of “production”
ished in the Third Party Punishment Game. that was costly to them and beneficial to the
The incentives implicit in the Third Party principal (the employer). The agent’s choice
Punishment Game appear to have substi- effectively determined the distribution of
tuted economic motivations for moral con- gains between the two, with the agent’s
cerns. These experiments are also consistent maximum payoff occurring if he ­produced
with our model of state-dependent prefer- nothing. Before the agent’s decision, the
­
ences, in which crowding out operates via an principal could elect to leave the choice of
effect of incentives on the behavior of those the level of production completely to the
with preexisting social preferences. agent’s discretion, or impose a lower bound
on the agent’s production (three bounds
were varied by the experimenter across
7.  Control Aversion: Incentives May
treatments, the principal’s choice was sim-
Compromise Intrinsic Motives
ply whether or not to impose it). The prin-
and Self-Determination
cipal could infer that a self-interested agent
Recent experiments by economists sur- would perform at the lower bound or, in
veyed in table 5 as well as nonexperimental the absence of the bound, at zero, and thus
studies in economics (surveyed in Frey and imposition of the bound would maximize the
Jegen 2001) provide evidence for a third principal’s payoffs.
reason why social preferences may be state But in the experiment, agents provided a
dependent in ways leading to crowding lower level of production when the principal
out. Table 5 does not include the original imposed the bound. Apparently anticipating
“­overjustification” experiments done by psy- this response, fewer than a third of the prin-
chologists (referred to in the introduction). cipals opted for its imposition in the moder-
Unlike the experiments by psychologists ate or low-bound treatments. This minority
where incentives are typically implemented of “untrusting” principals earned on average
by the experimenter, economists often model half of the profits of those who did not seek to
strategic interactions in which the same control the agents’ choice in the low-bound
apparently control averse reaction occurs, treatment, and a third less in the intermedi-
so these experiments could also fall under ate bound condition.

02_Bowles_502.indd 398 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 399

Table 5
Control Aversion: Incentives May Compromise Intrinsic Motives and Self-Determination (S)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[21] Gneezy Israeli 50 IQ test •  Different levels of A discontinuity in the The presence of the
and students questions monetary rewards effect of incentives at incentive substitutes
Rustichini (160 for the (plus a for correct IQ test zero. Small rewards extrinsic for intrinsic
(2000b) main Principal response (very low, degrade performance; motivation. Categorical
experiment) Agent low, high and none) large rewards enhance it. crowding out. See Gneezy
Game) (2003) F
[22] Gneezy Israeli Collected •  Different levels of Discontinuity at zero. The presence of the
and students donations monetary rewards for Performance with small incentive substitutes
Rustichini (180) from the voluntary work rewards is lower than extrinsic for intrinsic
(2000b) households (low, high and none) performance with high motivation. Categorical
rewards and both are crowding out. See Gneezy
lower than performance (2003)
with no rewards.
[23] Rustrom U.S. Creative •  Two forms of external Penalties degrade Explicit incentives have
(2002) students task (‘tower enforcement (a penalty performance; large a detrimental effect on
(110) of Hanoi’) or a reward) rewards induce better performance, but only in
•  Different levels of the performance than small the case of penalties, not
external enforcement (but no better than the in the case of rewards.
(none, weak, strong) noincentive treatment) Penalties ‘distract’ subjects.
Categorical crowding out.
[24] Falk and Swiss Gift- •  Different levels of Most agents perform Imposing a lower bound
Kosfeld students Exchange a lower bound of minimally (namely at the compromises subject’s
(2006) (804) Game performance selected lower bound) in response sense of autonomy and
by the experimenter to the principals’ signals distrust and low
(low, medium, and controlling decision. expectations that diminish
high) Majority of the principals agents’ reciprocity and
•  The principal could anticipate this and do good will towards the
choose whether to impose the bound, principal. Categorical
impose the minimum earning higher profits as crowding out.
level or not a result. (See text) I
•  The principal chooses
the agent’s wage and
whether to impose the
bound
[25] Xiao and U.S. Public •  Exogenous Private punishment Weak incentives crowd
Houser students Goods punishment: None, induces lower levels out cooperation when
(2011) (72) Game private (only the of contribution than implemented privately, but
punished subject public punishment. the same incentives when
knows when a round implemented publicly (but
is monitored and anonymously) promote
the amount of the cooperation.
resulting punishment),
public (all members of
a group are told that
information)

(continued)

02_Bowles_502.indd 399 5/25/12 11:42 AM


400 Journal of Economic Literature, Vol. L (June 2012)

Table 5.
Control Aversion: Incentives May Compromise Intrinsic Motives and Self-Determination (S)
(continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)
[26] Gneezy U.S. Proposer- •  The responder Non-monotonic effects Extrinsic incentives
(2003)* students Responder has three forms of explicit incentives undermine intrinsic
(400) Game of enforcement (a (fines and rewards) on motivation: a small fine or
punishment at a given performance reward changes the mode
cost, a reward at a (a W-shaped function). of behavior from “moral”
given cost and nothing) Offers are highest with to “strategic”. See Gneezy
•  Different levels large incentives (fine and Rustichini (2000a,
of the responder’s and reward), and lowest 2000b) and Mulder
enforcement (weak, with small incentives. et al. (2006). Categorical
strong) The no incentive case, crowding out. F
when proposers simply
dictate allocation, is
intermediate.

Notes: The bold entries in the comments column—I, F, S, E, and C—indicate that the experiment in question
could also have been included in tables 3 (Information about the principal), 4 (Framing), 5 (Self-determination),
6 (Endogenous preferences), or 7 (Complementary relations between incentives and social preferences). All the
papers but those marked with an * are published or forthcoming in a publication. The entries for each table are
organized as follows: First, those studies that are published in a journal, ordered by year and first author. Second,
working papers, ordered by year and first author. In a multiperiod experiment, the “stranger” treatment is one in
which the pairing of subjects randomly changes from period to period. The “partner” treatment is one in which the
pairing of subjects continues from period to period.

Control aversion and the desire for self- priors about the agents, a population with
determination are not the only effects of preferences similar to these experimental
the principal’s seeking to bind the agent. As subjects could support both a trusting and
anticipated by our discussion of the infor- an untrusting (Pareto-inefficient) equilibria.
mation content of incentives above, the Thus results in the Falk and Kosfeld experi-
imposition of the minimum in this experi- ment appear to be the result of both com-
ment gave the agents remarkably accurate promised self-determination and negative
information about the principals’ beliefs information about the incentive designer.
about them. In post-play interviews, most
agents agreed with the statement that the
8.  The Economy Produces People:
imposition of the lower bound was a signal
Incentives Alter How New Preferences
of distrust; and the principals who imposed
Are Learned
the bound in fact had substantially lower
expectations of the agents. The untrusting As in the Hwang and Bowles model
principals’ attempts to control the agents’ introduced in section 4, incentives may
choices induced over half of the agents (in all also affect long-term change in motiva-
three treatments) to contribute minimally, tions because they alter key aspects of how
thereby affirming the principals’ pessimism. we acquire our motivations, influencing
Depending on the distribution of principal’s both the range of alternative preferences

02_Bowles_502.indd 400 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 401

to which one is exposed and the economic survey in table 6 a number of experiments
rewards and social status of those with pref- that are consistent with durable learning
erences different from one’s own (Bisin and effects of incentives. (We have placed all
Verdier 2001; Bowles 2004; Bar-Gill and of the experiments consistent with prefer-
Fershtman 2005). ence endogeneity in this table; of course
Experiments of at most a few hours dura- many of them also provide evidence of the
tion are unlikely to uncover the causal mechanisms we have identified as affecting
mechanisms involved in this process of state-dependent preferences.) We take as
durable preference change. This is because evidence for this the fact that the apparent
adopting new preferences is often a slow effect of incentives on preferences persists
process more akin to acquiring an accent even when, in later stages of an experiment,
than to choosing an action in a game. As in incentives are withdrawn, suggesting that the
the Hwang and Bowles model above, the prevalence of social preferences in a popula-
developmental processes involved typically tion may depend on exposure to incentives in
include population-level effects such as con- the past, as in the Hwang and Bowles model.
formism, schooling, religious instruction, An example follows. In the public goods
and other forms of socialization that are not experiment designed by Falkinger et al.
readily captured in experiments. Acquiring (2000), an incentive mechanism induced
new preferences (like a new accent) often subjects to contribute almost exactly the
takes place early in life and the learning amount predicted for a self-interested indi-
process is strongly attenuated thereafter. vidual, while in the absence of the incen-
However, historical, anthropological, tive subjects contributed significantly more
social psychological and other data (­surveyed than would have been optimal for a self-
in Bowles 1998) provide evidence for interested individual. But, consistent with
­endogenous preferences, showing that eco- a change in preferences due to exposure
nomic structures affect parental child rear- to incentives, in the absence of incentives,
ing values, personality traits rewarded by subjects who had previously experienced
higher grades in school, and other develop- the incentive system contributed 26 per-
mental influences. Additional evidence that cent less than those who had never expe-
preferences are endogenous comes from rienced it.
the experimental studies of fifteen small While the cultural diversity and variety of
scale societies with extraordinarily varied games appearing in table 6 are substantial,
economic structures, ranging from farming and we think the preference learning effects
to hunting and gathering. In these studies, that we have detected in these experiments
cross subject pool comparisons showed a are indeed at work, we do not yet have
strong association between the nature of the experiments capable of testing the mecha-
diverse economic tasks required to secure nism underlying the models of the influence
a livelihood—participating in large coop- of incentives on the evolution of preferences
erative hunting teams in contrast to solitary proposed by Hwang and Bowles, Bar-Gill
work in forest slash and burn horticulture, and Fersthman, and others.
for example—and its members’ experimen-
tally measured generosity and fair-minded-
ness in the Ultimatum Game (Henrich et al. 9.  Crowding In
2005, 2010).
Despite the limitations of experiments for In section 2, we identified a number of
the investigation of preference change, we cases in which crowding in may occur. For

02_Bowles_502.indd 401 5/25/12 11:42 AM


402 Journal of Economic Literature, Vol. L (June 2012)

Table 6
The Economy Produces People: Incentives Alter How New Preferences are Learned (E)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[27] Falkinger, Swiss Public Incentive compatible Subjects implement By rewarding


et al. students Goods (Falkinger (1996)) the self-interested level contributions and
(2000) and (196) Game mechanism and no of contribution under penalizing shirkers the
personal mechanism; large and the mechanism, but mechanism may have
communi- small group size; contribute substantially relieved subjects’ sense of
cation from Interior and corner more than the self- moral responsibility and
Gächter Nash equilibria. interested level in its legitimated the pursuit
18 February absence (until late in the of self-interest. The
2008. 20 period experiments) effects persisted after
(e.g., figure 5). After the withdrawal of the
experiencing the mechanism. F
mechanism, subjects
contribute 26 percent
less when it is withdrawn
than those who have not
experienced it.
[28] Gneezy and Parents from •  An explicit Tardiness doubles in the The modest fine may have
Rustichini ten day care enforcement (i.e., six treatment centers signaled ‘how bad’ lateness
(2000a) centers in fine) is imposed for and persists even after is and/or is perceived as
Haifa, Israel lateness in six of the fine is removed. a price of a service and
these centers. No change in the four displaces an
control centers. ethical frame by a strategic
one: “A fine is a price.”
I, F, S
[29] Bohnet, U.S. Contract •  Different legal The probability of “If there is enough
Frey, students Enforce- institutions (low, enforcement and/or time for the crowding
and Huck (154) ment medium or high the cost of breach in dynamics to unfold,
(2001) Game contract enforcement the early rounds have a environments with low
(finitely probability) non-monotonic effect contract enforcement
repeated) •  Low contract on contract performance can produce outcomes
enforcement in the in the later rounds: as efficient as high levels
last rounds for all intermediate levels of of enforcement” (p. 141)
sessions. contract enforcement “by affecting behavior,
decrease trustworthiness, institutions affect
low levels and high preferences.” (p. 142) F
levels of legal contract
enforcement increase
trustworthiness.
[30] Meier Swiss Contribu- •  Matching donations: Matching increases The negative matching
(2007) students tions to For a single semester contributions when effect is probably not
(11,379) two funds subjects’ contributions they are in force. But due to the information it
to support are not matched or those who experience conveys on the neediness
financially matched matching are of the funds (larger effect
needy other •  Matching donations substantially less likely for the smaller matching
students. at high or low rates. to make a contribution rate) or to the subjects’
No matching in to either fund in desire to compensate for
subsequent periods subsequent periods; higher matching induced
average contributions contributions in the
show a small, treatment period. F
insignificant negative net
effect of the incentive.

(continued)

02_Bowles_502.indd 402 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 403

Table 6
The Economy Produces People: Incentives Alter How New Preferences are Learned (E) (continued)
Subjects Games or Institutional environments Results relevant to Comment (quotes are
Citation (number) activities (treatments) separability from the cited paper)
[31] Reeson Australian Public •  Three external forms While the regulation Suasion enhances and
and Students Goods of enforcement: is in place (during imposed minimum
Tisdell (98) Game   - moral suasion in the middle stage) contribution reduces other
(2008) the form of a single contributions are regarding preferences.
sentence to the effect significantly higher Categorical crowding out.
that the payoff to all than in the initial F, S
would be higher if stage in which only
all contributed (all suasion occurs. After
periods); the regulation is
  - a binding minimum removed, contributions
contribution are 20 percent lower
unexpectedly than in the initial stage.
introduced during The suasion treatment
4 periods and then dramatically increases
removed voluntary contributions
  - none compared to a no
suasion control.

[32] Burks, et Swiss (139) Sequential Messenger exposure In a restricted sample The fact that the effects
al. (2009) and U.S. Prisoners’ to performance based unlikely to be affected are from a game having
(113) Dilemma pay in their work place by selection bias, no obvious connection
bike Game or not second movers’ with the job suggests that
messengers exposure to preferences learned under
performance pay the incentive conditions
is associated with of the work place are
between 12 and adopted outside the
15 percent greater workplace.
likelihood of defection
on a cooperative
first mover.

[33] Irlenbusch German Gift- •  Two internal forms of Incentives reduce Incentives (price rate)
and Sliwka students Exchange enforcement: cooperation (i.e. effort alter principals’ and
(2005)* (84) Game The principal can level) and the effect agents’ perception of the
choose persists after the situation: “they lead agents
- a trust (pure fixed incentive is removed. to adopt an individual
wage) contract Where principals are maximization frame . . .
- compensation constrained to offer rather than a cooperative
contract (i.e., a fixed wages, the effort frame,” “agents have a
variable piece rate) levels of agents are stronger concern for the
•  Two different considerably higher principal’s wellbeing in the
sequences for the than when employers pure fixed wage setting.”
contracts can choose an (p. 23) F
incentive contract.

(continued)

02_Bowles_502.indd 403 5/25/12 11:42 AM


404 Journal of Economic Literature, Vol. L (June 2012)

Table 6
The Economy Produces People: Incentives Alter How New Preferences are Learned (E) (continued)
Subjects Games or Institutional environments Results relevant to Comment (quotes are
Citation (number) activities (treatments) separability from the cited paper)

[34] Herrmann British Tullock •  Two different Players cooperate more Subjects may perceive
and Orzen students Rent- sequences (strategic when they previously the interaction in the
(2008)* (116) Seeking vs. individual): played an individual rent-seeking contest as a
Game or •  First stage: the two- choice task than negative one. “. . . an
individual player Tullock Rent- when the previous individual’s attitude
choice task Seeking Game (with game is competitive towards others undergoes
and then a a different subject) or –strategic, one (i.e. changes between different
Prisoners’ an individual choice the Rent-Seeking types of situations because
Dilemma task (with the same Game) Cooperation they evoke different
Game incentives). Second and reciprocity rates contextual cues”. (p. 3)
stage: a Prisoners’ decrease after subjects “the experience of over-
Dilemma are exposed to rent- competitiveness in the
seeking competition. contest game creates a
disposition of rivalry in
subjects that some cannot
immediately “turn off”
when the experiment ends”
(p. 26)

[35] Gächter, Swiss Gift- •  Three external forms Under incentive Incentives may have a
Kessler, and students Exchange of enforcement: a contracts agents choose a lasting negative effect on
Konigstein (500) Game Trust (pure fixed wage self-interested best voluntary cooperation. F
(2010)* contract), a deduction reply (effort) i.e.,
(i.e. fine) contract there is no voluntary
and a bonus incentive cooperation.
contract Experiencing well-
•  Stranger and Partner designed contracts
•  Different sequences reduces voluntary
cooperation even
after incentives are
withdrawn.

Notes: The bold entries in the comments column—I, F, S, E, and C—indicate that the experiment in question
could also have been included in tables 3 (Information about the principal), 4 (Framing), 5 (Self-determination),
6 (Endogenous preferences), or 7 (Complementary relations between incentives and social preferences). All the
papers but those marked with an * are published or forthcoming in a publication. The entries for each table are
organized as follows: First, those studies that are published in a journal, ordered by year and first author. Second,
working papers, ordered by year and first author. In a multiperiod experiment, the “stranger” treatment is one in
which the pairing of subjects randomly changes from period to period. The “partner” treatment is one in which the
pairing of subjects continues from period to period.

example, the incentive may provide good make incentives and social preferences syn-
news about the principal or it may lead to ergistic (that is complements) rather than
moral engagement rather than its opposite. substitutes, but also because it appears that
In table 7, we survey a number of studies crowding in occurs more often in games
that show this result. These experiments are with more than three players (Public Goods
of special interest to the social planner not [41, 42, 44, 47–  49] and Common Pool
only because they would ideally point the Resource [14, 43] games) a common char-
way to the design of policies which would acteristic of public policy settings. In the

02_Bowles_502.indd 404 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 405

Table 7
Incentives Crowd in Social Preferences (C)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[36] Falk, Hungarian Gift- •  Stranger and Partner Partner treatment Repeated interactions
Gächter, students Exchange •  Two social approval increased effort levels; provide powerful
and Kovacs (126, 38) Game treatments (face to social pressure has little incentives while enhancing
(1999) face, social pressure) effect. Wage effort both intrinsic reciprocity
relationship (based on motives and concerns for
reciprocity) is steeper equitable shares (social
under partner than pressure adds little).
under stranger.
[37] Gächter Austrian Gift- • Stranger and Partner With repetition, effort Repeated interaction
and Falk students Exchange levels are higher than strengthens reciprocity
(2002) (116) Game one shot interaction and norms and induces
some self-interested ‘imitated’ reciprocity. “The
subjects act strategically social norm of reciprocity
as reciprocators and and the repeated
then choose the minimal game incentives are
effort level in the last complementary” (p. 18).
period
[38] Masclet U.S. (96) Public •  Two forms of Both monetary Cooperation can be
et al. and Goods Punishment with and non-monetary enhanced by nonmonetary
(2003) French (44) Game different levels of sanctions induce sanctions for reasons
students disapproval (from 0 higher and similar that are not strategic and
(140) to 10 points received levels of contributions. may require repeated
by a subject from Individuals tend interaction. It appears
any other agent): to make higher that nonmonetary
Monetary punishment contributions relative punishment, while not
(subjects can reduce to the preceding period affecting the best response
the monetary payoff of the higher punishment of a payoff maximizing
others after observing they have received subject, nonetheless
their decisions) and the lower their raised contributions by
and nonmonetary contribution was enhancing the salience
punishment (subjects relative to the group of social motives such as
express disapproval of average. When the shame or external peer
others’ decisions with Punishment device pressure. See Lopez
no effect on others’ is removed, subjects et al. (2011).
earnings) having been the target
•  Stranger and Partner of previous monetary
•  Three stages: In sanctions show higher
the first and third contributions than
stages without the those targeted by
punishment. In the nonmonetary sanctions.
second stage, with
punishment
[-]* Cardenas Colombian Common •  Different levels of Deviation from the Individuals consider the
(2004) users of Pool external enforcement self-regarding Nash norm of cooperation that
rural Resource (low and high fines) extraction level was 29% is proposed externally [the
ecosystems Game with announcement greater under the small announced optimal level]
(265) of socially optimal fine than with no fine. when extracting (p. 238).
extraction level Categorical Crowding In.
and without (See text 15 –16 and also
communication table 3; where marginal
•  Communication crowding out also occurs.)
without fines and
announcement.

(continued)

02_Bowles_502.indd 405 5/25/12 11:42 AM


406 Journal of Economic Literature, Vol. L (June 2012)

Table 7
Incentives Crowd in Social Preferences (C) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[39] Henrich et Foragers, Ultimatum •  Differences between Substantial cross cultural Mutually beneficial
al. (2005) herders, Game societies in the level of covariation between interactions in market
others in 15 market integration and the degree of market interactions with
small-scale the potential payoffs integration (engagement strangers may support
societies to cooperation in market exchange) and the evolution of cultures
(1,128) both average UG offers of fair-mindedness
and (unpublished) the toward strangers; “doux
propensity to reject low commerce”? Hirschman
offers. (1977). This study also
presents evidence of
incentives alter how new
preferences are learned E
[40] Falk, Fehr, Swiss Labor •  With and without a The introduction of a “Minimum wages
and students Market minimum wage. legal minimum wage [may] affect [subjects’]
Zehnder (240) Game (one •  Two different affects workers’ fairness fairness perceptions”
(2006) employer, sequences preferences leading to a (p. 1376) creating
three rise in their reservation moral “entitlements.”
workers) wages (which persists Obligations activate
even after the minimum and or enhance social
wage has been removed). preferences. See Galbiati
and Vertova (2008),
Vertova and Galbiati
(2010)
[41] Tyran and Swiss Public Levels of sanctions: Experimenter imposed Experimenter imposed
Feld (2006) students Goods none, mild and severe mild sanctions do not sanctions raised the
(102) Game •  Enforcement: external significantly affect expected cost of freeriding
or self-imposed (by average contributions without affecting behavior;
referendum) to the public good. a possible explanation
Compliance is much is that only referendum
improved if mild law is imposed sanctions
endogenously chosen. conveyed a signal of moral
disapproval by peers.
[42] Herrmann 16 student Public •  Monetary Costly Cooperation is higher Punishment is socially
et al. pools around Goods Punishment in the punishment beneficial only if
(2008a) the world Game condition. However, complemented by
(1120) (Partner) the average payoff strong social norms of
with the punishment cooperation with strangers
condition is lower than so that peer punishment
the average without induces shame rather
punishment in many than resentment. The
countries. Weak norms quality of the formal law
of civic cooperation enforcement institutions
and the weakness of and informal sanctions are
the rule of law in a complements.
country are significant
predictors of antisocial
punishment (targeting
high contributors), which
reduces the net benefits
to the group.

(continued)

02_Bowles_502.indd 406 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 407

Table 7
Incentives Crowd in Social Preferences (C) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[43] Rodriguez- Rural Common •  Three different Under all treatments When fines are rejected,
Sickert, Colombians Pool forms of external other than no fine, the implied affirmation
Guzmán, from 5 Resource enforcement (A fine groups start at high of social norms may have
and communities Game regime imposed, a levels of cooperation. temporarily increased
Cárdenas (128) fine proposed to the Cooperation remains cooperation; reciprocal
(2008) players and rejected high only when a preferences (anger at
or accepted by them, fine, be it high or low contributors)
none) low, is in force. If may account for the
•  Different levels of the players reject subsequent erosion of
external enforcement the fine, cooperation cooperation. Small fines
(low, and high) for the slowly unravels. enhance unconditional
imposed fine Presence of low fines cooperation by relieving
prevented unraveling cooperators of the need
of cooperation. to retaliate against
defectors by withdrawing
their own cooperation.
[44] Carpenter U.S. Public •  Costly punishment: Shirkers are Altruistically motivated
et al. students Goods subjects can punish punished by peers mutual monitoring, by
(2009) (172) Game noncooperators at a and they respond by enhancing shame-
cost to themselves contributing more, induced cooperation,
•  Different team’s even in the last round supports high levels
residual claim unless the frequency of team performance.
(marginal per capita of reciprocators is Synergistic effects of
return on the public too low or the group social preferences and
good) is too large. High peer-imposed incentives.
•  Different group size contributors who are This study also presents
punished subsequently evidence of incentives
contribute less. alter how new
(Unpublished results preferences are
not reported in paper). learned E
[45] Carpenter U.S. Dictator •  Exogenous variation Small monetary The effect of image
and volunteer Game in the presence and incentives increase concerns increases
Myers firefighters level of small stipends turnout to fighting with the visibility of
(2010) (217) and paid to volunteer fires for firemen the activity (training
nonvolunteer firefighters unconcerned about is a less visible activity
community image but have no than fighting fires). For
members effect on image- firefighters with image
(189) concerned firemen concerns the positive
(the estimated direct effect of small
negative effect is not extrinsic incentives is
significant). offset by the negative
indirect effect of
incentives on the image-
value of fighting fires.

(continued)

02_Bowles_502.indd 407 5/25/12 11:42 AM


408 Journal of Economic Literature, Vol. L (June 2012)

Table 7
Incentives Crowd in Social Preferences (C) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[46] Gächter, British Gift- •  Employees move A homogeneous wage Unequal wages
Nosenzo, students Exchange sequentially does not affect effort conditional on worker
and (84) Game with (Employee 1 has when an employee is type may induce high
Sefton 3-member pay comparison matched with a coworker levels of reciprocity based
(2010) firms (one information (i.e., that provides less effort. effort; unconditional
employer information about Reciprocity toward employer generosity
and two what coworker earns) the employer is more fails to recognize the
employees) and Employee 2 pronounced when the ‘deserving’ worker, and
additionally has coworker is hard-working, is not reciprocated.
effort comparison as effort is strongly and Incentives and
information positively related to social preferences as
(information about own wage and when the complements. Workers
how coworker employer pays unequal respond to employers’
performs) wages to the employees. recognition of their
•  Employers can offer Exposure to pay work effort and hence
high wages to both comparison information deservingness, not to
employees, a high in isolation from effort employer generosity.
wage to Employee 1 comparison information
only, a high wage to does not appear to
Employee 2 only and affect reciprocity toward
low wages to both employers.
[47] Lopez et Colombian Public •  Public reminder about Priming subjects to Results suggest the
al. (2011) fishermen Goods benefits of cooperation feel guilty about low importance of moral
(180) Game plus 1/5 chance of contributions did framing and that the
receiving private not affect average fine did not work as an
reminder of the social contributions, but incentive but rather as
losses resulting from the random public a signal highlighting
the individual’s non- revelation of one’s the salience of the
cooperative behavior contributions (inducing ethical dimension of the
(Guilt); receiving public shame) substantially problem. Categorical
reminder of the social increased contributions. crowding in.
losses resulting from Experimenter’s
the individual’s non- imposition of the fine
cooperative behavior further increased
(Shame), facing an ontributions but the
external low penalty for level of the fine has no
not contributing to the effect.
public good, facing an
external high penalty
for not contributing.
[48] Vertova Italian Public •  Different levels There is a positive effect Incentives not only
and students Goods of a stated (non- of the obligation, which influence material
Galbiati (210) Game (and binding) obligation to is greater when it is payoffs but also frame
(2010)* a Lottery contribute (zero, low combined with a weak recommended high
Game) and high) monetary incentive contributions as
•  A symmetric incentive than when no incentives obligations. Incentives
structure (a level are offered. A stronger and obligations affect
of contribution less monetary incentive does people’s behaviors by
(more) than the not result in an increase in activating values and/or
minimum contribution contributions. The strong coordinating individuals’
could be subject to a monetary incentive also beliefs. See also Galbiati
probabilistic penalty has no effect on behavior and Vertova (2008).
(reward)) of low and in the absence of the stated
medium size obligation.

(continued)

02_Bowles_502.indd 408 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 409

Table 7
Incentives Crowd in Social Preferences (C) (continued)

Subjects Games or Institutional environments Results relevant to Comment (quotes are


Citation (number) activities (treatments) separability from the cited paper)

[49] Barr Zimbabwean Public •  Two external forms After the introduction The fact that nonmaterial
(2001)* villagers Goods of nonmonetary of the public punishment raises
(602) Game punishment i)Public announcement and contributions suggests
announcement: each public criticism that it induces shame or
player announces her subjects contribute other social emotions (the
level of contribution more. best response for a self-
to everyone present interested individual was
in the session, ii) unaffected). See Gächter
Subjects could and Fehr (1999) and
make public Mulder, et al. (2006).
verbal statements
about each
other’s decisions:
lighthearted criticism
or the withholding
of praise during
informal gatherings
[50] Serra British Bribery •  Three different Under the “Nonmonetary costs
(2008)* students Game forms of external accountability system, activated by the
(180) (public enforcement (no fewer officials engage bottom-up component
official– monitoring; top-down in corruption. The of the combined system
citizen) auditing, and an presence of only top- had a significant impact
accountability system down auditing did not on the public official’s
which gives citizens affect the number of decision to engage in
the opportunity officers who demanded bribery” (p. 17)
to report corrupt a bribe but induced
officials) corrupt officials to
demand a higher bribe
than no monitoring.

Notes: The bold entries in the comments column—I, F, S, E, and C—indicate that the experiment in question
could also have been included in tables 3 (Information about the principal), 4 (Framing), 5 (Self-determination),
6 (Endogenous preferences), or 7 (Complementary relations between incentives and social preferences). All the
papers but those marked with an * are published or forthcoming in a publication. The entries for each table are
organized as follows: First, those studies that are published in a journal, ordered by year and first author. Second,
working papers, ordered by year and first author. In a multiperiod experiment, the “stranger” treatment is one in
which the pairing of subjects randomly changes from period to period. The “partner” treatment is one in which
the pairing of subjects continues from period to period.

­ enultimate and final sections, we will return


p and Gächter 2000). Of course, crowding in
to these questions when we consider the pol- need not have been involved; individuals
icy implication of nonseparability. might have simply best-responded to the
Synergy between incentives and social anticipated loss in payoffs associated with low
preferences may explain why fines imposed contributions. But more than this is at work.
on free riders by altruistic peers in a Public Consistent with the interpretation that incen-
Goods Game induce higher levels of contri- tives imposed by peers activate shame or other
bution in subsequent rounds of play (Fehr social preferences, purely verbal messages of

02_Bowles_502.indd 409 5/25/12 11:42 AM


410 Journal of Economic Literature, Vol. L (June 2012)

disapproval have a substantial positive effect (Henrich et al. 2005, 2010), individuals
on free riders’ subsequent contributions from the more market-integrated societies
(Barr 2001; Masclet et al. 2003). When those gave more in the Ultimatum Game. The
who have contributed more than others are authors conjecture that this may be due
punished (as sometimes occurs, Herrmann, to the fact that more market exposed sub-
Thöni, and Gächter 2008), they subsequently jects had the experience of mutually ben-
contribute less, and costly retaliatory punish- eficial exchanges with strangers, much like
ment sometimes results (Bowles and Gintis in the anonymous experimental settings. A
2006; Carpenter et al. 2009; Hopfensitz and very different piece of evidence consistent
Reuben 2009). This appears to occur because with this interpretation is that subjects who
the targets of the punishment feel hostility were exposed to unobtrusive priming with
rather than shame. words relating to markets and exchange
There are also other mechanisms at prior to playing a Trust Game were more
work. The incentives and constraints typi- likely to trust their partner than were sub-
cal of the rule of law and other institutional jects exposed to primes unrelated to mar-
designs that limit the most extreme forms of kets (Al-Ubaydli et al. 2011).
antisocial behavior and facilitate mutually A distinct mechanism underlying crowd-
beneficial interactions on a large scale may ing in was apparently at work in a public
enhance the salience of social preferences by goods experiment by Vertova and Galbiati
assuring people that those who conform to (2010). Consistent with the Cardenas experi-
moral norms will not be exploited by their ment described in section 2, they found that
self-interested fellow citizens (Bowles 2011). the effect of a stated (nonbinding) obliga-
This may explain the Hokkaido University tion to contribute a certain amount was
subjects who cooperated more in a public greater when it was combined with a weak
goods experiment when assured that oth- monetary incentive than when no incentives
ers who did not cooperate would be pun- were offered. A stronger monetary incen-
ished (Shinada and Yamagishi 2007) despite tive did not result in an increase in contri-
the fact that this had no effect on their own butions. The strong monetary incentive also
material incentives (those told this were not had no effect on behavior in the absence
subject to the punishment). They apparently of the stated obligation. The authors’ inter-
wanted to be cooperative but wished even pretation (like that of Cardenas) is that the
more to avoid being exploited by defectors. explicit incentives enhanced the salience of
According to this interpretation, the fine the stated obligation. In our taxonomy, it is
imposed by the experimenter on any free a case of categorical crowding in (see also
riding liberated the individual to act pro- Galbiati and Vertova 2008).
socially without fearing being exploited by
less cooperative players. The respondents
10.  The Lab and the Street: Can One
may have exhibited what Bohnet and her
Generalize from Experimental Evidence?
coauthors call “betrayal aversion,” which was
attenuated by knowing that betrayal would The experimental evidence for nonsepa-
be punished by a third party (Bohnet et al. rability would not be very interesting if it
2008). did not reflect real-life behavior. Testing for
Market interactions may also favor the separability in natural settings is difficult,
endogenous evolution of social preferences. but generalizing directly from experiments
In two sets of experiments in small-scale even for phenomena much simpler than
societies in Africa, Asia, and Latin America separability is a concern in any ­empirical

02_Bowles_502.indd 410 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 411

study (Falk and Heckman 2009) and is provide evidence for the salience of social
often unwarranted (Levitt and List 2007). preferences are deliberately structured as
Consider, for example, the Dictator Game: strategic interactions like the Ultimatum or
typically more than 60 percent of the dicta- the Public Goods Game that give scope for
tors allocate a positive sum to the recipient, ethical or other-regarding behavior that may
and the average given is about a fifth of the be absent in competitive markets and other
endowment. We would be sadly mistaken important real-world settings (Sobel 2010).
if we inferred from this that 60 percent of It is impossible to know whether these
individuals would spontaneously transfer four aspects of behavioral experiments bias
funds to an anonymous passerby, or that the experimental results in ways relevant to the
same subjects would offer a fifth of the bills question of separability. For example, the fact
in their wallet to a person who is homeless that in most cases subjects are paid a “show
asking for help. Another example: while pro- up fee” to participate in an experiment might
social behavior in an experiment by Benz and attract the more materially oriented who
Meier (2008) was correlated with nonexperi- may be less motivated by social preferences
mental behavior, subjects who reported that subject to crowding out; or knowing that the
they had never given to a charity allocated topic of the experiment was cooperation the
65 percent of their endowment to a named subjects might be atypically civic minded.
charity in a lab experiment. We can do more than speculate about these
A possible explanation of these discrepan- problems. Baran and her coauthors asked if
cies between experimental and real-world University of Chicago Graduate School of
behavior is that most individuals are strongly Business students who were more recipro-
influenced by the cues of appropriate behav- cal in the Trust Game (those who as trustees
ior offered by the situation in which an action most generously reciprocated large trans-
is taken (Ross and Nisbett 1991), and there fers by the investor) were also those most
is no reason to think that experiments are an likely to contribute to the University upon
exception to this context-dependent aspect graduation. They were (Baran, Sapienza,
of individual behavior. External validity and Zingales 2010). Fehr and Lorenz Goette
concerns arise from four aspects of human found that in a group of bicycle messenger
behavioral experiments that do not arise in workers in Zurich, those who exhibited loss
most well-designed natural science experi- aversion in a laboratory experiment explor-
ments. First, experimental subjects typically ing the subjects’ preferences over lotter-
know they are under an unknown research- ies also exhibited loss aversion when faced
er’s microscope, possibly inducing different with real-life wage rate changes (Fehr and
behaviors than would occur under total ano- Goette 2007). Karlan (2005) implemented a
nymity or under the scrutiny of neighbors, Trust Game among Peruvians participating
family, or workmates. Second, experimental in a micro-credit program; those who were
interactions with other subjects are typi- least trustworthy (transferred less back to
cally anonymous and without opportunities the “investor”) in the experiment were less
for ongoing face to face communication, likely to repay their real world loans. Cohn
unlike many social interactions of interest to and his coauthors (Cohn, Fehr, and Goette
economists and policymakers. Third, subject 2011) found that reciprocators in the lab
pools may be quite different from the real- (measured by play in a sequential PD
world populations of interest, in part due to Game) responded positively to a randomly
the process of recruitment and self-selec- awarded fixed wage increase in their work,
tion. Finally, many of the experiments that while those who played the sequential PD

02_Bowles_502.indd 411 5/25/12 11:42 AM


412 Journal of Economic Literature, Vol. L (June 2012)

in a payoff maximizing way did not respond shrimpers. They found that both patience
to the wage increase. and ­cooperativeness in the game predicted
Among the Japanese shrimp fishermen larger trap holes (Fehr and Leibbrandt
that Carpenter and Seki (2010) studied, 2011). The effects, controlling for a large
those who contributed more in a public number of other possible influences on
goods experiment were more likely to be hole size, were substantial. A shrimper
members of cooperatives that shared costs whose experimentally measured patience
and catch among many boats than to fish and cooperativeness is a standard deviation
under the usual private boat arrangements. greater than the mean is predicted to cut
A similar pattern was found among fisher- holes in his traps that are half a standard
men in the Brazilian north east, where some deviation larger than the mean.
fish offshore in large crews whose success Additional evidence of external validity
depends on cooperation and coordina- comes from a set of experiments and field
tion, while those exploiting inland waters studies with 49 groups of herders of the
fish singly. The ocean fishers were signifi- Bale Oromo people in Ethiopia who were
cantly more generous (in Public Goods, engaged in forest commons management.
Ultimatum, and Dictator Games) than the Rustagi and his coauthors implemented
inland fishers (Leibbrandt, Gneezy, and public goods experiments with a total of
List 2010). 679 herders. They also studied the success
A better test of the external validity of of the herders’ cooperative forest projects.
experiments would include a behavior- The most common behavioral type in the
based measure of how cooperative the experiments, constituting a bit more than
individuals were, not simply whether they a third of the subjects, were “conditional
took part in a cooperation-sensitive produc- cooperators” who responded positively to
tion process. The Brazilian fishers provide higher contributions by others. Controlling
just such a test. Shrimp are caught in large for a large number of other influences
plastic bucket-like contraptions; holes are on the success of the forest projects, the
cut in the bottom of the traps to allow the authors found that groups with more con-
­immature shrimp to escape, thereby pre- ditional cooperators were more successful,
serving the stock for future catches. The in terms of number of new trees planted,
fishermen thus face a real-world social than groups with fewer conditional coop-
dilemma: the present value of expected erators. This was in part because members
income of each would be greatest if they of groups with more conditional coopera-
cut only small holes in their own traps while tors spent significantly more time monitor-
others cut large holes in theirs. Small trap ing the use of the forest by others. As in the
holes are a form of defection, and just as in case of the Brazilian shrimpers, the effects
the Public Goods Game it is the dominant of group composition were large. A 10 per-
strategy for a self-interested individual. But cent increase in the fraction of experimen-
a shrimper might resist the temptation to tally identified conditional cooperators in
defect if he were both public spirited toward a group was associated with an increase in
the other fishers and sufficiently patient to trees planted or time spent monitoring by
value the future opportunities that they all members of the group of about 3 percent
would lose were he to use traps with smaller (Rustagi, Engel, and Kosfeld 2010).
holes. Fehr and Leibbrandt implemented The available evidence suggests that stu-
both a Public Goods Game and an experi- dents volunteering for experiments are not
mental measure of impatience with the more pro-social in their orientations than

02_Bowles_502.indd 412 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 413

other students (Falk, Meier, and Zehnder v­iolated—faces a challenge that has yet to
2011); nor are student subjects more pro- be addressed in the public economics lit-
social than nonstudents, indeed the reverse erature: how to design optimal taxes, fines,
seems to be the case. (Fehr and List 2004; or subsidies when the preferences that will
List 2004; Cardenas 2005; Carpenter, Burks, determine citizen’s responses depend on
and Verhoogen 2005; Bellemare, Kröger, the incentives deployed. Thus, the designer
and van Soest 2008; Carpenter, Connolly, must consider the effects—whether state-
and Myers 2008; Burks, Carpenter, and dependent or endogenous—of the instru-
Goette 2009; Baran, Sapienza, and Zingales ments under consideration on individuals’
2010; Cleave, Nikiforakis, and Slonim 2010; social preferences and evaluate alternative
Cardenas 2011; Falk, Meier, and Zehnder policies on the basis of the resulting joint
2011. For a description of these studies, see equilibrium of these preferences and eco-
appendix A9 of Bowles and Gintis 2011.) nomic allocations.
Levitt and List (2007) are right that care The problem facing the planner is quite
should be taken in generalizing experimen- a bit more difficult than the one we faced
tal behavior to the real world. But none of writing this paper. We studied the effects of
the external validity concerns we have con- incentives in experiments and natural set-
sidered is sufficient to dismiss the experi- tings and then sought ex post to determine
mental evidence that social preferences the kinds of nonseparability—categorical
are important behavioral motivations and or marginal crowding out or in—that might
that these preferences may be affected by explain the results. The planner, however,
explicit incentives. This is especially the must determine ex ante whether the sepa-
case when experimental subjects exhibit rability assumption is likely to be violated,
motives such as reciprocity, generosity and and if so, how. The challenge is even greater
trust that allow a consistent explanation of because the nature and extent of nonsepara-
otherwise anomalous real world examples bility itself is not given but (as we will see)
of crowding in or out, such as those men- may be influenced by the overall policy pack-
tioned at the outset. age of which the incentives are a part.
We begin with the more modest way
of addressing the planner’s problem and
11.  Optimal Incentives for the consider as exogenously given the nature
­
Sophisticated Social Planner and degree of the indirect effects of incen-
tives on social preferences (that is, the signs
There are multiple plausible interpreta- and the size of the c­rowding parameters
tions of the mechanisms underlying non- ​λc​​, ​λ​m​, ​Λ​c​   , and Λ
​ ​m​). Given the nature and
separability in the experiments we have extent of nonseparability, we then seek
presented, as is clear from the substantial to determine the optimal level or mix of
size of the intersections among the hypoth- incentives taking account of their effects on
esized crowding out mechanisms that is preferences (Fershtman and Heifetz 2006;
evident in figure 2. It would nonetheless be Heifetz, Segev, and Talley 2007; Bowles and
difficult, in light of these data, to sustain the Hwang 2008; Hwang and Bowles 2011a).
implicit separability assumption adopted in Here, two results may guide the social
many economic models. planner. The first is that, in the presence of
A sophisticated social planner (or mech- crowding out, incentives and social prefer-
anism designer)—one who knows that ences are substitutes, so the deleterious indi-
the separability assumption is likely to be rect effect of incentives will be least where

02_Bowles_502.indd 413 5/25/12 11:42 AM


414 Journal of Economic Literature, Vol. L (June 2012)

individual social preferences are modest or would use them less than would the naïve
nonexistent (as will be the case in the endog- planner. But the prescription that incentives
enous preference model if there are few or no are overused does not follow from the (cor-
public spirited citizens or in the state depen- rect) observation that crowding out occurs:
dent model where the citizen’s baseline social it is readily shown that when crowding
preferences are modest or zero). Societies in out occurs the sophisticated planner may
which social preferences are more preva- make either greater or lesser use of explicit
lent not only may be able to afford less use incentives than would her naïve counter-
of incentives but will find them less effective part (Bowles and Hwang 2008; Hwang and
(when both direct and indirect effects are Bowles 2011a).
accounted for) than would be the case in a The reason why the planner may make
less civic minded culture. By the symmetry greater use of incentives when incentives
of the definition of substitutes (see equations crowd out social preferences is that if incen-
6 and 13) in the presence of crowding out, tives work less well than would be the case
policies to enhance social preferences (that under separability, then there are two offset-
is raising γ  or ​λ​0​) will be more effective in ting influences on their optimal use. The one
promoting contributions the public good that forms the basis of the Titmuss critique
where incentives are little used. is that crowding out reduces the marginal
In a cultural–institutional dynamic setting effect of the subsidy on the target’s behav-
where economic incentives and socialization ior; and if this were the only effect Titmuss
practices to promote civic mindedness are would be right. But there is a second often
adopted as alternative measures to enhance overlooked effect. Because the incentive is
public goods provision, this substitutability less effective (either categorically or mar-
property of incentives and social prefer- ginally), the under provision of the public
ences may support at least two evolution- good will be exacerbated (compared to what
arily stable equilibria. In one, extensive use would occur were crowding out absent) and
of incentives is coupled with relatively low if the benefits of the public good are concave
levels of civic mindedness in the population. in the amount provided, the marginal benefit
In this state, there is little reason to incul- of altering the target’s behavior is therefore
cate social preferences, the effect of which correspondingly greater.
would be modest given the crowding phe- The intuition is transparent: the doctor
nomenon resulting from the extensive use who discovers that a treatment he has been
of incentives. In the other cultural–insti- prescribing is less effective than he thought
tutional equilibrium, a social planner serv- may opt for stronger doses rather than weaker
ing a civic minded population makes more or for abandoning the treatment. As long as
modest use of incentives due to their lim- there are diminishing marginal returns to the
ited effectiveness, once their crowding out public good and crowding out is categorical
effects are accounted for. (and not too large) the naïve social planner
The second result for the social plan- will make too little use of the incentive. The
ner takes us back to Titmuss and others reason is that in this case crowding does not
who concluded that if incentives crowd out change the marginal effect of the incentive
social preferences then incentives will be on the citizens’ contribution level; but the
overused by a naïve planner who is unaware reduction in the public good resulting from
of the effects of incentives on preferences. crowding means that the marginal benefits
As a result, in these cases the sophisticated to increasing its supply rise. (If categori-
planner would either not use incentives, or cal crowding is sufficiently large, the naïve

02_Bowles_502.indd 414 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 415

­ lanner will overuse the incentive because


p Palestinian students in 2006 were asked how
the sophisticated planner will choose no angry and disgusted they would feel or how
incentive at all in this case.) But the sophis- supportive of violence they might be if their
ticated planner may make greater use of political leaders were to compromise on con-
incentives even when only marginal crowd- tested issues between the groups. Those who
ing out occurs, if the benefit function is suf- regarded their group’s claims (on Jerusalem,
ficiently concave. for example) as reflecting “sacred values”
A less modest approach to the design of (about half in each of the three groups)
appropriate incentives where separabil- expressed far greater anger, disgust and sup-
ity may not hold is to recognize that the port for violence if the compromise were
extent of the nonseparability problem (that accompanied by a monetary compensation
is, the magnitudes of the crowding param- for their own group than if no compensation
eters in the models of section 3 and 4, were offered (Ginges et al. 2007). Similar
namely ​λ​c​, ​λ​m​, ​Λ​c​, and ​Λ​m​) is not exogenous, results were fund in a survey of the willing-
but can be affected by the nature of the ness of Swiss citizens to accept environmen-
incentives and the manner in which they tal hazards (Frey and Oberholzer-Gee 1997).
are deployed. Designing policies that can (For a discussion on environmental motiva-
convert incentives from being substitutes tion and crowding effects, see Frey and
for social preferences to being their comple- Stutzer 2008.)
ments, however, requires an understanding The importance of constitutive rather
of why crowding out occurs. than acquisitive motives may be at work
The most plausible explanation for the in the negative response to incentives that
failure of the separability assumption is that convey adverse information about the indi-
when people engage in trade, produce goods vidual imposing the incentives. Recall that
and services, save, and invest, they are not in the Trust Game implemented by Fehr
only attempting to get things, they are also and Rockenbach (2003) the investor’s threat
trying to be someone, both in their own eyes to fine the trustee if the back transfer was
and in the eyes of others (Cooley 1902; Yeung not sufficient resulted in a lesser level of
and Martin 2003; Akerlof and Kranton 2010; reciprocity of the trustee: conditional on the
Bloom 2010). We refer to the second—the investor’s transfer to the trustee, back-trans-
being or becoming motives—as constitutive. fers were less under the fine condition. This
Incentives addressed to our acquisitive desires was especially the case when it appeared
sometimes appear to dampen or impede the that the intent of the fine was to induce the
pursuit of our constitutive aspirations. Among trustee to grant most of the joint surplus to
the reasons, we have seen, are that in addition the investor. Where the announced desired
to affecting the costs and benefits of an action, back-transfer would have allowed the inves-
incentives also provide information about the tor to capture most of the surplus had the
person imposing the incentive, suggest appro- trustee complied, use of the threat of the fine
priate behavior by framing decision situations, was very counterproductive.
may compromise the target’s sense of auton- The back-transfers from those facing an
omy, and alter the environments in which we investor who declined use the fine averaged
learn new preferences. a generous 60 percent of what the trustee
This may explain why incentives for settle- could have kept; while those threatened with
ment of conflicts may fail. Representative a fine transferred just 22 percent (almost half
samples of Jewish West Bank settlers in gave nothing and paid the fine!). But where
2005, Palestinian refugees in 2005, and the investor announced a desired back

02_Bowles_502.indd 415 5/25/12 11:42 AM


416 Journal of Economic Literature, Vol. L (June 2012)

t­ransfer such that the trustee who complied is an altruistic act as it benefits others at the
would retain half or more of the joint sur- expense of the punisher and hence it cannot
plus, back transfers from those threatened be interpreted as a signal of unfair intent. In
with fine for noncompliance were slightly this setting there is a strong positive response
lower but not significantly so than among by low contributors.
those facing an investor who chose not to use Although there is no direct evidence, a
the fine. It appears that the use of the fine in plausible explanation of the effectiveness of
these conditions signaled the unfair intent of incentives in this case is that, when punished
the investor, rather than simply his distrust by a peer who had nothing to gain by doing
of the trustee. so, those who have contributed less than oth-
Interpretations other than this “unfair ers interpret the punishment as a signal of
intent” result are possible, however, for the public-spirited social disapproval by fellow
larger was the desired back transfer, the group members seeking to uphold a social
more costly was compliance. Thus for larger norm and willing to sacrifice payoffs to do so.
demands, simply returning nothing and pay- As a result, targeted free riders and even free
ing the fine (as many of the subjects did) riders who escaped punishment feel shame,
might have been attractive to self-interested which they redress by subsequently contrib-
subjects, who had they been faced with a uting more. In this case the incentive (pros-
lower demand would have maximized pay- pect of peer imposed fines) has crowded in
offs by complying. It seems from this and social preferences. These experiments illus-
similar experiments that fines may have trate the opposite of the “bad news about the
negative effects even if imposed to imple- principal” results in section 5. The princi-
ment a fair outcome and even when the pals here are the peers who punish free rid-
decision to use the fine was not made by the ing fellow group members; and the positive
investor, but rather by chance (Fehr and List response to the fines in this case may reflect
2004; Houser et al. 2008). But the experi- the fact that the willingness to pay to pun-
ments also are consistent with the idea that ish defectors with no expectation of personal
threats deployed in self-interested ways can gain is good news about the person imple-
backfire. menting the incentive.
This helps explain the very different effect Consistent with the interpretation that
of incentives imposed by peers who do not crowding out does not follow from the use
stand to benefit personally. An example of incentives per se but rather from the
is the Public Goods experiment in which meaning that the incentives convey to the
fellow group members have the opportu- participants, is an extension of the “control
nity to reduce their own payoffs in order aversion” experiment of Falk and Kosfeld
to punish (reduce the payoffs of) others in (2006) described in section 5. Schnedler
their group once each member’s contribu- and Vadovic (2011) found that when agents
tions are revealed (Fehr and Gächter 2000, themselves implemented controls (rather
2002a; Masclet et al. 2003). One treatment than the principal) the negative response did
in these public goods experiments is par- not occur. A large number of experiments
ticularly revealing: group membership is have found positive effects of incentives
shuffled after each period so that in subse- imposed by the decision of the targets of the
quent periods a punisher will not be in the incentives rather than by the experimenter
same group with the target of his or her pun- or by a principal (Cardenas 2005; Tyran and
ishment, and thus cannot benefit from the Feld 2006; Kroll, Cherry, and Shogren 2007;
target’s response. Punishment in this case Ertan, Page, and Putterman 2009; Kosfeld,

02_Bowles_502.indd 416 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 417

Okada, and Riedl 2009; Mellizo, Carpenter, regulations is to stigmatize antisocial behav-
and Matthews 2011; Sutter, Haigner, and ior and thereby to influence citizens’ values
Kocher 2011). and behavioral codes (Hirschman 1985, 10).
John Stuart Mill (whose definition of the The fact that punishments are not only
restrictive boundaries of our discipline we incentives but also “moral lessons” that
mentioned at the outset) and economists “stigmatize antisocial behavior” may help
since have recognized that the purposes of resolve one of the puzzles in the literature
individual economic action are constitutive we have just surveyed. In a widely cited
as well as acquisitive (Akerlof and Kranton natural experiment, the imposition of fines
2010). But what some have missed is that our on parents arriving late to pick up their chil-
acquisitive and constitutive motivations may dren at day care centers in Haifa resulted
not be separable. in a doubling of the number of tardy pick-
Some of the founders of economics knew ups (Gneezy and Rustichini 2000a). But the
this. Jeremy Bentham’s Introduction to the small tax on plastic grocery bags enacted in
Principles of Morals and Legislation (1789), Ireland in 2002 had the opposite effect: in
is arguably the first text in what we now call two weeks it resulted in a 94 percent decline
public economics. In it he explained how in their use and appeared to crowd in social
proper incentives should harness self-inter- preferences (Rosenthal 2008).
ested objectives for public ends by making The contrast is instructive. In the Haifa
“it each man’s interest to observe . . . that case, the experimenters (respecting stan-
conduct which it is his duty to observe.” In dard experimental protocols) provided no
other words, make sure that doing his duty is justification for the introduction of the fine
incentive compatible. But he also understood on the tardy parents. Moreover the parents’
the constitutive side of action and the need occasional lateness could have occurred for
to design incentives that are complements of reasons beyond their control, rather than
the moral sentiments rather than substitutes: as the result of a deliberate disregard for
“A punishment may be said to be . . . a moral the inconvenience it caused the teachers.
lesson, when by reason of the ignominy it Finally, lateness was not so common as to
stamps upon the offence, it is calculated to be widely broadcast to the other parents. By
inspire the public with sentiments of aver- contrast, the introduction of the Irish plas-
sion towards those pernicious habits and dis- tic bag tax was preceded by a substantial
positions with which the offence appears to publicity campaign, and the use of the bags
be connected; and thereby to inculcate the required a deliberate choice made in a highly
opposite beneficial habits and dispositions” public condition. In the Irish case, as in the
(Bentham 1970 [1789], 26). experiment by Vertova and Galbiati (2010)
Few economists followed Bentham in this. mentioned in section 9, the monetary incen-
An exception is Albert Hirschman, who noted tive was introduced jointly with a message of
that economists seek “to deal with unethical explicit social obligation, and it apparently
or antisocial behavior by raising the cost of served as a reminder of the larger social costs
that behavior rather than proclaiming stan- of the use and disposal of the bags.
dards and imposing prohibitions and sanc- The same message comes from a vot-
tions. The reason is probably that they think ing study. In Switzerland, the removal of a
of citizens as consumers with unchanging or negligible fine for not voting significantly
arbitrarily changing tastes in matters civic as reduced voting turnout; but a considerable
well as commodity-related behavior. . . A prin- reduction in the cost of voting (by allowing
cipal purpose of publicly proclaimed laws and balloting by mail) had no effect on turnout.

02_Bowles_502.indd 417 5/25/12 11:42 AM


418 Journal of Economic Literature, Vol. L (June 2012)

The ­implication is that the fine for not voting the information the incentive provides, and
encouraged turnout not as an incentive (by the preexisting normative frameworks of
affecting the costs of not voting) but rather the actors. This is the message of the con-
as a message of the importance of one’s civic trast between the Irish grocery bag tax and
duty (Funk 2007). the Haifa fines for tardiness, along with the
The fact that fines often work more as fact that fines imposed on low contributors
messages than as incentives poses a problem by peers in Public Goods Games have posi-
for the sophisticated planner because the tive effects while fines imposed by principals
same intervention may bear radically differ- on agents sometimes backfire. In addition,
ent messages in different cultures. Bohnet incentives chosen by agents themselves (for
and her coauthors implemented a Trust example, by majority rule of team members),
Game in which in one treatment the inves- may have a more positive effect on individ-
tor had the option of reducing the payoffs ual performance than if they are imposed
of trustees who betrayed their trust (Bohnet (Mellizo, Carpenter, and Matthews 2011).
et al. 2010). Compared to the treatment in Fines deployed either to exploit or to con-
which this so-called “revenge” option was trol the target (or that give this appearance
not available, when they had the revenge or that have this effect) are likely to be less
option a substantially larger fraction of Saudi effective than they would under separabil-
investors trusted their partner, while a sub- ity and may even be counterproductive. The
stantially smaller fraction of American inves- reason, we think, is that they activate the
tors trusted. Making trust more incentive target’s desire to constitute himself or her-
compatible thus had diametrically opposed self as a dignified and autonomous individual
effects in the two cultures. who is treated fairly by others. It is this con-
stitutive motive that sometimes trumps the
acquisitiveness tapped by the incentive, and
12.  Conclusion: Are Incentives to Blame?
that leads to a contrary response. The same
Is there a simple lesson for public pol- incentives deployed by individuals who do
icy? We think there is. Titmuss was right not stand to benefit personally, and that are
that incentives sometimes crowd out non- intended to foster pro-social behavior are
economic motives, and this may degrade more likely to be complements rather than
economic performance. But Titmuss and substitutes for social preferences, crowd-
the literature that followed him targeted ing them in rather than out. They do this by
­incentives per se as the cause of crowding ­activating rather than diminishing the tar-
out and recommended a reduced role for get’s constitutive motives such as the desire
incentives in the governance of economic to be treated fairly and to treat others fairly,
interactions. to be a good member of a community, and
Both the diagnosis and the policy impli- the feeling of shame when others regard one
cation are wrong. Crowding out, as we have as having failed in this.
seen, may require greater, not lesser use of Present experimental and other evidence
incentives. And perhaps more important: give insufficient guidance to the planner who
fines, subsidies, and other monetary incen- wishes to know ex ante, the effects of the
tives per se may not be the culprit. What incentives he is considering implementing.
accounts for crowding out, we believe, is But on the basis of what we do know a good
the meaning of the fines or subsidies to the rule might be the following: The policy pack-
target of the incentives; and this depends on age of which the incentives are part should
the social relationships among the actors, let the target understand that the desired

02_Bowles_502.indd 418 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 419

modification in her actions will serve to Lab? Evidence from the Trust Game.” National
implement an outcome that is socially ben- Bureau of Economic Research Working Paper 15654.
Bar-Gill, Oren, and Chaim Fershtman. 2005. “Public
eficial so that that the target is more likely to Policy with Endogenous Preferences.” Journal of
endorse the purpose of the incentive, rather Public Economic Theory 7 (5): 841–57.
than being offended by it as either unjust or Barr, Abigail. 2001. “Social Dilemmas and Shame-Based
Sanctions: Experimental Results from Rural Zimba-
a threat to her autonomy or in some other bwe.” University of Oxford Centre for the Study of
way reflecting badly on the intentions of the African Economies Working Paper 2001.11.
planner. Barr, Abigail. 2004. “The Effects of Birthplace and
Current Context on Other-Regarding Preferences in
ACCRA.” Unpublished.
References Barr, Abigail, et al. 2009. “Homo Aequalis: A Cross-
Society Experimental Analysis of Three Bargaining
Aaron, Henry J. 1994. “Distinguished Lecture on Eco- Games.” University of Oxford Department of Eco-
nomics in Government: Public Policy, Values, and nomics Discussion Paper 422.
Consciousness.” Journal of Economic Perspectives 8 Becker, Gary S. 1976. “Altruism, Egoism, and Genetic
(2): 3–21. Fitness: Economics and Sociobiology.” Journal of
Akerlof, George A. 1984. An Economic Theorist’s Book Economic Literature 14 (3): 817–26.
of Tales. Cambridge and New York: Cambridge Uni- Belkin, Douglas. 2002. “Boston Firefighters Sick—or
versity Press. Tired of Working 15-day Allowance Seen Fueling
Akerlof, George A., and Rachel E. Kranton. 2010. Iden- Call-Ins.” Boston Globe, January 18, B1.
tity Economics: How Our Identities Shape Our Work, Bellemare, Charles, Sabine Kröger, and Arthur van
Wages, and Well-Being. Princeton and Oxford: Princ- Soest. 2008. “Measuring Inequity Aversion in a Het-
eton University Press. erogeneous Population Using Experimental Deci-
Al-Ubaydli, Omar, Daniel Houser, John Nye, Maria sions and Subjective Probabilities.” Econometrica 76
Pia Paganelli, and Xiaofei (Sophia) Pan. 2011. “The (4): 815–39.
Causal Effect of Market Participation on Trust: An Benabou, Roland, and Jean Tirole. 2003. “Intrinsic and
Experimental Investigation Using Randomized Con- Extrinsic Motivation.” Review of Economic Studies 70
trol.” Unpublished. (3): 489–520.
Andreoni, James. 1990. “Impure Altruism and Dona- Benabou, Roland, and Jean Tirole. 2006. “Incentives
tions to Public Goods: A Theory of Warm-Glow Giv- and Prosocial Behavior.” American Economic Review
ing?” Economic Journal 100 (401): 464–77. 96 (5): 1652–78.
Angrist, Joshua, and Victor Lavy. 2009. “The Effects Bentham, Jeremy. 1970. An Introduction to the Prin-
of High Stakes High School Achievement Awards: ciples of Morals and Legislation. London: Athlone
Evidence from a Randomized Trial.” American Eco- Press, 1789.
nomic Review 99 (4): 1384–1414. Benz, Matthias, and Stephan Meier. 2008. “Do People
Ariely, Dan, Anat Bracha, and Stephan Meier. 2009. Behave in Experiments as in the Field? Evidence
“Doing Good or Doing Well? Image Motivation and from Donations.” Experimental Economics 11 (3):
Monetary Incentives in Behaving Prosocially.” Ameri- 268–81.
can Economic Review 99 (1): 544–55. Besley, Timothy J., Gwyn Bevan, and Konrad B. Bur-
Ariely, Dan, Uri Gneezy, George Loewenstein, and chardi. 2009. “Naming & Shaming: The Impacts of
Nina Mazar. 2009. “Large Stakes and Big Mistakes.” Different Regimes on Hospital Waiting Times in
Review of Economic Studies 76 (2): 451–69. England and Wales.” Center for Economic and Policy
Arrow, Kenneth J. 1971. “Political and Economic Evalu- Research Discussion Paper 7306.
ation of Social Effects and Externalities.” In Frontiers Bewley, Truman F. 1999. Why Wages Don’t Fall During
of Quantitative Economics, edited by Michael Intrili- a Recession. Cambridge and London: Harvard Uni-
gator, 3–23. Amsterdam: North-Holland. versity Press.
Arrow, Kenneth J. 1972. “Gifts and Exchanges.” Phi- Birch, Leann L., and D. W. Marlin. 1982. “I Don’t Like
losophy and Public Affairs 1 (4): 343–62. It; I Never Tried It: Effects of Exposure on Two-Year-
Bandiera, Oriana, Iwan Barankay, and Imran Rasul. Old Children’s Food Preferences.” Appetite 3 (4):
2005. “Social Preferences and the Response to Incen- 353–60.
tives: Evidence from Personnel Data.” Quarterly Bisin, Alberto, and Thierry Verdier. 2001. “The Eco-
Journal of Economics 120 (3): 917–62. nomics of Cultural Transmission and the Dynamics
Bandura, Albert. 1991. “Social Cognitive Theory of of Preferences.” Journal of Economic Theory 97 (2):
Moral Thought and Action.” In Handbook of Moral 298–319.
Behavior and Development, Volume 1: Theory, Bisin, Alberto, and Thierry Verdier. 2011. “The Eco-
edited by William M. Kurtines and Jacob L. Gewirtz, nomics of Cultural Transmission and Socialization.”
45–103. Hillsdale, NJ: Erlbaum. In Handbook of Social Economics, Volume 1A, edited
Baran, Nicole M., Paola Sapienza, and Luigi Zingales. by Jess Benhabib, Matthew O. Jackson, and Alberto
2010. “Can We Infer Social Preferences from the Bisin, 339–416. Amsterdam and San Diego: Elsevier,

02_Bowles_502.indd 419 5/25/12 11:42 AM


420 Journal of Economic Literature, Vol. L (June 2012)

North-Holland. and the Evolutionary Process. Chicago and London:


Bliss, C. J. 1972. “Review of R. M. Titmuss ‘The Gift University of Chicago Press.
Relationship: From Human Blood to Social Policy.’” Burks, Stephen, Jeffrey Carpenter, and Lorenz Goette.
Journal of Public Economics 1 (1): 162–65. 2009. “Performance Pay and Worker Cooperation:
Bloom, Paul. 2010. How Pleasure Works: The New Sci- Evidence from an Artefactual Field Experiment.”
ence of Why We Like What We Like. New York and Journal of Economic Behavior and Organization 70
London: Norton. (3): 458–69.
Bohnet, Iris, and Yael Baytelman. 2007. “Institutions Cardenas, Juan Camilo. 2004. “Norms from Outside
and Trust: Implications for Preferences, Beliefs and and from Inside: An Experimental Analysis on the
Behavior.” Rationality and Society 19 (1): 99–135. Governance of Local Ecosystems.” Forest Policy and
Bohnet, Iris, Bruno S. Frey, and Steffen Huck. Economics 6 (3–4): 229–41.
2001. “More Order with Less Law: On Contract Cardenas, Juan Camilo. 2005. “Groups, Commons and
­Enforcement, Trust, and Crowding.” American Politi- Regulations: Experiments with Villagers and ­Students
cal Science Review 95 (1): 131–44. in Colombia.” In Psychology, Rationality and Eco-
Bohnet, Iris, Fiona Greig, Benedikt Herrmann, and nomic Behavior: Challenging Standard Assumptions,
Richard Zeckhauser. 2008. “Betrayal Aversion: Evi- edited by Bina Agarwal and Alessandro Vercelli,
dence from Brazil, China, Oman, Switzerland, Tur- 242–70. Houndmills, U.K. and New York: Palgrave
key, and the United States.” American Economic Macmillan.
Review 98 (1): 294–310. Cardenas, Juan Camilo. 2011. “Social Norms and
Bohnet, Iris, Benedikt Herrmann, Mohamad Al-Ississ, Behavior in the Local Commons as Seen through
Andrea Robbett, Khalid Al-Yahia, and Richard Zeck- the Lens of Field Experiments.” Environmental and
hauser. 2010. “The Elasticity of Trust: How to Pro- Resource Economics 48 (3): 451–85.
mote Trust in the Arab Middle East and the United Cardenas, Juan Camilo, John K. Stranlund, and Cleve
States.” Harvard University Kennedy School of Gov- Willis. 2000. “Local Environmental Control and Insti-
ernment Research Working Paper 10-031. tutional Crowding-Out.” World Development 28 (10):
Borges, Georg, and Bernd Irlenbusch. 2007. “Fair- 1719–33.
ness Crowded Out by Law: An Experimental Study Carpenter, Jeffrey, Samuel Bowles, Herbert Gintis,
on Withdrawal Rights.” Journal of Institutional and and Sung-Ha Hwang. 2009. “Strong Reciprocity and
Theoretical Economics 163 (1): 84–101. Team Production: Theory and Evidence.” Journal of
Bowles, Samuel. 1989. “Mandeville’s Mistake: Markets Economic Behavior and Organization 71 (2): 221–32.
and the Evolution of Cooperation.” Presented to the Carpenter, Jeffrey, Stephen Burks, and Eric Verhoo-
September Seminar, University College London Phi- gen. 2005. “Comparing Students to Workers: The
losophy Department. Effects of Social Framing on Behavior in Distribu-
Bowles, Samuel. 1998. “Endogenous Preferences: The tion Games.” In Field Experiments in Economics,
Cultural Consequences of Markets and Other Eco- edited by Jeffrey Carpenter, Glenn W. Harrison,
nomic Institutions.” Journal of Economic Literature and John A. List, 261–89. Amsterdam and San
36 (1): 75–111. Diego: Elsevier.
Bowles, Samuel. 2004. Microeconomics: Behavior, Carpenter, Jeffrey, Cristina Connolly, and Caitlin
Institutions, and Evolution. Princeton and Oxford: Knowles Myers. 2008. “Altruistic Behavior in a Rep-
Princeton University Press; New York: Russell Sage resentative Dictator Experiment.” Experimental Eco-
Foundation. nomics 11 (3): 282–98.
Bowles, Samuel. 2008. “Policies Designed for Self- Carpenter, Jeffrey, and Caitlin Knowles Myers. 2010.
Interested Citizens May Undermine ‘the Moral Sen- “Why Volunteer? Evidence on the Role of Altruism,
timents’: Evidence from Economic Experiments.” Image, and Incentives.” Journal of Public Economics
Science 320 (5883): 1605–09. 94 (11–12): 911–20.
Bowles, Samuel. 2011. “Is Liberal Society a Parasite on Carpenter, Jeffrey, and Erika Seki. 2010. “Do Social
Tradition?” Philosophy & Public Affairs 39 (1): 46–81. Preferences Increase Productivity? Field Experimen-
Bowles, Samuel, and Herbert Gintis. 2006. “Prosocial tal Evidence from Fishermen in Toyama Bay.” Eco-
Emotions.” In The Economy as a Complex Evolv- nomic Inquiry 49 (2): 612–30.
ing System III: Current Perspectives and Future Cavalli-Sforza, Luigi, and Marcus W. Feldman. 1981.
Directions, edited by Lawrence Blume and Steven Cultural Transmission and Evolution: A Quantitative
Durlauf, 339–66. Oxford and New York: Oxford Uni- Approach. Princeton and Oxford: Princeton Univer-
versity Press. sity Press.
Bowles, Samuel, and Herbert Gintis. 2011. A Coopera- Cervellati, Matteo, Joan Esteban, and Laurence
tive Species: Human Reciprocity and Its Evolution. Kranich. 2010. “Work Values, Endogenous Senti-
Princeton and Oxford: Princeton University Press. ments and Redistribution.” Journal of Public Eco-
Bowles, Samuel, and Sung-Ha Hwang. 2008. “Social nomics 94 (9–10): 612–27.
Preferences and Public Economics: Mechanism Cleave, Blair L., Nikos Nikiforakis, and Robert Slonim.
Design When Social Preferences Depend on Incen- 2010. “Is There Selection Bias in Laboratory Experi-
tives.” Journal of Public Economics 92 (8–9): 1811–20. ments?” University of Sydney Economics Working
Boyd, Robert, and Peter J. Richerson. 1985. Culture Paper 2010-1.

02_Bowles_502.indd 420 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 421

Cohn, Alain, Ernst Fehr, and Lorenz Goette. 2011. Sciences.” Science 326 (5952): 535–38.
“Fair Wages and Effort: Evidence from a Field Falk, Armin, and Michael Kosfeld. 2006. “The Hidden
Experiment.” Unpublished. Costs of Control.” American Economic Review 96 (5):
Cooley, Charles Horton. 1902. Human Nature and the 1611–30.
Social Order. New York: Charles Scribner’s Sons. Falk, Armin, Stephan Meier, and Christian Zehnder.
Cooter, Robert. 2000. “Do Good Laws Make Good Citi- 2011. “Did We Overestimate the Role of Social
zens? An Economic Analysis of Internalized Norms.” Preferences? The Case of Self-Selected Student
Virginia Law Review 86 (8): 1577–1601. Samples.” Institute for the Study of Labor Discussion
Deci, Edward L. 1975. Intrinsic Motivation. New York: Paper 5475.
Plenum Press. Falkinger, Josef. 1996. “Efficient Private Provision of
Deci, Edward L., Richard Koestner, and Richard M. Public Goods by Rewarding Deviations from Aver-
Ryan. 1999. “A Meta-Analytic Review of Experi- age.” Journal of Public Economics 62 (3): 413–22.
ments Examining the Effects of Extrinsic Rewards on Falkinger, Josef, Ernst Fehr, Simon Gächter, and Rudolf
Intrinsic Motivation.” Psychological Bulletin 125 (6): Winter-Ebmer. 2000. “A Simple Mechanism for the
627–68. Efficient Provision of Public Goods: E ­ xperimental
Deci, Edward L, and Richard M. Ryan. 1985. Intrinsic Evidence.” American Economic Review 90 (1):
Motivation and Self-Determination in Human Behav- 247–64.
ior. New York: Plenum Press. Fehr, Ernst, and Simon Gächter. 2000. “Cooperation
DellaVigna, Stefano. 2009. “Psychology and Economics: and Punishment in Public Goods Experiments.”
Evidence from the Field.” Journal of Economic Lit- American Economic Review 90 (4): 980–94.
erature 47 (2): 315–72. Fehr, Ernst, and Simon Gächter. 2002a. “Altruistic Pun-
Dickinson, David, and Marie-Claire Villeval. 2008. ishment in Humans.” Nature 415: 137–40.
“Does Monitoring Decrease Work Effort? The Com- Fehr, Ernst, and Simon Gächter. 2002b. “Do Incentive
plementarity between Agency and Crowding-Out Contracts Undermine Voluntary Cooperation?” Uni-
Theories.” Games and Economic Behavior 63 (1): versity of Zurich Institute for Empirical Research in
56–76. Economics Working Paper 34.
Drago, Francesco, Roberto Galbiati, and Pietro Vertova. Fehr, Ernst, Simon Gächter, and Georg Kirchstei-
2009. “The Deterrent Effects of Prison: Evidence ger. 1997. “Reciprocity as a Contract Enforcement
from a Natural Experiment.” Journal of Political Device: Experimental Evidence.” Econometrica 65
Economy 117 (2): 257–80. (4): 833–60.
Dufwenberg, Martin, Paul Heidhues, Georg Kirch- Fehr, Ernst, and Lorenz Goette. 2007. “Do Workers
steiger, Frank Riedel, and Joel Sobel. 2011. “Other- Work More If Wages Are High? Evidence from a
Regarding Preferences in General Equilibrium.” Randomized Field Experiment.” American Economic
Review of Economic Studies 78 (2): 613–39. Review 97 (1): 298–317.
Ellingsen, Tore, and Magnus Johannesson. 2008. Fehr, Ernst, Alexander Klein, and Klaus M. Schmidt.
“Pride and Prejudice: The Human Side of Incen- 2007. “Fairness and Contract Design.” Econometrica
tive Theory.” American Economic Review 98 (3): 75 (1): 121–54.
990–1008. Fehr, Ernst, and Andreas Leibbrandt. 2011. “A Field
Ellingsen, Tore, Magnus Johannesson, Johanna Möller- Study on Cooperativeness and Impatience in the
ström, and Sara Munkhammar. 2011. “Social Framing Tragedy of the Commons.” Journal of Public Econom-
Effects: Preferences or Beliefs?” http://www2.hhs.se/ ics 95 (9–10): 1144–55.
personal/ellingsen/pdf/FramingGEB1.pdf. Fehr, Ernst, and John A. List. 2004. “The Hidden
Elster, Jon. 1998. “Emotions and Economic Theory.” Costs and Returns of Incentives—Trust and Trust-
Journal of Economic Literature 36 (1): 47–74. worthiness among CEOs.” Journal of the European
Ertan, Arhan, Talbot Page, and Louis Putterman. 2009. ­Economic Association 2 (5): 743–71.
“Who to Punish? Individual Decisions and Majority Fehr, Ernst, and Bettina Rockenbach. 2003. “Detri-
Rule in Mitigating the Free Rider Problem.” Euro- mental Effects of Sanctions on Human Altruism.”
pean Economic Review 53 (5): 495–511. Nature 422 (6928): 137–40.
Falk, Armin, Ernst Fehr, and Urs Fischbacher. 2005. Fehr, Ernst, and Klaus M. Schmidt. 1999. “A Theory of
“Driving Forces Behind Informal Sanctions.” Econo- Fairness, Competition, and Cooperation.” Quarterly
metrica 73 (6): 2017–30. Journal of Economics 114 (3): 817–68.
Falk, Armin, Ernst Fehr, and Christian Zehnder. 2006. Fehr, Ernst, and Klaus M. Schmidt. 2007. “Adding a
“Fairness Perceptions and Reservation Wages—the Stick to the Carrot? The Interaction of Bonuses and
Behavioral Effects of Minimum Wage Laws.” Quar- Fines.” American Economic Review 97 (2): 177–81.
terly Journal of Economics 121 (4): 1347–81. Fershtman, Chaim, and Aviad Heifetz. 2006. “Read My
Falk, Armin, Simon Gächter, and Judit Kovacs. 1999. Lips, Watch for Leaps: Preference Equilibrium and
“Intrinsic Motivation and Extrinsic Incentives in a Political Instability.” Economic Journal 116 (508):
Repeated Game with Incomplete Contracts.” Journal 246–65.
of Economic Psychology 20 (3): 251–84. Fiske, Alan P. 1992. “The Four Elementary Forms of
Falk, Armin, and James J. Heckman. 2009. “Lab Experi- Sociality: Framework for a Unified Theory of Social
ments Are a Major Source of Knowledge in the Social Relations.” Psychological Review 99 (4): 689–723.

02_Bowles_502.indd 421 5/25/12 11:42 AM


422 Journal of Economic Literature, Vol. L (June 2012)

Frey, Bruno S. 1993. “Does Monitoring Increase Work Grant, Ruth W. 2011. Strings Attached: Untangling the
Effort? The Rivalry with Trust and Loyalty.” Eco- Ethics of Incentives. Princeton and Oxford: Princeton
nomic Inquiry 31 (4): 663–70. University Press.
Frey, Bruno S. 1997. “A Constitution for Knaves Greenberger, Scott S. 2003. “Sick Day Abuses Focus of
Crowds Out Civic Virtues.” Economic Journal 107 Fire Talks.” Boston Globe, September 17, B7.
(443): 1043–53. Güth, Werner, and Hartmut Kliemt. 1994. “Compe-
Frey, Bruno S., and Reto Jegen. 2001. “Motivation tition or Co-operation: On the Evolutionary Eco-
Crowding Theory.” Journal of Economic Surveys 15 nomics of Trust, Exploitation and Moral Attitudes.”
(5): 589–611. Metroeconomica 45 (2): 155–87.
Frey, Bruno S., and Felix Oberholzer-Gee. 1997. “The Healy, Kieran. 2006. Last Best Gifts: Altruism and the
Cost of Price Incentives: An Empirical Analysis of Market for Human Blood and Organs. Chicago and
Motivation Crowding-Out.” American Economic London: University of Chicago Press.
Review 87 (4): 746–55. Heifetz, Aviad, Ella Segev, and Eric Talley. 2007. “Mar-
Frey, Bruno S., and Alois Stutzer. 2008. “Environ- ket Design with Endogenous Preferences.” Games
mental Morale and Motivation.” In The Cambridge and Economic Behavior 58 (1): 121–53.
­Handbook of Psychology and Economic Behaviour, Helsley, Robert W., and William C. Strange. 2000.
edited by Alan Lewis, 406–28. Cambridge and New “Social Interactions and the Institutions of Local
York: Cambridge University Press. Government.” American Economic Review 90 (5):
Fryer, Roland G. 2011. “Financial Incentives and Stu- 1477–90.
dent Achievement: Evidence from Randomized Tri- Henrich, Joseph, et al. 2005. “‘Economic Man’ in Cross-
als” Quarterly Journal of Economics 126 (4): 1755–98. Cultural Perspective: Behavioral Experiments in 15
Funk, Patricia. 2007. “Is There An Expressive Func- Small-Scale Societies.” Behavioral and Brain Sciences
tion of Law? An Empirical Analysis of Voting Laws 28 (6): 795–815.
with Symbolic Fines.” American Law and Economics Henrich, Joseph, et al. 2010. “Markets, Religion, Com-
Review 9 (1): 135–59. munity Size, and the Evolution of Fairness and Pun-
Gächter, Simon, and Armin Falk. 2002. “Reputation and ishment.” Science 327 (5972): 1480–84.
Reciprocity: Consequences for the Labour Relation.” Herrmann, Benedikt, and Henrik Orzen. 2008. “The
Scandinavian Journal of Economics 104 (1): 1–26. Appearance of Homo Rivalis: Social Preferences and
Gächter, Simon, and Ernst Fehr. 1999. “Collective the Nature of Rent Seeking.” University of Notting-
Action as a Social Exchange.” Journal of Economic ham Centre for Decision Research and Experimental
Behavior and Organization 39 (4): 341–69. Economics Discussion Paper 2008-10.
Gächter, Simon, Esther Kessler, and Manfred König- Herrmann, Benedikt, Christian Thöni, and Simon
stein. 2011. “The Roles of Incentives and Voluntary Gächter. 2008. “Antisocial Punishment across Societ-
Cooperation for Contractual Compliance” Univer- ies.” Science 319 (5868): 1362–67.
sity of Nottingham Centre for Decision Research Heyman, James, and Dan Ariely. 2004. “Effort for Pay-
and Experimental Economics Discussion Paper ment: A Tale of Two Markets.” Psychological Science
2011-06. 15 (11): 787–93.
Gächter, Simon, Daniele Nosenzo, and Martin Sef- Hirschman, Albert O. 1977. The Passions and the Inter-
ton. 2010. “The Impact of Social Comparisons on ests: Political Arguments for Capitalism before Its
Reciprocity.” University of Nottingham Centre for Triumph. Princeton and Oxford: Princeton University
Decision Research and Experimental Economics Press.
Discussion Paper 2010-10. Hirschman, Albert O. 1985. “Against Parsimony: Three
Galbiati, Roberto, and Pietro Vertova. 2008. “Obliga- Easy Ways of Complicating Some Categories of Eco-
tions and Cooperative Behaviour in Public Good nomic Discourse.” Economics and Philosophy 1 (1):
Games.” Games and Economic Behavior 64 (1): 7–21.
146–70. Hoffman, Elizabeth, Kevin McCabe, Keith Shachat,
Ginges, Jeremy, Scott Atran, Douglas Medin, and Khalil and Vernon Smith. 1994. “Preferences, Property
Shikaki. 2007. “Sacred Bounds on Rational Resolu- Rights, and Anonymity in Bargaining Games.” Games
tion of Violent Political Conflict.” Proceedings of the and Economic Behavior 7 (3): 346–80.
National Adademy of Sciences 104 (18): 7357–60. Holmås, Tor Helge, Egil Kjerstad, Hilde Lurås, and
Gneezy, Uri. 2003. “The W Effect of Incentives.” Odd Rune Straume. 2010. “Does Monetary Punish-
Unpublished. ment Crowd Out Pro-social Motivation? A Natural
Gneezy, Uri, Stephan Meier, and Pedro Rey-Biel. 2011. Experiment on Hospital Length of Stay.” Journal of
“When and Why Incentives (Don’t) Work to Modify Economic Behavior and Organization 75 (2): 261–67.
Behavior.” Journal of Economic Perspectives 25 (4): Hopfensitz, Astrid, and Ernesto Reuben. 2009. “The
191–210. Importance of Emotions for the Effectiveness of
Gneezy, Uri, and Aldo Rustichini. 2000a. “A Fine Is a Social Punishment.” Economic Journal 119 (540):
Price.” Journal of Legal Studies 29 (1): 1–17. 1534–59.
Gneezy, Uri, and Aldo Rustichini. 2000b. “Pay Enough Houser, Daniel, Erte Xiao, Kevin McCabe, and Vernon
or Don’t Pay at All.” Quarterly Journal of Economics Smith. 2008. “When Punishment Fails: Research on
115 (3): 791–810. Sanctions, Intentions and Non-cooperation.” Games

02_Bowles_502.indd 422 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 423

and Economic Behavior 62 (2): 509–32. dence of Social Preferences.” Economic Journal 114
Hwang, Sung-Ha, and Samuel Bowles. 2011a. “Opti- (492): 121–49.
mal Incentives with State-Dependent Preferences.” Loewenstein, George. 2000. “Emotions in Economic
Unpublished. Theory and Economic Behavior.” American Eco-
Hwang, Sung-Ha, and Samuel Bowles. 2011b. “The nomic Review 90 (2): 426–32.
Sophisticated Planner’s Dilemma: Optimal Incentives Lopez, Maria Claudia, James J. Murphy, John M. Sprag-
with Endogenous Preferences.” Unpublished. gon, and John K. Stranlund. 2012. “Comparing the
Irlenbusch, Bernd, and Gabriele K. Ruchala. 2008. Effectiveness of Regulation and Pro-social Emotions
“Relative Rewards within Team-Based Compensa- to Enhance Cooperation: Experimental Evidence
tion.” Labour Economics 15 (2): 141–67. from Fishing Communities in Colombia.” Economic
Irlenbusch, Bernd, and Dirk Sliwka. 2005. “Incentives, Inquiry 50 (1): 131–42.
Decision Frames, and Motivation Crowding Out— Lucas, Robert E. 1976. “Econometric Policy Evalua-
An Experimental Investigation.” Institute for the tion: A Critique.” Carnegie–Rochester Conference
Study of Labor Discussion Paper 1758. Series on Public Policy 1: 19–46.
Karlan, Dean S. 2005. “Using Experimental Econom- MacLeod, W. Bentley. 2007. “Can Contract Theory
ics to Measure Social Capital and Predict ­Financial Explain Social Preferences?” American Economic
­Decisions.” American Economic Review 95 (5): Review 97 (2): 187–92.
1688–99. Masclet, David, Charles Noussair, Steven Tucker, and
Kessler, Esther. 2008. “Behavioural Economics of Per- Marie-Claire Villeval. 2003. “Monetary and Non-
formance Incentives.” University of Nottingham monetary Punishment in the Voluntary Contributions
School of Economics. Mechanism.” American Economic Review 93 (1):
Kosfeld, Michael, Akira Okada, and Arno Riedl. 2009. 366–80.
“Institution Formation in Public Goods Games.” Meier, Stephan. 2007a. “Do Subsidies Increase Chari-
American Economic Review 99 (4): 1335–55. table Giving in the Long Run? Matching Donations
Kreps, David M. 1997. “Intrinsic Motivation and Extrin- in a Field Experiment.” Journal of the European Eco-
sic Incentives.” American Economic Review 87 (2): nomic Association 5 (6): 1203–22.
359–64. Meier, Stephan. 2007b. “A Survey of Economic Theo-
Kroll, Stephan, Todd L. Cherry, and Jason F. Shogren. ries and Field Evidence on Pro-social Behavior.” In
2007. “Voting, Punishment, and Public Goods.” Eco- Economics and Psychology: A Promising New Cross-
nomic Inquiry 45 (3): 557–70. Disciplinary Field, edited by Bruno S. Frey and Alois
Laffont, Jean-Jacques, and Mohamed Salah Matoussi. Stutzer, 51–88. Cambridge and London: MIT Press.
1995. “Moral Hazard, Financial Constraints and Mellizo, Philip, Jeffrey Carpenter, and Peter Hans Mat-
Sharecropping in El Oulja.” Review of Economic thews. 2011. “Workplace Democracy in the Lab.”
Studies 62 (3): 381–99. Unpublished.
Lazear, Edward P. 2000. “Performance Pay and Produc- Mellstrom, Carl, and Magnus Johannesson. 2008.
tivity.” American Economic Review 90 (5): 1346–61. “Crowding Out in Blood Donation: Was Titmuss
Leibbrandt, Andreas, Uri Gneezy, and John A. List. Right?” Journal of the European Economic Associa-
2010. “Ode to the Sea: The Socio-ecological Under- tion 6 (4): 845–63.
pinnings of Social Norms.” Unpublished. Mill, John Stuart. 1867. Principles of Political Economy
Leider, Stephen, Markus M. Möbius, Tanya Rosenblat, with Some of Their Applications to Social Philosophy.
and Quoc-Anh Do. 2009. “Directed Altruism and New York; Bombay and Culcutta: Longmans, Green
Enforced Reciprocity in Social Networks.” Quarterly and Co., 1848.
Journal of Economics 124 (4): 1815–51. Mulder, Laetitia B., Eric van Dijk, David De Cremer,
Lepper, Mark R., Gerald Sagotsky, Janet L. Dafoe, and and Henk A. M. Wilke. 2006. “Undermining Trust
David Greene. 1982. “Consequences of Superflu- and Cooperation: The Paradox of Sanctioning Sys-
ous Social Constraints: Effects on Young Children’s tems in Social Dilemmas.” Journal of Experimental
Social Inferences and Subsequent Intrinsic Interest.” Social Psychology 42 (2): 147–62.
Journal of Personality and Social Psychology 42 (1): Murphy, Sheila T., Jennifer L. Monahan, and Robert
51–65. B. Zajonc. 1995. “Additivity of Nonconscious Affect:
Levine, David K. 1998. “Modeling Altruism and Spite- Combined Effects of Priming and Exposure.” Journal
fulness in Experiments.” Review of Economic Dynam- of Personality and Social Psychology 69 (4): 589–602.
ics 1 (3): 593–622. Murphy, Sheila T., and Robert B. Zajonc. 1993. “Affect,
Levitt, Steven D., and John A. List. 2007. “What Do Cognition, and Awareness: Affective Priming with
Laboratory Experiments Measuring Social Prefer- Optimal and Suboptimal Stimulus Exposures.”
ences Reveal about the Real World?” Journal of Eco- Journal of Personality and Social Psychology 64 (5):
nomic Perspectives 21 (2): 153–74. 723–39.
Li, Jian, Erte Xiao, Daniel Houser, and P. Read Mon- Rabin, Matthew. 1993. “Incorporating Fairness into
tague. 2009. “Neural Responses to Sanction Threats Game Theory and Economics.” American Economic
in Two-Party Economic Exchange.” Proceedings of Review 83 (5): 1281–1302.
the National Academy of Sciences 106 (39): 16835–40. Rabin, Matthew. 1998. “Psychology and Economics.”
List, John A. 2004. “Young, Selfish and Male: Field Evi- Journal of Economic Literature 36 (1): 11–46.

02_Bowles_502.indd 423 5/25/12 11:42 AM


424 Journal of Economic Literature, Vol. L (June 2012)

Reeson, Andrew F., and John G. Tisdell. 2008. Markets with Other-Regarding Preferences.” Inter-
“­
Institutions, Motivations and Public Goods: An national Review of Economics 56 (1): 3–16.
Experimental Test of Motivational Crowding.” Jour- Sobel, Joel. 2010. “Markets and Other-Regarding
nal of Economic Behavior and Organization 68 (1): Preferences.” http://federation.ens.fr/ydepot/semin/
273–81. texte0910/SOB2010MAR.pdf.
Rigdon, Mary. 2009. “Trust and Reciprocity in Incen- Solow, Robert M. 1971. “Blood and Thunder: Review of
tive Contracting.” Journal of Economic Behavior and the Gift Relationship: From Human Blood to Social
Organization 70 (1–2): 93–105. Policy by Richard M. Titmuss.” Yale Law Journal 80
Rodriguez-Sickert, Carlos, Ricardo Andrés Guzmán, (8): 1696–1711.
and Juan Camilo Cardenas. 2008. “Institutions Influ- Stanca, Luca, Luigino Bruni, and Luca Corazzini.
ence Preferences: Evidence from a Common Pool 2009. “Testing Theories of Reciprocity: Do Motiva-
Resource Experiment.” Journal of Economic Behav- tions Matter?” Journal of Economic Behavior and
ior and Organization 67 (1): 215–27. ­Organization 71 (2): 233–45.
Rosenthal, Elisabeth. 2008. “Motivated by a Tax, Irish Stiglitz, Joseph E. 1987. “The Causes and Conse-
Spurn Plastic Bags.” New York Times, February 2. quences of the Dependence of Quality on Price.”
Ross, Lee, and Richard E. Nisbett. 1991. The Person Journal of Economic Literature 25 (1): 1–48.
and the Situation: Perspectives of Social Psychology. Sunstein, Cass R. 1996. “On the Expressive Function
Philadelphia: Temple University Press. of Law.” University of Pennsylvania Law Review 144
Rustagi, Devesh, Stefanie Engel, and Michael Kosfeld. (5): 2021–53.
2010. “Conditional Cooperation and Costly Monitor- Sutter, Matthias, Stefan Haigner, and Martin
ing Explain Success in Forest Commons Manage- G. Kocher. 2011. “Choosing the Carrot or the
ment.” Science 330 (6006): 961–965. Stick? Endogenous Institutional Choice in Social
Rutstrom, E. Elisabet. 2002. “Sparing the Rod Does Dilemma Situations.” Review of Economic Studies
Not Spoil the Child: An Experimental Study of Incen- 77 (4): 1540–66.
tive Effects.” Unpublished. Taylor, Michael. 1987. The Possibility of Cooperation.
Schnedler, Wendelin, and Radovan Vadovic. 2011. Cambridge and New York: Cambridge University
“Legitimacy of Control.” Journal of Economics and Press.
Management Strategy 20 (4): 985–1009. Tenbrunsel, Ann F., and David M. Messick. 1999.
Schotter, Andrew, Avi Weiss, and Inigo Zapater. 1996. “Sanctioning Systems, Decision Frames, and Coop-
“Fairness and Survival in Ultimatum and Dictatorship eration.” Administrative Science Quarterly 44:
Games.” Journal of Economic Behavior and Organi- 684–707.
zation 31 (1): 37–56. Tirole, Jean. 1999. “Incomplete Contracts: Where Do
Seabright, Paul B. 2009. “Continuous Preferences and We Stand?” Econometrica 67 (4): 741–81.
Discontinuous Choices: How Altruists Respond to Titmuss, Richard M. 1971. The Gift Relationship:
Incentives.” B.E. Journal of Theoretical Economics 9 (1). From Human Blood to Social Policy. New York:
Serra, Danila. 2008. “Combining Top–Down and Bot- Pantheon Books.
tom–Up Accountability: Evidence from a Bribery Tversky, Amos, and Daniel Kahneman. 1981. “The
Experiment.” University of Oxford Centre for the Framing of Decisions and the Psychology of
Study of African Economies Working Paper 2008-25. Choice.” Science 211 (4481): 453–58.
Shinada, Mizuho, and Toshio Yamagishi. 2007. “Punish- Tyran, Jean-Robert, and Lars P. Feld. 2006. “Achiev-
ing Free Riders: Direct and Indirect Promotion of ing Compliance When Legal Sanctions Are Non-
Cooperation.” Evolution and Human Behavior 28 (5): deterrent.” Scandinavian Journal of Economics 108
330–39. (1): 135–56.
Shu, Lisa L., Francesca Gino, and Max H. Bazerman. Vertova, Pietro, and Roberto Galbiati. 2010. “How
2011. “Dishonest Deed, Clear Conscience: When Laws Affect Behavior: Obligations, Incen-
Cheating Leads to Moral Disengagement and Moti- tives and Cooperative Behavior. http://ssrn.com/
vated Forgetting.” Personality and Social Psychology paper=1615349.
Bulletin 37 (3): 330–49. Warneken, Felix, and Michael Tomasello. 2008.
Sliwka, Dirk. 2007. “Trust as a Signal of a Social Norm “Extrinsic Rewards Undermine Altruistic Tenden-
and the Hidden Costs of Incentive Schemes.” Ameri- cies in 20-month-olds.” Developmental Psychology
can Economic Review 97 (3): 999–1012. 44 (6): 1785–88.
Sloof, Randolph, and Joep Sonnemans. 2001. “The Xiao, Erte, and Daniel Houser. 2011. “Punish in
Interaction between Explicit and Relational Incen- ­Public.” Journal of Public Economics 95 (7–8):
tives: An Experiment.” Games and Economic Behav- 1006–17.
ior 73 (2): 573–94. Yeung, King-To, and John Levi Martin. 2003. “The
Sobel, Joel. 2002. “Can We Trust Social Capital?” Jour- Looking Glass Self: An Empirical Test and Elabora-
nal of Economic Literature 40 (1): 139–54. tion.” Social Forces 81 (3): 843–79.
Sobel, Joel. 2005. “Interdependent Preferences and Young, H. Peyton, and Mary A. Burke. 2001. “Compe-
Reciprocity.” Journal of Economic Literature 43 (2): tition and Custom in Economic Contracts: A Case
392–436. Study of Illinois Agriculture.” American Economic
Sobel, Joel. 2009. “Generous Actors, Selfish Actions: Review 91 (3): 559–73.

02_Bowles_502.indd 424 5/25/12 11:42 AM


Bowles and Polanía-Reyes: Economic Incentives and Social Preferences 425

Zajonc, Robert B. 1968. “Attitudinal Effects of Mere Behavior.” Pschological Science 21 (3): 311–14.
Exposure.” Journal of Personality and Social Psy- Zhong, Chen-Bo, Jeffrey Loewenstein, and J. Keith
chology 9 (2 Part 2): 1–27. Murnighan. 2007. “Speaking the Same Language:
Zhong, Chen-Bo, Vanessa K. Bohns, and Francesca The Cooperative Effects of Labeling in the Pris-
Gino. 2010. “Good Lamps Are the Best Police: oner’s Dilemma.” Journal of Conflict Resolution 51
Darkness Increases Dishonestly and Self-Interested (3): 431–56.

02_Bowles_502.indd 425 5/25/12 11:42 AM

Das könnte Ihnen auch gefallen