Incentives and Prosocial Behavior - Princeton University - Home

Incentives and Prosocial Behavior

By ROLAND BENABOU AND JEAN TIROLE*

We develop a theory of prosocial behavior that combines heterogeneity in individualaltruism and greed with concerns for social reputation or self-respect. Rewards orpunishments (whether material or image-related) create doubt about the true motivefor which good deeds are performed, and this “overjustification effect” can inducea partial or even net crowding out of prosocial behavior by extrinsic incentives. Wealso identify the settings that are conducive to multiple social norms and, moregenerally, those that make individual actions complements or substitutes, which weshow depends on whether stigma or honor is (endogenously) the dominant reputa-tional concern. Finally, we analyze the socially optimal level of incentives and howmonopolistic or competitive sponsors depart from it. Sponsor competition is shownto potentially reduce social welfare. (JEL D11, D64, D82, Z13)

People commonly engage in activities that arecostly to themselves and that primarily benefitothers. They volunteer, help strangers, vote, giveto political or charitable organizations, donateblood, join rescue squads, or even sacrifice theirlife for strangers. In experiments, many subjectsalso display altruistic or reciprocal behaviors. Atthe same time, a number of important phenomenaand puzzles cannot be explained by the sole pres-ence of individuals with other-regarding prefer-ences. What is, therefore, the broader set ofmotives that shape people’s social conduct, andhow do these motives interact with each other andthe economic environment?

A first puzzle is that providing rewards andpunishments to foster prosocial behavior some-

times has a perverse effect, reducing the totalcontribution provided by agents. Such a crowdingout of “intrinsic motivation” by extrinsic incen-tives has been observed in a broad variety of socialinteractions (see Bruno S. Frey, 1997, and Freyand Reto Jegen, 2001, for surveys). Studyingschoolchildren collecting donations for a charita-ble organization, Uri Gneezy and Aldo Rustichini(2000b) thus found that they collected less moneywhen given performance incentives (see also Freyand Lorenz Gotte, 1999, on volunteer work sup-ply). These findings are in line with the ideas inRichard Titmuss (1970), who argued that payingblood donors could actually reduce supply. On thepunishment side, George A. Akerlof and WilliamT. Dickens (1982) suggested that imposing stifferpenalties could sometimes undermine individuals’“internal justification” for obeying the law. Frey(1997) provided some evidence to that effect withrespect to tax compliance, and Gneezy and Rus-tichini (2000a) found that fining parents for pick-ing up their children late from day-care centersresulted in more late arrivals. In experiments onlabor contracting, subjects provided less effortwhen the contract specified fines for inadequateperformance than when it did not (Fehr et al.,2001; Fehr and Gachter, 2001) and they behavedmuch less generously when the principal had sim-ply removed from their choice set the most selfishoptions (Armin Falk and Michael Kosfeld, 2006).These findings extend a large literature in psychol-ogy documenting how explicit incentives can leadto decreased motivation and unchanged or re-

* Benabou: Department of Economics and Woodrow Wil-son School, Princeton University, Princeton, NJ 08544, Cana-dian Institute for Advanced Research, Centre for EconomicPolicy Research, Institute for the Study of Labor, and NationalBureau for Economic Research (e-mail: [email protected]); Tirole: Institut d’Economie Industrielle, 21 Allees deBrienne, 31000 Toulouse, France, Groupe de Recherche enEconomic Mathematique et Quantitative, Centre d’Enseigne-ment et de Recherche en Analyse Socioeconomique, and Mas-sachusetts Institute of Technology (e-mail: [email protected]). Wethank George Akerlof, Samuel Bowles, Armin Falk, RolandFryer, Timur Kuran, Bentley MacLeod, Eric Rasmusen, TomRomer, participants at various seminars and conferences, andthree anonymous referees for useful comments. We are espe-cially indebted to Ian Jewitt for valuable suggestions. Benabougratefully acknowledges support from the John SimonGuggenheim Memorial Foundation in 2004 and from the Na-tional Science Foundation (SES-0424015).

1652

duced task performance (see, e.g. Edward Deci,1975; Deci and Richard Ryan, 1985). In studyingthis class of phenomena, however, one cannotsimply assume that rewards and punishments sys-tematically crowd out spontaneous contributions.Indeed, there is also much evidence to support thebasic premise of economics that incentives work,for instance in workplace contexts (e.g., RobertGibbons, 1997; Canice Prendergast, 1999; Ed-ward P. Lazear, 2000a, b). A more discriminatinganalysis is thus required.

A second set of issues relates to the fact thatpeople commonly perform good deeds and re-frain from selfish ones because of social pres-sure and norms that attach honor to the formerand shame to the latter (e.g., Dan Batson, 1998;Richard B. Freeman, 1997). Charitable and non-profit institutions make ample use of donors’desire to demonstrate their generosity and self-lessness (or at least the appearance thereof),with displays ranging from lapel pins and T-shirts to plaques in opera houses or hospitals,and buildings named after large contributors.Patricia Funk (2005) finds that the introductionof mail voting in Switzerland, which allowedcitizens to vote at a lower cost but simulta-neously made unobservable who did their “civicduty” and who did not, failed to raise the ag-gregate voting rate and actually resulted in adecline in small communes. The presence of asocial signalling motive for giving is also evi-dent in the fact that anonymous donations areboth extremely rare—typically, less than 1 per-cent of the total number1—and widely consid-ered to be the most admirable. Conversely,boasting of one’s generous contributions is of-ten self-defeating. Codes of honor, whose strin-gency and scope vary considerably across timeand societies, are another example of normsenforced largely through feelings of shame orglory. To understand these mechanisms, it isagain important not to posit exogenous socialconstraints, but rather to model the inferencesand market conditions involved in sustaining orinhibiting them.

Finally, as much as people care about theopinion others have of them, they care about

their self-image. In the words of Adam Smith(1759), they make moral decisions by assessingtheir own conduct through the eyes of an “im-partial spectator,” an “ideal mate within thebreast”: “We endeavour to examine our ownconduct as we imagine any other fair and im-partial spectator would examine it. If, uponplacing ourselves in his situation, we thor-oughly enter into all the passions and motiveswhich influenced it, we approve of it, by sym-pathy with the approbation of this supposedequitable judge. If otherwise, we enter into hisdisapprobation, and condemn it.”

In more contemporary terms, psychologistsand sociologists describe people’s behavior asbeing influenced by a strong need to maintainconformity between one’s actions, or even feel-ings, and certain values, long-term goals, oridentities they seek to uphold.2 Recent studiesconfirm the importance of such self-image con-cerns in explaining prosocial behavior in anon-ymous settings.3 A very telling experiment byJason Dana et al. (2003) thus shows that whenpeople are given the opportunity to remain ig-norant of how their choices affect others, or oftheir precise role in the outcome (as with firingsquads, which always have one blank bullet),many “altruists” choose not to know and revertto selfish choices.4

To examine this broad array of issues, wedevelop a theory of prosocial behavior thatcombines heterogeneity in individuals’ degreesof altruism and greed with a concern for social

1 See, e.g., the studies reported in Amihai Glazer and KaiA. Konrad (1996, p. 1021). Note that anonymous contribu-tions have the same tax-deduction benefits as nonanony-mous ones.

2 Thus Batson (1998) writes, “The ability to pat oneselfon the back and feeling good about being a kind, caringperson, can be a powerful incentive to help”; he also dis-cusses the anticipation of guilt. Daniel Kahneman and JackKnetsch (1992) find that subjects’ stated willingness to payfor alternative public goods is well predicted by indepen-dent assessments of the associated “moral satisfaction.”Michele Lamont (2000) documents the importance attachedby her interviewees to the presence or absence of the “car-ing self” not just in others, but also in themselves.

3 For instance, in an anonymous transportation-relatedsurvey of about 1,300 individuals, Olof Johansson-Stenmanand Peter Martinsson (2006) find that people who are askedwhich attributes in a car are most important to them sys-tematically put environmental performance near the top andsocial status near the bottom; but when asked about the truepreferences of their neighbors or average compatriots, theygive dramatically reversed rankings. Interviews with cardealers show intermediate results.

4 For evidence of self-image management in dictatorgames, see also J. Keith Murnighan et al. (2001).

1653VOL. 96 NO. 5 BENABOU AND TIROLE: INCENTIVES AND PROSOCIAL BEHAVIOR

reputation or self-respect. The key property ofthe model is that agents’ prosocial or antisocialbehavior reflects an endogenous and unobserv-able mix of three motivations: intrinsic, extrin-sic, and reputational, which must be inferredfrom their choices and the context. We obtainfour main sets of results.

Rewards and Punishments.—The presence ofextrinsic incentives spoils the reputational valueof good deeds, creating doubt about the extentto which they were performed for the incentivesrather than for themselves. This is in line withwhat psychologists term the “overjustificationeffect” (e.g., Mark R. Lepper et al., 1973), towhich we give a formal content in terms of asignal-extraction problem.5 Rewards act like anincrease in the noise-to-signal ratio, or evenreverse the sign of the signal, and the resultingcrowding out of the reputational (or self-image)motivation to contribute can make aggregatesupply downward-sloping over a wide range,with possibly a sharp drop at zero.

Publicity, Praise, and Shame.—A greaterprominence and memorability of contributionsstrengthens the signaling motive and thus gen-erally encourages prosocial behavior. When in-dividuals are heterogeneous in their imageconcerns, however, it also acts like an increasein the noise-to-signal ratio: good actions be-come suspected of being motivated by appear-ances, which limits the effectiveness of policiesbased on “image rewards” such as public praiseand shame. The same concern can lead peopleto refrain from turning down any rewards thatare offered.

Social and Personal Norms.—The inferencesthat can be drawn from a person’s actions de-pend on what others choose to do, creatingpowerful spillovers that allow multiple normsof behavior to emerge as equilibria. More gen-erally, individuals’ decisions will be strategiccomplements or substitutes, as will legal andsocial sanctions, depending on whether reputa-

tional concerns are (endogenously) dominatedby the avoidance of stigma or the pursuit ofdistinction. The first case occurs when there arerelatively few types with low intrinsic altruism,and when valid excuses for not contributing aremore rare than events that make participationinevitable, or unusually easy. The second caseapplies in the opposite circumstances.

Welfare and Competition.—When setting in-centives, sponsors such as charities, nonprofitorganizations, or government agencies will ex-ploit these complementarities or substitutabili-ties, which respectively increase or decrease theelasticity of the supply curve. Because they donot internalize the reputational spillovers thatfall on nonparticipants or on those who contrib-ute through other sponsors, however, their pol-icies will generally be inefficient. Thus, even amonopoly sponsor may offer rewards and“perks” (preferred seating, meetings with fa-mous performers, valuable social networkingopportunities, naming rights to a building, sta-dium, or professorial chair, etc.) that are toogenerous from the point of view of social wel-fare, and sponsor competition may further ag-gravate this inefficiency. The socially optimalincentive scheme, by contrast, subtracts fromthe standard Pigouvian subsidy for public goodsprovision a “tax” on reputation-seeking, which,per se, is socially wasteful. In the market forprosocial contributions, finally, a form ofholier-than-thou competition can also leadsponsors to offer agents opportunities for repu-tationally motivated sacrifices that will againreduce social welfare, without any increase inthe supply of public goods.

While a number of related themes have beenexamined in the literature, none of the existingmodels provides a unified account of this broadrange of phenomena. Standard models of publicgoods provision or altruistic behavior, whetherbased on a concern for others’ welfare, a purejoy of giving, or reciprocity, are not consistentwith a (locally) downward-sloping response ofprosocial behavior to incentives, nor with peo-ple choosing not to know how their actions willaffect others and reverting to selfish behaviorwhen such ignorance is feasible. Models of giv-ing as a signal of wealth explain monetary do-nations but not in-kind prosocial acts such as

5 It is also consistent with the informal explanation pro-vided by Frey and Jegen (2001): “An intrinsically motivatedperson is deprived of the chance of displaying his or herown interest and involvement in an activity when someoneelse offers a reward, or orders him/her to do it.”

1654 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

volunteering, helping, giving blood, etc. (theseshould instead be avoided, as they signal a lowopportunity cost of time), the greater admirationreserved for anonymous contributions, or peo-ple’s choosing to be modest about their gooddeeds. Models that postulate a reduced-formcrowding out (or in) of intrinsic motivation byincentives do not really explain its source and missits dependence on the informational environment,such as the observability of actions and rewards orthe distribution of preferences in the population.The same is true for models of social norms thatassume complementarities in payoffs.

The papers most closely related to the presentone take a signaling approach to social interac-tions, although none share with it the structureof multidimensional uncertainty that is essentialto generating overjustification effects and netcrowding out. In Benabou and Tirole (2003), apotential conflict between extrinsic and intrinsicmotivation arises from the fact that giving anagent high-powered incentives may convey badnews about the task or his ability. The idea thatthe principal has private information aboutthese variables applies well to child rearing,education, and empowerment versus monitoringof employees, but not to activities such as con-tributing to a charitable cause, donating blood,voting, etc.6 In B. Douglas Bernheim (1994),individuals take actions designed to signal thattheir tastes lie close to “the mainstream,” lead-ing to conformity in behavior and multiple so-cial norms. When reputation bears on prosocialorientation, however, what is valuable is not toresemble the average but to appear as altruisticas possible. Such is the case in Giacomo G.Corneo’s (1997) signaling model of union

membership, with which our analysis of socialnorms shares some important insights. On theother hand, Corneo’s model does not give riseto crowding out, and while Bernheim does notconsider the effects of incentives, the similarlyunidimensional structure of his model will alsolead to a standard upward-sloping response.Jerker Denrell (1998) shows how the presenceof monetary or side benefits in some activity candestroy the separating equilibrium that wouldotherwise obtain. While this again does not leadto crowding out, a principal may obtain higherprofits with a zero reward than with a positiveone.7 Closest to our paper is that of Paul Sea-bright (2002), where individuals derive fromparticipating in a “civic activity” both a directbenefit that depends on their private type and areputation that will make them more desirablepartners in a later matching market. Under asorting condition that makes high types caremore about reputation, a “payment discontinu-ity” arises at zero, in that total participation canbe greater when no reward is offered than witha small positive one.8

The paper is organized as follows. Section Ipresents the model and an intuitive illustrationof the image-spoiling effect of rewards. SectionII formally demonstrates the crowding-out phe-nomenon, as well as a related form of the over-justification effect. Section III deals with social

6 The informed-principal approach to rewards remainsapplicable when agents know their own type, however, ifthey care about the principal’s perception of it. Thus, inFlorian Herold (2004), strong incentives can signal that theprincipal does not trust the agent, which is bad news forother aspects of the (multitask) relationship. In Tore Elling-sen and Magnus Johannesson (2006), agents derive utilityfrom the principal’s ultimate view of their ability or taste forthe activity. Depending on the curvature of this “esteem”function, strong incentives, which signal unfavorable priors,may then damage or enhance the expected return to effort.Whereas all the papers above focus on the ex ante choice ofincentives, Anton Suvorov and Jeroen van de Ven (2005)show that ex post (discretionary) bonuses may serve toenhance motivation by functioning as a credible feedbackmechanism.

7 Funk (2005) shows how a model of voting that incor-porates a motive to signal oneself as a “good citizen” canvery plausibly account for her empirical findings concerningthe Swiss policy change discussed earlier. The phenomenonthus captured is also not an instance of crowding out, asboth the cost of voting and its visibility are changed simul-taneously and it is the latter that causes participation to fall.

8 Our paper naturally also ties in to the large literature ongifts and donations, such as James Andreoni (1993, 2006),Glazer and Konrad (1996), William Harbaugh (1998), An-drea Buraschi and Francesca Cornelli (2002), and Prender-gast and Lars A. Stole (2001). Other related papers includeRonit Bodner and Drazen Prelec (2003) and Benabou andTirole (2004b) on self-signaling; Akerlof and Rachel E.Kranton (2000) on identity; Kjell Arne Brekke et al. (2003)on moral motivation; Maarten Janssen and Ewa Mendys-Kamphorst (2004) on rewards and the evolution of socialnorms; and Wolfgang Pesendorfer (1995) and Laurie SimonBagwell and Bernheim (1996) on ostentatious consump-tions as signaling devices. Our work is also technicallyrelated to a recent literature on signals that convey divergingnews about different underlying characteristics (AloisioPessoa de Araujo et al., 2004; Philipp Sadowski, 2004;David Austen-Smith and Roland G. Fryer, 2005).


norms and more generally the strategic comple-mentarity or substitutability of individual deci-sions. Section IV examines the possibility foragents to turn down rewards. Section V exam-ines the setting of incentives by public or pri-vate sponsors and the welfare effects ofcompetition. Section VI concludes. All proofsare gathered in the Appendix.

I. The Model

A. Preferences and Information

We study the behavior of agents who choosethe extent of their participation in some pro-social activity: contributing to a public goodor worthy cause, engaging in a friendly ac-tion, refraining from imposing negative exter-nalities on others, etc. Each selects aparticipation level a from some choice setA � � that can be discrete (voting, blooddonation) or continuous (time or money vol-unteered, fuel efficiency of car purchased).Choosing a entails a utility cost C(a) andyields a monetary or other material rewardya. The incentive rate y � 0 may reflect aproportional subsidy or tax faced by agents inthis economy, or the fact that participationrequires a monetary contribution; note alsothat a subsidy to a is equivalent to a tax orfine on �a. The incentive rate is set by aprincipal or “sponsor” and, for now, individ-uals take it as given.

Denoting by va and vy an agent’s intrinsicvaluations for contributing to the social good(discussed further below) and for money (con-sumption of market goods), participation atlevel a yields a direct benefit

(1) �va � vy y�a � C�a�.

Each individual’s preference type or “identity”v � (va, vy) � �2 is drawn independently froma continuous distribution with density f(v) andmean (v�a, v�y). Its realization is private informa-tion, known to the agent when he acts but notobservable by others.

Social Signaling.—In addition to these directpayoffs, decisions carry reputational costs andbenefits, reflecting the judgements and reactions

of others—family, friends, colleagues, employ-ers. The value of reputation may be instrumen-tal (making the agent a more attractive match,as in Denrell, 1998, Herbert Gintis et al., 2001,or Seabright, 2004) or purely affective (socialesteem or shame as a hedonic good). For sim-plicity, we assume that it depends linearly onobservers’ posterior expectations of the agent’stype v, so that the reputational payoff fromchoosing a, given an incentive rate y, is

(2)

R�a, y� � x��a E�va�a, y� � �y E�vy�a, y��,

with �a � 0 and �y � 0.9

The signs of �a and �y reflect the idea that peoplewould like to appear as prosocial (public-spirited)and disinterested (not greedy), while the factorx � 0 measures the visibility or salience of theiractions: probability that it will be observed byothers, number of people who will hear about it,length of time during which the record will bekept, etc. Defining �a � x�a and �y � x�y, anagent with preferences v � (va, vy) and reputa-tional concerns � � (�a, �y) thus solves

(3) maxa�A

�va � vyy�a � C�a�

� �a E�va�a, y� � �y E�vy�a, y�}.

In the basic version of the model, � is taken tobe common to all agents and thus public knowl-edge. In the full version, we also allow for unob-served heterogeneity in image-consciousness,with � distributed independently of v. Finally,while we shall generally cast the analysis interms of effortful or time-consuming prosocialactions such as volunteering and voting, it isequally applicable to monetary (e.g., charitable)

9 This payoff is defined net of the constant (1 �x)(�av�a � �yv�y), which corresponds to the case where aremains unobserved. The restriction to payoffs that arelinear in the posterior distribution over v is without muchloss of generality, since introducing (monotonic) nonlinearpayoffs of the form E[�(va)�a, y] would be essentiallyequivalent to redefining the density of va (see also footnote34). The more restrictive assumption, which we make fortractability, is that the coefficients in (2) are independent ofthe agent’s type v.


donations, with only a simple relabelling of thevariables a (dollars contributed) and y (goods orservices used as rewards).10

Self-Signaling, Identity, and Moral Senti-ments.—The model admits an important rein-terpretation in terms of self-image. Whenmaking a decision affecting others’ welfare, anindividual will often engage in a self-assessment:“How important is it for me to contribute to thepublic good? How much do I care aboutmoney? What are my real values?” Later on,however, this information may no longer beperfectly “accessible” in memory—in fact,there will often be strong incentives to recall itin a self-serving way. Actions, by contrast, aremuch easier to remember than their underlyingmotives, making it rational to define oneselfpartly through one’s past choices: “I am thekind of person who behaves in this way.” Sup-pose, therefore, that the exact feelings or signalsunderlying the participation decision becomeinaccessible with some probability proportionalto x and that, later on, the agent cares about“what kind of a person he is.”11 If, for sim-

plicity, this utility from self-image is linear inbeliefs, with weights �a and ��y on perceivedsocial orientation and greediness, the model isformally equivalent to the social-signaling one.

Relation to Altruism and Public Goods.—Anagent’s intrinsic motivation to behave prosocially,va, can stem from two sources. First, he may careabout the overall level of a public good to whichhis action contributes, such air quality. Let thiscomponent of utility be wa(na�/n�), where a� repre-sents the average contribution, n the size of thegroup, and � � 0 the degree of congestion; wathen measures the intensity of the individual’s“pure” altruism.12 Second, he may experience a“joy of giving” ua (independent of social- or self-esteem concerns) that makes him value his owncontribution to na� more than someone else’s.13

Combining these “pure” and “impure” forms ofaltruism (Andreoni, 1988) yields va ua � wa /n�;in large groups with � � 0, the second termbecomes vanishingly small. The simplest interpre-tation of our model is thus one with a continuumof agents (so va ua) in which the averagecontribution generates a public good (� 1),which an individual values as waa�. The modelapplies equally well, however, to finite groups ofany size n and value �. All that matters is thatthere be heterogeneity in the intrinsic propensityto contribute or reciprocate, va, no matter itssource, and that agents value being perceived, orperceiving themselves, as having a high va. Thissocial (self) esteem benefit, �aE(va�a, y), is per-haps what corresponds best to the idea of a “warmglow” of giving: gaining social approval, feeling

10 Let a be the dollar amount contributed by an individ-ual with a known utility for income, the concavity of whichis represented by �C(a). Each dollar generates one unit ofpublic good and entitles the contributor to y units of gifts,perks, and privileges (meeting with performers, gala events,networking, etc.), a “currency” for which he has utility vy.The case where the sponsor offers matching funds instead ofperks, i.e., rewards contributors in the same currency, cor-responds to vy � va and �y � 0. In the discrete specificationused in Sections IIIA to V (a � {0, 1} and vy � 1), it canalso be represented as the sponsor’s reducing an agent’smonetary cost of providing a unit of public good from c �C(1) to c � y.

11 This may reflect pure feelings of pride or guilt fromseeing oneself as generous or selfish (e.g., Akerlof andDickens, 1982; Botond Koszegi, 2000), an instrumentalvalue of providing the motivation to undertake and perse-vere in long-term relationships (e.g., Juan D. Carrillo andThomas Mariotti, 2000; Benabou and Tirole, 2002), or both.The idea that individuals take their actions as diagnostic oftheir preferences originated in psychology with Daryl J.Bem (1972) and relates closely to cognitive dissonancetheory (Leon Festinger and James Carlsmith, 1959). Whilepsychologists would generally view people as unable todiscern precisely their own motives even at the time they act(responding only to the overall mix), this is formally equiv-alent to our specification in which preference states becomeinaccessible after some (possibly very brief) period. The linkbetween imperfect recall and intertemporal self-signaling isanalyzed in Benabou and Tirole (2004b), while Bodner and

Prelec (2003) examine contemporaneous self-signaling in adual-self model.

12 Since we abstract from decreasing marginal utilityover the total supply of public goods, it is worth noting thatthe standard substitution effect that it would generate (“ifothers give more, I should give less”) can never causeequilibrium aggregate supply to be downward sloping. Notealso that, at the cost of some additional complexity, onecould make agents care about social welfare (which is thendefined as a fixed point) rather than about the level of thepublic good per se.

13 Such would be the effect of feelings of empathy (em-phasized by Batson, 1998) or reciprocity. Equivalently, themarginal cost of participation may include an individualcomponent equal to �ua. The term ua could also arise fromagents’ following the Kantian imperative to evaluate theiractions as if they would lead everyone to make those samechoices (Brekke et al., 2003).


good about oneself, etc. The important point,however, is the need to go beyond the standarddichotomy between “pure” and “impure” altruismand distinguish, within the latter, between fixedpreferences (ua) and motives that relate to what aperson’s behavior says about him or her, whichwill depend on the informational and economiccontext, including what others are doing. Finally,note that the action a chosen by agents and givingrise to reputation could be their reaction to some-one else’s behavior, such as cooperation or defec-tion. The model is thus applicable to reciprocity aswell as to unconditional prosocial behavior.

We now turn to the terms in (3) relating tomaterial compensation. That in vyy requires noexplanation, except to note that if the individualbelieves that his receiving y reduces the re-sources available to the sponsor for supportingother activities he cares about, it will be atten-uated by an “eviction effect.”14 Consider nextthe potential negative reputation attached to“greed” or money-orientation, ��yE(vy�a, y).Note first that all the paper’s results but one(Proposition 3) obtain with �y � 0 just as well.It is, nonetheless, natural to allow for such aneffect: “greedy” is no compliment, and indeedsomeone who has a high valuation for moneyrelative to effort and/or public goods is not avery attractive partner in friendship, marriage,hiring to a position of responsibility, electing tooffice, or other situations where it is difficult toalways monitor behavior or write complete con-tracts. Demonstrating a low marginal utility formoney vy can also be valuable because it signalshigh wealth, a motive that figures prominentlyin the literatures on charitable contributions andon conspicuous consumption (e.g., Glazer andKonrad, 1996; Bagwell and Bernheim, 1996).

B. The Image-Spoiling Effect of Rewards:Basic Insights

We begin with an intuitive presentation ofsome key mechanisms. Consider the first-ordercondition for an agent’s choice of a, assuming awell-behaved decision problem over a continu-ous choice set. By (3), an individual with type(v, �) who faces a price y equates

(4) C��a� � va � vy y � r�a, y; ��,

where the last term is his (marginal) reputa-tional return from contributing at level a:

(5) r�a, y; �� a

E�va�a, y�

a� �y

E�vy�a, y�

a.

Three important points are apparent from (4).First, observing someone’s choice of a reveals thesum of his three motivations to contribute (at themargin): intrinsic, extrinsic, and reputational. Ingeneral, all three vary across individuals, so thatlearning about va or vy corresponds to a signal-extraction problem. Second, a higher incentiverate y reduces the informativeness of actions aboutva, while increasing it about vy. Third, heteroge-neity in agents’ image concerns � represents anadditional source of noise which makes inferencesabout both va and vy less reliable, and which isamplified when actions become more visible(higher x).

To gain further insight into the impact of incen-tives on inferences and behavior, let us now focuson the benchmark case where va and vy are inde-pendent random variables, while �a and �y arefixed and will be omitted from the notation. Fig-ure 1 then shows, for any a � 0, how the set ofagents who contribute at least a varies with thereward y. This group, which we shall term “highcontributors,” comprises all agents with

(6) va � vy y � C��a� � r�a, y�,

so its boundary is a straight line corresponding to(4), along which agents choose exactly a. Thesame condition applies when the participation de-cision is discrete, a � {0, 1}, as will be the case inthe second half of the paper, provided we denoteC�(1) � C(1) � C(0) and r(1, y) � R(1, y) � R(0,y). Along the boundary, agents are now indifferentbetween participating and abstaining.

14 In experiments on charitable giving (e.g., Gneezy and Rus-tichini, 2000b), it is typically emphasized to subjects that anyrewards will come from an entirely separate research budgetand therefore do not reduce the amount actually donated. In thereal world, the presence and magnitude of an eviction effectwill depend on individuals’ beliefs about the level at which thebudget constraint binds and how they value the alternative usesof funds. Suppose, for instance, that a charity has a fixedbudget and will use any funds left over to hire “professionals”who produce units of a per dollar, or some other public goodof equivalent value. An individual’s valuation of a reward y forhis contribution will now be (vy � wa/n�)y. This simplyamounts to a redefinition of vy, in a way that contributes tomaking it negatively correlated with va.


When no reward is offered, y 0, the sepa-rating locus is vertical: an agent’s contributionreveals nothing about his vy, but is very infor-mative about his va. In the continuous case,prosocial orientation is learned perfectly; in thediscrete case one learns whether it is above orbelow a known cutoff.

When a reward y � 0 is introduced, the slopeof the separating locus becomes �1/y 0. Ifwe ignore, in a first step, any changes in theinferences embodied in the intercept, the origi-nal boundary simply pivots to the left, as shownin Figure 1 (everything works symmetrically fora fine or penalty, y 0). The set of agentscontributing at least a thus expands, as types inthe hatched area (A � B) are drawn in. Sincethis occurs at every level of a, the distribution ofcontributions shifts up (stochastically), resultingin a higher total supply; this is the standardeffect of incentives. In equilibrium, however,there are two reputational effects:

● The new members of the high-contributors’club have lower va’s than the old ones, sothey drag down the group’s reputation forprosocial orientation. The reputation of thelow-contributors’ group also declines, how-ever, so in the discrete-choice case, the neteffect on the reputational incentive to partic-

ipate can clearly go either way. Similarly, inthe continuous case, the reputation E(va�a, y)attached to contributing exactly a declines (asthat locus pivots to the left), but so does thereputation attached to contributing exactly a� a � da, where da is small; the effect on themarginal return E(va�a, y)/a is thus a prioriambiguous.

● The new high contributors are “greedy” types(have a vy above the mean), whereas those whostill contribute below a after the reward is in-troduced reveal that they care less about moneythan average. This unambiguously reduces thereputational incentive to participate, as is clearin the discrete case. In the continuous case, thisfollows from the fact that, after the rotation, thelocus for contributing at a � da lies below thatfor contributing a.15

If the overall impact of these changes ininferences is negative, r(a, y) r(a, 0), asdrawn in Figure 1, the reward attracts somenew participants (more greedy agents in area B)to contributing a or more, but repels some ex-isting ones (more public-spirited agents in area

15 This is due to the fact that C�(a) � r(a, y) is increasingin a, by the second-order condition for (3).

FIGURE 1. THE EFFECTS OF REWARDS ON THE POOL OF PARTICIPANTS


C).16 Overall, the number of agents who con-tribute at least a may increase or decrease, de-pending on the weights given to B and C by thedistribution f(v). If a net decrease occurs atevery a, the distribution of contributions shiftsdown (stochastically) and total supply actuallydeclines when a reward y � 0 is introduced,starting from a no-reward situation.

II. The Overjustification Effect andCrowding Out

We now turn to the formal analysis, establish-ing three main results. First, we show how the“overjustification effect” discussed by psycholo-gists can be understood as a signal-extractionproblem in which rewards amplify the noise, lead-ing observers (or a retrospecting individual) toattribute less of a role to intrinsic motivation inexplaining variations in behavior. We then iden-tify the conditions under which monetary incen-tives crowd out reputational motivation—or,equivalently, legal sanctions undermine socialones—resulting in a supply curve that is down-ward-sloping over a potentially wide range, orexhibits a sharp drop at zero. Finally, we assessthe effectiveness of nonmaterial rewards and pun-ishments such as public praise and shame, show-ing in particular that it is also limited by a form ofoverjustification effect.

We use here a specification of the modelthat builds on and extends the familiar normal-learning setup. Let actions vary continuouslyover A �, with cost C(a) ka2/2.17 Agents’valuations v � (va, vy) are distributed in thepopulation as

(7) �va

vy� � N �v� a

v� y, � �a

2 �ay

�ay �y2 �� ,

v� a 0, v� y � 0,

and at first we continue to focus on the case

where everyone has the same reputational con-cerns, � � (�� a, �� y). We then extend the anal-ysis to the case where � is also normallydistributed across individuals.18

A. Material Rewards

With fixed �’s, the reputational return (5) isconstant across agents and equal to

(8) r��a, y� � �� a

E�va�a, y�

a

� �� y

E�vy�a, y�

a.

Thus, by (4), an agent’s choice of a reveals hisva � yvy, equal to C�(a) � r�(a, y). Standardresults for normal random variables then yield

(9) E�va�a, y� � v� a � ��y�

� �ka � v�a � v�yy � r��a, y��,

(10) E�vy�a, y� � v� y � ��y�

� �ka � v�a � v�yy � r��a, y��,

where

(11) ��y� ��a

2 � y�ay

�a2 � 2y�ay � y2�y

2

and

y��y� � 1 � ��y�.

Intuitively, the posterior assessment of anagent’s intrinsic motivation, E(va�a, y), is aweighted average of the prior v�a and of themarginal cost of his observed contribution, netof the average extrinsic and reputational incen-tives to contribute at that level.

16 This matches William Upton’s (1973) findings thatwhile offering a monetary reward for giving blood predict-ably brought in new donors, it reduced donations by thosewho had regularly been giving for free.

17 The case of a general convex function C(a) is treatedin Benabou and Tirole (2004a). Both here and there, wefocus attention on equilibria in which the reputation vector,E(v�a, y), is differentiable in a.

18 As is often the case, normality yields great tractability atthe cost of allowing certain variables to take implausible neg-ative values. By choosing the relevant means large enough,however, one can make the probability of such realizationsarbitrarily small; but (7) and (17) below should really beinterpreted as local approximations, consistent with the linear-ity of preferences assumed throughout the paper.


Finally, substituting (8) into (9) and (10) showsthat an equilibrium corresponds to a pair of func-tions E(va�a, y) and E(vy�a, y) which solve asystem of two linear differential equations.

PROPOSITION 1: Let all agents have thesame image concern (�� a, �� y). There is a unique(differentiable-reputation) equilibrium, in whichan agent with preferences (va, vy) contributes atthe level

(12) a �va � vy y

k� �� a�� y� � �� y�� y�,

where �( y) and �( y) are defined by (11). Thereputational returns are E(va�a, y)/a �( y)k and E(vy�a, y)/a �( y)k, resultingin a net value r�( y) k(�� a�( y) � �� y�( y)),independent of a.

The effects of extrinsic incentives on infer-ences and behaviors can now be analyzed.While a higher y increases agents’ direct payofffrom contributing, va � vyy, it also tends toreduce the associated signaling value along bothdimensions. In the benchmark case of no corre-lation (�ay 0), for instance,

(13) ��y� �1

1 � y2�y2 /�a

2

and

��y� �y�y

2 /�a2

1 � y2�y2 /�a

2 ,

so a higher y acts much like an increase in thenoise-to-signal ratio � � �y /�a, leading observ-ers who parse out the agent’s motives todecrease the weight attributed to social orienta-tion, �(y), and increase its counterpart forgreediness, �(y).19 When �ay � 0, a positivecorrelation tends to amplify the decline in �(y);a negative one works to weaken or even reverseit. Indeed, the more va and vy tend to move

together, the less observing a high contributiona, or equivalently a high va � vyy, representsgood news about the agent’s intrinsic valuationva; and the larger is y, the stronger is this “dis-counting” effect.20

Summing (12) over agents yields the (percapita) aggregate supply of the public gooda�(y), whose slope,

(14) a� ��y� �v� y

k� �� a��y� � �� y��y�,

reflects both the standard effect of incentivesand the crowding out or in of reputational mo-tivation they induce. Since the general expres-sion (provided in the Appendix) is a bitcomplicated, we focus here on two benchmarkcases that make clear the main factors at play.The first one is that of independent values, forwhich we show that as long as the reputationalconcern over either prosocial orientation ormoney-orientation is above some minimumlevel, there exists a range over which incentivesbackfire.

PROPOSITION 2 (Overjustification and crowd-ing out): Let �ay 0 and define � � �y/�a.Incentives are counterproductive, a��(y) 0, atall levels such that

(15)v�y

k� �� a �

2y�2

�1 � y2�2�2 � �� y ��2�1 � y2�2�

�1 � y2�2�2 .

Consequently, for all �� a above some threshold�*a � 0, there exists a range [y1, y2] such thata�(y) is decreasing on [y1, y2] and increasingeverywhere else on �. If �� y v�y /k�2, then �*a �0 and 0 y1 y2; as �� a increases, y1 rises andy2 falls, so [y1, y2] widens. If �� y � v�y /k�2, then�*a 0 and y1 0 y2; as �� a increases bothy1 and y2 rise and, for �� a large enough, [y1, y2]again widens.

19 More precisely, y�(y) 1 � �(y) rises with y every-where, but the same is true of �(y) only for �y� � 1/�.

20 Thus, as the correlation between va and vy rises from�1 to 0 to 1, the function �(y) pivots downward over therange 0 y 1/�, from 1/(1 � �y) to 1/(1 � �2y2) and thento 1/(1 � �y). The effect of �ay on the slope ��(y) is morecomplex, as it depends on �ay

2 ; see (A2) and (A3) in theAppendix.


As illustrated in Figure 2A, crowding out canoccur over a fairly wide range, making all but verylarge rewards inferior to none.21 Most interestingare the comparative statics on �� a and the cross-effects between �� a and y, two predictions forwhich a recent experiment provides a strikingmatch. Studying the willingness of 238 subjects tojoin a blood-donor program, Carl Mellstrom andJohannesson (2005) found that: (a) absent mone-tary rewards, women contributed significantly

more than men: 52 percent versus 28 percent;(b) introducing a monetary payment (of about $7)caused a moderate, statistically insignificant, in-crease in men’s participation rate (to 37 percent),but led to a dramatic collapse in that of women,which fell to 30 percent; (c) when subjects had theopportunity to turn over their fee to a (cancer-related) charity, men’s participation remained es-sentially unchanged (33 percent), but that ofwomen went right back to 53 percent. If onegrants that, for easily understood reasons, it ismore important for women than for men to beperceived (and think of themselves) as caring andcompassionate human beings—that is, if theyhave a higher �a—then Proposition 2 (or Figure2A) predicts both that they will contribute more inthe absence of rewards and that they will be theones most likely to respond negatively to mone-tary incentives.22 By the same logic, they will alsorespond the most to the option of turning down orgiving away the reward, which restores to theblood donation its original, unsullied meaning.23

The second case we highlight is that of “smallrewards,” which is interesting for two reasons. First,some studies find crowding out (a�(y) decreasing) tooccur mostly at relatively low levels, and it is some-times even suggested that the main effect is a dis-continuity at zero in subjects’ response to incentives(Gneezy and Rustichini, 2000b; Gneezy, 2003). Isthere something qualitatively different between“unrewarded” and “rewarded” activities that couldcause rational agents to behave in this way? Weshow that there is, and explain when it will matter.The second reason why “small rewards” are ofinterest is that in real-world situations where timehas an opportunity cost, they will actually corre-spond to substantial values of y.

PROPOSITION 3 (Small net incentives andsignal-reversal): (1) Small rewards or punish-ments are counterproductive, a��(0) 0, whenever

(16)v� y

k� �� a��ay

�a2 � � �� y��y

2 � 2�ay2 /�a

2

�a2 � .

21 The values used in Figure 2A are k 1, v�a 4, v�y 3, �� y 0, � 0.2, and �� a � {0, 20, 25, 30, 36, 44}. InFigure 2B, they are k 1, v�a 3, v�y 1, �� a �� y 1, and� � {0, 1, 2, 3, 5}.

22 By contrast, the experimental results described abovecannot be explained by what would have been the standardinterpretation of condition (a) alone, namely that womenare, on average, more prosocial than men (have a higher v�a).

23 This case is analyzed, in a simpler version of themodel, in Section IV, where we show that such options or“menus” may not always be so effective.

A. Varying �� a (with �� y 0). The straight linecorresponds to �� a 0 (no reputation concern).

B. Varying � �y /�a. The lower straight linecorresponds to �� a �� y 0 (no reputation concern),the upper one to � 0 (standard one-dimensionalsignaling model).

FIGURE 2


(2) Let �� y � 0 and assume that va and vy areuncorrelated, or more generally not toocorrelated. Then, as �a /�y becomes small,the slope of the supply function at y 0tends to ��.

(3) Suppose that participation entails a unitopportunity cost with monetary value y.Then a��( y) 0 and a��( y)3 �� under theconditions stated in (1) and (2) above, re-spectively.

The first term on the right-hand side of (16)reflects the intuition given earlier about the roleof correlation in generating crowding out—orcrowding in. Most important is the second term,whose dependence on the noise-to-signal ratiois illustrated in Figure 2B: letting �ay 0, forinstance, shows that a��(0) v�y /k � �� y(�y /�a)2.Thus, in situations where there is much moreuncertainty (hence more to learn) about individ-uals’ desire for money than about their motiva-tion for the specific task at hand, even a minimalconcern about appearing greedy (a small �� y �0) is sufficient to cause a sharply negative re-sponse to small incentives and, in the limit, adownward discontinuity in the supply response.This result, moreover, applies whether or notthe task has any prosocial dimension (�� a mayequal zero), thus also explaining why adverseeffects of small rewards have been found bothin experiments involving private, puzzle-solvingtasks and others involving public-goods provi-sion. The intuition for why “zero is special” isthat, at that point, participation switches frombeing an “unprofitable” to a “profitable” activityand thus comes to be interpreted as a signal ofgreed rather than disinterestedness. This signal-reversal effect, operating specifically around azero net reward, creates an additional source ofcrowding out on top of the general signal-jamming effect (decrease in �( y)) that wasshown to operate at all levels of y.24

If the empirical validity of this signal reversalwere restricted to very small prizes and fines, itwould be of somewhat limited interest. Thethird result shows, however, that the relevant“tipping point” is not zero (except in laboratoryexperiments, where subjects, once there, haveno profitable alternative uses of their time) butagents’ monetary value of time, which can bequite substantial. This also suggests that exist-ing experiments may not have been focusing onthe most relevant scale of costs and benefits,and that future empirical work should involvesituations in which opportunity costs are non-trivial and vary across subjects.

B. Image Rewards

Public authorities and private sponsors aimingto foster prosocial behavior make heavy use ofboth public displays and private mementos con-veying honor or shame. Nations award medalsand honorific titles, charitable organizations senddonors pictures of “their” sponsored child, non-profit organizations give bumper stickers and T-shirts with logos, and universities award honorary“degrees” to scholars.25 Conversely, the ancientpractice of the pillory has been updated in theform of televised arrests, posting on the Internetthe names of parents who are delinquent on childsupport and those of sexual offenders, and pub-lishing in local newspapers the license plate num-bers of cars photographed in areas known for drugtrafficking or prostitution.26

24 When the two effects are combined, it is easy to getsupply curves that have a sharp local minimum at y 0, sothat neither offering rewards (up to a point) nor requiringsacrifices raises supply. Note also that whereas the signal-reversal effect (lim�a/�y30[a��(0)] ��) is a robust andeconomically intuitive phenomenon, the fact that the am-plitude �a�(y)� near zero also becomes unbounded in the limitis only an artefact of the linear-quadratic specification. InBenabou and Tirole (2004a), we thus obtain a similar dis-

continuity in a�(y) at y 0 with bounded actions (a � {0, 1})and �a /�y 0.

25 Our previous results may also help one understand cer-tain common features of the items that charities, public radio ortelevision stations, etc., offer in their mass fundraising cam-paigns. The relevant interpretation of the model here is that inwhich a is a monetary donation and y a reward rate in terms of“thank-you gifts” (see footnote 10). Equation (13) then showsthat in order to minimize the image-spoiling effect and maxi-mize contributions, the items should not only be cheap com-pared to the donation (low y) but also have little variance in theprivate value that individuals attach to them (low �y

2); hencethe offering of standardized goods with commercially availablesubstitutes, such as mugs, umbrellas, etc., rather than originalor personalized ones. The only dimension in which the itemsare unique is the logo they bear, which allows the contributorto “automatically” display a token of his generosity by usingthem (relatively high x).

26 Peer groups also play an important role by creating arehearsal mechanism: if acquaintances all contribute to a cause,


Formally, greater publicity or prominencecorresponds to a homothetic increase in (�a,�y). Our model then confirms the intuitionsabove, but also delivers important caveats. Inparticular, when agents are heterogeneous intheir reputational concerns, giving greater scru-tiny to their behavior may not work that well, asgood actions come to be suspected of beingimage-motivated. To analyze these issues, wenow allow agents’ image concerns, like theirvaluations, to be normally distributed:

(17) ��a

�y� � N �� a

�� y, � �a

2 �ay

�ay �y2 �� ,

�� a � 0, �� y � 0,

with v and � independent. In the first-ordercondition (4), the reputational return r(a, y; �)is now also normal and independent of v (con-ditionally on a), with mean r� (a, y) given by (8)and variance

(18) ��a, y�2 � �E(va�a, y)

a�

E(vy�a, y)

a �

� ��a2 �ay

�ay �y2� � �

E(va�a, y)

a

�E(vy�a, y)

a.

The signal-extraction formulas (9) and (10) thusremain unchanged, except that the updating co-efficients �(y) and �(y) are respectively re-placed by

(19) ��a, y� ��a

2 � y�ay

�a2 � 2y�ay � y2�y

2 � ��a, y�2

and

��a, y� �y�y

2 � �ay

�a2 � 2y�ay � y2�y

2 � ��a, y�2 .

An equilibrium then corresponds again to a pairof functions E(va�a, y) and E(vy�a, y) whichsolve the differential equations (9) and (10), butthis system is now nonlinear, due to the term�(a, y)2 in � and �. We are able to solve it forthe intuitive and important class of solutionswhere � is independent of a, so that reputationsremain linear in a. We cannot a priori excludethe existence of other, nonlinear, equilibria.

PROPOSITION 4: (1) A linear-reputation equi-librium corresponds to a fixed-point �(y), so-lution to:

(20) �� y�2/k2 � �a2�� y�2

� 2�ay�(y)�(y)��y2�(y)2,

where �( y) and �( y) are given by (19) with�(a, y) � �( y). The optimal action chosenby an agent with type (v, �) is then

(21) a �va � vy y

k� �a�� y� � �y�� y�

and the marginal reputations are E(va�a,y)/a �( y)k and E(vy�a, y)/a �( y)k,with a net value of r( y; �) (�a�( y) ��y�( y))k for the agent.

(2) There always exists such an equilibrium,and if �ay 0 it is unique (in the linear-reputation class).

A greater variability of image motives,�(y)2 Var(r(y; �)), makes individuals’ be-havior a more noisy measure of their true un-derlying values (va, vy), reducing both �(y) and�(y). This variance is itself endogenous, how-ever, as agents’ reputational calculus takes intoaccount how their collective behavior affectsobservers’ signal-extraction problem. This is re-flected in the fixed-point nature of equation(20).27

Proposition 4 allows us to demonstrate howincreased publicity gives rise to an offsetting

one is constantly reminded of one’s generosity, or lack thereof.People indeed volunteer more help in response to a directrequest to do so, especially when it comes from a friend, acolleague, or family (Freeman, 1997), whose opinion of themthey naturally care about more than that of strangers.

27 When �ay � 0, there could be multiple equilibria, withdifferent degrees of informativeness. Since the generaltheme of multiplicity is investigated in Section IIIA, we donot pursue it here.


overjustification effect. Let all the reputationalweights � (�a, �y) be scaled up by someprominence or memorability factor, x; the ma-terial incentive y remains constant. Aggregatesupply is then

(22) a� �y, x� �v� a � v� yy

k

� x�� a��y, x� � �� y��y, x��,

where the dependence on x indicates that all thecovariance terms (�a

2, �ay, �y2) in the original

equation (20), corresponding to x 1, are nowmultiplied by x2. A greater visibility of actions(and of any rewards attached to them) thus hastwo offsetting effects on the reputational incen-tive to contribute:

● A direct amplifying effect, the sign of which isthat of �a�(y, x) � �y�(y, x) for an individualand �� a�(y, x) � �� y�(y, x) on average. Forpeople who are mostly concerned about appear-ing socially minded (�a � �y), this increasesthe incentive to act in a prosocial manner,whereas for those most concerned about notappearing greedy (�y � �a), it has the reverseeffect.28

● A dampening effect, as reputation becomesless sensitive to the individual’s behavior,which observers increasingly ascribe to im-age-seeking. Formally, the “effective noise”�(y, x) increases with x (in any stable equi-librium) and �(y, x) and �(y, x) consequentlytend to decrease with it.

This tradeoff implies that giving increased pub-licity to prosocial or antisocial behavior may be ofsomewhat limited effectiveness, even when it isrelatively cheap to do. Consider, for instance, thecase where �y is known (�y 0), possibly equalto zero. As x becomes large (more generally,xk�a

2 � 1), equation (20) yields

(23) ��y, x� ��a2 � y�ay

k2�a2 � 1/3

x�2/3.

The aggregate social benefit from publicity�� a x�( y, x) thus grows only as x1/3, implyingthat it is optimal to provide only a finite levelof x even when it has a constant marginal cost,or even a marginal cost that declines slowerthan x�2/3.29 Policies by parents, teachers, gov-ernments, and other principals that rely on the“currency” of praise and shame are thus effec-tive up to a point, but eventually self-limiting.

III. Honor, Stigma, and Social Norms

The second main issue we explore is that ofsocial and personal norms. We first show howmultiple standards of “acceptable” behavior canarise from the interplay of honor and shame,then examine what characteristics of the “mar-ket,” such as the distribution of social prefer-ences, the availability of excuses, or theobservability of action and inaction, facilitate orimpede their emergence.

For the remainder of the paper we focus onthe case of a binary participation decision, A {0, 1}, in which the notions of honor and stigmaare most sharply apparent. Unless otherwisespecified, we also assume that all agents sharethe same reputational concern � � (�a, �y) andthe same valuation for money, which we nor-malize to vy � 1. Their prosocial orientation va,by contrast, is distributed on some interval [va

�,va

�].30 Indeed, whereas two-dimensional uncer-tainty is essential to the overjustification andbackfiring-incentives effects analyzed earlier, itis not needed for most of the other results. Thissimplification also removes any potential incen-tive for agents to “burn money” in order tosignal a low vy.

We again denote r(y) � R(1, y) � R(0, y) andlet c � C(1) � C(0). Thus, an agent now par-ticipates if va � c � y � r(y) � v*a(y). Todetermine this equilibrium threshold of altru-ism, let us define, for any candidate cutoff va,the conditional means in the upper and lowertails:

28 For y � 0. We are focusing this discussion, for simplic-ity, on the “natural” case where � and � are both positive,which occurs as long as �ay is not too negative; see (19).

29 On the other hand, there cannot be full crowding out,namely x�(y, x) actually decreasing with x: otherwise, by (19)and (20), �(y, x) would be increasing in x, a contradiction.

30 The results generalize to the case where va and vy areindependently distributed and reputation bears only on theformer (�y 0).


(24) M��va � � E�va�va � va �,

(25) M��va � � E�va�va � va �.

The first expression governs the “honor” con-ferred by participation, which is the differencebetween M�(va) and the unconditional meanv�a. The second one governs the “stigma” fromabstention, which is v�a � M�(va). Since bothare nondecreasing functions, the net reputa-tional gain M�(va) � M�(va) and the marginalagent’s total nonmonetary return to contributing,

(26) ��va � � va � �a �M��va � � M��va ��

� va � ��va �,

may increase or decrease with overall participa-tion, [va, va

�]. The slopes of these two functionswill play central roles in what follows.31

A. Endogenous Social Norms

What makes a given behavior socially ormorally unacceptable is often the very fact that“it is just not done,” meaning that only peoplewhose extreme types make them social outlierswould not be dissuaded by the intense shameattached to it. In other places or times, differentnorms or codes of honor prevail, and the factthat “everyone does it” allows the very samebehavior to be free of all stigma. Examples in-clude choosing surrender over death, not going tochurch, not voting, divorce, bankruptcy, unem-ployment, welfare dependency, minor tax evasion,and conspicuous modes of consumption.

We show here that such interdependenciesbetween agents’ choices arise endogenouslythrough the inferences made from observed be-haviors, creating the potential for multiplenorms of social responsibility. In particular, noassumption of complementarity in payoffs (e.g.,

between va and the average contribution a� , rep-resenting a form of “reciprocity”) or other valueof “conformity” is required to explain the com-mon finding that individuals contribute more topublic goods when they know that others arealso giving more.32

The following results, illustrated in Fig-ure 3, characterize the set of equilibria of theparticipation game and the associated supplycorrespondence.33

PROPOSITION 5: (1) When � is increasing,there is a unique equilibrium, with no partici-

31 Recall also that, in the discussion of Figure 1, it wasargued that the reputation for prosociality of contributorsmay worsen either more or less than that of noncontributorswhen the separating locus pivots to the left due to thepresence of a reward y � 0. Indeed, for any given value ofvy (over which one then integrates), these reputations re-spectively correspond to M�(v*

a � vyy) and M�(v*a � vyy),

whose difference may increase or decrease with y depend-ing on the slope of M� � M�.

32 For instance, James H. Bryan and Mary A. Test (1967)found that motorists were more likely to stop and helpsomeone with a flat tire, and walkers-by more likely to putmoney into a Salvation Army kettle, when they had ob-served someone else (a confederate) doing so a few minutesbefore. See also Jan Potters et al. (forthcoming) on charities’frequent strategy of publicly announcing “leadership” con-tributions and on the higher yields achieved when donorsact sequentially rather than simultaneously.

33 To pin down off-the-equilibrium-path beliefs whenthere is full participation, we make the standard assumptionthat the support of beliefs is weakly increasing in the levelof contribution off the equilibrium path, as is necessarily thecase on the equilibrium path: if a � a� and v�(a�) and v�(a)denote the sup and the inf of the two supports, respectively,then v�(a�) � v�(a).

A. Unique equilibrium

B. Multiple equilibria

FIGURE 3


pation (v*a va�) for y ca � �(va

�), partic-ipation increasing in y for y � (ca � �(va

�),ca � �(va

�)), and full participation (v*a va�)

for y � ca � �(va�).

(2) When � is decreasing, there are three equi-libria for all y � (ca � �(va

�), ca ��(va

�)): full participation, no participation,and an unstable interior equilibrium de-fined by �(v*a) ca � y. For y � (ca ��(va

�), ca � �(va�)), there is again a

unique, corner equilibrium.(3) When � is nonmonotonic, there exists a

range of values of y for which there are atleast two stable equilibria, of which one atleast is interior.

When va is uniformly distributed on [0, 1],for instance, �(va) va � �a /2 so the supplycurve is a standard upward-sloping one, as inFigure 3A. When va has density g(va) 2va on[0, 1], by contrast, �(va) va � (2�a /3)(1 �va)�1 is decreasing for all �a � 6, resulting inthree equilibria, as in Figure 3B. For �a � (3/2,6), � is hump-shaped, making the higher-participation equilibrium interior.

B. Strategic Complementarity andSubstitutability

The intuition for the results is that agents’actions will (endogenously) be strategic com-plements or substitutes, depending on whetherit is stigma or honor that is most responsiveto the extent of participation. This same condi-tion turns out to play a key role in other issues,such as the socially optimal level of incentives(see Section VA) and the disclosure or confi-dentiality of rewards (see Benabou and Tirole,2004a).

DEFINITION 1: Participation decisions ex-hibit strategic complementarities if ��(va) ��a(M� � M�)�(va) 0 for all va.

When �� 0, a wider participation (dva 0)worsens the pool of abstainers more than that ofcontributors, so that the stigma from abstentionv�a � M�(va) rises faster than the honor fromparticipation M�(va) � v�a fades. When ��

�1, or �� 0, the resulting net increase inreputational pressure is strong enough that themarginal agents in [v*a � dva, v*a], who initiallypreferred to abstain, now feel compelled to con-tribute. This further increases participation andconfines abstention to an even worse pool, etc.,leading to corner solutions as the only stableequilibria, as in Figure 3B. When �� (�1, 0),complementarity is weak enough that the mar-ginal agents still prefer to stay out, hence sta-bility obtains. This is a fortiori the case whenthere is substitutability, �� 0.

Equipped with this general intuition, we nowinvestigate the main factors that make strategiccomplementarity—and thus the existence ofmultiple social norms—more or less likely.

Distribution of Social Preferences.—One ex-pects that stigma considerations will be domi-nant when the population includes only a few“bad apples” with very low intrinsic values,which most agents will be eager to differentiatethemselves from. Formally, an increasing den-sity g(va) makes it more likely that M� � M�

is declining: a rise in va hardly increasesE(va�va � va) but substantially increasesE(va�va � va), since the weight reallocated atthe margin is small relative to that in the uppertail, but large relative to that in the lower tail.Conversely, honor will dominate when there areonly a few heroic or saintly types, whom themass of more ordinary individuals would like tobe identified with.34

PROPOSITION 6: (1) (Jewitt, 2004) If the dis-tribution of va has a density that is (a) decreas-ing, (b) increasing, (c) unimodal, then (M� �M�)(va) is, respectively, (a) increasing, (b)decreasing, (c) quasi convex.

34 Corneo (1997) provides related insights (but formalresults only in a quadratic case), based on whether the valueof reputation is assumed to be a concave (“conformist”) ora convex (“elitist”) function of someone’s perceived rank(which, by definition, is uniformly distributed) in the dis-tribution of altruism. For any such function s(ra) of rankr(va) � G(va), we can define va � s(ra) (s � G)(va), whichhas density g � 1/(s� � s�1)(va) 1/(s�(ra)). Thus all theresults in Proposition 6.1 on increasing, decreasing, andunimodal densities immediately carry over to concave, con-vex, and convex-concave payoff functions.


(2) If the distribution of va has a log-concavedensity (more generally, a log-concave distri-bution function), then for all �a � [0, 1], thesupply function is everywhere upward-sloping.

Part 1 provides sufficient conditions for themonotonicity of M� � M�, which definescomplementarity or substitutability. What ulti-mately matters for uniqueness or multiplicityand the slope of the supply curve, however, isthe behavior of �(va) va � �a(M� � M�)(va), for which the strength of reputational con-cerns, �a, is also relevant. In part 2 we thusshow that, for all �a � [0, 1], uniqueness ob-tains as long as g does not increase too fast—amuch weaker condition than 1(b). No simpleanalogue is available for the case of multiplic-ity, but it is clear that it corresponds to situa-tions where complementarity obtains and �a ishigh enough (as in the example given earlier).

Excuses, Forced Participation, and Observ-ability.—Thus far, we have assumed that ob-servers (other agents, future “self”) know forsure that the individual had an opportunity tocontribute and, if so, whether he did. This isoften not the case, however.

Suppose that with probability � � [0, 1], anindividual faces (unverifiable) circumstancesthat preclude participation: not being informed,having to deal with some emergency, etc. Forany potential cutoff va, the honor conveyed byparticipation is unchanged, MP(va) M�(va),while the stigma conveyed by nonparticipationis lessened, taking the form of a weightedaverage

(27) MNP�va; �� v�a � �1 � ��G�va�M��va�

� � �1 � ��G�va�.

The same expressions are easily seen to apply ifabstention never gives rise to a signal that theindividual contributed, but a contribution maygo unnoticed (fail to generate such a signal)with probability �.

Conversely, suppose that with probability �� [0, 1], an individual is forced to contribute, ordraws a temporarily low cost c. The stigmafrom abstention is now unchanged, MNP(va) M�(va), but the distinction conveyed by par-ticipation is dulled, and given by

(28) MP�va ; ��

��v�a � �1 � ��1 � G�va��M��va�

�� 1 � ��1 � G�va��.

The same expressions apply if participation al-ways gives rise to a signal suggesting that theindividual contributed, but nonparticipation cango undetected (also lead to such a signal) withprobability ��.

PROPOSITION 7: (1) An increase in theprobability of unobserved forced participationfacilitates the emergence of strategic comple-mentarities and multiple social norms, whereasan increase in the probability of (unobserved)involuntary nonparticipation inhibits it.

(2) The same results hold for, respectively, anincrease in the probability that abstentionmay escape detection and an increase inthe probability that a good deed goes un-noticed.

Empirical and Policy Implications.—The re-sults of this section have a number of interestingimplications. First, for behaviors such as crime,from which most people are deterred by either astrong intrinsic distaste (the density of va isincreasing) or strong extrinsic constraints (ahigh ��), stigma avoidance will be the dominantreputational concern (by contrast, having nocriminal record is not particularly glorious) andactions will be strategic complements, poten-tially leading to substantial variations over timeand space. Conversely, opportunities to engagein heroic behaviors (risking one’s life for some-one else, donating an organ or significantwealth) are relatively rare (high �) and fewpeople are intrinsically motivated to such greatfeats of abnegation. The signaling motive willtherefore be dominated here by the pursuit ofdistinction, making noble acts strategic substi-tutes and their prevalence much less variablethan that of (comparably rare, on average) crim-inal acts.

Second, even absent multiplicity, the twotypes of behaviors will respond quite differentlyto public intervention. Since

(29) a� ��y� � �1 � ��v*a �y��1g�v*a �y��,


we see that for crime-like behaviors the effect oflegal rewards and punishments (y) is amplifiedby the response of social pressure (crowdingin), whereas for self-sacrifices it is dampened byit (partial crowding out). We shall come back tothis point when analyzing the socially optimallevel of incentives.

IV. Turning Down Rewards

An agent may be eager to engage in a proso-cial action, but concerned that the value of hisgood deed will be sullied by an inference thatmaterial considerations played a role in the de-cision. In some situations, he may then want toturn down part or all of the reward (provided theincentive scheme is indeed a payment for goodbehavior rather than a penalty for bad behavior),or even supplement his participation with anet monetary contribution.35

Naturally, the issue does not arise if give-backs are not observable by the audience towhom agents are trying to signal, or if thesponsor can reward them secretly. On the otherhand, taking secret rewards does not help withself-image, and may even damage it.

Suppose now that the realized transfer fromthe sponsor to the agent is effectively observ-able. When the uncertainty is only about va, thenet reputational gain from participating for y� �y, relative to not participating, is r(y�) �a(E(va�1, y�) � E(va�0, y�)). The agent there-fore cannot signal his type by taking less than yor even giving money to the sponsor: the loss ofincome, vy(y � y�), and the net reputationalbenefit, r(y�) � r(y), are both type-independent.Consequently, the equilibria studied in SectionIII (where vy � 1) are still equilibria of theenlarged game in which agents can turn down part

or all of the reward.36 For the same reason, offer-ing menus of rewards cannot benefit the sponsor.

By contrast, when the uncertainty is (also)about vy, which is needed to obtain net crowd-ing out, turning down the reward or part of itcould be used to signal the absence of greed.The idea that offering such “menus” may be agood strategy for increasing contributions (as inthe blood-donation experiment discussed ear-lier) is consistent with both our information-based approach to prosocial behavior, whichemphasizes individuals’ concerns with the in-ferences attached to their contributions, andwith the general principle that a principal al-ways (weakly) benefits from being able toscreen agents along more dimensions.

Yet, even in this case, it may be that allagents either just accept y or do not participate,but never turn down rewards, so that there is nogain to introducing the option. The intuition isthat doing so could lead the audience to ques-tion an agent’s motivation along another dimen-sion: is he genuinely disinterested, or merelyconcerned about appearances? It is thus linkedto the general idea that good deeds that are “tooobvious” may backfire, which was first encoun-tered when studying public prominence in Sec-tion IIB.37

To capture this idea, we allow again uncer-tainty about v (va, vy) to combine with un-certainty about agents’ degree of image-consciousness � (�a, �y), but focus here ona very simple case, to avoid what would other-wise be a rather technical analysis. Suppose that(�a, �y) ( x�a, x�y), where (�a, �y) is fixedand thus known to the audience, whereas x isindependently distributed from (va, vy) andtakes one of two extreme values: agents areeither image indifferent ( x 0) or image driven( x ��). Image-indifferent individuals partic-ipate if and only if va � c � vyy � 0; when theydo, they clearly never turn down the reward (or

35 Alternatively, sponsors may respond to contributors’desire to appear intrinsically rather than extrinsically moti-vated by publicly announcing low rewards. In Benabou andTirole (2004a) we show that: (a) with strategic substitutes(�� 0), a sponsor would indeed like to do so, but thiscreates a commitment problem: if it can later on secretlyrenegotiate with the agents, both will agree to raising y;(b) with strategic complements (�1 �� 0), on thecontrary, the sponsor offers a higher fee under publicdisclosure than under confidentiality, and this is renego-tiation-proof since agents will not agree to secret cuts intheir rewards.

36 It can also be verified that these equilibria satisfy theNever-a-Weak-Best-Response criterion of In-Koo Cho andDavid M. Kreps (1987).

37 The same intuition implies that people may want to be“modest” about their generosity. Thus one can show, in asimple extension of the model (with again heterogeneity in�a), that agents may refrain from disclosing their gooddeeds, hoping that the audience will come to learn of themthrough other channels.


part of it), as this would be a strictly dominatedstrategy. We shall assume that if the populationconsisted only of image-indifferent individuals,participation would be reputation enhancing(E(�ava � �yvy�va � vyy � c) � 0, whichalways holds for y below some threshold). Turn-ing now to image-driven individuals, they all poolon the actions that yield the highest reputation,choosing an a � {0, 1} and a reward y� � y thatmaximize R(a, y�) �aE(va�a, y�) � �yE(vy�a, y�).If, in equilibrium, a positive fraction of themchose to participate and receive y� y, theywould be identified as image-driven types, andso their reputation would correspond to the priormean (v�a, v�y).

38 But they would then be strictlybetter off pooling with those image-indifferentagents who participate at price y. The uniqueequilibrium thus consists in participation, at theoffered price y, by all image-driven individualsand by those image-indifferent individuals forwhom va � c � vyy � 0.

PROPOSITION 8: Agents may never turndown the reward, or part of it, even when thiswould be publicly observed and there is uncer-tainty about money orientation, vy.

It is worth pointing out that in deriving thisresult, we did not assume any social opprobriumon image consciousness; presumably, thiswould only reinforce agents’ reluctance to turndown rewards.39

V. Welfare and Competition

We now examine the way in which public orprivate sponsors (social planner, governmentagency, nongovernmental organization (NGO),

religious organization) will set incentives andthe welfare properties of the resulting equilib-rium. For these purposes, we first need to makeexplicit again the public-good aspects of agents’contributions, then specify different sponsors’objective functions.

Recall from Section IA that an individual’sintrinsic motivation can, in general, have twocomponents: va ua � wa /n�, where ua is a pure“joy of giving,” whereas wa is the marginal utilityof a public good na�/n� generated by total contri-butions na�. To simplify the analysis, we take hereua and wa to be independently distributed (withagain vy � 1) and denote the mean of wa as w� a.

Given an incentive rate y, an equilibrium(unique or not) is determined by a cutoff v*a.Agents’ expected per capita welfare is thus

(30) U� �v*a ; y�

� E�wa�na�/n�� E �a�ua � c � y� � �ava�

� �v *a

va�

��n � 1��w� a /n��

� va � c � y]g�va � dva � �av� a .

This expression embodies three effects. First,each agent who contributes enjoys a direct util-ity va � c � y and additionally generates for then � 1 others a positive spillover, equal to w� a /n�

on average. Second, the pursuit of esteem is azero-sum game: the average reputation in soci-ety remains fixed at �av�a, reflecting the martin-gale property of beliefs.40 Third, because anagent’s participation decision is based on theprivate reputational return rather than the socialone (which is zero), it inflicts an externalityonto others. Thus, starting from equilibrium, thewelfare impact of a marginal increase in partic-ipation is38 If they pooled at multiple values y�, these would all

need to deliver the same average reputation, which wouldtherefore again correspond to the prior mean.

39 Note also that, while Proposition 8 focuses for sim-plicity on the extreme case where x 3 ��, the effect itbrings to light is much more general. One can thus showthat: (a) for all finite x, there always exists an equilibrium inwhich no one turns down the reward; (b) even in the bestequilibrium for the sponsor, the fraction of image-consciousagents who do so remains bounded away from one across allvalues of x, thus limiting the profitability of introducing thisform of price discrimination.

40 That is, E[E[va�a, y]] v�a. It thus does not matterwhether we include agents’ utilities from reputation (e.g., van-ity) in the definition of social welfare. Note that the zero-sumproperty also relies on the linearity of the reputational payoffand the independence of �a from va. When these assumptionsdo not hold, the distribution of reputation across agents willhave allocative and efficiency consequences—for instance,through subsequent matching patterns.


(31) �U� �v*a ; y�

v*a

��n � 1��w� a /n�� v*a � c � y�g�va�

��n � 1��w� a /n�� v*a��g�v*a�.

The first term is the standard public-goods exter-nality, which we shall denote e� � (n � 1)(w� a/n�).The second term reflects the fact that each mar-ginal participant brings down the “quality” of thepool of contributors as well as that of noncontribu-tors: by the martingale property, the reputationallosses of inframarginal agents on both sides mustadd up to the gains of the marginal participant,�(v*a) r(y). Equivalently, we can think of (31)as the difference between a free-riding effect anda reputation-stealing effect.

A. Sponsors’ Choice of Incentives and theSocial Optimum

Consider now a public or private sponsor thatinternalizes some fraction � � [0, 1] of agents’welfare and also derives from each one’s par-ticipation a private benefit, with equivalentmonetary value B. We focus first on the case ofmonopoly or differentiated public goods, thenconsider competition. The sponsor’s expectedpayoff (normalized by population size n) is thus

(32) W� �y� � �U��v*a�y�; y� � �B � y�a��y�.

For a social planner whose preferences mirrorthe ex ante utility of the n potential contributorsand who has access to lump-sum taxes, � 1and B 0. More generally, B � 0 could reflecta different discounting of the welfare of futuregenerations (e.g., with pollution or biodiversity)and � � 1 the presence of a shadow cost ofpublic funds: clearly, replacing B � y by B �(1 � �)y in �(y) is equivalent to dividing bothB and � in (32) by 1 � �. For other actors suchas charities, NGOs, or specialized governmentagencies, B may reflect the premium placed ona public good by a particularly motivated con-stituency (friends of the arts, environmental-ists), or some purely private benefits tied to thechanneling of donations or the delivery of pub-lic goods: rents appropriated in the process by

the organization, bundling of a religious mes-sage with schooling or poverty relief, or (inreduced form) the sponsor’s own signaling orcareer concerns.41 Both B and the weight �placed by the sponsor on social welfare areagain normalized by the opportunity cost offunds that it faces.42

Since rewards that lead to net crowding out,a��(y) 0, are never optimal, we assume that�� 0, resulting in a unique equilibrium v*a(y)which, for simplicity, we take to be interior, anda supply curve na�(y) n[1 � G(v*a(y))], withelasticity �(y) � ya��(y)/a�(y) � 0. We alsoassume that W� is strictly quasiconcave in allcases (it always is for � 1). Using (31) andnoting that a��(y) �(v*a)�(y) � g(v*a(y)), wehave

(33) W� ��y� � ��e� � ��v*a �y�� B � y�

� a��y� � �1 � ��a��y�.

For (symmetric) competitive sponsors, theprivate-payoff term in (32) is replaced by (B �yi)a� i(y), where a� i(y) is the share of total contri-butions specifically channeled through sponsori; in equilibrium, all rewards are then driven toB.43 We shall denote the values of �, B, y, andW� for the social planner, monopolistic, and

41 Sponsors also often care about the quality of partici-pation, not just total enrollment, in cases where it is subjectto adverse selection or moral hazard. Thus, one argumentfor relatively low pay for the military is to select truepatriots rather than mercenaries whose main loyalty is towhoever pays more. Similarly, it is often argued that notpaying for blood reduces the fraction of donors with hepa-titis and other diseases. These ideas could be captured byintroducing a hidden action (beyond a � A, which is ob-served) whose marginal cost to the individual decreaseswith va, leading to a benefit for the sponsor B(va), with B� �0. For instance, a purely private sponsor (� 0) would nowmaximize Ev,�[(B(va) � y)a(v, �; y)].

42 It is worth recalling here that the model also applies tomonetary donations, with sponsors offering either a match-ing rate or “perks” and other goods or services (in additionto the publicity); see footnote 10.

43 While this is the standard result, it depends here cru-cially on the fact that vy 1 is known. Otherwise, there isa reputational payoff to participating for a lower fee, andsponsor competition will then lead to rewards being biddown rather than up, leaving firms with positive profits. This“reversal” of Bertrand competition is analyzed in Benabouand Tirole (2004a) and shares important similarities withBagwell and Bernheim’s (1996) analysis of the pricing ofconspicuous-consumption goods.


competitive sponsors by the superscripts s, m,and c respectively, with �s � max{�m, �c}.

PROPOSITION 9: (1) The socially optimal in-centive rate is always strictly less than thestandard Pigouvian subsidy yP � e� � Bs whichleads agents to internalize the full public-goodvalue of their contribution. When taxation isnondistortionary (�s 1), it equals ys yP ��(c � yP); more generally, it is given by

(34) ys ��s�e� � ��v*a � ys�� Bs

1 � �1 � �s�/�� ys�.

(2) A monopoly sponsor with �m �s mayoffer contributors a reward ym that is toogenerous (or require of them too low amonetary donation) from the point of viewof social welfare, resulting in excess par-ticipation. This is true even when the ben-efit it derives from agents’ participationcoincides exactly with the gap betweentheir social and private contributions to thepublic good (Bm � �me� Bs � �se�).

(3) Competition between sponsors increasesrewards (or reduces required monetarycontributions) and may thus reduce socialwelfare, compared to a monopoly (with thesame �c �m and Bc Bm).

The first result shows that the optimal incen-tive scheme should include a tax that correctsfor the reputation-seeking motive to contribute,which in itself is socially wasteful. This repu-tational rent is endogenous to the reward, how-ever. Thus with �s 1, when individualcontributions are complements (respectively,substitutes), ys yP � �(c � yP) responds less(respectively, more) than yP to changes in Bs

(which leave the function � unchanged). Simi-larly, the optimal penalty for antisocial activi-ties such as littering, polluting, etc., should“leave space” for the effect of opprobrium,which itself depends on the fine. As to a highershadow cost of public funds (a proportionalreduction in �s and Bs), it naturally tends toreduce ys; when contributions are substitutes,some of this reduced public intervention ismade up by increased social pressure, as �rises in response to the decline in participa-tion. With complements, however, the repu-

tational incentive to contribute is also weakened.These results provide both some support andan important qualification to arguments (e.g.,Geoffrey Brennan and Philipp Pettit, 2004)calling for a shift in public policy from theuse of fines and other costly sentences to agreater reliance on public praise and shame.Esteem-based incentives can adequately re-place material rewards and punishments inspheres where gaining distinction is the dom-inant reputational concern (self-sacrifice, her-oism, great inventions), but not in those whereavoiding stigma is most important (crime, wel-fare dependency).

The intuition for the second result in Propo-sition 9 is that a monopolist setting ym does notnot internalize the reputational losses of infra-marginal agents to the same extent a plannerwould. This gives it an incentive to attract toomany “customers,” which works against thestandard monopolistic tendency to serve toofew. The tension between these two forces canbe seen from the fact that (W� s)�(ym) 0 if

(35) ��s � �m��ym/��ym� � ��v*a �ym��

� Bs � �se� � Bm � �me� � 0.

A low supply elasticity � causes the monopolistto offer too low a price, as usual. When repu-tational concerns are important enough, how-ever (a high �a and therefore a high �), theinformational externality can dominate, makingthe monopolist too “generous” or not demand-ing enough in the standards it sets for monetarydonations. The last two terms in (35), finally,represent the total benefits (private benefit plusinternalized contribution to social welfare) de-rived by each sponsor from a marginal agent’sparticipation, each normalized by the corre-sponding shadow cost of funds. The effect oftheir difference on the sign of ym � ys isstraightforward, and part 2 of the propositionnormalizes it to zero as a benchmark.

Sponsor competition, finally, further exacer-bates the inefficiency above, because each firmnow has a much higher incentive to raise itsoffer than a monopolist (it takes the whole mar-ket), but still inflicts the same reputational coston all inframarginal noncontributors. This sug-gests, for instance, that universities may sell the


naming rights to professorial chairs and build-ings too cheaply, relative to the social optimum.

B. Holier-than-Thou Competition

We saw that competition may reduce welfareby inducing excessive participation in prosocialactivities that generate only moderate public-good benefits but have a high visibility. We willnow see that it can reduce welfare (relative to amonopolist) even without any change in partic-ipation, by leading sponsors to screen contrib-utors in inefficient ways. This result formalizesin particular the idea of religions and sects com-peting on orthodoxy, asceticism, and othercostly requirements for membership (e.g., EliBerman, 2000). Another important example isthat of charities sponsoring events whereagents, instead of simply donating or raisingmoney (or on top of it), engage in time-intensive,strenuous activities such as a day-long walk,marathon, or other tests of endurance often re-quiring months of preparation.44

To capture this phenomenon most simply, let vatake values va

H with probability �, or vaL va

H withprobability 1 � �, while maintaining vy � 1.Assume, furthermore, that the nonmonetarycost of contributing is c (possibly zero) unlessthe sponsor demands a “sacrifice,” which it isable to verify and publicly certify. The cost thenbecomes cH for the high type and cL for the lowtype, where

(36) c � cH � cL.

A sacrifice is a pure deadweight loss, whoseonly benefit is to help screen agents’ motiva-tion. The assumption that cL � cH reflects theidea that such a sacrifice is less costly to a moremotivated agent. For simplicity, we will assume

that cL is so large that the low type is neverwilling to sacrifice and will focus on determin-istic contracts offered by sponsors seeking tomaximize their private payoff �(y); that is, weset � 0 (the results would extend to any � 1).

PROPOSITION 10: In the two-type case de-scribed above, a monopoly sponsor who wantsboth types to contribute does not screen con-tributors inefficiently. By contrast, competingsponsors may require high-valuation individu-als to make costly sacrifices that represent puredeadweight losses, thereby reducing socialwelfare.

The intuition for this result is that nonpricescreening imposes a negative externality onlow-type agents, the cost of which a monopolistmust fully bear but which competitive sponsorsdo not internalize. Indeed, screening throughcostly sacrifices has two effects: (a) it inflicts adeadweight loss cH � c on the high type, whichthe sponsor must somehow pay for; (b) it booststhe high type’s reputation and lowers that of thelow type. When the high-type’s reputationalgain exceeds the cost of sacrifice, the sponsorthrough which he contributes can appropriatethe surplus, in the form of a lower reward. If thissponsor is a monopolist who finds it profitableto serve the whole market (which is always thecase when � is low enough), he must also com-pensate the low type for his reputational loss.By a now-familiar argument, these losses mustexactly offset the high type’s reputation gains,so the net effect of (b) on agents’ average util-ity, as well as on the monopolist’s payoff, is nil.This leaves only the net cost corresponding to(a), implying that a sponsor serving the wholemarket will never require sacrifices.

Things are quite different under free entry.First, since vy is known, price competition againdrives all sponsors to offer B. Second, by re-quiring a sacrifice, entrants can now attract thehigh types away from competitors who imposeno such requirement, leaving low types (or theirsponsors) with the resulting reputational loss.This “cream skimming” leads inevitably to anequilibrium where a proportion � of the con-tracts offered by active sponsors requires aninefficient sacrifice and attracts only high types,while the remaining 1 � � requires only the

44 Camille Sweeny (“The Latest in Fitness: Millions forCharity,” New York Times, July 7, 2005) documents that: (a)many large, health-related charities in the United States nowderive over a third of their revenues from endurance pro-grams and challenges; (b) most sponsored participants arenot athletic types or even regular exercisers (leading to ahigh rate of injury); (c) while their motivations vary, thefundraising/doing-good aspect is the dominant one (theyoften have themselves been, or are personally close to,victims of the disease which the funds they are raising willbe dedicated to combat).


normal contribution c and attracts the lowtypes.45

Turning finally to welfare, one can show thatboth types of agents are better off under compe-tition than under monopoly (see the Appendix).The sponsors or their underlying beneficiaries,however, must necessarily lose more than all con-tributors gain: total participation remains un-changed (both types still contribute), the same istrue of average reputation (by the martingale prop-erty), and rewards are pure transfers. There isnow, however, a deadweight loss of �(cH � c),corresponding to the wasteful sacrifices made bythe high types to separate. Therefore, competitionunambiguously reduces welfare.

VI. Conclusion

To gain a better understanding of prosocialbehavior, we sought, paraphrasing Adam Smith,to “thoroughly enter into all the passions andmotives which influence it.” People’s actionsindeed reflect a variable mix of altruistic moti-vation, material self-interest, and social or self-image concerns. Moreover, this mix variesacross individuals and situations, presenting ob-servers seeking to infer a person’s true valuesfrom his behavior (or an individual judginghimself in retrospect) with a signal-extractionproblem. Crucially, altering any of the threecomponents of motivation, for instance throughthe use of extrinsic incentives or a greater pub-licity given to actions, changes the meaningattached to prosocial (or antisocial) behaviorand hence feeds back into the reputational in-centive to engage in it.

This simple mechanism leads to many newinsights concerning individuals’ contributionsto public goods, the interactions between formalincentives and social norms, and the strategicdecisions of public or private sponsors seekingto increase or capture contributions. This line ofresearch could be extended in several interest-ing directions. A first one concerns organizations,where high-powered incentives or performancepay could conflict with agents’ signaling mo-tives that arise from teamwork or career concerns.A second relates to the role and objectives of

sponsors, who in practice often have their ownsignaling concerns. A third one, linked to theself-image interpretation of the model and pur-sued in Benabou and Tirole (2006), is the topicof identity and the many instances where peoplerefuse transactions that seem to be in their besteconomic interest, but which they judge to beinsulting to their dignity.

APPENDIX

PROOF OF PROPOSITION 1:Since y here is a fixed parameter, in what

follows we will temporarily omit from the no-tation the dependence of all functions on thisargument. Differentiating (9) and (10) with re-spect to a yields

(A1)dE�va�a�

da� ��k � r��a��

and

dE�vy�a�

da� ��k � r��a��.

Therefore, r�(a) is a solution to the linear differ-ential equation r�(a) �(k � r��(a)), where � �� a� � �� y�. The generic solution is r�(a) k(� � �e�a/�), where � is a constant of integra-tion. For � � 0, however, the objective functionof every agent is not globally concave and isactually maximized at a �� (depending onthe sign of ��). The only well-defined equilib-rium is thus for � 0.

PROOF OF PROPOSITIONS 2 AND 3:From (11), we have

(A2) �� y� � �2y�a

2�y2 � �ay��a

2 � y2�y2�

��a2 � 2y�ay � y2�y

2�2 ,

(A3)

�� y� ��y

2��a2 � y2�y

2� � 2�ay � y�y2 � �ay �

��a2 � 2y�ay � y2�y

2�2 .

Substituting into (14) immediately yields part1 of Proposition 3 in the case y 0, and part1 of Proposition 2 when �ay 0. This last

45 As long as � is not too large, this is the only equilib-rium that is robust to the Cho-Kreps (1987) criterion.


inequality can be rewritten as

(A4) Q� y� � �v� y /k��1 � y2�2�2

� �� y�4y2 � 2�� a�

2y

� �� y�2 � L� y�.

The left-hand side is a second-order polyno-mial in y2, hence convex and symmetric overall of �, with value Q(0) v� y /k � 0 at theorigin. The right-hand side is an increasinglinear function with L(0) �� y�2. Conse-quently, if L(0) � Q(0), then for any �� a � 0,L( y) intersects Q( y) once at some y1 0 andonce at some y2 � 0. On the other hand, ifL(0) Q(0), there exists a unique �*a � 0for which L( y) has a (single) tangency pointy* � 0 with Q( y). For all �� a �*a, Q( y) �L( y) on all of �*, so a� �( y) � 0 everywhere.For all �� a � �*a, however, L( y) intersectsQ( y) twice, at points 0 y1 y2. Theseproperties, together with the linearity of L in�� ay and the convexity of Q( y), conclude theproof of Proposition 2.

Part 2 of Proposition 3 follows from the factthat, given part 1, as � �y /�a 3 ��, thedominant term in a��(0) is asymptotically equiv-alent to �� y�

2[1 � 2(�ay /�a�y)2], which tends

to �� as long as the correlation between va andvy is less than 1/�2 in absolute value.

PROOF OF PROPOSITION 4:The only difference with Proposition 1 is

the presence of the term �( y)2 k2Var[r( y;�)] in the denominator of � and � (see (19)),leading to the fixed-point equation defining�( y):

(A5) �2 � k2

� Var��a� �a2 � y�ay

�a2 � 2y�ay � y2�y

2 � �2�� y� y�y

2 � �ay

�a2 � 2y�ay � y2�y

2 � �2� �� Z��2�.

Since Z(�2) is always positive but tends tozero as �2 becomes large, there is always atleast one solution. When �ay 0, moreover,Z(�2) is the sum of two squared terms thatare decreasing in �2, so the solution isunique. When �ay � 0, one cannot rule outmultiple equilibria; note, however, that thosethat are stable (in a standard, tatonnementsense) are those where Z cuts the diagonalfrom above. Therefore, in any stable equilibrium,� is increasing in k, which in turn implies that�(y) and �(y) are decreasing in k, as long as �ayis not too negative. Finally, multiplying all the(�a, �y)’s by a common “publicity factor” x hasthe same effect on (A5) as multiplying k2 by x,which concludes the proof.

PROOF OF PROPOSITION 6:Part 1 is due to Jewitt (2004). To show part 2,

we can write:

va � �a �M��va� � M��va�� va � M��va�

� �aM��va� � �1 � �a�M��va�,

then observe that both M� and M� are in-creasing functions, and so is va � M�(va) (��

va G(v) dv)/G(va) if the integral of G islog-concave. Since log-concavity is preservedby integration over convex sets, it suffices thatG itself be log-concave. In turn, a sufficientcondition for this is that g be log-concave.

PROOF OF PROPOSITION 7:To show part 1, rewrite (MP � MNP)(va;

�) [M�(va) � v�a]/[1 � (1 � �)(1 � G(va))]and observe that if (MP � MNP)�(va; �) � 0, thisexpression is also positive for all �� , since

1

�MP � MNP��va; ��

1

�MP � MNP��va; ��

�� 1 � G�va��

M��va� � v�a,

and the last term is clearly decreasing in va. Sim-ilarly, to show part 2, note that in this case(MP �MNP)(va; �) [v�a � M�(va)]/[1 � (1 ��)G(va)] and that if (MP � MNP)�(va; �) 0, itis also negative for all �� .


PROOF OF PROPOSITION 9:The general formula in part 1 follows from (33)

and the assumed strict quasiconcavity of W� s. For�s 1, we have (W� s)�(y) [Bs � e� � y ��(v*a(y))] � a��(y) and v*a(y) � �(v*a(y)) c � y(interior equilibrium), so (W� s)�(y) has the sign ofBs � e� � c � v*a(y). Therefore, W� s(y) is strictlyconcave and maximized at the point ys such thatv*a(ys) � Bs � e� c, which is the standardSamuelson condition for efficient public-goodsprovision. Substituting v*a(ys) c � yP into theequilibrium condition yields ys yP � �(c � yp).

For part 2, note that (35) holds for all Bm ��me� � Bs � �se� as long as �(v*a(ym)) � a�(y)/a��(y), or

(A6)��v*a �

��v*a � � g(v*a)

1 � G(v*a)� � 1,

where v*a stands for v*a( ym). For instance, for�m 0 and va uniformly distributed on [0,1], we have v*a( ym) (c � �a/ 2 � 1 � B)/2 � (0, 1) and ym (B � 1 � c � �a/ 2)/2 B as long as ��a/2 1 � B � c 2 ��a/2. Thus, ym � ys B � e� � �a/2 whenever�a � 1 � B � c � 2e�, which is consistent withthe previous inequalities as long as �a � 2e�.Part 3, finally, is implied by part 2 as long asym Bm Bc yc (which is always the case aslong as �m is not too large), since W� s is declin-ing to the right of ys.

PROOF OF PROPOSITION 10:(1) As long as � is not too small, it is optimal

for the monopolist to get both types on board. Ifhe does not demand any sacrifice, he sets y so asto make the low type indifferent: y c � va

L ��a(v�a � va

L), where v�a � �vaH � (1 � �)va

L is theprior mean. The sponsor’s payoff is then �1 �B � y B � c � va

L � �a(v�a � vaL). Suppose

now that the high type is asked to sacrifice.Rewards are then yL c � va

L and (from incen-tive compatibility) yH yL � cH � c � �a(va

H �va

L). The sponsor’s payoff is then only �2 B ��yH � (1 � �)yL �1 � �(cH � c) �1.

(2) Under free entry, all sponsors offer, andall contributors accept, y B. Moreover, ifcH � c � �a(va

H � vaL), it is now an equilibrium

for the high type to separate from the low typeby opting for a sponsor who requires a sacrifice.In the resulting equilibrium (described in the

text), both types of agents are better off thanunder monopoly: the low type’s payoff risesfrom �ava

L to �avaL � va

L � c � B, while the hightype’s payoff increases by at least va

L � c � B,which is positive from the condition that themonopoly prefers to enlist both types. The factthat sponsors must necessarily lose more thanthe agents gain, resulting in a net welfare lossfrom competition, was established in the text.

REFERENCES

Akerlof, George A., and William T. Dickens.1982. “The Economic Consequences of Cog-nitive Dissonance.” American Economic Re-view, 72(3): 307–19.

Akerlof, George A., and Rachel E. Kranton. 2000.“Economics and Identity.” Quarterly Journalof Economics, 115(3): 715–53.

Andreoni, James. 1989. “Giving with ImpureAltruism: Applications to Charity and Ricar-dian Equivalence.” Journal of Political Econ-omy, 97(6): 1447–58.

Andreoni, James. 2006. “Philanthropy.” InHandbook of the Economics of Giving, Rec-iprocity and Altruism, ed. Serge-ChristopheKolm and Jean M. Ythier, Vol. 2, chap. 18.Amsterdam: Elsevier Science North-Holland.

Araujo, Aloisio Pessoa de, Daniel Gottlieb, andHumberto Moreira. 2004. “A Model of MixedSignals with Applications to Countersignal-ing and the GED.” Getulio Vargas Founda-tion Economics Working Paper 553.

Austen-Smith, David, and Roland G. Fryer, Jr.2005. “An Economic Analysis of ‘ActingWhite’.” Quarterly Journal of Economics,120(2): 551–83.

Bagwell, Laurie Simon, and B. Douglas Bernheim.1996. “Veblen Effects in a Theory of Con-spicuous Consumption.” American EconomicReview, 86(3): 349–73.

Batson, Dan. 1998. “Altruism and Prosocial Be-havior.” In Handbook of Social Psychology.Vol 2, ed. Daniel T. Gilbert, Susan Fiske, andGardner Lindzey, 282–316. New York:McGraw Hill.

Bem, Daryl J. 1972. “Self-Perception Theory.”In Advances in Experimental Social Psychol-ogy. Vol. 6, ed. Leonard Berkowitz, 1–62.New York: Academic Press.

Benabou, Roland, and Jean Tirole. 2002. “Self-


Confidence and Personal Motivation.” Quar-terly Journal of Economics, 117(3): 871–915.

Benabou, Roland, and Jean Tirole. 2003. “Intrin-sic and Extrinsic Motivation.” Review of Eco-nomic Studies, 70(3): 489–520.

Benabou, Roland, and Jean Tirole. 2004a. “Incen-tives and Prosocial Behavior.” Centre for Eco-nomic Policy Research Discussion Paper 4633.

Benabou, Roland, and Jean Tirole. 2004b. “Will-power and Personal Rules.” Journal of Polit-ical Economy, 112(4): 848–86.

Benabou, Roland, and Jean Tirole. 2006. “ACognitive Theory of Identity, Dignity, andTaboos.” Unpublished.

Berman, Eli. 2000. “Sect, Subsidy, and Sacri-fice: An Economist’s View of Ultra-OrthodoxJews.” Quarterly Journal of Economics,115(3): 905–53.

Bernheim, B. Douglas. 1994. “A Theory of Con-formity.” Journal of Political Economy,102(5): 841–77.

Bodner, Ronit, and Drazen Prelec. 2003. “Self-Signaling and Diagnostic Utility in EverydayDecision Making.” In The Psychology ofEconomic Decisions. Volume I: Rationalityand Well-Being, ed. Isabella Brocas and JuanD. Carrillo, 105–26. Oxford: Oxford Univer-sity Press.

Brekke, Kjell A., Snorre Kverndokk, and KarineNyborg. 2003. “An Economic Model ofMoral Motivation.” Journal of Public Eco-nomics, 87(9–10): 1967–83.

Brennan, Geoffrey, and Philip Pettit. 2004. TheEconomy of Esteem: An Essay on Civil andPolitical Society. Oxford: Oxford UniversityPress.

Bryan, James H., and Mary A. Test. 1967. “Mod-els and Helping: Naturalistic Studies in Aid-ing Behavior.” Journal of Personality andSocial Psychology, 6(4): 400–07.

Buraschi, Andrea, and Francesca Cornelli. 2002.“Donations.” Centre for Economic PolicyResearch Discussion Paper 3488.

Carrillo, Juan D., and Thomas Mariotti. 2000.“Strategic Ignorance as a Self-DiscipliningDevice.” Review of Economic Studies, 67(3):529–44.

Cho, In-Koo, and David M. Kreps. 1987. “Sig-naling Games and Stable Equilibria.” Quar-terly Journal of Economics, 102(2): 179–221.

Corneo, Giacomo G. 1997. “The Theory of the

Open Shop Trade Union Reconsidered.” La-bour Economics, 4(1): 71–84.

Dana, Jason D., Roberto A. Weber, and Jason X.Kuang. 2003. “Exploiting Moral WriggleRoom: Behavior Inconsistent with a Prefer-ence for Fair Outcomes.” Carnegie MellonUniversity Behavioral Decision ResearchWorking Paper 349.

Deci, Edward L. 1975. Intrinsic Motivation.New York: Plenum Press.

Deci, Edward L., and Richard M. Ryan. 1985.Intrinsic Motivation and Self-Determinationin Human Behavior. New York: Plenum Press.

Denrell, Jerker. 1998. Essays on the EconomicEffects of Vanity and Career Concerns.Stockholm: Stockholm Institute of Interna-tional Economics Press.

Ellingsen, Tore, and Magnus Johannesson. 2006.“Pride and Prejudice: The Human Side ofIncentive Theory.” Unpublished.

Falk, Armin, and Michael Kosfeld. 2006. “TheHidden Costs of Control.” American Eco-nomic Review, 96(5): 1611–30.

Fehr, Ernst, and Simon Gachter. 2001. “Do In-centive Contracts Crowd Out Voluntary Co-operation?” University of Zurich Institute forEmpirical Research in Economics WorkingPaper 34.

Fehr, Ernst, Alexander Klein, and Klaus Schmidt.2001. “Fairness, Incentives and ContractualIncompleteness.” University of Munich Dis-cussion Papers in Economics 18.

Festinger, Leon, and James Carlsmith. 1959.“Cognitive Consequences of Forced Compli-ance.” Journal of Personality and Social Psy-chology, 58(2): 203–10.

Freeman, Richard B. 1997. “Working for Noth-ing: The Supply of Volunteer Labor.” Jour-nal of Labor Economics, 15(1): S140–66.

Frey, Bruno S. 1997. Not Just for the Money: AnEconomic Theory of Personal Motivation.Cheltenham: Edward Elgar.

Frey, Bruno S., and Lorenz Goette. 1999. “DoesPay Motivate Volunteers?” University of Zu-rich, Institute for Empirical Research in Eco-nomics Working Paper 7.

Frey, Bruno S. and Reto Jegen. 2001. “Motiva-tion Crowding Theory.” Journal of EconomicSurveys, 15(5): 589–611.

Funk, Patricia. 2005. “Theory and Evidence on theRole of Social Norms in Voting.” Unpublished.


Gibbons, Robert. 1997. “Incentives and Careers inOrganizations.” In Advances in Economics andEconometrics: Theory and Applications. Vol.2, ed. David M. Kreps and Kenneth F. Wallis,1–37. Cambridge: Cambridge University Press.

Gintis, Herbert, Eric Smith, and Samuel Bowles.2001. “Costly Signaling and Cooperation.”Journal of Theoretical Biology, 213(1): 103–19.

Glazer, Amihai, and Kai A. Konrad. 1996. “ASignaling Explanation for Charity.” Ameri-can Economic Review, 86(4): 1019–28.

Gneezy, Uri. 2003. “The W Effect of Incentives.Unpublished.

Gneezy, Uri, and Aldo Rustichini. 2000a. “A FineIs a Price.” Journal of Legal Studies, 29(1):1–17.

Gneezy, Uri, and Aldo Rustichini. 2000b. “PayEnough or Don’t Pay at All.” Quarterly Jour-nal of Economics, 115(3): 791–810.

Harbaugh, William T. 1998. “What Do Dona-tions Buy? A Model of Philanthropy Basedon Prestige and Warm Glow.” Journal ofPublic Economics, 67(2): 269–84.

Herold, Florian. 2004. “Contractual Incomplete-ness as a Signal of Trust.” Unpublished.

Janssen, Maarten C. W., and Ewa Mendys-Kamphorst. 2004. “The Price of a Price: Onthe Crowding Out and In of Social Norms.”Journal of Economic Behavior and Organi-zation, 55(3): 377–95.

Jewitt, Ian. 2004. “Notes on the Shape of Dis-tributions.” Unpublished.

Johansson-Stenman, Olaf, and Peter Martinsson.2006. “Honestly, Why Are You Driving aBMW?” Journal of Economic Behavior andOrganization, 60(2): 129–46.

Kahneman, Daniel, and Jack L. Knetsch. 1992.“Valuing Public Goods: The Purchase ofMoral Satisfaction.” Journal of Environmen-tal Economics and Management, 22(1): 57–70.

Koszegi, Botond. 2004. “Utility from Anticipa-tion and Personal Equilibrium.” Unpublished.

Lamont, Michele. 2000. The Dignity of WorkingMen: Morality and the Boundaries of Race,Class and Immigration. New York: RussellSage Foundation.

Lazear, Edward P. 2000a. “Performance Payand Productivity.” American Economic Re-view, 90(5): 1346–61.

Lazear, Edward P. 2000b. “Personnel Econom-

ics and Economic Approached to Incentives.”HKCER Letters, 61.

Lepper, Mark, David Greene, and Richard Nis-bett. 1973. “Undermining Children’s Inter-est with Extrinsic Rewards: A Test of the’Overjustification Hypothesis’.” Journal ofPersonality and Social Psychology, 28(1):129 –37.

Mellstrom, Carl, and Magnus Johannesson. 2005.“Crowding Out in Blood Donation: Was Tit-muss Right?” Goetborg University WorkingPapers in Economics180.

Murnighan, J. Kieth, John M. Oesch, andMadan Pillutla. 2001. “Player Types andSelf-Impression Management in DictatorshipGames: Two Experiments.” Games and Eco-nomic Behavior, 37(2): 388–94.

Pesendorfer, Wolfgang. 1995. “Design Innova-tion and Fashion Cycles.” American Eco-nomic Review, 85(4): 771–92.

Potters, Jan, Martin Sefton, and Lise Vesterlund.Forthcoming. “Leading-by-Example and Sig-naling in Voluntary Contribution Games: AnExperimental Study,” Economic Theory.

Prendergast, Canice. 1999. “The Provision ofIncentives in Firms.” Journal of EconomicLiterature, 37(1): 7–63.

Prendergast, Canice, and Lars Stole. 2001. “TheNon-Monetary Nature of Gifts.” EuropeanEconomic Review, 45(10): 1793–1810.

Sadowski, Philipp. 2004. “Overeagerness.”Unpublished.

Seabright, Paul. 2004. “Continuous PreferencesCan Cause Discontinuous Choices: An Ap-plication to the Impact of Incentives on Al-truism.” Institut d’Economie IndustrielleWorking Paper 257.

Smith, Adam. The Theory of Moral Sentiments.1759. Amherst, NY: Prometheus Books.

Suvorov, Anton, and Jeroen van de Ven. 2005.“Discretionary Bonuses as a Feedback Mech-anism.” Unpublished.

Titmuss, Richard. 1970. The Gift Relationship:From Human Blood to Social Policy. Lon-don: Allen and Unwin.

Upton, William. 1973. “Altruism, Attributionand Intrinsic Motivation in the Recruitmentof Blood Donors.” In American Red Cross,Selected Readings in Donor Motivation andRecruitment. Vol. 2I, 7–38. Washington,DC: American Red Cross.


Incentives and Prosocial Behavior - Princeton University - Home

Documents