Intrinsic and Extrinsic Motivation Robert Östling Outline

Intrinsic and Extrinsic Motivation Robert Östling

Outline • • • Experimental results Theoretical considerations Bénabou & Tirole (2003, RES) Bénabou & Tirole (2006, AER) Ellingsen & Johannesson (2006) Herold (2005)

Frey and Oberholzer-Gee (1997, AER) • • Question: Tests reaction to monetary compensation for the acceptance of a nuclear waste repository in their community. Setting: Hypothetical referendum voting question was posed to subjects about if they were willing to permit the construction of a nuclear waste repository in their community. The same question was then repeated with a specific monetary compensation for accepting the nuclear waste repository (varied between $2175, $4, 350 and $6, 525 per individual and year in three sub-samples). Everyone who rejected the compensation then received another question with a higher compensation ($3, 263, $6, 525, or $8, 700). Subjects: 305 interviews of randomly selected subjects in two communities in Switzerland (potential sites for a nuclear waste repository). Results: Fraction of acceptance without monetary compensation: 50. 8%. Fraction of acceptance with monetary compensation: 24. 6% (the amount of compensation had no significant effect on the acceptance probability).

Gneezy & Rustichini (2000, QJE) • • • Setting: Subjects were asked to answer a quiz consisting of 50 problems taken out of a psychometric test used to scan university applicants (questions involving reasoning and computation). Treatments: 0 NIS (except show-up), 0. 10 NIS, 1 NIS or 3 NIS per correct answer Subjects: 160 subjects

Gneezy & Rustichini (2000, QJE) • • • Setting: High-school students collected donations from the public to some charity (done in Israel a few days a year). Students divided into pairs of students who collect money together as a team. Treatments: Baseline (speech), baseline + 1% of collected money, baseline + 10% of collected money. Subjects: 180 Israeli high-school students.

Gneezy & Rustichini (2000, JLS) • • • Setting: A study of day-care centers in Israel. The day care centers operates between 7. 30 and 16. 00. Before the study there was no fine if parents came late to pick up their children. Treatments: Control (only record late parents) and treatment (recorded first 4 weeks, then a fine of 10 NIS for late pick-up, removed fine in 17 th week). Subjects: The study was carried out on 10 day-care centers in Israel (center 1 -6 in the test group and center 7 -10 in the control group). Between 28 and 37 children in each day care center.

Fehr & Rockenbach (2003, Nature) • • • Setting: A trust game is played anonymously between two individuals (both endowed with 10 MU). The amount sent is tripled. Investors always state desired back-transfer. Treatments: Trust condition (no fine possible) and incentive condition (4 MU fine if less than stated back-transfer). Subjects: 138 subjects.

Mellström & Johannesson (2005) • • • Setting: Subjects are given the opportunity to become blood donors (complete a health examination consisting of a health declaration and a physical examination at a blood donation center). Subjects are offered different incentive schemes to become a blood donor depending on the experimental treatment. Treatments: No compensation, 50 SEK compensation, a choice between SEK 50 or donating SEK 50 to charity. Subjects: 262 subjects.

Falk & Kosfeld (2006, AER) • • • Setting: Principal has zero endowment, agent has 120 MU endowment. Agent choose how much to transfer to principal knowing that the transer is doubled. The principal has choice whether to restrict the agents choice set or not. Used strategy method. Treatments: C 5: Principal can impose restriction of at lest 5 MU transfer or choose not to do so. C 10: Same but 10 MU as minimum. C 20: Same but 20 MU as minimum. Subjects: 418 subjects.

Theoretical considerations • Extrinsic and intrinsic motivation (e. g. Frey and Oberholzer -Gee 1997): Individuals have an intrinsic value of performing some activities (e. g. helping others, civic duty) and external monetary compensation reduce the intrinsic value. • Incomplete contracts (e. g. Gneezy and Rustichini 2000): The introduction of a monetary payment changes the perception of the incomplete contract (the introduction of a fine for coming late changes the perception of the consequences of frequent delays). • Reciprocity (e. g. Fehr & Rockenbach 2003): The introduction of a fine is viewed as an hostile/unfair act, and is therefore punished by reduced generosity. • Signaling: The decision to introduce a fine or incentive signals some private information on part of the principal. Can incorporate all of the above explanations.

? b ? e B&T ’ 03 Private information about A (or task). Imperfect self-knowledge (from private signal): ”I’m not trusted, so probably I am not trustworthy and won’t work so hard. ” B&T ’ 06 No private information. Private information about own prosociality and greed with reputational concerns: ”The principal trusted me so now I can really show off that I am trustworthy rather than greedy. ” E&J ’ 06 Private information about own expectations about A (resulting from private information about own characteristics). Private information about own characteristics and esteem concerns: ”The principal is not trusting so I don’t need to work so hard to live up to expectations. ” H ’ 05 Private signal about A’s type. Private information about own characteristics (trustworthy/untrustworthy) and uncontractible part of relationship: ”He didn’t trust me, so there is no point in doing something extra outside our contract. ”

Bénabou & Tirole (2003, RES) • A choose e=0 or e=1 at a cost c. Task succeeds with probability θ if undertaken. • P knows c, whereas A only gets a signal σ that is informative about c. • If task succeeds, A gets V>0 and P gets W>0. If it fails, both get zero. • P chooses whether to reward effort with a bonus b≤W or not. • If A knew c, he would exert effort only if θ(V+b)≥c. • With two-sided private information we look for a PBE of the two-stage game.

Bénabou & Tirole (2003, RES) • Reminder: A PBE is a strategy profile and belief system such that strategies are sequentially rational at each node of the game tree and beliefs are consistent. • At the second stage A forms an expectation E(c|b, σ) and exerts effort only if θ(V+b)≥ E(c|b, σ) which implies that there is a threshold σ* for the agent to exert effort. • P choose b in order to maximize expected payoff: Eσ[UP]= θ[1 -G(σ*(b)|c))[W-b]. • Ignoring degenerate equilibria.

Proposition 1 (B&T ’ 03) • Bonus is a short term reinforcer: if b 2>b 1, then σ*(b 1)>σ*(b 2). • Bonus is bad news, if c 2>c 1 then b 2≥b 1. • Bonus undermine A’s perception about task attractivness: if b 1<b 2, then E(c|b 1, σ) < E(c|b 2, σ). • Conclusion: Extrinsic motivation can crowd out intrinsic motivation in the long term.

Bénabou & Tirole (2006, AER) • P has no private information, only A knows his identity (ve and vb). A engages in social signaling. • Agent effort level e to maximize the following: (ve+vbb)e – C(e) + µe. E(ve|e, b) – µb. E(vb|e, b). Intrinsic immaterial valuation (prosociality) Intrinsic material valuation (greed) Material cost of effort Reputational effect of observer’s expectation of prosociality Reputational effect of observer’s expectation of greed

Bénabou & Tirole (2006, AER) The first order condition is: C’(e) = ve+vbb + µe∂E(ve|e, b)/∂e – µb∂E(vb|e, b)/∂e. 1. 2. Observing e reveals the sum of three marginal motivations (intrinsic, extrinsic and reputational). Higher b increases informativeness of vb and reduces it of ve. Assume µe and µb are identical and known to all.

vb b=0 + Potential crowding out! b>0 without reputation b>0 with reputation − C’(e)-r(e, 0)=ve* C’(e)-r(e, b)=ve** Reputational effects: 1. New contributors have lower ve which may diminish the reputational effect of e. 2. New contributors are greedy, i. e. have higher vb, which reduces the reputational effect of e. ve

Bénabou & Tirole (2006, AER) • Interpretation of Mellström & Johannesson (2005): Women may not be more prosocial, but care more about appearing prosocial! • Argument does not rely on private information of P, but requries multidimensional characteristics. • Implies that P can be replaced by computer or experimenter (blood donation okay, but what about workplace relationships? )

Pride and Prejudice (E&J ’ 06) • P private information about expectation of A’s type. • A private information about own type, but cares intrinsically about P’s expectation (esteem). Depending on the situation, esteem from P might depend on P’s type (salience weights). • Requires heterogeneity among both P and A. • Heterogeneity of P’s expectations violates the assumption of common priors. • So first let us derive these beliefs from common “metaprior” in a simple two-type case.

Agreeing to Disagree (E&J ’ 06) • • Two types: θH>θL Two distributions: P(θH|G)=h>P(θL|B)=l. Metaprior of θH: h/(h+l) Prior for distribution of θH type: P(G|θH)=h/(h+l)>1/2 Prior for distribution of θL type: P(G|θL) = (1−h) / (2−h−l)<1/2 Calculating the prior for θH given θH and θL shows that a high type attaches a higher probability to θH than a low type. Common metapriors and type heterogeneity delivers different priors A high type is optimistic, a low type is pessimistic.

Ellingsen & Johannesson (2006) • • Crowding out: High-powered incentives may signal low expectations, implying that A does not need to work so hard to get P’s esteem. This argument requires that esteem is worth less from a pessimistic, low type, P. Gift exchange: If esteem from a high type, optimistic P is worth more to A, then P may signal he is a high type by paying a higher wage and A will work harder to get the esteem benefit Hidden cost of control: Exogenous imposition does not matter (since it does not carry expectations). However, a controlling P is a relief for A since esteem becomes cheaper – asymmetric information may consist of different beliefs about what is appropriate behavior in a situation. Social salience channel: Similar effects might arise from different views of what appropriate behavior is. P’s actions might signal what he believes is appropriate and A will be interested in this because it signals how to achieve esteem. The above mechanisms can be restated in this way, but it more generally implies that esteem-seeking agents are sensitive to cues about appropriate behavior. May explain how different types of rewards (e. g. small bonus, gold star) can be important, but more generally about framing (PD as Wall Street Game vs Community Game).

Example: Trust Game (E&J ’ 06)

Example: Trust Game (E&J ’ 06) • Relevant characteristic: Fehr-Schmidt inequality aversion. • One equilibrium for “reasonable” parameters: Only high type P 1 are trusting, and high type P 2 rewards trust only in the voluntary trust game. • Idea: Esteem more valuable from a high type P 1 and it is only in the voluntary trust game that P 2 can be sure that P 1 is a high type. • Key assumption: Low type esteem for P 2 be sufficiently low compared to high type esteem for P 2.

Into the future… • Relation to psychological game theory • Bounded rationality (e. g. Bayesian updating) • More than two players (too complex? ) Bye, bye! • Framing, social preferences, symbolic interactionism, meaning et c.