«Ulrike Malmendier; Klaus M. Schmidt: You Owe Me Munich Discussion Paper No. 2012-30 Department of Economics University of Munich Volkswirtschaftliche ...»
Ulrike Malmendier; Klaus M. Schmidt:
You Owe Me
Munich Discussion Paper No. 2012-30
Department of Economics
University of Munich
Online at http://epub.ub.uni-muenchen.de/14241/
You Owe Me*
Ulrike Malmendiera) and Klaus M. Schmidtb)
November 8, 2012
In many cultures and industries gifts are given in order to influence the recipient,
often at the expense of a third party. Examples include business gifts of firms and lobbyists. In a series of experiments, we show that, even without incentive or informational effects, small gifts strongly influence the recipient’s behavior in favor of the gift giver, in particular when a third party bears the cost. Subjects are well aware that the gift is given to influence their behavior but reciprocate nevertheless.
Withholding the gift triggers a strong negative response. These findings are inconsistent with the most prominent models of social preferences. We propose an extension of existing theories to capture the observed behavior by endogenizing the “reference group” to whom social preferences are applied. We also show that disclosure and size limits are not effective in reducing the effect of gifts, consistent with our model. Financial incentives ameliorate the effect of the gift but backfire when available but not provided.
Keywords: Gift exchange; externalities; lobbyism; corruption; reciprocity; social preferences.
JEL: C91, D73, I11.
*) We would like to thank Yan Chen, Rachel Croson, Robert Dur, Ernst Fehr, Rudolf Kerschbamer, Matthias Kräkel, George Loewenstein, Cristina Nistor and seminar participants in the 2011 ASSA meetings, the Third Behavioral Economics Annual Meeting (BEAM) and at the universities of Aix-Marseilles, Berkeley, Emory, Innsbruck, Maastricht, MIT, Munich, Pennsylvania, Princeton, and Toulouse for comments and suggestions. Leonie Giessing, Jana Jarecki, Sara Kwasnick, Eddie Ning, and Slobodan Sudaric provided excellent research assistance. Financial support by Deutsche Forschungsgemeinschaft through SFB-TR 15, the Excellence Initiative of the German government, and the Alfred P. Sloan Foundation is gratefully acknowledged.
a) Ulrike Malmendier, Department of Economics, University of California at Berkeley, 501 Evans Hall, Berkeley, CA 94720-3880, USA, email : firstname.lastname@example.org b) Klaus M. Schmidt, Department of Economics, University of Munich, Ludwigstrasse 28, D-80539 Munich, Germany, email: klaus.schmidt@LMU.de 1 Introduction Social preferences influence individual behavior in many economically important settings, including bargaining, incentive provision at the workplace, cooperation and competition in small groups, and donations to charity (Camerer, 2003; Fehr and Gächter, 2000; Sobel 2005). A prominent example is gift exchange. Following Akerlof’s (1982) seminal paper on gift exchange in labor markets, a large experimental literature has shown that gifts induce cooperation, both in the laboratory and in the field (see, e.g., Fehr, Kirchsteiger and Riedl, 1993; Gneezy and List, 2006;
Falk, 2007). We also have a well-developed body of social-preference theories that explain the observed gift-exchange behavior, such as intention-based reciprocity (Rabin, 1993) or inequity aversion (Fehr and Schmidt, 1999).
Much less attention has been devoted to the dark side of pro-social behavior, the presence of negative externalities. Pro-social behavior towards one person may come at the expense of another person. In the case of gift exchange, the purpose of the gift is to influence behavior, often to the detriment of a third party. Consider, for example, the wide-spread use of business gifts. In a typical scenario, a procurement manager receives gifts (ranging from small “tokens of appreciation” such as pens or coffee mugs to precious bottles of wine or event tickets) from a supplier, who hopes to get favorable treatment relative to his competitors, even if his competitors offer better or cheaper products. 1 Similarly, politicians and regulators receive gifts or campaign contributions from lobbyists trying to affect their decisions in favor of special interest groups. In both examples the recipient of the gift makes a decision on behalf of a “client” who is often anonymous: the shareholders of the procurement manager and the general public. Such practices have raised concerns – and stirred a regulatory debate – about the influence of gifts. Gift giving has been blamed as a major contributor to weak corporate governance, to the dramatic rise of health care costs, and to wasteful pork barrel politics. 2 Nevertheless, these externalities have been largely ignored in the theoretical and experimental literature.
In this paper we use a controlled laboratory study to explore why and to what extent gifts An extreme example is the pharmaceutical industry that has been estimated to spend USD 8,000-15,000 per year on each physician in the US for marketing, including luxurious dinners, conferences at attractive locations, and generous honoraria (Blumenthal, 2004, p. 1885).
See e.g. Katz et al (2003), Blumenthal (2004), Susman (2008). Policy initiatives addressing these problems range from voluntary codes of conduct (see e.g. Murphy  on corporate ethics statements of large U.S. corporations and Grande  on self-regulation in the pharmaceutical industry) to regulatory reforms and laws limiting the possibilities for gift giving and requiring disclosure, such as the Lobbying Disclosure Act of 1995 and the Honest Leadership and Open Government Act of 2007 in the US.
with negative externalities are effective – and what can be done to mitigate their effects. Standard economic theory can explain the effectiveness of gifts in the case of a repeated relationship or if the gift has informational content. For example, a physician may prescribe more drugs of a pharmaceutical company after attending a conference sponsored by that company because he wants to get more sponsoring in the future or because of scientific information provided at the conference.
The main contribution of our paper is to show that there is an additional and powerful effect of the gift per se. Subjects reciprocate to (small) gifts even if there are no monetary incentives for doing so and if the gift does not convey positive information about the product. The laboratory setting allows us to exclude future interaction, informational content, or any (other) monetary incentives as explanations for such a response. We show a significant effect even for small-scale gifts that amount to little compared to the income of the recipient. We also find that the effect is significantly stronger when it comes at the expense of a third party, compared to the classic giftexchange situation with two parties.
We consider a situation in which a decision maker has to make a decision on behalf of a client. Before making the decision he may receive a gift from an interested party. We show that existing theories, including theories of social preferences, are inconsistent with the recipient’s response to the gift. In fact, the parameters of the experiment have been chosen such that most theories, including altruism, inequality aversion, maximin preferences, and various theories of reciprocity, predict that the recipient does not favor the gift giver at the expense of a third person.
Nevertheless, we find that the effect of gifts is large. It is also significantly stronger than in the classic gift-exchange game without a third party. Furthermore, we observe that the gift is given with the intention to affect the decision of the recipient at the expense of a third party. The recipients are fully aware of this intention but reciprocate nevertheless.
In our experiment, a decision maker has to buy one of two possible products on behalf of a client. The products are simple 50/50 lotteries. The decision maker is instructed to choose the product that is best for her client and is paid a fixed wage, independent of her choice. Before she makes the decision she may receive a small gift from one of the two producers. The gift is given unconditionally and before the producers learn the payoffs of their products so that the gift cannot contain any information about the quality of the product. The setting is anonymous, and players are re-matched after each round. Hence, the gift does not provide any monetary incentives to favor the gift giver. Moreover, in the main treatment (Gift Treatment), neither client nor producers find out which product the decision maker picks. Nevertheless, gifts strongly affect behavior.
After sending a gift, the gift giver’s product is picked twice as often than in a Baseline Treatment in which there is no possibility of gift giving. Even if the product of the gift giver is first-order stochastically dominated by the other product (and its expected value is much lower), almost 50 percent of decision makers choose the product of the gift giver, compared to less than 10 percent in the Baseline. The Baseline Treatment also reveals that decision makers have no problem figuring out what the best product for the client is – their choices coincide with the product the clients prefer in 92 percent of all cases. Hence, the distortion of behavior must be due to the gift.
We also find that not giving the gift has a strong effect. If a producer could have given a gift but chose not to do so, the decision maker often punishes him by refusing to buy his product.
Even if the product is the better product (higher expected value), not giving the gift reduces the likelihood that it is chosen from more than 90 percent in a setting without gifts to less than 60 percent if a gift could have been given but the potential gift giver chose not to do so.
Our experimental design allows us to test whether decision makers are aware how strongly the gift affects their behavior. This question is much debated in practice. For example, a questionnaire study by Steinman et al. (2001) found that only 39 percent of medical residents believe that gifts from pharmaceutical companies affect their prescription behavior, but 84 percent believe that other physicians are influenced. In our experiment, we asked decision makers at the end to estimate how often their decisions coincided with the client’s preferences, and used a quadratic scoring rule to induce unbiased estimates. On average, their estimates are highly accurate. However, as in Steinman et al (2001), they believe that other decision makers are more strongly affected by the gift than they are.
We then compare the results to a situation without third parties. In the No Externality Treatment, there is no client; the decision maker buys the product for herself. We test whether the decision maker reciprocates to the same extent as she does when acting for a client. In that case, gift giving would not reduce efficiency but only redistribute income. We find, however, that the effect of the gift becomes significantly smaller. In particular, when the product of the gift giver is much worse than that of his competitor the effect of the gift vanishes completely.
The most prominent economic theories of other-regarding behavior cannot explain the observed phenomena. Outcome-based theories of social preferences (e.g. altruism or inequity aversion) fail to predict an effect of the gift since favoring one producer harms another producer by the same amount and, in addition, a third person (the client). Theories of type-based or intentionbased reciprocity assume that actions affect social preferences by signaling the “type” or the “intentions” of the gift giver. However, in our experiments gift giving sends a negative signal and should not be reciprocated according to these theories.
How can we explain the observed behavior? Our questionnaire evidence suggests that a gift triggers an obligation to repay the gift, independently of the intentions of the gift giver and the distributional consequences. The gift seems to create a special bond between the giver and the recipient, in line with a large anthropological literature documenting that gifts create obligations.
Similarly, sociologists argue that many forms of social exchange are based on a universal social norm that gifts have to be reciprocated.
In this paper, we propose to extend existing models of social preferences by endogenizing the reference group. The weight that individual i attaches to the welfare of individual j depends on the actions of j that affect i, relative to the expected behavior of j. A favorable act such as giving a gift strengthens the bond between the giver and the recipient, and the recipient will reciprocate. The more the favorable act exceeds expectation, the stronger the positive response. This simple model is consistent with the observations in our experiments, including the fact that decision makers punish the potential gift giver for not giving the gift.
Finally, we conduct several policy treatments to evaluate how the problem of gift giving can be mitigated. Most remedies that have been proposed, and sometimes been implemented, fall in two broad categories, disclosure and size limits. In our experimental set-up, we find that disclosure on its own has no effect on behavior. When we inform clients which producer is the potential gift giver, whether he does give the gift, and which product the decision maker chooses, decision makers’ behavior remains very similar even though they know that everything is disclosed. This finding is consistent with our proposed model. If recipients reciprocate because the gift has created (or strengthened) a bond, then disclosure should have no “shaming” effect.
Varying the size of the gift, we find that larger gifts have less of an impact. When the gift is three times as large, decision makers favor the gift giver in 50 percent of all cases, compared to 68 percent before.
This may be surprising at first glance, and is contrary to the logic of size limits, but it is consistent with our theory of social preferences with an endogenous reference group:
In our set-up decision makers have higher expectations that the producer will send the gift if it benefits them more, and the reward for favorable acts is smaller if they are expected.