Human Motivation and Social Cooperation:
Experimental and Analytical Foundations†
Ernst Fehr1 and Herbert Gintis2
1Institute for Empirical Research in Economics, University of Zurich, Blumlisalpstrasse 10, CH – 8006 Zurich, Switzerland; email: firstname.lastname@example.org
2Santa Fe Institute and Central European University, 15 Forbes Avenue, Northampton, MA 01060, USA; email: email@example.com
Since Durkheim, sociological explanations of social cooperation emphasize the internalization of values that induce norm compliance. Since Adam Smith, economic explanations of social cooperation emphasize incentives that induce selfish individuals to cooperate. Here we develop a general approach – the Beliefs, Preferences, and Constraints approach – showing that each of the above models is a special case. Our approach is based on evidence indicating that pure Homo Sociologicus and pure Homo Economicus views are wrong. We show that self-regarding and norm-regarding actors coexist and that the available action opportunities determine which of these actor types dominates the aggregate level of social cooperation. Our approach contributes to the solution of long-standing problems, including the problems of social order and collective action, the determinants and consequences of social exchanges, the micro-foundations of emergent aggregate patterns of social interactions, and the measurement of the impact of cultural and economic practices on individuals’ social goals.
†To appear in the Annual Review of Sociology (2007)
Over the last 20 years, prominent sociologists such as Abbott (1998), Boudon (2003), and Coleman (1986) have expressed serious reservations about the explanatory power of sociological theory and research. Boudon (2003), for example, writes: “Skepticism toward sociology has grown over the years. … To many sociologists the state of the discipline is unsatisfactory”. We believe that one important reason for this dissatisfaction is that despite important analytical contributions such as Coleman (1990), Hedstrom (2005), and the work summarized in Macy and Willer (2002) and Hedstrom and Swedberg (1998), sociological theory has not developed a coherent, broadly accepted framework that facilitates cumulative scientific progress and explains the emergent aggregate patterns of social behavior in terms of individuals’ preferences, their beliefs, and the social and economic constraints they face. Nor has sociological research developed a parsimonious, empirically grounded view of the basic motivational driving forces of human behavior, which may be due to the limited role that controlled experiments played in the development of the discipline.
Decades ago, sociologists criticized the “oversocialized conception of man” (Wrong 1961) that played a prominent role in the work of Durkheim (1938) and Parsons (1937). They rightly questioned Homo Sociologicus, a creature who follows prevailing social norms without regard to self-interest. But they did not develop an alternative, empirically grounded, and widely accepted conception of the basic motivational driving forces of humans. This contrasts sharply with the approach taken by mainstream economics which rests on the notion of Homo Economicus, a creature who is rational and purely self-regarding. However, the Homo Economicus approach is also erroneous, as the assumption that humans are exclusively self-regarding has been decisively rejected by the evidence (Camerer 2003; Fehr and Fischbacher 2003; Gintis et al. 2003).1 Thus, while the lack of a model of human social behavior leaves sociology without an anchor, mainstream economics is hitched to the wrong anchor, i.e., adheres to a biased view of human nature.
There is a rich tradition of experimental research in sociology (e.g., Bonacich and Light 1978; Cook and Hegtvedt 1983; Kollock 1998; Yamagishi, Jin and Kyonari 1999); controlled experiments played, in particular, an important role at the boundaries between sociology and social psychology (e.g., Marwell and Ames 1979; Yamagichi 1986; Lawler and Yoon 1993; Molm 1997; Stolte, Fine and Cook 2001; Buskens and Raub 2002; Simpson 2004; Horne 2004; Diekman 2004). However, the experimental method is generally not prominently featured in sociological research, and experimental results are rarely published in leading journals like the American Journal of Sociology and the American Sociological Review. Most empirical work in sociology is based on non-experimental methods; experimental work by sociologists has rarely entered the mainstream of the discipline. This is unfortunate because the experimental method permits cumulative empirical and theoretical progress through the rigorous testing of alternative theories, the establishment of causal relationships between key variables, and the replicability of results. In contrast, non-experimental field data rarely permit a clean discrimination between theories; the inference of causality is typically not possible from this data, and replicability often proves difficult or impossible. The rigor and cumulative knowledge generated in the natural sciences is undoubtedly based upon their ability to conduct controlled experiments. Of course, the experimental method faces particular challenges in the social sciences, but the development of experimentation in psychology, sociology, and economics over the last two decades has clearly demonstrated the huge scientific benefits that accrue from laboratory experiments. Recent years also have witnessed the proliferation of field experiments that enable researchers to observe how experimentally controlled changes in key variables affect individual behavior in a natural social environment (Carpenter, Harrison and List 2005).
During the last two decades, there has been a surge in experimental research within economics. Much of this research has focused on topics of considerable importance for sociology, but it is our impression that this research is not widely known among sociologists. The purpose of this article is to highlight some of the research in this tradition. In addition, we will describe our own approach - the Beliefs, Preferences and Constraints model - which is characterized by a combination of micro-level experimental research and theoretical models informed by the experimental results. We will show, in particular, that one of the classic problems addressed by sociology – the problem of social order – can be better understood through this approach. In addition, we document, based on important work of sociologists such as Homans (1958), Blau (1965), Emerson (1972) and Cook (1987), how our approach can contribute to a better understanding of social exchanges. We show experimentally and in theory how social exchanges are enforced, how they shape trading relationships and competitive markets, how fairness norms shape them, how they limit the impact of supply and demand forces on price formation, and how they affect the distribution of the gains from exchange.
In the final section of our paper, we point out that experimental tools are critical for answering one of sociology’s deepest questions: To what extent does society shape individuals’ preferences, and how does it do so? Perhaps the foremost feature distinguishing sociology from the other social sciences consists in sociologists’ emphasis on the role of socialization in general, and of the internalization of norms in particular, in constituting human behavior. However, we cannot acquire reliable knowledge about the effects of socialization without carefully controlled experiments. Questionnaires, surveys, and observation of behaviors in the field do not enable the researcher to make reliable inferences concerning motivation. In particular, one can attribute a self-regarding motive to virtually any real-life behavior, however deeply it appears driven by altruistic concerns, because observed prosocial acts can almost invariably be attributed to the selfish motive of acquiring a good reputation. By contrast, in laboratory experiments we can rule out the kind of repeated interactions on which reputational incentives are based.
Our approach rests on a large body of experimental evidence that refutes an important assumption of mainstream economics, namely, that all or most people are exclusively self-regarding (Camerer 2003; Fehr and Fischbacher 2003; Gintis et al. 2003). We define a self-regarding actor to be an agent in a social situation who maximizes his own payoff. A self-regarding actor thus cares about the choices and payoffs to other individuals only insofar as these influence his own payoff. The experimental evidence not only rejects the selfishness assumption routinely made in economics, but also suggests an alternative view about a basic predisposition of humans: strong reciprocity. Strong reciprocity is the behavioral predisposition to cooperate conditionally on others’ cooperation and to punish violations of cooperative norms even at a net cost to the punisher. We will show that a substantial proportion of experimental subjects typically exhibits strongly reciprocal behavior. In addition, the evidence and our theoretical approach suggest that the interaction between strongly reciprocal and self-regarding actors drives the emergent patterns of social cooperation and social exchange. Finally, theory and evidence also indicate that the social structure of interaction plays a decisive role in shaping the emergent aggregate patterns of behavior by affecting how strongly reciprocal and self-regarding individuals interact.
Throughout the paper, the BPC approach in combination with motivational assumptions inferred from laboratory experiments guides our arguments. The BPC model (see Gintis, in press) may be considered as a variant of a game-theoretic approach. It is based on the assumption that people have consistent (transitive) preferences and beliefs about other people’s behavior and about the consequences of their choices. Behavior in this approach can be represented as choices that best satisfy people’s preferences, given their beliefs and the constraints they face. In contrast to mainstream economics, our motivational assumptions are firmly grounded in empirical evidence. This evidence suggests the existence of a heterogeneous population of strong reciprocators and self-regarding individuals.
Subjects face economic incentives in all experiments discussed in this paper: they can earn money – sometimes in substantial amounts. It has been shown, for example, that strongly reciprocal behavior persists even when subjects can earn up to three months’ income in a two hour experiment (Cameron 1999). The experiments also typically implement anonymous interactions between the subjects to rule out reputation effects, and sometimes even full anonymity between the experimenter and the subjects. Many experiments investigating strong reciprocity were first based on a student subject pool, but the results have since been extended to adult samples in advanced industrial societies (Falk 2004), and in a variety of cross cultural contexts (Henrich et al. 2001, 2005, 2006). Moreover, data based on nationally representative samples (Fehr et al. 2002; Bellemare and Kröger 2003) have largely replicated the original experiments with student subject pools.
the Problem of Social Order and Cooperation
The problem of social order goes back to at least Thomas Hobbes who argued that “the life of man (is) solitary, poore, nasty, brutish, and short” in the state of nature. Hobbes concluded that social order is the product of powerful social institutions, including property rights, codified law, and a strong state. Hobbes’ approach in modern times has been strongly espoused by neoclassical economic theory, which has applied general equilibrium and repeated game theory to show that these institutions permit large-scale cooperation among unrelated self-interested individuals. However, in an evolutionary time frame, these order-producing institutions came into place only very recently. Humans had to solve the problem of social order long before they invented and implemented these institutions. In fact, the very existence of these order-producing institutions is itself a result of foregoing social cooperation. We therefore must search for more basic mechanisms that could already generate social order in much simpler societies. Unfortunately, conventional repeated game theory – which is based on the assumption of Homo Economicus – has failed to produce plausible analytical models of social cooperation in a state of nature because these models do not have the required properties of dynamical stability and informational robustness (Gintis 2004)
Of course, theorists working in the sociological tradition have long been skeptical of the neoclassical model of society based on self-interest alone, and are hence not at all surprised at this failure of economic theory. At least since Durkheim (1938), sociologists have referred to the civilizing power of the internalization of social values that restrain individuals’ self-interest. The role of internalized social values in the constitution of social order is most clearly stated in the work of Talcott Parsons (1937). Parsons argued that individuals internalize social values and feel a need for social approval to such a degree that there is little conflict between self-interest and social values, except for a minority of “social deviants”.
The Parsonian solution to social order, however, fails to integrate the mechanisms of norm internalization and the need for social approval into a coherent model of individual choice and social interaction (Wrong, 1961; Gintis 1975). In particular, it neither explains how individuals adjudicate between satisfying personal material needs and social obligations, nor does it clarify the conditions under which individuals accept and reject alternative normative principles. Rather than attempting to repair this lacuna in the Parsonian framework, sociologists generally rejected the approach completely. Sociology is now in the difficult position of embracing the internalization of norms as a fundamental aspect of social life, but without a model of individual behavior to which this concept can be appropriately attached.
Integrating the internalization of norms into decision theory can only be accomplished based on extensive empirical research. The sorts of armchair speculation often found in the discussion of “human nature” simply will not suffice. Fortunately, experimental approaches in psychology (e.g., van Lange 1997, 1999), political science, (e.g., Ostrom, Walker and Gardner 1992), sociology (e.g., Marwell and Ames 1979; Yamagishi 1986; Kollock 1998; Raub and Snijders 1997, Simpson 2004) and economics (e.g., Andreoni 1988; Ledyard 1995) have already planted the seeds for a solution to this problem.
Social order as a public good
How do self-regard and social norms interact to determine individual behavior? Laboratory experiments provide a powerful tool for answering this question. One key aspect of social order can be captured by public goods experiments, in which self-interest and the social good are counter posed with great clarity. In a public goods experiment, each member of a group of N 2 people is endowed with $Y. Each group member can keep this money or invest up to Y into a group account which represents the public good. The experimenter multiplies every dollar invested on the group account by a factor M (which exceeds 1 but is smaller than N, 1 < M < N). When all group members have made their contributions to the group account, and when the experimenter has multiplied this amount by M, the multiplied amount is equally distributed among the group members. Thus, if a subject contributes, say, $5 to the group account, the group as a whole earns 5M from this contribution, and each individual, including the contributor of the $5, receives 5M/N from the group account. Note that because M < N, the investment of 5 exceeds the return of 5M/N, i.e., the investing individual always decreases his or her economic payoff when contributing to the public good. Thus, a self-interested individual will never contribute anything to the public good. However, if all N group members contribute their individual endowment of Y, each individual earn NYM/N = YM. For example, if M = 2 and N = 10, each individual can double his or her income by contributing everything to the public good relative to a situation where everybody keeps the endowment. But each individual also has a strong private incentive to free ride on the contributions of others because every $ invested yields only a private return of $2/10.
The incentive structure captured by the public goods experiment is ubiquitous in reality, and captures goods like public security (i.e., the absence of violence and crime), environmental pollution problems, contract enforcement, and team compensation problems, as well as the enforcement of all kinds of rules that are beneficial for the common good but costly for individuals to obey. If a stable group of individuals has the chance to repeat the experiment several times, and if the group receives feedback about the individual contributions at the end of each repetition, cooperation then typically converges to very low levels [see Figure 1 from Fehr and Gächter (2000) which is based on six groups with N = 4, Y = 20, M = 1.6 and ten repetitions (periods)]. In period ten, roughly 55% of the subjects contribute nothing to the public good and the remaining subjects contribute very little. Thus, it seems that – at least towards the end – individual self-interest largely dominates behavior and the view that internalization of cooperative values in itself suffices to generate social cooperation is clearly wrong. The experimental evidence (Ledyard 1995) strongly suggests that Homo sociologicus, who is socialized to obey the prevailing social order without regard to self-interest, is indeed a rare creature.
However, the first 10 periods presented in Figure 1 tells only half of the story. After period 10, the experimenters implemented the following variation of the experiment. They gave subjects the opportunity to punish each other at the end of each period, after each group member’s contribution in that period was revealed. Each subject could assign punishment points to each of the other group members which decreased both the punisher’s economic payoff and that of the punished subject. Each punishment point assigned to another group member reduced the punisher’s payoff by 1 money unit and the punished subject’s payoff by 10 percent (i.e, by roughly 3 money units). Note that the punishment opportunity should not change the cooperation incentives within a group if all group members are purely self-interested because self-interested subjects will never punish if it is costly – at least towards the final round of the experiment. Therefore, adding this punishment opportunity should not generate stable cooperation if all subjects are only self-interested and internalized values play no role for behavior.
Figure 1: Cooperation in the absence and the presence of private punishment opportunities
The subjects’ actual cooperation behavior shows a radically different pattern. In the presence of the punishment opportunity, behavior converges to almost full cooperation (see squares in Figure 1). In period 20, 83% of the subjects contribute their whole endowment to the public good, and most of the remaining subjects contribute close to the maximum, so that the average cooperation rate reaches almost 100%. Note that exactly the same subjects generated both the behavioral pattern of unraveling cooperation over the first ten periods as well as that of high levels of cooperation in the second ten periods. In fact, there were groups in which every group member contributed nothing to the public good in period 10 and everything in period 20. Several other researchers have documented similar results to Figure 1 in somewhat different settings (Anderson and Putterman 2006, Ostrom et al. 1992, Yamagishi 1986)
How is it possible that the same subjects exhibit such radically different behavioral patterns? Whereas the observations in period 10 suggest that humans are largely self-interested, the observations in period 20 suggest that the Parsonian view of social order has an element of truth because it is hard to explain why people cooperate at almost maximal levels even in round 20 in the absence of internalized social values. The subjects knew that the experiment ends in round 20. Therefore, purely self-interested subjects will never punish in period 20 because this is costly for them and generates no future economic benefits. But if nobody punishes in the final period, why should self-interested subjects then cooperate in this period? In addition, what should we infer from the fact that a cooperation rate of roughly 50% is maintained for the first few periods in the absence of a punishment opportunity?
Social cooperation and STrong reciprocity
The puzzle posed by Figure 1 can be resolved if one takes the fact that the subject population is heterogeneous in specific ways into account. Ample evidence from different types of experiments (Fehr and Fischbacher 2003; Gintis et al. 2003) indicates that a substantial share of the subjects exhibits strong reciprocity. However, the same experiments also show that a large share of subjects is indeed completely self-regarding. Recall that strong reciprocity is a behavioral propensity to cooperate conditionally on other group members’ cooperation and to punish the violations of social norms even though this is costly for the punisher and causes him an economic net loss. Thus, strong reciprocators must have internalized cooperative social values because not only their self-interest shapes their behavior, but conditional cooperation and punishment motives as well. This means that strong reciprocity is not simply long-term, enlightened self-interest because the strong reciprocator cooperates and punishes even in anonymous one-shot interactions where the cooperative and punishing acts obviously reduce his economic net gain. The notion of strong reciprocity also differs from Gouldner’s (1960) concept of reciprocity; in particular, Gouldner’s definition does not include the punishing aspect of strong reciprocity.
A large body of research has documented important motivational forces behind strong reciprocity. Two of the most prominent forces have been termed “reciprocal fairness” (Rabin 1993, Falk and Fischbacher 2006) and “inequity aversion” (Fehr and Schmidt 1999). A reciprocally fair subject is motivated by the desire to respond to kind acts with kindness and to hostile acts with hostility. An inequity averse subject is motivated by the desire to avoid inequity and to implement equitable outcomes. Inequity averse and reciprocally fair subjects are no saints who resist unfair outcomes and punish unfair behavior under all circumstances. Rather, these subjects value equity and reciprocal fairness in addition to their economic self-interest, implying that if the costs of maintaining equity or of reciprocally fair behaviors increase they are less likely to engage in these behaviors (Anderson and Putterman 2006, Carpenter in press). Inequity aversion and reciprocal fairness are also motives that lead to purposeful behavior that can be mathematically represented by the appropriate utility functions. This possibility has the great advantage that the powerful apparatus of modern game theory can be applied to our problem at hand and to many other related problems such as social exchange or competitive market behavior. This enables the development of a unifying model that explains behavior in a wide variety of domains.
Models of inequity aversion (Fehr and Schmidt 1999) and reciprocal fairness (Falk and Fischbacher 2006) can, in particular, explain why we observe little cooperation in the absence of a punishment opportunity and why cooperation flourishes when punishment is possible. In the absence of a punishment opportunity, strongly reciprocal subjects will initially cooperate if they believe that others will also cooperate. However, they notice over time that other group members – the self-regarding ones – free ride. As strong reciprocators are only willing to cooperate if most others also cooperate, they cease to cooperate in the later periods of the experiment. Strong reciprocators also have a desire to punish free riders because they perceive free riding on their cooperation to be unfair. However, stopping cooperation is the only way to punish other group members in the absence of a direct punishment opportunity that enables the subjects to target the punishment on the free riders. Thus, both the free riders and the strong reciprocators contribute little or nothing to the public good towards the end, albeit for different reasons. The self-regarding subjects contribute nothing because this maximizes their economic payoff. The strong reciprocators contribute nothing because they are only willing to cooperate if sufficiently many others also cooperate, and because free riding is the only way available to punish the self-regarding subjects.
The argument presented above shows that the self-regarding subjects ultimately induce the reciprocators to free ride as well in the absence of a direct punishment opportunity. This contrasts sharply with the behavior observed in the presence of a direct punishment opportunity. Here, virtually all subjects ultimately cooperate despite the fact that there is a substantial share of self-regarding subjects. The reason is that the strong reciprocators can now punish the defectors directly, creating an economic incentive for the self-regarding subjects to cooperate. Moreover, the strong reciprocators will also cooperate because they need not fear others’ defection, as the self-regarding individuals are disciplined. Thus, the strong reciprocators induce the self-regarding subjects to cooperate in the presence of a direct punishment opportunity.