آیا پاداش بیشتر به تلاش کمتر منجر می شود؟ انگیزه معکوس در گروه ها

Conventional wisdom suggests that a global increase in monetary rewards should induce agents to exert higher effort. In this paper we demonstrate that this may not hold in team settings. In the context of sequential team production with positive externalities between agents, incentive reversal might occur, i.e., an increase in monetary rewards (either because bonuses increase or effort costs decrease) may induce agents that are fully rational, self-centered money maximizers to exert lower effort in the completion of a joint task. Incentive reversal happens when increasing one agent's individual rewards alters her best-response function and, as a result, removes other agents’ incentives to exert effort as their contributions are no longer required to incentivize the first agent. Herein we discuss this seemingly paradoxical phenomenon and report on two experiments that provide supportive evidence.

Most economists would presumably agree to the statement that, basically, economics is all about incentives.1 The statement is regularly understood to be about monetary payments, in the sense that high monetary rewards equal strong incentives, and vice versa. This simplification applies to many economic situations. However, it does not necessarily apply to environments in which individuals interact in groups and their individual rewards are affected by others’ actions, e.g., in team production settings. In the context of sequential team production, incentive reversal might occur – in particular for rational individuals whose main objective is the maximization of their own monetary income. In this paper, we illustrate under which circumstances this might happen and report corresponding experimental results. Following Winter (2009), who introduced the theoretical foundations for incentive reversal, we consider simple strategic environments involving team production with moral hazard. In this context, incentive reversal refers to situations in which an increase of promised rewards to all team members results in fewer agents exerting effort. Incentive reversal is caused by the existence of externalities among peers that arise from the team's production technology, and builds on two properties that are descriptive of many team environments: (i) some agents have internal information about the effort level of others (which requires a certain extent of sequencing in the production process) and (ii) agents’ efforts are complements in the team's production technology. Given these assumptions, the line of reasoning behind incentive reversal is straightforward. Since the underlying production technology involves complementarity in terms of team members’ efforts, moderate rewards can generate an implicit threat against shirking, in the sense that agent i chooses to exert effort only if his peer, agent j (whose effort is observable by i) has done so as well. A substantial increase to agent i's rewards may induce this agent to exert effort as a dominant strategy (regardless of what agent j is doing). This in turn eliminates the implicit threat that was present in the outset and induces agent j to shirk even though his promised reward increased as well. Simple as it may seem, it is not clear whether the argument for incentive reversal is empirically sound for at least two reasons: cognitive limitations and other-regarding preferences. We tend to think about monetary rewards and motivation as moving in the same direction. Thus, when the rewards of all agents in a team are increased, they may respond “heuristically” with high effort to the increase in their own reward, without considering the strategic implications of the increase in their peers’ rewards. Such heuristic responses might be facilitated if individuals are not able, or expect others not to be able, to follow the backward induction reasoning underlying incentive reversal.2 Even if cognitive limitations do not apply, other-regarding preferences (and in particular the presence of reciprocity) may eliminate incentive reversal. If an individual who detects the shirking of his peer is inclined to retaliate by shirking as well, regardless of the monetary incentives, the observed individual (anticipating reciprocal behavior) would be reluctant to shirk. In this event, incentive reversal will be washed out.3 Given the above considerations, whether incentive reversal in teams actually occurs or not ultimately remains an empirical question. Moreover, theoretical predictions strongly rely on having sufficiently precise knowledge about the shape of the production technology, the move structure and information set of each agent, as well as the potential rewards and individuals’ costs of exerting effort. We conducted two separate experiments that allowed a sufficient degree of control over these factors to clearly test for incentive reversals. Both experiments involve teams of agents who work on a joint team project. Agents decide on their individual effort level (with effort being costly) and are paid as a function of the team's joint effort. In both experiments we look at situations that are susceptible to incentive reversal. Keeping the environment (in particular the production technology) fixed, we explore how subjects behave under high, respectively under low rewards. The two experiments differ in several respects, allowing us to establish the behavioral validity of the incentive reversal phenomenon across specific features of the decision environment. The first experiment implemented a two-agent game in a laboratory setting using a labor framing. Subjects in this experiment were provided with the explicit payoff structure of the game and second movers made their decisions after observing the decisions of first movers. The level of incentives was manipulated within subjects between rounds by setting different reward levels, as in Winter (2009). The second experiment implemented a three-agent game in a classroom environment. The game was presented as a money game, the payoff structure of which was not explicitly provided, but could be extrapolated from the basic rules. All decisions were collected simultaneously, with second- and third-mover strategies conditional on previous movers’ decisions. The incentive level manipulation was implemented by varying the costs of effort, rather than the rewards, in a between-subjects design. Put together, the two experiments provide a robust test for the existence of incentive reversal. In order to be able to ascertain that any observed incentive reversal effect is indeed driven by the hypothesized mechanism, we take two complementary approaches. In the first experiment, we add two control treatments that correspond to the experimental treatments in all but one aspect: the subjects choose their actions simultaneously rather than sequentially. Thus, while we retain the payoff structure, the strategic structure which gives rise to incentive reversal in the sequential games is eliminated in the simultaneous games. In the second experiment, we use a strategy method instead of a direct-response method to obtain counterfactual data. By observing subjects’ decisions in each node of the game tree we can test for incentive reversal by looking at behavior along and off the theoretical equilibrium path. Our experimental data provide clear evidence that incentive reversal in teams can occur. When the comparative-static predictions of the theoretical analysis predict incentive reversal, we do observe it. In the first experiment, increasing the second-mover's rewards has the negative effect of reducing the first-mover's incentive to exert effort as this agent chooses to free-ride on the second-mover's effort – but only in sequential games and not in simultaneous games. These behavioral patterns are indeed observed: the average effort provided by the first-movers drops by almost 50% when rewards are increased under the sequential protocol, whereas the average effort stays constant in the simultaneous protocol. Incentive reversal is observed in our second experiment as well. The average team output is significantly higher under high costs than under low costs. For example, first-movers’ average effort is increased by almost 130% when costs are increased. Moreover, subjects’ subsequent choices along the equilibrium path are well in line with the predictions from incentive reversal. Interestingly, this holds true even though we observe reciprocal behavior in both treatments, which underlines the relative importance of incentive reversal in such an environment.

In this paper we report on two experiments designed to directly test for incentive reversal – the seemingly paradoxical inverse relationship between monetary rewards and incentives. Our results provide strong support for the emergence of incentive reversal. In particular, we observe in both experiments that when rewards increase (or costs decrease), late movers become less responsive to the observed history. Under these circumstances, early movers seem to anticipate that late movers will always exert effort. As a consequence, first movers shirk and free ride on the effort of late movers. Although our lab experiments establish the empirical validity of incentive reversal, they abstract from features that are potentially important for team environments in the field, but are outside the scope of the theoretical model in Winter (2009). Most importantly, we did not allow for learning and reputation building, which come with repeated interactions. Our experimental results on reciprocal behavior suggest that incentive reversal depends on one-shot interactions and might not persist in a stable environment that allows for repeated interactions and concerns about reputations. In Experiment 1, we find that the optimal actions of first movers given the observed choices made by second movers do not entail incentive reversal. With repetition, first movers would be able to adapt to the behavior of second movers, thus eliminating incentive reversal. In particular, second movers in the high rewards treatment were willing to forgo part of their payoff in order to punish shirking by first movers. This behavior is likely to be strengthened if it can serve to build a ‘tough’ reputation, leading to high efforts by first movers regardless of the reward level. In Experiment 2, a similar pattern emerges. The fact that participants who choose reciprocal strategies as late movers do not anticipate reciprocity from others when deciding as first movers suggests that, with learning, behavior could converge to high effort choices even with low costs. Our findings complement the existing literature studying the impact of monetary rewards on individuals’ behavior. There is substantial evidence based on laboratory and field experiments showing that individuals’ willingness to exert effort may not monotonically increase with monetary rewards. For example, parents’ late pickup at daycare centers turns more severe after imposing a fine on late arrival, and scouts’ performance in door-to-door collection of donations deteriorates when these children are offered to keep a share of the raised donations for themselves (Gneezy and Rustichini, 2000a and Gneezy and Rustichini, 2000b). Similarly, opting to fine untrustworthy behavior actually increases such behavior (Fehr and List, 2004 and Houser et al., 2008). These results, however, build on the behavioral dissonance between intrinsic and extrinsic motivations (see also Bowles, 2008 and Bowles, 2009 for brief overviews or Frey and Jegen, 2001 for a comprehensive survey of empirical evidence for motivation crowding-out). Whereas in the articles above it is the absence of money-maximizing individuals that causes incentives to ‘backfire’, the incentive reversal phenomenon described in our paper is due to the presence of rational, self-centered, money-maximizing individuals. Another difference is that the crowding-out literature highlights the potential adverse effects of small monetary rewards, which crowd out intrinsic motivation but are not large enough to substitute for it. The phenomenon studied here, in comparison, emerges when potential rewards are large enough to eliminate the role of actions taken by previous movers. Along these lines, there exist also some closely related studies that analyze dysfunctional behavioral responses without relying on the discrepancy between intrinsic and extrinsic rewards. For example, Camerer et al. (1997) find a negative elasticity of New York City cabdrivers’ number of working hours with respect to realized earnings per hour. They argue that this is due to income effects, i.e., drivers having daily income targets (but see also Farber, 2008 and Crawford and Meng, 2011). Another example would be Fehr and Schmidt (2004), who demonstrate that in an environment with multidimensional effort where only one effort dimension is contractible, piece-rate contracts are outperformed by fixed-wage contracts. In contrast to our work, these studies usually focus on individual decision problems rather than on team relationships. Moreover, they put forward different reasons for the occurrence of incentive reversal. In essence, incentive reversal in teams is a manifestation of second (or higher) degree incentives. It follows a long tradition in game theory by highlighting the fact that individuals respond not only to direct incentives but also take into account the incentives of others with whom they interact. As such, potential implications of incentive reversal go beyond the workplace and the labor market. It applies to a variety of team environments and suggests that increasing all team members’ stakes in the success of the joint activity may (though not necessarily shall) be counter effective. Political campaigns, commercial ventures, fundraising, joint decisions of committees and allocation in public–private partnerships (Athias and Soubeyran, 2013) are all relevant environments in which incentive reversal may emerge. While incentive reversal is a “rational” phenomenon, our findings also have “behavioral” implications. Substantial experimental and empirical evidence reveals the role of reciprocity in teams (e.g., Ichino and Maggi, 2000, Fehr and Fischbacher, 2003, Falk and Ichino, 2006, Gould and Winter, 2009 and Mas and Moretti, 2009). Team members are psychologically reluctant to exert effort or contribute when they detect shirking by their peers. This reluctance is, in fact, very important for the functioning of teams, as it generates an implicit threat against shirking. Our findings about incentive reversal suggest that high-powered monetary incentives may be counter-effective as they may jeopardize the credibility of this implicit threat. We believe that the findings reported here are of interest for theorists and practitioners alike. They show that even the well intentioned introduction of (additional) rewards may occasionally backfire. For example, granting a pay rise to the workforce or offering job-training opportunities that reduce workers’ effort costs might not always lead to an increase in performance (with the caveat regarding repeated interactions discussed above). On the contrary, actions that are meant to motivate workers may actually lead to incentive reversal – resulting in an effort reduction and higher costs to the principal. While this possibility depends on the exact characteristic of the environment at hand, principals should be aware of it and take it into account when designing contracts.