A. Team Incentives and Peer Evaluations
In the last 20 years, more intensive use of teamwork in organizations has aroused interest with regard to the factors affecting the success of teams (Devine et al., 1999; Kozlowski et al., 1999; Lawler, Mohrman, and Ledford 1995). One of the crucial elements in the success of teams relates to the choice of the sharing rule for the joint outcome. On the one hand, paying team workers according to their individual contributions instead of assigning them a fixed share of the joint outcome reduces free-tiding behaviors (Alchian and Demsetz 1972; Holmstrom 1982). (1) On the other hand, rewarding workers according to their individual performance is likely to be costly as individual contributions are typically difficult to evaluate.
In this paper, we analyze the impact on team performance of using peer assessments to share a jointly produced outcome in a laboratory experiment in which three subjects are randomly matched to work on a real-effort task. Subjects had to complete a real team task for which individual effort and abilities determined joint profits. This allowed us to introduce a dimension of merit in the subjects' decisions concerning peer evaluations. We focus on a small-team context in which the different partners are likely to observe each others' levels of effort and could, in principle, use this information to evaluate their team partners' relative contributions. (2) In our setting, subjects use peer evaluations to determine team members' respective shares of the joint outcome. We impose that the allocation rule for the joint outcome is budget balanced so that subjects cannot allocate an amount of money greater than their team output. In the case of budget-balanced allocation rules, one can derive from Holmstrom (1982) that purely self-interested partners will have incentives to lie and undermine the achievements of their coworkers preventing group members from being rewarded according to their relative contribution. (3) As a result, even in a context in which team members are able to assess the contribution of their partners without errors they will have no incentives to offer a balanced and truthful assessment and peer evaluations will affect neither the motivation of team members nor team performance.
However, peer ratings have been used in many disciplines such as engineering, management, and medical sciences as a mechanism with which to gather information about individual contributions (Clark, Davies, and Skeets 2005: Conway et al., 1993; Dochy, Segers, and Sluijsmans 1999; Ramsey and Wenrich 1999; Thomas, Gebo, and Hellmann 1999; Tu and Lu 2005; Van Rosendaal and Jennett 1992), suggesting that reliable information, not accessible to outsiders, can be extracted from such assessments. Our objective is to provide an experimental analysis that helps us assess the impact of such practices on team performance. There are two possible effects at play here. First, if team members effectively report an accurate estimate of the achievements of others, subjects will end up being paid according to a measure of their relative contribution to the joint outcome so that free riding behaviors may be reduced. Second, the use of peer evaluations, by focusing the attention of team partners on each others' contribution may increase peer pressure and reduce free-riding behaviors in teams (Falk and Ichino 2006; Kandel and Lazear 1992). Evidently, peer evaluations are likely to be imperfect as they can be subject to the influence of team partners through politicking activities or be driven by social norms or pure self-interest. However, peer evaluations allow for the implementation of payment schemes that are based, even though imperfectly, on individuals' contributions without the need for external monitoring. In sum, peer ratings can be seen as an inexpensive mechanism to collect information about workers' contributions.
In this paper, we investigate experimentally the effect on team performance of an allocation mechanism based on peer evaluations. …