Optimal Incentive Contracts for Teams

许多现有的激励理论描述的是一种只持续一次交易的关系。除了最近的几篇文章外，这些静态模型的观点促进了对员工个人绩效衡量标准敏感的激励计划，并通过锦标赛或相对绩效评估鼓励员工之间的竞争。这些模型的静态性质不仅不切实际，因为大多数现实世界的组织都有很长的、往往是无限期的生命周期，而且它们的预测似乎与许多(长期的)代理关系不一致。对许多员工的激励往往很弱，公司越来越依赖于“自我管理的团队”，这些团队通过相互问责、团队激励和授权来鼓励成员之间的合作关系。虽然还没有确凿的证据证明它们的好处，但许多知名公司将它们的生产率提高和高业绩记录归功于工作团队的成功运营。最近的几位作者，如Holmstrom、Milgrom和Itoh，展示了通过基于团队的激励来鼓励工人之间合作的好处。然而，这些文件的静态性质没有解释如何实现代理人之间的合作。这个项目通过关注代理关系的重复性质来弥补这一限制。这样做不仅使我们能够将代理之间的合作解释为自我强制行为，而且它还可以对内部组织的各个方面产生更丰富、相互关联的预测。提议的模型中的新元素，而静态模型中没有的，是创造出一种“隐性激励”的可能性。在委托人设计的明确激励的适当激励下，代理人可以被诱使监督和惩罚彼此的逃避。本研究的目的是探索隐性激励如何被显性激励所诱导，并与显性激励相结合。该项目将就按业绩计酬计划的设计、任务分配、权力分配以及沟通的作用和价值等重要问题提供有益的见解。初步研究表明，最优激励方案表现出团队安排的许多可观察特征。具体地说，它预测，寿命长和相互问责的关系的特点可能是权力较低、集体激励和权力高度下放，而寿命较短或没有相互问责的关系的特点是权力强大、竞争激烈的激励机制，如相对业绩评价和锦标赛，以及高度集中控制和监督。这些发现表明，这一理论不仅有助于发展委托代理文献，而且有助于理解动态代理关系中出现的一般激励问题。该项目在确定基于团队的组织的好处方面具有直接应用，但它也将揭示其他机构，如基于团队的贷款计划和日本的分包关系。

Much of the existing theory of incentives describes a relationship that lasts for just one transaction. Except for a few recent articles, the view emerging from these static models promotes incentive schemes that are sensitive to employees' individual performance measures and encourage competition among employees through tournaments or relative performance evaluation. The static nature of these models is not only unrealistic, given that most real world organizations have long, often open-ended, life spans, but their predictions appear to be at odds with many (long-term) agency relationships. Incentives for many employees are often muted, and firms increasingly rely on `self-managed teams` that encourage cooperative relationships among their members through mutual accountability, group incentives and delegated authority. While there has been no firm proof of their benefit, many prominent firms attribute their productivity gains and high performance records to the successful operation of work teams. Several recent authors such as Holmstrom, Milgrom and Itoh demonstrated the benefits of inducing cooperation among workers through team-based incentives. The static nature of these papers leaves unexplained how cooperation among agents can be achieved, however. This project remedies this limitation by focusing on the repeated nature of agency relationships. Doing so not only enables us to explain cooperation among agents as self-enforcing behavior, but it also yields much richer, interrelated predictions on various aspects of internal organization. The new element in the proposed model, absent in the static models, is the possibility of an `implicit incentive` being created. Agents can be induced to monitor and punish one another's shirking, when appropriately motivated by the explicit incentives designed by the principal. The objective of the study is to explore how the implicit incentives can be optimally induced by, and combined with, explicit incentives. This project will provide useful insights on important issues such as the design of pay-for-performance scheme, task assignment, allocation of authority, and the role and value of communication. The preliminary research suggests that the optimal incentive scheme displays many observed features of team arrangements. Specifically, it predicts that relationships with long life spans and mutual accountability are likely to be characterized by low-powered, group incentives and by a high degree of delegation of authority, whereas relationships with short life spans or no mutual accountability are characterized by high-powered, competitive incentive schemes such as relative performance evaluation and tournaments and by a high degree of centralized control and supervision. These findings show a great deal of promise for a theory that will contribute not only to the development of the principal-agent literature but also to the understanding of general incentive issues arising in dynamic agency relationships. The project has direct applications in identifying the benefit of team-based organizations but it will also shed light on other institutions such as group-based lending programs and Japanese subcontracting relationships.