Content area
Full text
The strength of incentives used in an organization and the productivity of employees that results from these incentives depend to a large degree on the characteristics of the performance measures available to the organization. Some employees work under high-powered explicit incentive contracts, while others have no explicit incentive contracts at all. The ability of agency theory to predict the pattern of incentive provision in organizations has not been very impressive, and the divergence between the prescriptions of agency theory and the actual practice of firms has been widely noted.
This paper (along with both other papers in this session [Edward P. Lazear, 2000; Canice Prendergast, 2000]) attempts to fill some of the gap between theory and practice. I examine the characteristics of performance measures (those data on which explicit incentive contracts are based) to understand how firms use incentive contracting, and to predict the use of incentives in practice.
I. Risk and Distortion in Performance Measures
Agency theory before 1991 was chiefly concerned with the "much studied trade-off between incentives and insurance" (Robert Gibbons, 1998 p. 115). This literature stressed the problem of risk-bearing, and the fact that risk neutral firms would insure risk-averse employees against random variation in output by reducing the slope of their incentive contracts. Beginning with Bengt Holmstrom and Paul Milgrom (1991) and continuing with Baker (1992), Gerald Feltham and Jim Xie (1994), and many others, concern began to focus on another problem: the distortion of incentives that resulted when firms "rewarded for A while hoping for B" (Steven Kerr, 1975 p. 769). The use of distorted performance measures induces gaming, noncooperative behavior, sabotage of coworkers, and generally unintended and dysfunctional consequences of all sorts in organizations.
In a recent paper (Baker, 2000), I develop a simple two-parameter characterization of performance measures that captures both distortion and risk. The model in this paper allows the problems of distortion and risk to be examined simultaneously and reveals and explains many puzzles presented by the design of compensation systems. A key benefit of my model is its transformation of the "multi-tasking problem" from a complex multidimensional effort-choice problem into one that can be described in simple geometric terms, which greatly simplifies the characterization of the distortion in performance measures.
Baker (2000) shows that...