Relative Performance Evaluation and Project Selection

Abstract

Relative performance evaluation (RPE) occurs when one person's compensation depends on the relation of his output to that of others. Agency theory (e.g., Lazear and Rosen [1981] and Holmstrom [1979; 1982]) predicts that managerial compensation will use relative performance when other firms' outputs reveal information about the manager's actions which is not knowable from the manager's output alone. Tests of the theory measure other firms' outputs using the performance of other firms in the same industry or some macroeconomic measures of performance. These two approaches are designed to separate industry and economy effects from the effects of the manager's actions. Empirical tests by Antle and Smith [1986] and Janakiraman, Lambert, and Larcker [1992] found limited evidence that managerial compensation is adjusted for industry effects; Gibbons and Murphy [1990] found some adjustment in CEO compensation for industry performance, and much more adjustment for changes in market-wide rates of return. This paper provides an explanation for these empirical results by illustrating how the benefits of RPE-based schemes depend on the absence of project choice decisions in the canonical agency model.1