Effects of Accounting-Method Choices on Subjective Performance-Measure Weighting: Experimental Evidence on Precision and Error Covariance

Abstract

Performance-measure weights for incentive compensation are often determined subjectively. Determining these weights is a cognitively difficult task, and archival research shows that observed performance-measure weights are only partially consistent with the predictions of agency theory. Ittner, Larcker and Meyer (2003) have concluded that psychology theory can help to explain such inconsistencies. In an experimental setting based on Feltham and Xie (1994), we use psychology theories of reasoning to predict distinctive patterns of similarity and difference between optimal and actual subjective performance-measure weights. The following predictions are supported. First, most individuals in our setting use the measures' error variance (precision) and error covariance to determine performance-measure weights. Second, directional errors in the use of these measurement attributes are relatively frequent, resulting in a mean under-reaction to the accounting-method change. Third, individuals seem insufficiently aware that a change in the accounting for one measure has spillover effects on the optimal weighting of the other measure in a two-measure incentive system, and in consequence, they make performance-measure weighting choices that are likely to result in misallocations of agent effort.