Performance evaluations for workers are typically subjective impressions held by supervisors rather than easily quantifiable measures of output. We argue that perhaps the most important aspect this is that it gives supervisors the opportunity to exercise their personal preference towards their employees in a way that is detrimental for performance. both for incentive reasons and through misallocation of workers to jobs. We illustrate that firms will respond to this problem in a number of ways. First, they will make compensation less sensitive to performance, even when workers are risk neutral. Furthermore, firms will typically use bureaucratic procedures for allocating rewards, even though these are known to be ex post inefficient. In addition, firms may tie wages to jobs as a means of credibly rewarding the best performers. These organizational changes are used because directly monitoring supervisors' decisions is fraught with problems, among them the creation of 'yes men,' so that the indirect mechanisms described above are likely to be optimal responses to favoritism.
