Achievement effects of individual performance incentives in a teacher merit pay tournament

This paper examines the role of worker performance feedback and measurement precision in the design of incentive pay systems, specifically in the context of an individual teacher value-added merit pay tournament. We first build a model in which workers use proximity to an award threshold to update their beliefs about their own ability, which informs their expected marginal return to effort. The model predicts that effort will be maximized at some point proximal to but above the award threshold in order to maximize the likelihood of winning an award when effort translates into value-added with noise. As the noise in the value-added measure increases, teacher effort becomes less responsive to prior value-added because the value-added score becomes a less reliable measure of ability. Using administrative teacher-student linked data from Houston, we test the central prediction of the model that there should be excess achievement gains near award thresholds. The data strongly reject the existence of such excess gains. We argue that a likely reason for this lack of responsiveness is that the value-added measures used to determine awards were too noisy to provide informative feedback about one's ability.