Principal Incentives and Student Success: Evaluating Pay-for-Performance Policies

School principals play a pivotal role managing teachers and students. Yet, unlike managers in the private sector, principal salary is typically tied to education and experience rather than measures of school performance. This paper studies the effects of a principal incentive pay that was introduced in North Carolina public schools in 2017 on principal effort, teacher effectiveness (measured by value-added) and student test scores. First, we use latent factor analysis applied to principal self-report data on hours worked as well as teacher reports about school leadership. Second, we estimate test score models to measure teacher value-added (VA), allowing effects to depend on whether the performance pay program was in effect. Third, we explore the relationship between estimated teacher VA and observable teacher, principal and school characteristics. Fourth, we use a difference-in-difference strategy to estimate causal effects of the policy on principal effort, comparing changes over time in public schools to changes in charter schools that did not institute the policy. Fifth, we develop and estimate a Stackelberg model of principal and teacher effort decisions, where principal effort reduces the effort cost for teachers, and student knowledge accumulation depends, in part, on the equilibrium effort allocations. The model allows teachers to heterogeneously respond to principal effort as well as principal and teacher heterogeneity in preferences over student knowledge. The estimated model provides insight into the mechanisms underlying responses to principal pay-for-performance schemes. The model will also be used to simulate the effects of several hypothetical pay policies: one rewarding principals with more challenging student populations, one awarding retention bonuses based on tenure within a particular school rather than overall experience, varying the level of the incentive pay, and adjusting the state's calculation that determines principal performance pay to depend also on student and school characteristics rather than only test score history. The empirical analysis is based on rich administrative data to which we have recently obtained access from the North Carolina Education Research Data Center as well as linked survey data from the North Carolina Teacher Working Conditions Survey.