Holding educators more accountable for the academic achievement of their students has been a central feature of recent education reforms. In several prominent instances, accountability schemes have set pecuniary performance targets that condition on prior scores as a means of controlling for student heterogeneity. Yet doing so introduces a potential dynamic distortion in incentives: teachers may be less responsive to the reform today in an effort to avoid more onerous targets in future — an instance of the so-called `ratchet effect.' The main chapters of this thesis consider possible dynamic distortions from three distinct but related vantage points. Chapter 2 builds a theoretical foundation for understanding the origin of ratchet effects in an educational context. Given an environment where school-level targets depend on student prior scores, I show that such dynamic gaming behaviour depends crucially on variation in the grade horizon of students, with teachers distorting their effort less when their decision affects fewer future scores within the same school. Chapter 3 analyzes the ratchet effect from an empirical perspective. Making use of rich educational panel data from North Carolina, I exploit variation in the grade span of schools to identify ratchet effects, finding compelling evidence of dynamic distortions using a difference-in-differences approach. I then directly estimate the structural parameters of the corresponding model, allowing for complementarities in production between teacher effort and student ability. Using these estimates, the grade five score in K-5 schools would be about 1.25 standard deviations lower under a counterfactual setting without any accountability scheme and 4.6% of a standard deviation higher if ratchet effects were eliminated via a prescribed reduction in each school's incentive target. Chapter 4 explores a potentially important class of mechanisms through which the identified dynamic effects might arise. In particular, school principals may re-allocate teachers across classrooms differentially by grade. By examining the yearly change in teacher assignments to tested subjects, I find evidence that principals re-sort higher quality teachers to higher grades and focus more costly re-sorting on lower grades in response to the reform, which is consistent with the overarching dynamic objective and results of the prior chapter.

Holding educators more accountable for the academic achievement of their students has been a central feature of recent education reforms. In several prominent instances, accountability schemes have set pecuniary performance targets that condition on prior scores as a means of controlling for student heterogeneity. Yet doing so introduces a potential dynamic distortion in incentives: teachers may be less responsive to the reform today in an effort to avoid more onerous targets in future — an instance of the so-called `ratchet effect.' The main chapters of this thesis consider possible dynamic distortions from three distinct but related vantage points. Chapter 2 builds a theoretical foundation for understanding the origin of ratchet effects in an educational context. Given an environment where school-level targets depend on student prior scores, I show that such dynamic gaming behaviour depends crucially on variation in the grade horizon of students, with teachers distorting their effort less when their decision affects fewer future scores within the same school. Chapter 3 analyzes the ratchet effect from an empirical perspective. Making use of rich educational panel data from North Carolina, I exploit variation in the grade span of schools to identify ratchet effects, finding compelling evidence of dynamic distortions using a difference-in-differences approach. I then directly estimate the structural parameters of the corresponding model, allowing for complementarities in production between teacher effort and student ability. Using these estimates, the grade five score in K-5 schools would be about 1.25 standard deviations lower under a counterfactual setting without any accountability scheme and 4.6% of a standard deviation higher if ratchet effects were eliminated via a prescribed reduction in each school's incentive target. Chapter 4 explores a potentially important class of mechanisms through which the identified dynamic effects might arise. In particular, school principals may re-allocate teachers across classrooms differentially by grade. By examining the yearly change in teacher assignments to tested subjects, I find evidence that principals re-sort higher quality teachers to higher grades and focus more costly re-sorting on lower grades in response to the reform, which is consistent with the overarching dynamic objective and results of the prior chapter.