Do Schools Learn: Response to North Carolina’s School Accountability Program

The paper tests the predictions of a Bayesian model of learning by doing in North Carolina’s school accountability program. The program aims to provide incentives for higher performance by offering monetary rewards if schools meet a specified test score target. Schools are assumed to exert effort based on their expectations of reaching the target. At the end of each year, they can observe how they performed relative to the target and draw inference on the optimal level of effort needed to reach the target. To test the model’s predictions, I use test score data from grades 3 and 4 from before the program started and exploit the fact that, while schools were tested prior to the program, they were not informed of their test score performance. The schools’ difference in performance before and after they receive the new information is partially consistent with learning by doing. In particular, I find that the highest gains in performance come from schools who performed poorly the year before. However, I also find that schools which passed the target the year before experience an additional gain in performance, which rejects the predictions of the learning model . This finding suggests that there may be an additional effect on incentives if schools meet the target, consistent with a decrease in the disutility of effort following a successful year. Finally, I test for the differential effect of the program across different parts of the test score distribution, and find that the schools’ response benefited low achieving students the most. ∗I am particularly grateful to my advisors Andrew Foster, John Tyler and Kenneth Chay for their help and support in this project. I am also grateful to Sriniketh Nagavarapu, Anna Aizer, Pedro Dal Bo, and participants at Brown University’s Applied Micro Lunch Seminars for valuable comments. I also thank Gary Williamson and Lou Fabrizio from the North Carolina Department of Public Instruction for providing additional information on the implementation of the accountability program. Remote access to the data was generously provided by the North Carolina Education Research Data Center (NCERDC). I gratefully acknowledge the financial support received from Brown University Graduate School.