Do accountability rules affect public school retention practices? Using a simple model of grade retention, I show that an administrator will retain students differently depending on the accountability ratings criteria he seeks to maximize. I show that an administrator whose school is rated based on student standardized exam passing rates has a strategic incentive to retain borderline students, while an administrator whose school is rated based on year-to-year growth in student standardized exam scores has incentive to retain only the lowest-scoring students. I further show that this effect is most pronounced in the final grade offered by a given school, when promotion of a student ensures her removal from the school’s pool of test-takers . I test the predictions of my framework using a novel dataset of school-grade level retention rates for 7 states in the U.S. and an event study design. I find that about 18% fewer students are retained on average each year when a state adds student growth to the accountability criteria by which schools are evaluated. This number roughly corresponds to around 100,497 fewer retained students each year nationwide, and $1.4 billion saved in public school expenditures. I further find that administrators do retain significantly fewer students in the last grade offered by their schools, implying that administrators use retention strategically. This paper is the first to show evidence that school administrators are willing to use retention as a tool for optimizing their schools' accountability ratings, and demonstrates that the individual components of accountability systems alter administrator behaviors.