State and federal accountability policies are predicated on the ability to estimate valid and reliable measures of school impacts on student learning. The typical spring-to-spring testing window potentially conflates the amount of learning that occurs during the school year with learning that occurs during the summer. We use a unique dataset to explore the potential for students' summer learning to bias school-level value-added models used in accountability policies and research on school quality. The results of this paper raise important questions about the design of performance-based education policies, as well as schools' role in the production of students' achievement.