I use a detailed panel of data and a unique modeling specification to explore how public schoolteachers respond to the incentives embedded in North Carolina's retirement system. Like most public-sector retirement plans, North Carolina's teacher pension implicitly encourages teachers to continue working until they are eligible for their pension benefits, and then leave soon afterward. I find that teachers with higher levels of quality, as measured by a teacher's value-added to her students' achievement test scores, are more responsive to the "pull" of teacher pensions. Younger teachers, those with higher salaries, and nonwhite teachers are also more likely to stay during the pension "pull." All teachers show a strong response to the pension "push," with about a quarter of teachers leaving every year once they become eligible for their pension. I depart from other models of teacher retirement by using a Cox proportional hazard model. Given that salaries are generally fixed by the state, I find that the number of years a teacher must work before she is eligible for her full pension benefit is the major driver of variation in pension wealth. This specification has the benefit of a flexible baseline hazard that can easily capture the sharp incentives driving a teacher's retirement decision that are dependent on her proximity to retirement eligibility, and can flexibly account for differences driven by local labor market conditions. These analyses highlight important unintended effects that inform education policies going forward to ensure the retention of high-quality teachers in all types of schools. [This research received additional funding from the Center for Retirement Research at Boston College and the Bankard Fund for Political Economy at the University of Virginia.]