A number of teacher workforce policies implicitly assume that job performance is a relatively stable attribute within teachers. We use longitudinal data on 5th grade teachers in North Carolina to assess this using value‐added measures (VAMs) of job performance. We find that there is a permanent component of teacher performance that is stable in teachers over long periods—implying that workforce policies selecting teachers based on VAMs could effectively improve student achievement. But importantly, our time series model suggests that the permanent component of performance is considerably smaller than that which is often used to estimate workforce policy impacts.