Abstract An extensive literature finds that while teachers vary considerably in initial quality there are limited teacher quality dynamics: except for the first few years of teaching, teacher quality does not improve over the course of a teacher's career. This study evaluates the importance of various modeling restrictions to the key findings of this literature. Using data covering all 5th grade public school teachers from the state of North Carolina, I replicate the findings of the previous literature using their restrictive experience assumptions. However, using an unrestricted experience model I find that for mathematics achievement there are high returns to later career teaching experience, about twice as much dispersion in initial teacher quality as previously estimated, and a pattern of negative selection where high quality teachers are more likely to exit.