This study examines the impact of teachers on student mathematical achievement in Chicago public high schools using administrative data. The authors find that teachers significantly influence student performance, with a one-semester increase in teacher quality rated two standard deviations higher leading to a 0.3 to 0.5 grade equivalent improvement in math scores. The study also reveals that teacher quality ratings are stable over time, robust to controlling for non-math teachers, and not driven by classroom sorting or selective reporting of test scores. However, observable teacher characteristics, including those used for compensation, explain only a small portion of the variation in teacher quality. The findings highlight the disconnect between teacher pay and productivity and the challenges in developing compensation schedules that reward good teaching. The study uses data from 1996-97 to 1998-99, covering 27,000 to 30,000 students annually, and focuses on 9th-grade math achievement. The results are robust to various robustness checks, including adjustments for sampling error and the inclusion of additional student, family, and peer characteristics.This study examines the impact of teachers on student mathematical achievement in Chicago public high schools using administrative data. The authors find that teachers significantly influence student performance, with a one-semester increase in teacher quality rated two standard deviations higher leading to a 0.3 to 0.5 grade equivalent improvement in math scores. The study also reveals that teacher quality ratings are stable over time, robust to controlling for non-math teachers, and not driven by classroom sorting or selective reporting of test scores. However, observable teacher characteristics, including those used for compensation, explain only a small portion of the variation in teacher quality. The findings highlight the disconnect between teacher pay and productivity and the challenges in developing compensation schedules that reward good teaching. The study uses data from 1996-97 to 1998-99, covering 27,000 to 30,000 students annually, and focuses on 9th-grade math achievement. The results are robust to various robustness checks, including adjustments for sampling error and the inclusion of additional student, family, and peer characteristics.