We estimate the importance of teachers in Chicago public high schools using matched student‐teacher administrative data. A one standard deviation, one semester improvement in math teacher quality raises student math scores by 0.13 grade equivalents or, over 1 year, roughly one‐fifth of average yearly gains. Estimates are relatively stable over time, reasonably impervious to a variety of conditioning variables, and do not appear to be driven by classroom sorting or selective score reporting. Also, teacher quality is particularly important for lower‐ability students. Finally, traditional human capital measures—including those determining compensation—explain little of the variation in estimated quality.