Although much has been written about the importance of leadership in the determination of organizational success, difficulties separating the impact of leaders from other organizational components complicate the estimation of their productivity and effectiveness. Panel data on students and educators plus regular measurement of school output offer a window into the importance of leadership quality. High dimensional two-way fixed effects methods have been used widely to separate worker and firm effects, but the small number of schools with multiple principals who have worked in multiple schools limits their use in the such methods. We use school fixed effect and matching methods to estimate directly the variation in the effectiveness of public-school principals in the state of Texas and the city of Chicago. Effective principals have a large and significant impact on student achievement that is remarkably similar across the different samples.  Moreover, this effect is larger in schools with concentrations of disadvantaged students. We show that part of this impact comes through the effects of principal quality on the composition of exiting teachers. Finally, we combine the quantitative data with principal, teacher and student surveys to investigate the contributions of instructional leadership and data management skills to principal effectiveness. In this descriptive analysis, we find achievement growth to be strongly related to the principal’s self-reported focus on instructional leadership, to teacher impressions of instructional leadership, and to student interest in school and perceptions of safety.