Over past decade, districts throughout the United States have reformed their teacher evaluation systems in order to support teachers to improve teacher practices and to improve the composition of teachers in the workforce. Despite the widespread reform of evaluation systems, little research has examined the impacts these systems have on teacher composition or teacher performance. Using data from the Institute for Education Science's "Impact Evaluation of Teacher and Leader Performance Evaluation Systems," a multi-year, multi-site random assignment trial, I examine the impact of a reformed evaluation system on the composition of teachers measured by differential retention and mobility, the differential impacts on teachers attitudes and actions related to retention and mobility, and explore the heterogeneity of the impacts on teacher performance by retention status and baseline performance. I find substantial differences in treatment effects on retention by the end of the second year. Specifically, I find that the intervention increased the retention of high performing teachers by 7.3 percentage points (p=0.108), decreased the retention of low performing teachers by -11.5 percentage points (p=0.008), resulting in a differential impact on retention rate of 18.8 percentage points (p=0.003). The differential retention results were largely driven by differences in which teachers left the district; which itself is driven by high performing treatment teachers, who were less likely than high performing control teachers to leave the district (-9.8 percentage points, p=0.006). I also find substantial differential impacts on two measures of teachers attitudes and actions. High performing treatment teachers were 11.2 percentage points more likely than similar control teachers to report that they would teach as long as they are able (p=0.058), and low performing treatment teachers were 18.6 percentage points (p=0.008) less likely to report the same; again, the differential effect of almost 30 percentage points is substantial (p=0.001). Moreover, I find that high performing treatment teachers were also more likely to report that they did not seek another position during the 2012-13 school year compared to similar control teachers (12.3 percentage points, p=0.016). While the impact on low performing teachers (-6.3 percentage points) did not reach conventional levels of statistical significance (p=0.304), the differential impact was large and statistically significant (18.6 percentage points, p=0.032).Finally, I find that the positive impact on instructional practice, as measured by the Classroom Assessment and Scoring System (CLASS) rubric, was consistent across two subgroups of teachers -- Stayers, who remained teaching in a tested grade in the same school throughout the study, and Joiners, who entered into the school after random assignment. Additionally, I find that the impact on instructional practice as measured by the CLASS rubric is negatively and significantly moderated by teachers' prior performance. That is, the lower the baseline performance scores were, the larger the impact on the CLASS score. The results in this dissertation both confirm and flesh out the theory of action for teacher evaluation. This dissertation adds significantly to the small body of literature on teacher and leader evaluation systems and provides further evidence that reformed teacher evaluation systems work as intended. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com.bibliotheek.ehb.be/en-US/products/dissertations/individuals.shtml.]