Teacher Evaluation, Instructional Practice and Student Achievement: Evidence from the District of Columbia Public Schools and the Measures of Effective Teaching Project

In recent years, many states and districts have introduced new teacher evaluation policies aimed at improving the overall performance of the teacher workforce. While one goal of these systems is to provide an overall measure of teacher performance that can be used for accountability purposes, another, perhaps more important goal, is to use the individualized feedback on specific aspects of instruction to help teachers improve. However, we know little about teachers’ ability to respond to the evaluation process by changing their instructional practice. This dissertation provides evidence on the potential for performance evaluation systems to support improvements in instructional practice and student achievement in three independent studies.

Chapter 1 employs a regression discontinuity (RD) design to examine the effects of the incentives embedded in IMPACT, the District of Columbia Public Schools’ (DCPS) teacher evaluation program, on specific aspects of teaching practice captured through classroom observation. The RD design leverages a natural experiment in the design of IMPACT that creates a stark incentive contrast at two performance thresholds: one that is associated with the potential for dismissal if performance does not improve, and the other that implies a potential base salary increase. While I detect no positive effects for low-performing teachers following the first year of the program, results in the second and third years suggest that low-performing teachers who face strong incentives are able to improve some, but not all, aspects of instruction, and that these improvements are concentrated on instructional standards where the observation rubric details more specific strategies for success. In contrast, high-performing teachers who are eligible for a permanent base salary increase experience much smaller and less robust improvements in different areas of practice, suggesting that there may be differences in how low- and high-performers respond to incentives in evaluation systems.

Chapter 2 examines the effects of the same IMPACT incentives for low-performing teachers on student achievement outcomes in math and reading. Consistent with Chapter 1, there are no student achievement effects for low-performing teachers after the first year of the program, but we do observe positive effects of roughly seven percent of a standard deviation in each of the next two years. These average effects are driven by improvements in math, especially following the third year of IMPACT. The magnitude of the average effects we observe is educationally significant, and equates to between 12 and 20 percent of a year of learning.

Finally, Chapter 3 explores whether information on teaching practice captured by classroom observation can be used to identify typologies of teaching, or “profiles” of instruction. If there are groups of teachers who share similar characteristics in their instructional practice (e.g., weak in ability to provide strong content explanations, but strong in classroom management), it may be helpful to explicitly identify these profiles in order to coordinate professional support. I employ latent profile analysis to identify profiles of instruction using data from two contexts: DCPS IMPACT and the Measures of Effective Teaching (MET) research project. While the profiles of instruction in MET provide some information on teachers’ relative strengths and weaknesses, the defining characteristic of the profiles in DCPS is simply being more or less effective across all aspects of practice. This difference between the findings in MET and DCPS is driven by lower dimensionality in the information captured through classroom observation in DCPS. This limited dimensionality has implications not only for the construction of instructional profiles, but for the goal of providing useful formative feedback to teachers more broadly. Collectively, these studies underscore the importance of design considerations in evaluation systems that employ multiple measures of teacher performance, and begin to build our understanding of how teachers respond to these systems.