Panel Paper:
Does Teacher Effectiveness Translate Across School Contexts? Evidence from a Randomized Experiment
*Names in bold indicate Presenter
Previous research on whether teachers may be affected by a change in environment, students, and peers, is decidedly mixed. Several studies testing the validity of value-added measures, which capture teachers’ contributions to student performance on standardized tests, leverage teacher transfers to conduct event studies and find that teacher effectiveness is largely transferable (Xu et al., 2012; Chetty et al., 2014; Bacher-Hicks et al., 2015). Other studies, find evidence that some aspects of teacher effectiveness are specific to grade levels (Ost, 2015), school contexts (Kraft & Papay, 2014), and the quality of the match between teachers and schools (Jackson, 2013). These studies, however, are limited by the observational nature of the data they use. Teachers mostly choose to transfer to similar or higher performing schools, providing only a weak test of the portability of teacher effectiveness.
In this study, we examine the portability of teacher effectiveness using experimental data from the Talent Transfer Initiative (TTI) study conducted in 10 districts across seven states. In the TTI study, 81 high-performing teachers from high-achieving schools were incentivized with a $20,000 stipend to move and remain at low-achieving schools within their districts. Participating low-achieving elementary and middle schools with open positions were randomly assigned to be eligible to offer this incentive bonus. The transfer incentive further introduced a degree of exogeneity into who transferred and created large contrasts between transferring teachers prior and new schools.
We test the portability of teacher effectiveness in a difference-in-differences framework with teacher fixed effects. We find transferring from high-achieving to low-achieving schools decreased high-performing teachers’ contributions to student achievement by between 0.10 to 0.15 standard deviations in the first year. Our second difference, teachers who transferred into control schools, suggests that this is not simply due to the transfer itself or any loss of school-specific human capital, but the effect of a large change in school context. Estimates using a sub-sample of teachers for which a second-year of post-transfer data is available suggest that a substantial portion of this decline in effectiveness persists. These results are robust to the use of several alternative constructions of the comparison group. We conduct a range of exploratory analyses to examine whether this decline is due to differences in school contexts, poor match quality, or the loss of grade and/or school-specific human capital.