Association for Public Policy Analysis & Management

We have known that student learning in schools is a function of not one teacher, but of the combined effort of many teachers. Research on policy initiatives of placing effective teachers in disadvantaged schools are mainly based on a human capital explanation that effective teachers would benefit their own students’ learning, yet rarely from the perspective of leveraging peer effects to augment individual teachers’ influence.  The documentation of teacher peer effects has not spurred wide discussions on optimal grouping of teachers to maximize student learning in the aggregate. In this study, we leverage on teacher transfers between schools that create peer shocks to examine the extent to which the gain in student achievement varies depending on their teachers’ peer structure.

This study contributes to the literature on teacher peer effects by distinguishing four types of peer structure. The “linear- in- means” model assumes that the spillover effect of a new transfer is the same for all incumbent teachers; the “better-than-me” model assesses the degree to which the change in student test scores of focal (incumbent) teachers can be attributable to the arrival of a new peer who is more effective than themselves; in contrast, the “worse-than-me” model tests the effect of a new peer who is less effective than focal teachers themselves. Lastly, the “substitution” model evaluates which types of teachers are more/less responsive to the effectiveness of a new peer.

Our analyses draw on longitudinal data from Miami-Dade County Public Schools (M-DCPS), the fourth largest school district in the U.S., covering school years from 2003-04 through 2012-13. The data include 1.5 million student-year observations and 2,273 teacher-year transfer observations.  Our main analyses focus on grade-level peers as others who taught the same grade in the same school and year. We measure teachers’ effectives using their prior stable effectiveness of averaging value-added scores in past three years to avoid reflection issues in estimating teacher effects. We have also undertaken several important robustness tests to examine the threats of student sorting and teacher selection issues.

Although positive estimates of “linear-in-means” are statistically insignificant, we do find evidence of strong spillover effects associated with the introduction of “better-than-me” peers into a teacher group. If a student has a new peer teacher at the same grade level who is about one standard deviation more effective than that of own teacher, this student would have a 14 or 18 percent of a standard deviation increase in math test scores. This spillover effect is about 22% to 26% of this student’s own teacher effect. Interestingly, although teachers benefit from a “better-than-me” peer, their students are not particularly hurt by the presence of a “worse-than-me” peer. This finding is important because it implies that the optimal grouping of teachers to maximize all students’ learning is to mix teachers with diverse performance. We also find that low-performing teachers are more responsive to the composition of peers than high-performing teachers. Assigning a high performing peer is particularly beneficial for students of low-performing teachers.