Indiana University SPEA Edward J. Bloustein School of Planning and Public Policy University of Pennsylvania AIR American University

Panel Paper: Teacher Spillover Effects Across Four Subjects in Middle Schools

Thursday, November 12, 2015 : 4:10 PM
Flamingo (Hyatt Regency Miami)

*Names in bold indicate Presenter

Yuan Kun, RAND Corporation
Background, Research Questions, and Significance

Prior research has examined two types of teacher spillover effects (TSEs). The first type of TSE refers to a teacher’s influence on another teacher’s students through peer interactions between two teachers (Jackson & Breugmann, 2009). The second type of TSEs, which is the focus of this study, refers to a teacher’s influence on his/her students’ achievement on another subject taught by another teacher (Koedel, 2009).

In this study, I studied the second type of TSEs across four types of teachers, including math, ELA, science, and social studies, on student achievement in each of the four subjects. Specifically, I addressed the following two research questions:

  1. Do TSEs exist across teachers of math, ELA, science, and social studies on any of the four subjects at the middle school level?
  2. If so, how does controlling for TSEs affect the relative stance, variation, and precision of own-subject teachers’ value-added (VA) scores?

This is the first study on TSEs at the middle school level. Moreover, I investigated TSEs across four groups of teachers on student achievement in four subjects.


Data came from an urban school district in the Southern United States. This district serves a student population of about 70-80 thousand students annually. Data used for analysis followed students at grades 7-8 and their math, ELA, science, and social studies teachers from 2006-07 to 2008-09. In total, the analytical sample included 13,663 students at grades 7-8 and 636 teachers.


I analyzed data by grade-year groups. This is consistent with the common practice that focuses on estimating teachers’ contributions to student achievement gains in a single year rather than estimating the same teacher’s contributions to student achievement gains using multiple years of data. In total I analyzed six grade-year groups.

I investigated the existence of TSEs using two types of models. First, I implemented a value-added model with fixed teacher effects and used F-test to test the significance of teachers’ joint contributions to student achievement. Second, I implemented a value-added model with random teacher effects to examine the magnitude of TSEs.

I implemented the random teacher effect model with and without controlling for TSEs and examined differences in the quartile rankings of the same teachers’ value-added scores before and after controlling for spillover effects. I also examined changes in the estimated standard deviations of teachers’ effects, the standard error of individual teachers’ VA scores, and the quartile rankings of teachers’ VA scores before and after controlling for TSEs for own-subject teachers.


Results from both types of value-added models showed consistent evidences of TSEs at both grade levels in middle schools, especially from ELA teachers to students’ achievement on the other three subjects. Results also showed that controlling for TSEs affected the relative stance of value-added scores for up to a quarter of own-subject teachers. It reduced the variation but increased the estimation errors of own-subject teachers’ effects.