“Peer Effects on Student Achievement: An Instrumental Variable Approach Using School Transition Data” by Hongliang Zhang
The Chinese University of Hong Kong
In this paper, we examine peer effects on student achievement in middle school using a multi-cohort longitudinal data set from China. We use cohort-to-cohort variation in peer composition within schools to identify the impact of schoolmates. Our methods innovate on previous peer effects studies that employ the value-added models with lagged achievement. We show that estimates from the typical value-added specifications are biased if a student's old and new peer groups have members in common, and prose an empirical strategy to use the lagged achievement of new peers to instrument for that of all peers. Our within-school IV estimate of the linear-in-means peer effects coefficient is positive but insignificant. While we cannot reject the null hypothesis of no peer effects, we also cannot reject relatively large peer effects that have been found in the previous literature. Estimates of a heterogeneous peer effects model, however, show evidence in support of the single-crossing property: an increase in average peer ability benefits high-achieving students more than low-achieving students.