Is it socioeconomic or academic? Disentangling sources of peer effects on student achievement

Abstract

Studies regarding school composition have been highly influential in the design of policies. However, methodological and conceptual controversies have hindered the emergence of a consensus on the existence, size, and direction of peer effects. Drawing on four cohorts of Chilean students (n = 620,044), this work analyses the extent to which the socioeconomic and academic classmates’ characteristics are associated with student attainment. The findings suggest a positive medium- to low-magnitude effect of increases in the peers’ academic performance. At the same time, there are almost no effects associated with changes in the classmates’ socioeconomic status. These findings are stable across school types and are not driven by the sorting of students within them. The results are interpreted in the context of a new nationwide reform aiming to foster school integration and taking advantage of peer effects interaction.

Keywords:

• School composition
• tracking
• school choice

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 In the meta-analysis conducted by these authors, less than half of the studies analysed included measures of prior attainment. The studies that did not account for prior attainment almost doubled the size of the effects compared to those surveys including this covariate.

2 These results show only the main variables of interest. Tables including all the controls are available upon request.

¹ Two issues are checked as part of this analysis.

First is the allocation of students within schools (which may interfere with correctly identifying peer effects). Schools using academic sorting across classrooms were identified following Clotfelter, Ladd and Vigdor (2006) and later works (Shure 2021; Treviño, Valenzuela and Villalobos 2016). A Chi-squared test was performed for each school with two or more eighth-grade classes, using quartiles based on their fourth-grade SIMCE scores. All schools rejecting the null hypothesis of a similar distribution of students according to their academic performance (p<.05) were considered to be implementing sorting practices. According to this estimate, 15.5% of the students were enrolled in schools that sorted students across classes, 55.3% were in schools without sorting, and 29.2% were in schools with only one classroom per level. In addition, ANOVA was implemented to assess the differences in performance across classes (with SIMCE expressed as a continuous variable), leading to similar results.

Second is the presence of former peers in secondary school (where the outcome is observed). During the transition from primary to secondary school, most students are sorted into a classroom with a proportion of their former primary school peers. This lack of variation implies that the current peer group is affected by the former classroom group (including the student's own performance). One option to reveal this relationship is to estimate the effect of the peers only for students with no former primary peers at the secondary school, which helps observe whether the main results remain stable or if there are changes when there is an important change in the composition of the group. Column four shows the estimates for the subsample of students without former peers at the secondary school.

Additional information

Funding

This study was supported by Economic and Social Research Council and Agencia Nacional de Investigación y Desarrollo.