Panel Paper: Setting a Good Example? Examining the Heterogeneous Spillover Effects of Sibling Academic Performance Using Regression Discontinuity Design

Saturday, November 10, 2018
Harding - Mezz Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Krzysztof Karbownik, Northwestern University and Umut Ozek, American Institutes for Research


Social scientists and policymakers have long been interested in understanding the human capital production function and the role it plays in equality of opportunity. One question that gained a lot of interest in recent decades relates to potential externalities generated by changes in human capital. The two streams of research in this area relate to spillovers across generations (Black and Devereux, 2011) and spillovers due to societal interactions (Manski, 2000). Both intergenerational transmission and peer effects are important because if these hypothesized associations reflect causality, then positive (negative) shocks to an individual can propagate beyond what is typically considered in program evaluation or cost-benefit analyses, and lead to greater declines (increases) in inequality.

In this paper, we study a particular type of peer effect that received relatively little attention in the extant literature due to data limitations and identification challenges - a sibling spillover. In particular, we examine the spillover effects of a sibling's academic performance in elementary and middle school on a wide array of educational outcomes including test scores, non-test outcomes (e.g., grade retention, being identified as a special education student, juvenile detention, high school dropout), and postsecondary outcomes including enrollment, completion, and college selectivity.

To identify this causal link, we make use of the variation in academic performance of students driven by school starting policies in a large, anonymous district in the state of Florida. Previous work has documented large effects of this policy on children's development (Bedard and Dhuey, 2006; Dhuey et al., 2017), and here we confirm these estimates in our data. Using this policy-driven variation in academic performance and exact student birth dates, we examine sibling spillover effects by comparing the educational outcomes of students whose siblings were born right before and after the school starting cutoff in a regression discontinuity framework.

We find significant effects of the test performance of the older sibling on the younger sibling test scores, likelihood of identified as gifted, and college enrollment that are entirely driven by impoverished households and racial minorities. These findings might point to two underlying mechanisms through which sibling spillovers might be operating. First, a struggling older sibling might be more detrimental in resource-constrained households when they reallocate their resources to help the struggling kid. Second, high-poverty household structure might be more conducive to older siblings serving as role models since these households are more likely to be headed by single parents.

Three preliminary findings seem to suggest that the latter is a more plausible explanation for the uncovered externality. First, when investigating the role of gender complementarity across siblings, we find much larger effects for same-sex siblings. Second, identified spillovers are larger among closely spaced siblings rather than those spaced more than three years apart. Finally, had resource reallocation been at play, we would expect to observe significant spillovers from younger to older sibling as well, yet we find no such evidence irrespective of household affluence, gender composition or age distance between children.