Research consistently shows that college choice in an important predictor of college completion and labor market outcomes. Yet, with thousands of colleges that consist of countless attributes and heterogeneous student preferences and constraints, even the best models do a poor job at predicting how students sort into college. This paper explores the merits of using one of the most direct forms of peer effects to better understand how students choice where to enroll—the older sibling. To do so, we identify siblings in the millions of SAT test-takers between the 2004 and 2011 high school graduation cohorts and explore the enrollment patters both descriptively and causally.

In the descriptive section, we find that younger siblings enroll in the same college as their older sibling 21.2 percent of the time. Also, conditional on their own SAT scores, we find that younger siblings whose older siblings enrolled in four-year colleges and the most selective colleges are 17.4 and 21.3 percentage points, respectively, more likely to themselves enroll in four-year and the most selective colleges. Overall, adding characteristics and enrollment decisions of older siblings to standard college
choice models improves model fit and consequently, are valuable pieces of information for explanatory and predictive power. These results imply that using sibling information may be a good way to target interventions to improve college choice.

In the causal section, we exploit the fact that many colleges have SAT thresholds for admission. That is, students just above the threshold are much more likely to gain admission than those just below, despite being similar on all other characteristics. Using a set of colleges that we identify as clearly having these thresholds, we employ a regression discontinuity design. We first note that in the first stage, older siblings who are just above the admissions threshold are much more likely to attend the collegeand consequently, more likely to enroll in a four-year college. Restated, those older siblings below the
threshold are more likely to attend a two-year college or lower-quality four-year college. Then, we use this exogenous variation to determine how the older sibling’s enrollment choice affects where the younger sibling enrolls. This second stage shows that the younger sibling is not only more likely to attend that specific college the older sibling attends, but also more likely to attend any four-year college when the older sibling is above the threshold. This implies that legacy effects are not the only
mechanism.