Association for Public Policy Analysis & Management

If students do not always make optimal choices in their college attendance decisions, admissions policies that expand students’ college choice sets could have adverse effects on college performance for those students whose attendance decisions are impacted (see Arcidiacono and Lovenheim [2015] for a detailed discussion of the assumptions and existing evidence related to this theory). This mismatch hypothesis presumes that students vary in the campus that would be best for their college performance. In other words, colleges may specialize in a certain type of student: One campus might be best for under-prepared students and worse for well-prepared students, while another may be ideal for well-prepared students but worse those who are under-prepared. This paper aims to test whether data are consistent with this pattern of college heterogeneity.

To address this question, we focus on a particular policy in place in Texas. In 1998, the Texas legislature guaranteed admissions to any public university in the state for all Texas high school students with GPAs in the top 10 percent of their high school graduating class. In our analysis, we use the universe of applications to eight Texas universities (six public, two private) during the period 1998-2000, data available as part of the Texas Higher Education Opportunity Project. As a measure of college performance, we focus on five-year graduation, which is also available in these data.

Students who are just above the 10 percent cutoff have a larger choice set and, as a result, attend different colleges on average than those who are just below the cutoff. Under the presumption that nothing else related to graduation changes across this cutoff, we can estimate a causal effect of attending a given university using a regression discontinuity design. By repeating this exercise conditioning on SAT and ACT scores, we are able to obtain causal effects for students of differing levels of college preparedness and test whether the universities in our dataset specialize in graduating a certain type of student. This approach relies on the fact that high schools vary in the test scores of students who are at the 90th percentile of GPAs. We first test whether the two Texas flagship campuses specialize relative to the other colleges in our dataset, which we expect to be true if there is any degree of specialization present among the colleges we observe. To estimate distinct treatment effects for each university using the single discontinuity, we instrument using distance from student to college but only assume that effect magnitudes do not vary with distance, a weaker requirement than that of a typical instrumental variables estimation.

While past college mismatch work has assumed the existence of the type specialization we test for, we are the first paper to our knowledge to test for this type of heterogeneity across specific campuses. Our results have important implications for institutions considering policies similar to the Texas top 10 percent law, as well as for other policies that shift the college choice set of students, such as affirmative action.