Nearby College Enrollment and Geographical Skills Mismatch: (Re)conceptualizing Student Out-Migration in the American Higher Education System

ABSTRACT

Student out-migration is a well-studied topic covering more than 40 years of research. This literature has typically equated student out-migration to out-of-state enrollment and has classified all college attendance taking place in students’ state of residence as in-state enrollment. This study argues that failing to capture students’ out-migration within their states of residence may overestimate the positive returns of “out-migrating” by underestimating the effect of attending college “in state.” Accordingly, the purpose of the study was 2-fold. First, it relied on geographical network analysis to offer a framework that disaggregated the sole measure of in-state enrollment into (a) nearby college enrollment and (b) within-state out-migration. Second, it tested the impact of these newly proposed conceptualizations of students’ out-migration decisions on educational and financial outcomes. With the use of the Education Longitudinal Study (2002:12) and other individual-, institution-, geographic-, and state-level indicators, findings indicated that the typical in-state versus out-of-state definition exaggerates the assumed benefits of “migrating.” Indeed, within-state out-migrants attained similar academic and salary-based results while incurring significantly lower undergraduate loan debt compared with students who out-migrated out of state. The study offers evidence of geographical skills mismatch associated with students’ worsened outcomes when enrolling in their 5 closest options and offers a framework to minimize such a mismatch.

KEYWORDS:

• Academic outcomes
• college choice heterogeneity
• degree attainment and success
• geographical information systems and data visualization
• geographical network analysis
• geographical skills mismatch and undermatch
• nearby college enrollment
• student load debt
• quasiexperimental analysis and propensity score weighting with machine-learning boosted models to reduce overfitting

Notes

1. For example, assume a student who attended college locally was unable to graduate within 6 years of initial enrollment. Assume further that two students who out-migrated, one within state and another out of state, did graduate. If we aggregate the results of “in-state students,” their average graduation rate would be , which compared with that of the nonresident student will have an inflated difference of 0.5 (or 1 – 0.5).

2. This estimate is considered more precise than those previously used in the literature; for example, Jacob et al. (2018) used high school zip code as a proxy for college enrollment within close proximity to students’ parents’ homes, which in some cases could be farther away from a student’s residential home or in a different zip code.

3. The only study to date in which the authors looked at within-state migration is the study by Alm and Winters (2009); however, their study did not compare the effects of NCE and was limited to the state of Georgia.

4. The reference to Bourdieu (1986) alludes to differences in forms of capital and does not imply that this author referred to students’ out-migration decisions.

5. Estimates that controlled for unobservables using the Heckman control function rendered similar results. Due to space limitations, these models were omitted from this study but are available upon request.

6. Treatment and outcome variables were not imputed. Data points were assumed to be missing at random. Multiple imputations were based on a chained equations approach (Van Buuren & Groothuis-Oudshoorn, 2011). The conditional distribution of each variable was assigned depending on whether the variable was binary, ordered categorical, or continuous.

7. Please see https://bigfuture.collegeboard.org/find-colleges/how-to-find-your-college-fit/college-search-step-by-step

Additional information

Funding

This research was supported by a grant from the American Education Research Association/National Science Foundation (Grant #DRL-0941014). The opinions reflect those of the author and do not necessarily reflect those of the granting agencies.