Which college types increase earnings? Estimates from geographic proximity

Abstract

ABSTRACTThe question of why postsecondary institutions produce different labor market outcomes is difficult to answer due to unobserved student characteristics. Here, I leverage students' geographic proximity to three classifications of postsecondary institutions – earnings-enhancing, competitive, and Historically Black Colleges and Universities (HBCUs). Using a nationally representative sample, I estimate attainment and earnings effects of first attending each type. Attending an institution classified as earnings-enhancing increases humanities credit completion, degree attainment, and early-career wages. Among underrepresented students, living closest to an HBCU strongly predicts HBCU enrollment. This yields higher STEM credit completion but lower early-career wages, suggesting possible labor market bias.Abbreviations: Competitive: Barron's Top 3 Selectivity Tier Institution; HBCU:Historically Black College or University; HSI: High-Success Institution; STEM: Science; Technology; Engineering; and Mathematics; Underrepresented Minority (URM): Black; Indigenous; or Hispanic/LatinxHIGHLIGHTSNearest-college attributes predict college choice for many high school students, especially those living near HBCUs.Colleges previously linked to students' wage mobility yield higher earnings by students' mid-20s.Higher earnings effects coincide with higher humanities credit completion, bachelor's completion, and postbaccalaureate training.HBCU attendance relative to other options yields higher STEM credit completion, but lower early-career wages.HBCU attendance relative to no college also increases humanities credit completion and bachelor's degree completion.KEYWORDS: Human capitalsalary wage differentialsinstitutional effectsinstrumental variablescollege proximity Disclosure statementNo potential conflict of interest was reported by the author.Notes1 Chetty et al. (Citation2017) also found high variation in the ‘mobility rates’ of institutions, which they defined as the product of institutions' success rates and the fraction of bottom-quintile students enrolled in them.2 ELS:2002 provides cross-sectional base-year weights for each school and student to reflect both the inverse probability of selection, which is known from the sampling design, and the probability of nonresponse, which is estimated from student and school attributes at baseline. The dataset also includes panel weights for use in longitudinal analyzes across the other survey waves. I do not employ the ELS weights in this analysis because my identification strategy, instrumental variables analysis, in effect assigns greater weight to respondents who are sensitive to the set of geographic instrumental variables. Applying sampling and non-response weights may therefore distort the internal validity of the IV analysis (Solon, Haider, and Wooldridge Citation2015).3 The four HBCUs also classified as high-success institutions are Howard University, Morehouse College, Spelman College, and Xavier University of Louisiana.4 I attempted to instrument for entry into a high-STEM institution as well, but the set of geographic instruments did not predict entry into this type of institution.