Association for Public Policy Analysis & Management

As the costs of college increase, there is increasing scrutiny on the financial returns to college. The College Scorecard, release in October 2015, represented the first large-scale attempt to provide college-specific early-career earnings to establish a new metric of college quality. Yet the Scorecard raised concerns about whether student or college factors drive these earnings. The natural tendency to interpret such differentials as a value-add of the college may mislead prospective college applicants into making suboptimal college choices if student characteristics at the point of college-entry can explain the lion’s share of the variance in college-level earnings. Variation in major-specific earnings (Altonji et al. 2012; Carnevale et al., 2015) makes selection into major a relevant and policy actionable margin.

Previous research using nationally representative longitudinal datasets such as the NLSY (Arcidiacono, 2004; Black and Smith, 2006) do not have sufficient coverage to study the interaction of major choice and institutional quality. While the Census collects major choice since 2009, it does not collect graduating institution. Policy-makers are hungry to satisfy this need, working with states state-level analysis to study earnings across college type and major (Schneider, 2014).

In this study, we construct the largest, most comprehensive panel of data consisting of all students who participated in any College Board program (AP, PSAT/NMSQT, or SAT) from the high school graduation cohorts of 2004 through 2010, merged to college enrollment, college completion and college major data from the National Student Clearinghouse. To capture the early-career labor market outcomes of these students, we worked with TransUnion to assemble a new dataset, containing earnings, credit scores and student loans.

Unsurprisingly, we find that student characteristics such as SAT scores have strong associations with earnings and credit scores. Following the college-quality principal component analysis of Dillon and Smith (2017), we find that students attending higher quality colleges have better labor market outcomes, and that students choosing majors requiring advanced technical skills (e.g. engineering) similarly fare better early in their careers.

Student selection into colleges and majors drives much of the variation behind these basic descriptive findings. Even after controlling for college attended, student demographics, SAT scores, and high school GPA, and attempting to follow the Dale and Krueger (2011) approach of accounting for college application behavior, we find that the pre-professional majors of business and engineering are associated with earnings premiums of 10 percent relative to humanities majors. We also observe trivial differences in the earnings of humanities graduates relative to mathematics and natural science graduates.

We find that earnings for engineering graduates are independent of college quality. By contrast, business and social sciences majors from the top quintiles of institutions earn approximately 2 to 6 percent more than graduates from lower quintiles. We attribute variation for business and social science graduates to the professional and social networks available at the more selective institutions. Our results highlight the importance of technical skills for closing socio-economic gaps. Earnings related to majors requiring technical skills can be useful for students with limited access to higher quality colleges.