There’s little question that a bachelor’s degree can substantially boost a person’s lifetime earnings, but students in the United States who attend different colleges go on to earn vastly different amounts. The US Department of Education’s College Scorecard shows that students who attend MIT earn a median annual income of $111,222 10 years after enrolling, while those who attend Alabama State make only $32,084.
According to research by Chicago Booth’s Jack Mountjoy and Washington University in St. Louis’s Brent Hickman, these publicly available income statistics overwhelmingly reflect preexisting differences in the earnings potential of incoming students, rather than substantial variation in the value added by a school itself. Analyzing 10 years of data on 30 diverse state-supported universities in Texas, they demonstrate that better indicators of a college’s real benefit for students are its spending on instruction, emphasis on STEM majors, and graduation rates.
“These results suggest that both the ordering and the magnitudes of the raw outcome comparisons often highlighted in college guides, the popular press, and state funding formulas are driven largely by selection bias,” or the ability of top schools to enroll students who would do well no matter where they attend, the researchers write. This leaves “a smaller and more nuanced role” for the contributions of the educational institutions themselves.
Calculating the value-added of colleges and universities is a hot topic among students, parents, and policy makers. “With state legislators increasingly exploring funding models that directly tie public college appropriations to student outcomes, it [value-added] is also an important consideration for policy-makers deciding where and how much to invest,” Mountjoy and Hickman observe.
Carrying out such an analysis is no simple matter, the researchers argue. The obstacles include finding a big enough sample to estimate the educational benefits conferred by each institution, following a series of individual-level data linkages from student characteristics to enrollment to degree completion and earnings, and creating a research design that takes into account the decentralized, idiosyncratic nature of college admissions and enrollment in the US.
They addressed these challenges by tapping into several administrative registries in Texas, the second-largest US state by population behind California. This enabled them to track 422,949 students in 10 high-school graduating classes from 1999 through 2008 who enrolled in the state’s 30 public universities. Mountjoy and Hickman then employed a “matched applicant” research design that compared the outcomes of students who enrolled at different colleges but applied to and were admitted by the same set of institutions. This approach, they explain, succeeded in balancing precollege predictors of future outcomes, such as high-school test scores and family income, across the different colleges that students ended up attending, yielding fairer institutional comparisons than the raw, uncontrolled statistics.