In designing our study, we strive for a methodology for measuring students’ value-added earnings outcomes that rivals the rigor and accuracy of academic studies and that can also be implemented at scale in states with strong data systems. 

Academic Studies

Academic studies of value-added earnings outcomes in higher education are often based on regression discontinuity designs or admission lotteries. These academic studies are rigorous in their designs and accurate in their results. However, they typically pertain to, and draw conclusions about, small subsets of students, institutions and programs and are typically of limited use to practitioners and policymakers who need comprehensive information on value-added earnings outcomes.
 

Indices and Classifications

In another category of measurement work on earnings outcomes in US higher education are earnings-related indices and classifications published by advocacy organizations.

These efforts to measure value-added earnings outcomes pertain to all or many institutions, programs, and students, and they typically rely on, and maximize insights from, publicly available data, mainly issued by the US Department of Education and the US Census Bureau.

These efforts produce insights into general patterns in earnings outcomes in US higher education, often identify institutions with high-end or low-end outlier results, and provide valuable feedback to institutions interested in improvement.
 

PSC’s Texas Study and Multi-state Research Program

In this context, we seek in our Texas study and in our general research program to implement a methodology for measuring students’ value-added earnings outcomes that rivals the rigor and accuracy of academic studies and that can scale in states with strong data systems.

We hope that our methodology is of use to federal and state policymakers who are interested in implementing — and setting policy based on — rigorous, accurate, and scalable approaches to measuring economic outcomes in higher education.

We chose to conduct this study in Texas because of its exceptionally high-quality longitudinal data system. We can repeat our study in other states that have data systems with the well-developed, linked data systems.

We hope that the findings in our study inform decision makers in higher education, including institutions (in their internal improvement work), students (in their choices about pursuing postsecondary education), and policymakers (in their general understanding of outcomes in higher education).

We focus on students’ value-added earnings, not their absolute earnings. 

We assess the degree to which students’ choices to enroll improve their earnings over the counterfactual earnings they would experience in a scenario where they do not enroll at the institution in question or pursue further postsecondary education elsewhere. 

We do not focus merely on students’ absolute earnings after exiting or graduating from an institution. Absolute earnings are inaccurate indicators of the contribution that institutions and programs make to students’ earnings.

We measure outcomes for all entrants to institutions and programs, including both eventual completers and non-completers. 

We do not merely measure outcomes for completers, for first-time/full-time students, or for students who receive federal financial aid.

We generate sophisticated comparison group earnings estimates — also referred to as “counterfactual earnings” — that adjust extensively for attributes of students in our treatment cohorts. 

We produce comparison group earnings estimates by assembling comparison groups with individuals who closely resemble students in our treatment cohorts (minus the choice to enroll in postsecondary education during the relevant follow-up period).

Individuals in comparison groups are matched to students in treatment cohorts based on a wide range of attributes — including prior earnings, household income, high school test scores, age, local geography, and prior postsecondary attainment — to maximize resemblance between treatment cohorts and comparison groups and to maximize the validity of our comparison group earnings estimates.

We generate unique comparison group earnings estimates for each of the ~8,500 entering cohorts that we examine in this study.

We capture the opportunity cost to students of foregone earnings during enrollment because we estimate students’ value-added earnings from their point of entry (not from graduation).

Opportunity costs associated with foregone earnings constitute the majority of students’ true cost of attendance.

We generate high-quality estimates of students’ net cost of attendance. 

Our cost estimates include tuition and fees, net of federal and state grants, exemptions, and waivers. We estimate costs for each student in each semester, based on their enrollment intensity in each semester.

We make high-quality observations of earnings for individuals in our treatment cohorts and in our comparison groups, based on person-level quarterly earnings data in Texas’ unemployment insurance records.

Our study is large. We report findings for 935,767 students who enrolled in public institutions in Texas between 2008-09 and 2018-19 to pursue bachelor’s degrees, associate’s degrees, or certificates.

Our methodology is software-based and operates with minimal human bottlenecks. It can be repeated in any state data system with the requisite data and data linkages.