Panel Paper: Using Multiple Discontinuities to Estimate Broad Effects of Public Need-Based Aid for College

Thursday, November 2, 2017
Ogden (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Drew M. Anderson, University of Wisconsin - Madison


Succeeding in college is an outcome that affects and is affected by multiple government bodies and public policies. The various bodies may not always share data, leading to a fragmented view of the effects of any given policy. In Wisconsin, data on student success in college and the returns to college are housed in no fewer than five large state agencies: the university system, the technical college system, the public K-12 system, the labor department, and an agency that delivers financial aid. Each is partially funded by, or is involved in raising, general program revenue for the state. Each therefore has an interest in efficient use of taxpayer funds, and in the tax revenue generated by additional citizens with college degrees. However these agencies typically do not have access to each others' data or the capacity to analyze data.

Insufficient data leaves the state without a base of evidence for making policy decisions. One important policy is the Wisconsin Grant, a long-running financial aid program for state residents from low-income backgrounds. The program spends over $100 million each year, serving over 60,000 students, and covering one quarter of college tuition for the average recipient. The Wisconsin Grant is not fully funded, meaning not all eligible applicants can receive support. A key question is therefore whether students who receive grant offers achieve better outcomes, as compared to similar applicants who do not receive grant offers. This question can be credibly answered by implementing a regression discontinuity design, comparing students who receive different levels of support but have only slightly different eligibility criteria, in terms of family finances or the date they file a financial aid application.

An initial study implemented this causal analysis, by linking together administrative data from public high schools, the National Student Clearinghouse, and financial aid applications. Among recent high school graduates the Wisconsin Grant induced small increases in enrollment and eventual completion of two- and four-year degrees.

Given the modest educational impacts, it is still possible that the Wisconsin Grant had other meaningful impacts on students' lives. A follow-up study will link these same data, plus data on adult college applicants, to information from the Census Bureau data infrastructure. This will result in a ten-year panel including the universe of financial aid applicants from Wisconsin. These data will allow for estimation of the Wisconsin Grant's impacts on employment, earnings, occupation choice, use of social benefits, mobility, and household formation. The estimation will use a similar design to find comparison groups of students.

Financial aid could affect post-college outcomes directly by lowering financial constraints on students, potentially lowering their debt, or changing their labor supply during college. Financial aid could also affect each post-college outcomes indirectly through the additional education it pays for. The size and heterogeneity of the sample allow for descriptive subgroup analyses that could suggest which of these channels is more important.

These answers are particularly relevant as the state is considering increases to the Wisconsin Grant budget and changes to the program structure.