Predicting postsecondary attendance by family income in the United States using multilevel regression with poststratification

Abstract

Despite tens of billions of dollars in yearly public spending to fund grants for higher education for youth from low-income families, no government agency tracks how many young people from low-income families enroll in higher education by state. Proxy measures like the number of college students who receive Pell grants address the number of already enrolled students who come from low-income families rather than tracking the rate of enrollment among the overall number of young people who are from low-income families. Estimates of postsecondary enrollment among low-income young people from U.S. Census surveys likely overestimate enrollment among this population due to their design and administration. In this paper we use multilevel regression with poststratification to estimate postsecondary attendance rates by family income in 50 states and the District of Columbia for a recent cohort of young people. Our application of Bayesian techniques for estimation and inference allow us to make appropriate statements of uncertainty regarding our estimates, including the probability that low-income young people will attend higher education in a given state.

Keywords: college access, low income, multilevel regression with poststratification