Predicting Postsecondary Attendance by Income in the American States Using Multilevel Regression with Poststratification

Abstract

In this paper we introduce a novel technique to estimate postsecondary attendance rates by income in 50 states and the District of Columbia. We use parameter estimates based on the federally administered Educational Longitudinal Study of 2002 combined with known characteristics of state populations from the American Community Survey in order to estimate postsecondary attendance rates by income in each of the states. 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. We find substantial variation in the postsecondary attendance rates of low-income young people, ranging from 34 percent in some states to over 54 percent in others. Our results also provide evidence that the within-state difference between low-income and middle/high-income college attendance rates ranges from 20 to 28 percentage points. Our validity checks show that our estimates follow expected patterns based on estimates from other surveys.

Keywords: Postsecondary Education, Bayesian Estimation, Multilevel Regression with Poststratification, Policy