The Role of Information in Postsecondary Enrollment Outcomes

Eric Grodsky, University of California, Davis
Stefanie DeLuca, Johns Hopkins University

The decision about the kind of college to attend, if any, is a critical link in the status attainment process of young adults in the United States. This decision may have implications for their lifetime earnings, their occupational autonomy, the characteristics of the marriage market they confront, and their worldviews. Given the importance of this decision, it is not surprising that both economists and sociologists have contributed to a substantial research literature on the nature of postsecondary transitions and choice over the past quarter century. Despite the progress researchers have made in understanding the choice process, substantial empirical and theoretical contradictions in the research literature remain to be resolved. In order to estimate the models most germane to quantitative inquiries into postsecondary attainment, it is necessary to assume that errors in estimation are independently and identically distributed across students. This assumption is contradictory to some of the qualitative research on higher education which shows that economically disadvantaged students have less information about their options than do more advantaged students and may make their postsecondary decisions in a less predictable, more haphazard manner (McDonough, 1997; Hossler et al, 1999). If these qualitative findings hold true for the population, quantitative estimates of student disadvantage using standard models would be biased and the reliability of estimation lowest for the most disadvantaged students in our samples?students who are generally the focus of policy interventions. In this paper, we reconcile these disparate streams of research by estimating a series of multinomial logistic models of postsecondary attainment that explicitly incorporate these qualitative insights into the choice process. Specifically, we model the systematic portion of student postsecondary destinations as a function of both ascribed and achieved characteristics and estimate the variance in the error of estimation as a function of the quality of information available to students. By simultaneously estimating both the log-odds of attendance and variation in the error term, we empirically test the degree to which assumptions about uniform error variance lead to biased parameter results. We then present corrected parameter and probability estimates for student postsecondary attendance patterns.