About SES & educational expectations: interrelations in the determination of higher education baccalaureate attainment.

Community colleges and four-year colleges provide two differing alternatives to post secondary education. High school seniors face several options upon high school completion: entering the labor market, entering a community college (for a two-year degree or as a step towards a baccalaureate) or attending a four-year institution. Selection into each of these is clearly not a random process, but one related to previous educational experiences, family characteristics and social class, and educational expectations, among others. Attempting to address this issue of self selection, Rouse (1994 & 1995) explicitly posed the question of the democratization versus diversion effects of community colleges. Her work provides evidence of a rational behavior on the part of two-year college students who respond to price and proximity of such institutions (1994), and of the existence of primarily a democratization effect (1994 & 1995). This paper proposes a variation on the work by Rouse (1995) and Leigh and Gill (2003) by inquiring into the effect of SES in relation to students’ educational expectations. It builds on these two models. The underlying hypothesis is that expectations are not independent from SES and therefore examining social class differences and their interaction with educational expectations would support the theory of endogeneity between educational expectations and socioeconomic background. We directly control for expectations and interactions between SES and expectations, as well as considering variations to modeling SES and use alternative estimation methods for bounded probabilities (Logit and Biprobit) and compare these to their approaches. W find the effect of expectations on the probability for middle and middle-high class students’ proved steeper than for the everyone but low SES students. Effect on expected probabilities increasing with SES. These results disappear when estimation methods are improved, using logits and bivariate probit methods, instead of OLS and IVs (explicitly modeling expectations). However, the democratization effect remains significant through all estimations, and strengthens when estimation methods are improved. Out estimates therefore reinforced the findings by Rouse (1995) and Leigh and Gill (2003) of a positive democratization effect that outweighs any diversion effects.