ABSTRACT
Using well-defined notions of region, panel data, and an econometric approach suited to estimating parameter coefficients on time-invariant regressors, this study finds that region is a highly influential factor for state-level public higher education prices even after controlling for other well-known drivers. Additionally, prices track the historical development of higher education in the US. Based on the history of higher education and regional norms, these estimates may be evidence of suboptimal tuition and fee equilibria resulting from implicit gaming. Substantively, a greater awareness of these historical and competitive forces and their effects on the cost of public higher education can be a substantial resource for policy-makers, institutional decision-makers, and researchers alike.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 Tuition and fees are measured here as the average listed in-state price at state-supported four-year instituitions of higher education, covering the student portion of expenses as well as mandatory fees, but not including room and board and other elective fees.
2 These values were deflated and are reported in constant 2013–2014 dollars for appropriate comparison (see Baum and Ma Citation2013, 20).
3 It should be highlighted that based on information from the Bureau of Labor Statistics (Citation2014), inflation was also down, and in fact was negative in this year at −0.4.
4 Implicit gaming is understood in this context as a situation in which each party responds only to the price-setting behaviors of proximate competitors with incomplete information and an inability to collude in price-setting activities (Gu Citation2015; McMillen, Singell, and Waddell Citation2007).
5 See Renault (Citation2012, 2636) for an excellent mathematical exposition on the topic.
6 Here states refer to conditions, contexts, or circumstances and not the US states.
7 This test is very similar to the traditional Hausman test but differs in that it extends directly to heteroskedastic and cluster-robust versions (see https://kb.iu.edu/d/bcmq at Indiana University Citation2014).
8 The FEVD estimator has faced scrutiny from some scholars (see Political Analysis 19 (2), 2011) followed by a compelling rebuttal from the developers (Pluemper and Troeger Citation2011), and while some might not consider this a reasonable solution for the problem of time-invariant regressors, recent studies (Beck Citation2011; Hearn and Belasco Citation2015; Heinemann, Osterloh, and Kalb Citation2014; Serna Citation2015; Zhu Citation2013) suggest that though limited in some instances, the estimator's usefulness is still evident if the finite sample properties of the estimator can be met, if important theoretical variables are time-invariant, and the inclusion of fixed-effects is vital (Zhu Citation2013). This study meets all of these requirements.
9 Denoted as uit but which includes both w and p.
10 Readers should note that in this equation α is not equal to the unit effects as in Equation (3), but rather is the intercept since these effects were removed in previous stages.
11 Doyle employed a measure of the proportion of full-time equivalents in the state enrolled at private institutions and interacted this variable with governmental liberalism. It was under these conditions that tuition and fees were influenced by private four-year enrolments.