A two-stage research performance assessment of Turkish higher education institutions using data envelopment analysis and beta regression

Abstract

In this paper, we study the research efficiency of the Turkish higher education sector in a two-stage data envelopment analysis model with variable returns to scale. Using a sample of 50 private and public universities, the first stage of our model calculates the efficiency scores and determines the reference set of efficient institutions for each inefficient university. DEA efficiency scores are a prime example of fractional data - a fact that has been disregarded by many previous studies. Instead of popular Tobit regression, which was developed for censored data, we use beta regression for fractional data in the second stage of our analysis to estimate the effects that external factors, such as age, size and ownership status of the university, may have on efficiency scores. Our results indicate that research efficiency within the Turkish system of higher education is relatively high: DEA scores range from 0.576 to 1, with an average score of 0.878. We find 25 of the 50 universities in our sample to be research efficient, with a maximum score of 1. Beta regression summary statistics suggest that extra-large universities tend to be less research efficient than large universities, as the corresponding coefficient is statistically significant at the 10% confidence level. In turn, no statistically significant dependence on the research efficiency score is detected for ownership status and age of the institution. Bootstrapped hypothesis testing further corroborates the latter result.