Regression quantiles to assess higher education performance

Abstract

From 2001 Italian university system has adopted the credits to mea- sure the workload of the students. The weighted mean of marks with credits as weights is used to measure the their performance. In our opinion it does not seem a proper way to measure. We suggest to adopt the median of the weighted marks, because we are considering an ordinale variable, with a non-normal empir- ical distribution, and a®ected by outliers. Then, instead of using an OLS multiple regression, we investigate the determinants of the performance measured by the median using a regression quantile for ordinal response. A real dataset concerning 133 students of the Faculty of Economics of the University of Palermo is consid- ered. Results show how the quantile regression is more able than the OLS one to catch the e®ect of each covariate on student performance.