Benchmarking performance through efficiency analysis trees: Improvement strategies for colombian higher education institutions

Abstract

We introduce benchmarking analysis based on state-of-the-art machine learning techniques applied to the measurement of efficiency to assess the performance of Higher Education Institutions (HEIs). We rely on Efficiency Analysis Trees (EAT) and its Convexified frontier counterpart (CEAT) to assess the efficiency of 144 private HEIs in Colombia and compare the results with those achieved with classical Data Envelopment Analysis (DEA). Both EAT and CEAT show a higher discriminatory power than DEA when determining efficiency scores. Our results identify the different splits of the production frontier, corresponding to each node of the efficiency tree, which groups HEIs according to specific management models. By identifying relevant peers for inefficient observations at the node level, we show which strategic guidelines can be adopted to improve the performance of each HEI. This process encourages mutual learning and suggests potential changes within each node leading to efficiency improvements.