S. Abbes, M.A. Mezghani

Pages: 163-178

Abstract
The purpose of this paper is to examine technical efficiency in the Tunisian public transit system and its determinants through parametric and non-parametric techniques. Our analysis is based on a sample of a 12-bus transit company over the 2000-2017 period. We have conducted a two-stage analysis. First, we evaluated input-oriented Data Envelopment Analysis (DEA) scores under different assumptions. Then we regressed efficiency scores on potential contextual variables using the bootstrapped regression and right truncated regression. The DEA results show that the average efficiency scores range from 0.5 to 1. According to the right-truncated and the bootstrapped-truncated regression; the proportion of subsidies, vehicle age and population density were identified as factors that significantly deteriorate efficiency. Conversely, network density and the existence of universities in the considered region improve efficiency. The calculation of margin effects highlights the importance of vehicle age and network density, compared to the other variables. Our study makes a significant contribution to the literature by employing a new contextual variable, and by conducting the Simar-Wilson bootstrapping analysis using input-oriented DEA scores. The findings of this study provide also useful observations and insights for urban planners and policymakers across cities.
Keywords: public transport; Data Envelopment Analysis; contextual variables; bootstrapped truncated regression; Simar- Wilson; marginal effects