The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Transit productivity analysis in heterogeneous conditions using data envelopment analysis with an application to rail transit
Author:	Martinez, Manuel J.
View Online:	njit-etd2001-078 (xvii, 170 pages ~ 13.3 MB pdf)
Department:	Executive Committee for the Interdisciplinary Program in Transportation
Degree:	Doctor of Philosophy
Program:	Transportation
Document Type:	Dissertation
Advisory Committee:	Bladikas, Athanassios K. (Committee chair) Pignataro, Louis J. (Committee member) Spasovic, Lazar (Committee member) Chien, I-Jy Steven (Committee member) Lutin, Jerome M. (Committee member)
Date:	2001-05
Keywords:	Rail Transit Transit Productivity Analysis Data Envelopment Analysis (DEA)
Availability:	Unrestricted
Abstract:	This dissertation extends transit productivity analysis by developing a new method of Data Envelopment Analysis (DEA), the linear programming approach to productivity analysis. The new model analyzes productivity of transit working under heterogeneous operating conditions. It is named Two-Farrell DEA for it applies DEA in two stages, DEA (1), that calculates the productivity frontiers at given operating conditions and DEA (2), that uses inputs adjusted by multipliers calculated in DEA (l). The model Two Farrell DEA calculated productivity benchmarks for each rail transit agency and estimated its potential for higher revenue or lower expense improvement. Additionally, the results identify two production techniques of rail transit, the sources of increasing returns to scale, the degree of flexibility to changes in the shadow prices of the inputs, and a method to prioritize investment for expansion of operations. Its indirect contribution to transit operations planning consists of checking the consistency and feasibility of new rail projects. Moreover, this dissertation includes the first correlation analysis made between productivity and operating conditions related to network form, factor analysis of transit operating conditions, the comparison of results between the new model to four other methods, and the evaluation of the empirical accuracy of methods with cluster analysis.