Determining the best performing benchmarks for transit routes with a multi-objective model: the implementation and a critique of the two-model approach

Document Type

Journal Article

Publication Date


Subject Area

mode - bus, operations - performance, planning - service level


Benchmarking, DEA, Operational efficiency, Effectiveness, Transit routes, Two-model approach


In addition to their operational efficiency, the service level based performance dimensions of transit systems has begun to attract the attention of several researchers in recent years. The consideration of both operational efficiency and service level performance enables a more comprehensive performance evaluation for bus transit operators. This study aims to investigate the operational efficiency and service level performance of public transportation companies with data envelopment analysis (DEA). However, DEA might assign some highly efficiently operating routes as benchmarks to inefficient ones despite some unsatisfactory service level performance, and vice versa. This benchmarking might help to improve one performance dimension but can result in worsening the other. To overcome this problem, the two-model approach introduced by Shimshak and Lenard (INFOR 45(3):143–151, 2007) is utilized to determine the best performing routes. This approach removes the high operational efficient routes with low service level performance and high service level performers but operationally inefficient routes from the analysis and helps to define best performing benchmarks being able to enhance both operational efficiency and service level performance. At the end of the study, a critique of the two-model approach is presented. This paper makes three contributions to the practice of transit performance evaluation. First, it puts forward the necessity of multi-objectivity for the subunits of transit systems. Second, it demonstrates the applicability of the two-model approach in the transportation industry. And third, it points out that despite the usefulness of the two-model approach to determine the best benchmarks in a multi-objective model, the model suffers to assign input/output goals for inefficient decision making units where input/output variables of distinct objectives are the same or related to each other.


Permission to publish the abstract has been given by SpringerLink, copyright remains with them.