|Abstract: ||Metro rail provides a vital role towards facilitating the travel needs of major urban economies, and has contributed substantially in transporting the population within cities. However, implementing a safe service to meet with the statutory requirements of operation is fraught with difficulties. Due to high capital expenditures and need for public money, metros are politically sensitive and are subject to scrutiny. Consequently, understanding variation in metro performance continues to be a major research objective. This has proven to be far from straightforward due to the complex nature of the industry and that metro operators are generally monopolistic in nature, with no source of performance comparisons in the same region. This emphasises the need for an international comparison.
This thesis focuses on technical efficiency, which concerns the use of input factors (such as capital and labour) to produce metro services. The study is bolstered by using a high quality panel dataset, consisting of 27 metro systems for the period 2004 to 2012. Additional insight into the variation of metro performance is provided as shortcomings in the literature include the lack of appropriate data and insufficient application of statistical techniques.
Three empirical contributions are provided. Firstly, by assessing the relative performance of a group of metro systems by calculating technical efficiency scores using Stochastic Frontier Analysis, the study reveals a number of drivers of performance that affect output efficiency. Secondly, the study identifies reliability to be a key influence, and this is subsequently investigated further. Count data regression models are estimated to reveal determinants of incidents which cause a delay to service and provide a means for carrying out future forecasting of incident rates. Finally, given the growing capacity restrictions experienced by metros, the study investigates the causal impact of introducing a technological treatment (in this case, moving block signalling) on technical efficiency using a Propensity Score Matching approach.|