Understanding the costs of urban transportation using causal inference methods

With urbanisation on the rise, the need to transport the population within cities in an efficient, safe and sustainable manner has increased tremendously. In serving the growing demand for urban travel, one of the key policy question for decision makers is whether to invest more in road infrastructure or in public transportation. As both of these solutions require substantial spending of public money, understanding their costs continues to be a major area of research. This thesis aims to improve our understanding of the technology underlying costs of operation of public and private modes of urban travel and provide new empirical insights using large-scale datasets and application of causal econometric modelling techniques. The thesis provides empirical and theoretical contributions to three different strands in the transportation literature. Firstly, by assessing the relative costs of a group of twenty-four metro systems across the world over the period 2004 to 2016, this thesis models the cost structure of these metros and quantifies the important external sources of cost-efficiency. The main methodological development is to control for confounding from observed and unobserved characteristics of metro operations by application of dynamic panel data methods. Secondly, the thesis provides a quantification of the travel efficiency arising from increasing the provision of road-based urban travel. A crucial pre-condition of this analysis is a reliable characterisation of the technology describing congestion in a road network. In pursuit of this goal, this study develops novel causal econometric models describing vehicular flow-density relationship, both for a highway section and for an urban network, using large-scale traffic detector data and application of non-parametric instrumental variables estimation. Our model is unique as we control for bias from unobserved confounding, for instance, differences in driving behaviour. As an important intermediate research outcome, this thesis also provides a detailed association of the economic theory underlying the link between the flow-density relationship and the corresponding production function for travel in a highway section and in an urban road network. Finally, the influence of density economies in metros is investigated further using large-scale smart card and train location data from the Mass Transit Railway network in Hong Kong. This thesis delivers novel station-based causal econometric models to understand how passenger congestion delays arise in metro networks at higher passenger densities. The model is aimed at providing metro operators with a tool to predict the likely occurrences of a problem in the network well in advance and materialise appropriate control measures to minimise the impact of delays and improve the overall system reliability. The empirical results from this thesis have important implications for appraisal of transportation investment projects.Open Acces