Econometric cost analysis in vertically separated railways

This thesis is concerned with addressing the cost analysis challenges in vertically separated railways. Both the challenges in infrastructure management and passenger railway operations are considered. A hedonic cost function is applied to better incorporate measures of heterogeneity of output into passenger train operation cost analysis. This allows for a richer understanding of the cost structure of the industry, through explicitly making scale economies a function of output heterogeneity, which in turn allows for tenders to be specified in a cost minimising manner. Three example tender re-mappings are considered for Britain. It is estimated that two out of three actually increase costs, reflecting that the model implies that for very heterogeneous TOCs, returns to density can not be exploited. In addition, the thesis details methodological work in developing analytical frameworks to exploit a multi layer panel dataset comprising observations on regions of many individual infrastructure managers. As well as providing extra observations to estimate cost frontiers, the data structure permits estimation of a dual-level inefficiency model which separates sub-company persistent inefficiency from subcompany varying inefficiency. This decomposition gives an indication as to whether inefficiency predominantly varies within firm or between firms. The example shows statistically significant inefficiency variation at both levels, and importantly, failure to take into account the dual-level nature of inefficiency is shown to under predict inefficiency. The thesis also develops new techniques to quantify uncertainty in inefficiency predictions from stochastic frontier models. This has application across the applied efficiency analysis discipline and not just in railways. Overall, this thesis finds that robust cost and efficiency analysis can only be conducted through explicit allowance for heterogeneity in output (both observed and unobserved), ensuring sufficient data quantity and that data relates to the organisational level to which decisions are made and that, in any analysis, the impact of uncertainty is quantified.