Metro Station Operating Costs: An Econometric Analysis

This article develops an econometric analysis of metro station operating cost to identify factors that create variation in cost efficiency. Stations operating costs can be classified amongst the semifixed costs that a metro faces in the sense that they do not vary proportionately with metro output. They may therefore be important in determining the degree of returns to density. This article seeks to provide an improved understanding of some of the major factors driving these costs. Empirical results show that strong system-specific influences impact costs but over and above these we detect positive associations from a range of station characteristics, including the length of passageways, number of platforms, peak-level service frequency, interchange demand, and the provision of toilet facilities. In addition, we find that the presence of air-conditioning has a substantial effect in increasing expected station operating cost by as much as 40 percent. Introduction The cost structure of the mainline railway industry has received a great deal of attention in the academic literature (e.g., Caves et al. 1980; Caves et al. 1981a, 1981b; Freeman et al. 1985; Caves et al. 1985; Dodgson 1985; McGeehan 1993; Bookbinder and Qu 1993; Oum and Yu 1994; Cowie and Riddington 1996; Wunsch 1996; Tretheway et al. 1997; Oum et al. 1999; Cantos et al. 1999; Cantos et al. 2002). Research has demonstrated the very large variance in cost efficiency, or Journal of Public Transportation, Vol. 10, No. 2, 2007 94 productivity, that is often present within a sample of rail firms and has developed cost and production function approaches to analyze the factors underpinning this variance. A prominent theme in the rail efficiency literature is whether cost structures are subject to returns to scale (RTS) or returns to density (RTD). RTS describe the relationship between all inputs and the overall scale of operations, including both output and network size. RTD describe the relationship between inputs and outputs with the rail network held fixed. Evidence in the literature indicates that RTD are due to the prevalence of fixed costs in the rail industry and to a range of semifixed costs that do not vary proportionally with output. Less consistent evidence is available on the existence of scale economies, though the majority view is that railways operate under constant returns to scale. Few studies have been conducted on the costs structure of urban metros, though Graham et al. (2003) estimates increasing RTD and constant RTS. Station operations may provide an important source of increasing RTD in metro operations. Stations must remain staffed and functioning, with all the energy and other resources required, throughout the duration of the metro operating hours. Moreover, costs may differ quite substantially from one station to another due to the nature of engineering, the depth of station, its size and dimensions, the technology employed, and so on. So we can conceive of station operating costs as semifixed costs that do not vary proportionately with system throughput and therefore may be instrumental in giving rise to increasing RTD. In this study we develop an econometric model to analyze variance in station operating costs. An econometric model is essential to determine the effect of a particular characteristic of a metro station on its operating costs while controlling for all other factors affecting the metro station operating cost. The analysis of historical data fails to control for the effects of other factors while estimating the effect of a particular factor. We use data on 83 stations from 13 metro systems from around the world to estimate the main drivers of cost. Model specifications and the data used for estimation are discussed and results are presented. Model Specification and Data The data available for our analysis describe the total operating cost of each station and a range of station characteristics collected from a total of 13 metros (Buenos Aires, Dublin, Glasgow, Hong Kong KCR, Hong Kong MTR, Lisbon, London, Montreal, Naples, Sao Paulo, Singapore, Taipei, and Toronto). The analysis we develop Metro Station Operating Costs: An Econometric Analysis 95 below regresses the total operating costs against these station characteristics to determine their role in influencing variance in costs. It is important to stress that we do not adopt a conventional cost function approach. We do not have data on factor prices and therefore cannot estimate the cost function. However, another important consideration in this respect is that since the operating costs of any one particular station represents only a small fraction of total metro operating costs, individual stations cannot be regarded as the appropriate units over which cost decisions are made. For instance, metro operators do not demand factor inputs at the station level in accordance with prices but make rational decisions relating to costs and operations for the system as a whole. Furthermore, it would be wrong to ascribe any particular behavioral assumptions to individual stations (e.g., cost minimizing behavior). A metro may not seek to sustain a set level of station efficiency across the system but rather allow for disparities in efficiency to achieve some broader objectives relating to the appropriate level of system output given overall costs. In this respect, it is mainly how the station characteristics serve to influence total cost that is of interest in the present analysis. One important issue, however, relates to the absence of factor price data, because this will certainly be important in determining station costs. To control for these omitted variables, which we cannot observe, we estimate the station operating costs model with a set of dummy variables for the 13 metro systems. We assume that these dummies will capture unobserved system-specific effects including factor prices. A log linear model is used to identify the factors that influence the operating cost of a metro station. The model can be written as: 1n yi = +1nXi+Di+i (1)