The cost of bus travel time variability
The reliability of bus systems is a vital issue on the transport agenda, since urban areas are yearning for high quality alternatives for the private car. A key indicator of reliability is a low level of day-to-day travel time variability (TTV). To obtain funds for reducing TTV, it is necessary to give evidence for the benefits from such improvement, but current tools for estimating the cost of TTV are insufficient. This thesis covers issues that arise when analysts need to show that improved bus infrastructure brings benefits from reduced TTV. The first part of the thesis aims at understanding how the attitudes of travellers to TTV can be converted into monetary terms. The design of a survey is described, where respondents trade-off between TTV and other attributes. A modelling experiment, based on the survey responses, finds that the effect of TTV is best explained using variables that represent trip scheduling considerations. Following is a series of experiments that seek to estimate the willingness-to-pay for reduced TTV in a way that is sensitive to taste variation between travellers. Several Mixed Logit models are estimated, but some doubts about their credibility are raised, and hence the same willingness-to-pay estimates are also computed using nonparametric techniques. Some conclusions are drawn regarding the process of estimating heterogeneous willingness-to-pay and the ability to recognise the willingness-to-pay from survey data. The starting point for the second part of the thesis is the lack of tools for estimating the level of TTV in hypothetical scenarios. We, consider the case for using traffic microsimulation to estimate TTV by running a microsimulation model multiple times, and looking at the variation between runs as an estimate of the variation between different days. Such concept of estimation requires a special calibration methodology, which sets the level of simulated inter-run variability at a similar level to inter-day variability in the real network. A full calibration methodology is developed, tackling methodological, computational and statistical issues. Finally, the demand and supply methodologies are combined, and it is illustrated how the savings from improved bus infrastructure can be examined. The contribution of the entire study includes methodological and technical insights into modelling the attitudes to TTV, estimating the distribution of the willingness-to-pay and calibrating traffic microsimulation models; but it also brings up policy issues concerning the role of TTV in transport appraisal.