Modelling multi-period customer-searching behaviour of taxi drivers

Various taxi network models have been developed to investigate the consequences of implementing taxi regulation policies. These models rely on logit-based taxi customer-search models to depict the customer-searching behaviour of drivers. However, very few studies have sought to validate the hypotheses on which the logit models are based, or to determine the factors that affect drivers' customer-searching choices at different times of day. In this study, global positioning system (GPS) data from 460 urban taxis were used to validate the logit model and determine the factors underlying drivers' searching behaviour. The analysis reveals that cross-zonal travel distance, intra-zonal circulating distance, relative passenger demand and rate of return affect the customer-searching decisions of the drivers of vacant taxis and that their searching behaviour varies with the time of day. The data also reveal that the cross-district search behaviour of vacant taxi drivers is strongly related to passenger demand over time and space. The drivers of vacant taxis tend to travel to high-demand districts during the mid-night period when overall demand is low, but they avoid crossing districts during high-demand peak hours.