Taxis play important roles in densely developed metropolitan areas, supplementing mass transit by providing door-to-door service with unrestricted operating hours. Taxis cruising on urban streets to pick up passengers, however, add to the road traffic, sometimes quite substantially. Wong et al. (2001) report that taxis overall constitute 25%, and in some places as much as 50 to 60%, of traffic in Hong Kong. Taxis waiting for customers at a taxi bay may cause problems as well when the capacity of the bay is inadequate and the queue of taxis extends onto the street and obstructs road traffic, a phenomenon often observed in Japanese large cities. These problems call for the development of infrastructure, regulations and strategies for taxi operation. This in turn calls for a thorough understanding of the behavior of taxi drivers and the nature of the demand for taxi trips. Little is known on either of them at this point. This study addresses the former.
This paper is organized as follows. In the next section, the structure of the discrete-state, stochastic process model system of taxi behavior is described in its general form. In the following section, taxi drivers' choice between cruising and waiting after delivering passengers is formulated as a discrete choice model. A simplified stochastic process model is then formulated based on the probe data from Nagoya, and its properties are discussed. Results of a numerical analysis are presented in the following section to show that drivers' individualistic rational choice and system optimality are not consistent with each other. Finally, taxi drivers' choice is analyzed empirically to demonstrate that their choices are indeed rational, and act against system optimality.