Proceedings of the Institution of Civil Engineers Transport 163 November 2010 Issue TR4 Pages 203–210 doi: 10.1680/tran.2010.163.4.203 Paper 900019 Received 01/04/2009 Accepted 10/12/2009 Keywords: mathematical modelling/transport management/ transport planning R. M. N. T. Sirisoma Post-doctoral Fellow, Department of Civil Engineering, Schulich School of Engineering, University of Calgary, Alberta, Canada S. C. Wong Professor , Department of Civil Engineering, The University of Hong Kong, Hong Kong SAR, China William H. K. Lam Professor, Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong SAR, China Donggen Wang Professor, Department of Geography, Hong Kong Baptist University, Kowloon Tong, Hong Kong SAR, China Hai Yang Professor, Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Kowloon, Hong Kong SAR, China Peng Zhang Professor, Shanghai Institute of Applied Mathematics and Mechanics, Shanghai University, Shanghai, China Empirical evidence for taxi customer-search model R. M. N. T. Sirisoma BSc(Eng), PhD, MITE, AMIE, S. C. Wong MPhil, PhD, FCILT, FHKIE, FHKSTS, MASCE, MITE, MIHT, W. H. K. Lam MSc, PhD, FHKIE, FHKSTS, FIHT, MCILT, MICE, MITE, D. Wang MA, MSc, PhD, FHKSTS, H. Yang BSc, DEng, FHKSTS and P. Zhang MSc, PhD A mutinomial logit model of urban taxi services has been developed for the study of the operational characteristics of the taxi industry, in which it is hypothesised that the customer-searching behaviour of vacant taxis follows a multinomial logit choice model. Although the model is commonly used, little empirical evidence exists to validate the choice mechanism and determine the set of important factors that affect the choice. In the present study, a stated preference survey of 400 taxi drivers was conducted to analyse the customer-searching behaviour of vacant taxis. The results explain how the considered parameters of waiting time, journey time, travel distance and toll affect the driver behaviour when searching for customers. In addition, market segmentation analysis was carried out to study the effects of driver demographics and operational characteristics on the searching behaviour. The parameters that were considered in this section of the study were the age of the driver, taxi ownership, driver experience and marital status. NOTATION c sr generalised travel time via the shortest route from zone s to zone r LR log likelihood ratio L R log likelihood for the base model calibrated for the combined dataset with all market segments L U sum of the log likelihoods of the sub-models calibrated for different market segments P r =s probability that a vacant taxi originating in zone s will meet a customer in zone r U i utility function of a taxi driver in searching for the next customer V i deterministic component of the utility w r waiting (search) time in zone r x i D distance for individual i x i JT journey time for individual i x i T toll for individual i x i WT waiting time for individual i â JT journey time coefficient â D distance coefficient â T toll coefficient â WT waiting time coefficient ª weighting that a taxi driver places on waiting compared to journey time å i random component of the utility function for individual i Ł degree of uncertainty of customer demand and taxi services in the whole market í D value of distance í JT value of journey time í WT value of waiting time 1. INTRODUCTION Taxi services play a vital role in the transportation system as they offer door-to-door service to passengers, and are a more comfortable and convenient transportation mode than other public modes of transportation, especially for the travel requirements of disadvantaged groups such as the disabled and elderly (Hensher, 2007; Petzall, 1995; Pucher and Renne, 2003; Su et al., 2010). The main issue of taxi services is that although an occupied taxi has to take a direct route to the customer’s destination, there is no guarantee that the driver of the vacant taxi can find a new customer at the destination. This leads to inefficient movement of vacant taxis, which circulate on roads in search of customers and cause traffic congestion and air pollution in surrounding areas. The taxi drivers’ knowledge of the market is limited and updated only from their experience (Kim et al., 2005). Traditionally, studies use an aggregate approach to investigate the taxi industry, which is subject to various types of regulation, such as entry restriction and price control, and the economic consequences of regulatory restraints (Arnott, 1996; Beesley and Glaister, 1983; Cairns and Liston-Heyes, 1996; De Vany, 1975; Douglas, 1972; Foerster and Gilbert, 1979; Frankena and Pautler, 1986; Loo et al., 2007; Manski and Wright, 1976; Schroeter, 1983; Shreiber, 1975; Yang et al., 2000, 2005a, 2005b), but they do not take into account the spatial structure of the taxi market. In view of the necessity and importance of network modelling of taxi traffic, Yang and Wong (1998) made an initial attempt to characterise taxi movements in a road network for a given customer origin–destination demand pattern. They proposed a simultaneous system of equations to describe the movements of both vacant and occupied taxis, and solved the problem with a fixed-point algorithm. The model explicitly deals with the effects of taxi fleet size and the degree of taxi driver Transport 163 Issue TR4 Empirical evidence for taxi customer-search model Sirisoma et al. 203