Modelling user perception of taxi service quality Borja Alonso, Rosa Barreda, Luigi dell’Olio * , Angel Ibeas University of Cantabria, Grupo de Investigacion de Sistemas de Transportes (GIST), Av. De los Castros 44, 39005, Santander, Spain ARTICLE INFO Keywords: Taxi Service quality Ordered probit Random effect Systematic variation in taste ABSTRACT This article presents a research study on modelling taxi user perceived quality. Following a thorough review of the available international literature, relevant variables are individualised using focus groups of taxi users, and a satisfaction survey is designed. The resulting data are used to estimate two types of ordered probit models, which consider systematic and random variations in taste: on the one hand, a model that provides attributes and their influence on the perceived quality of taxi service when no previous information about the system is available and, on the other hand, a model that considers their changed perception after being informed about the attributes that could affect the system's perceived quality. The results mention the attributes that the users normally value and those that they would value after being asked to reflect on them. It is worth noting that waiting time is the most important factor for frequent users, and journey time is highly valued by almost all users, along with safety, accessibility and comfort. Surprisingly, the fare is not usually as important as at the first thought. The results of this type of study are valuable because they allow us to understand and improve the system depending on user preferences, thereby improving the quality of services and increasing the demand. 1. Introduction The constant growth in the number of private vehicles travelling on urban roads causes problems such as noise, congestion, pollution and traffic jams. All administrations address these issues by promoting sus- tainable mobility, a policy fundamentally based around the use of public transport. These goals can be achieved by using management tools to support policy application while optimising the use of available re- sources. The first step in guaranteeing sound resource management consists of applying user and journey characterisation studies to deter- mine how users perceive the quality of service when deciding to make their journeys by public transport. From a point of view of sustainable mobility, the taxi is not the best way to travel around an urban area; nevertheless, it has its environmental advantages in that people will generally use it to avoid problems asso- ciated with parking and for its speed and ease of use. Taxi use avoids the creation of ‘parasitic traffic’ due to cars cruising for a parking place, and it frees up public space because one vehicle provides a service to many users who have not used their own vehicle and who would otherwise be occupying street space and ultimately a parking space. To encourage continued public use, the service needs to operate in accordance with certain quality standards that would be better under public control. This article proposes an overall methodology for studying the quality of service perceived by taxi users. The methodology includes the design of a satisfaction survey and the use of ordered probit models on the data collected to propose marketing strategies aimed at increasing taxi use. The paper is structured as follows: the introduction is followed by an analysis of the state of the art, and then, the methodology used is pre- sented, which includes the design of the questionnaire and an explana- tion of the models used in the analysis of the data collected. A case study is then described, and the results are analysed. This is followed by the most important conclusions drawn from the work. 2. State of the art The perception and understanding of user behaviour relative to service quality is vital in researching any mode of transport (Tam et al., 2008; dell’Olio et al., 2011). An abundance of literature is available on the methodologies for evaluating the perceived quality of service of transport modes. These were based mainly on the debate and application of two of the most well known and most widely used methodologies: SERVQUAL (Parasuraman et al., 1988) and SERVPERF (Cronin and Taylor, 1992). Alternative approaches were applied later: the use of neuronal net- works (Behara et al., 2002; Garrido et al., 2014), discrete choice models * Corresponding author. E-mail address: delloliol@unican.es (L. dell’Olio). Contents lists available at ScienceDirect Transport Policy journal homepage: www.elsevier.com/locate/tranpol https://doi.org/10.1016/j.tranpol.2017.12.011 Received 11 January 2016; Received in revised form 11 October 2017; Accepted 6 December 2017 0967-070X/© 2017 Elsevier Ltd. All rights reserved. Transport Policy 63 (2018) 157–164