Evaluating Innovative Travel Modes and Services in Lisbon: A Stated Preference Approach

Evaluating Innovative Travel Modes and Services in Lisbon: A Stated Preference Approach


L Yang, C Choudhury, M E Ben-Akiva, Massachusetts Institute of Technology, US; J Abreu e Silva, Universidade Tecnica de Lisboa, PT


This paper analyzes a practical Stated Preference survey design with multidimensional choices of travel mode, departure time and occupancy. The organization of such a large number of alternatives and attributes posed a challenge.


This paper presents a stated preference (SP) study conducted in Lisbon, Portugal, as part of the SCUSSE (Smart Combination of passenger transport modes and services in Urban areas for maximum System Sustainability and Efficiency) initiative of the MIT-Portugal program. The survey has been conducted using Internet and computer-assisted personal interviews for evaluating the acceptability and willingness-to-pay for innovative travel modes, services and management. The design of the survey was remarkable from several methodological aspects and in this paper we particularly focus on the multi-dimensionality of the SP choice scenarios.

The survey involved a large number of candidate innovative modes and services that needed to be tested and compared simultaneously with existing modes. A focus group study was conducted first with three complementary objectives: to find aspects of public transport, car, the new alternative modes and services that could act as attraction or repulsion factors, to identify attributes characterizing the new services that may be used in the SP survey, and to identify potential attitudinal aspects that could be included in the SP survey. The new travel modes included collective taxis, express minibus, one-way car rentals, park-and-ride systems with a tutored delivery of children to their schools, and one-way car rentals with heavy mode. Alongside the five existing modes (car, regular taxi, bus, heavy mode, bus and heavy modes), this yielded a choice set of up to ten alternatives per respondent. The first dimensional choice is the choice consisted of these existing and new travel modes. The level-of-service of these alternatives varied substantially on the time of travel; in particular there were significantly long travel time and high costs (in the form of congestion charge and parking enforcement) for traveling in peak hours. This is expected to strongly influence the individual travel pattern and the choice of departure time intervals was included in the survey as a second dimension. In addition, it is expected that these radically different modes and level-of-service are likely to foster the sharing of trips. A third dimension has been added in the choice structure: the choice of occupancy for private car, one-way car rentals and regular taxi. Due to the deployment of real-time information services, a new attribute of travel time variability also needed to be considered in the survey.

By virtue of the fact that alternatives in a multidimensional choice set share unobserved attributes along various dimensions, there exists a significant amount of literature focusing on the modeling techniques, such as joint logit model and nested logit model for destination and mode choice (Ben-Akiva and Lerman 1985), multinomial probit model for brand choice (Raap and Franses 2000), mixed multinomial logit and ordered logit model for residential location and car ownership decision (Bhat and Guo 2007), error components logit model for time-of-day and mode choice (De Jong et al. 2003), and mixed logit model for vehicle choice (Hess et al. 2006). However, most of the research deals with revealed preference (RP) data or SP data with simplified and clustered alternatives. To our knowledge, there has not been much research on the practical design of complex SP scenarios with multidimensional choices.

The organization and presentation of a multidimensional choice exercise with such a large number of alternatives and attributes posed a challenge for our SP survey, since cognitive burden may force respondents to adopt a simple decision strategy based on only partial information (Caussade et al. 2005). To minimize the complexity, each respondent was presented with two SP scenarios, four choice tasks, at most four alternatives per choice task, and at most seven attributes per alternative (Hensher 2006). In this SP survey, travel modes were classified into three groups with similarities: car-based group, public transportation and multi-modal group. Each respondent was asked to select his/her preferred travel mode together with departure time and occupancy in each group, and three preferred alternatives were presented to the respondent again for a final choice.

To make the choice tasks realistic and credible, the alternative availability and level-of-service were anchored against their RP trip characteristics. A novel experimental design was used and verified by a Monte-Carlo micro-simulation.


Association for European Transport