The Potential of Online Respondent Data for Choice Modeling in Transportation Research: Evidence from Stated Preference Experiments using Web-based Samples

Abstract

The aim of this thesis is to analyze the potential of online survey services for conducting stated preference experiments in the field of transportation planning. Several web-products for hosting questionnaires are evaluated considering important features required when conducting a stated preference survey. Based on this evaluation, the open-source platform LimeSurvey is the most appropriated for this kind of research. A stated preference questionnaire about pedestrians’ route choice in a Singaporean urban context is submitted to three different samples: a commercial sample of Survey Sampling International (SSI), workers of Amazon Mechanical Turk (MTurk) and a convenience sample of urban planning specialists. Samples’ quality is investigated based on the answers’ collection time, the survey completion rate, the answer consistency, the representativeness of the Singaporean population and the reliability of the choice behavior. The last criterion is assessed by computing multinomial choice models. After evaluation, the SSI sample is elected as the best online sample investigated. Especially, it provides the only one valuable choice model. The reliability of the choice behavior of this sample is comparable to the one of a Singaporean field survey used as reference. Finally, effects of stated preference design variations on answer consistency and reported choice behavior are investigated. It is observed that respondents who answer surveys with pictorial representations perceive better the experiment attributes, especially the walking time.