|Ph.D Student||Tenenboim Reichenber Einat|
|Subject||Integrating Subjective Travel Time Estimates in Travel|
Behavior Modeling of Destination and Toll Route
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Yoram Shiftan|
|Dr. Nira Munichor|
|Full Thesis text|
Travel time is considered one of the main determinants of travel behavior and a key factor in travel demand modeling. Many efforts are constantly made to estimate useful models for forecasting travel behavior, yet recent reviews of forecasting outturns reveal poor estimation quality. Given that people base their travel decisions on their perceptions rather than objective attributes, the motivation of the present endeavor was to challenge the common practice of employing travel time in its objective form. A sensible integration of subjective time data within the modeling process is expected to yield improved forecasting accuracy. Focusing on pre-trip time estimates, we employed fundamental and recent psychological theories for promoting the understanding of subjective time estimation. In an initial experimentation, the effects of several factors on estimated travel time were assessed. According to one factor, the return trip effect, travelers estimate return trips as shorter in time compared to outbound trips. In a questionnaire, 174 respondents provided time estimates either from their home to eight local shopping areas (the outbound group), or from these shopping areas to their home (the return group). A return trip effect was found only for trips to/from poorly familiar areas, highlighting the role of familiarity. An additional factor identified was toll road travel, as drivers estimated toll trips via the Carmel tunnels as shorter than non-toll trips. Presumably, merely thinking about paying the toll had led drivers to form expectations of travel time savings in exchange (the payment account). Alternatively, it is possible that drivers estimated toll trips as shorter aiming to justify their route choice (the choice account). These two accounts were systematically studied in a field experiment, wherein shoppers departing a mall located in proximity to the tunnels were questioned about their imminent trip. All participants were paid 10 NIS (approximately 2.5 Euros), yet some were told it was to cover the cost of the toll (toll-paid group), whereas others were told it was a participation fee (control group). This payment manipulation and drivers’ (toll) route choice yielded a 2*2 experimental design. In a questionnaire, 386 drivers indicated their chosen route, alternative route, and estimated times for both. Actual travel data were obtained using a route tracker application installed on participants’ smartphones. Data analysis yielded support for the choice account, as drivers’ who chose the toll route substantially exaggerated in their time savings estimation, suggesting they attempted to justify their route choice. No support was found for the payment account. In both experiments, drivers’ choices were modeled using multinomial logit. Experiment 1 included shopping destination choice modeling, wherein estimated time showed a significant contribution, almost equivalent to that of actual time. A combination of both time measures yielded an even better fit. In Experiment 2, modeling of (toll) route choice with estimated time revealed a substantial improvement compared to actual time. Subjective time data seems to entail unique valuable information regarding travelers’ preferences. Understanding the way travelers estimate travel time is critical for gaining a better insight into travel behavior, thereby improving forecasting accuracy.