Estimating bicycle route choice model in value-of-time space, a case study for Gothenburg

Abstract

Previous studies that have estimated bicycle route choice models have focused on the preference valuations in the Value-of-distance space (VoD), i.e., the valuation of a potential investment in infrastructure relative to the reference infrastructure is measured in the unit of travel distance. That is to say, one meter of cycling on separated bike path is equivalent to x meters of cycling in mixed traffic (reference). This paper presents an effort to estimate a bicycle route choice model in Greater Gothenburg area where the utility is specified in the Value-of-Time space (VoT), which assumes that cyclists primarily aim to minimize travel time rather than distance per se. Estimation in VoT space also allows for the use of model estimation results in Cost-benefit analysis (CBA) given that most benefits in CBA are measured in terms of travel time savings. To achieve this, travel time, desired speed and delays at intersections are estimated from GPS data. A Path-size Logit model is developed where a utility function for each alternative (route) is formulated using travel time rather than distance. Desired speed and delay at intersections are estimated at trip level and intersection level. The estimated preference weights in the VoT space are compared with the estimates in the VoD space using the same data and from the literature. Results suggest that valuation of separated bike path in VoT space is comparable to VoD space using the same dataset but lower than the ones in VoD space in literature. The valuation of very steep slope (>5 %) in VoT space is substantially lower. Commuting cyclists slightly more prefer separated bike path and have a substantially stronger preference for avoiding signalized intersections than non-commuting trips.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.