The accumulated regret of trip chaining.

Abstract

Past research on commuting behaviour has tried to assemble the factors that effectuate mode choice. Thereby, trip chaining was found to be an important influence. The aim of this thesis is to apply the effect of trip chaining in a random regret minimization (RRM) model. In this thesis, a revealed preference data set will be used where individuals have registered their travel behaviour for one day. The attributes of work-related trip chains have been summed up such that mode choice relies on all trips instead of merely the work-related trip. The main question of this thesis is to what extent a regret-based model with trip chaining data would improve model fit. This thesis not only contributes to existing literature by testing a new methodology to process trip chaining data, but also by applying it to the RRM model, where it has never been incorporated before. Model estimation was based on the cost and time for each trip, and precipitation data from the Royal Dutch Meteorological Institute (KNMI). The results showed that aggregating trip chain data did not improve model fit. Furthermore, it was found that combining software- and human estimated travel times and distances led to faulty parameter estimates. On top of that, the random regret minimization model showed signs of low robustness; convergence strongly depended on the starting values of the parameters. Nevertheless, there is indication that private modes of transport are preferred for trip chains. Future research is encouraged to further explore this effect of trip chaining in regret-based models. In addition, it is recommended to first test new trip chaining methods on stated preference data to control for the travel time and distance on which the decision-maker bases its decision.