Contrarians do better: Testing participants' response to information in a simulated day-to-day route choice experiment

This paper reports the results of a laboratory-like experiment aimed at investigating some key aspects of day-to-day route choice behavior in a simple two-link network. Two different information scenarios (full vs. partial information about previous travel times) were implemented as sequential phases of the experiment. It was found that User Equilibrium conditions, though reached on several occasions, did not persist as a steady state. The temporal evolution of route changes showed no significant trend within each phase of the experiment, but a considerable "between-phase" decrease in route switching activity was detected. The analysis of participants' modes of response to information reveals a prevalence of direct ("naive") over contrary ("strategic") behavior, and shows that contrarian subjects are, on average, more successful in terms of individual travel time than those reacting in a direct manner. Memory depth (the number of past values of travel times typically considered by participants as the basis for their next route choice) was found to be significantly associated with individual performance and prevailing mode of response only under conditions of partial information. Finally, the implementation of a discrete-time, deterministic process model of day-to-day route choice dynamics on the experimental network enabled the derivation of empirically based estimates of the route choice reviewing rate, the weight of the most recent experience in the formation of travelers' forecasts of travel times, and the dispersion parameter of the underlying route choice model.