Estimating Value of Congestion and of Reliability from Observation of Route Choice Behavior of Car Drivers

Abstract

In recent years, a consensus has been reached about the relevance of calculating the value of congestion and the value of reliability for better understanding and therefore better prediction of travel behavior. The current study proposed a revealed preference approach that used a large amount of GPS data from probe vehicles to provide insight into actual behavior in choosing a route. Mixed path size correction logit models were estimated from samples of 5,759 observations in the peak period and 7,964 observations in the off-peak period, while a mean-variance model was specified to consider both congestion and reliability terms. Results illustrated that the value of time and the value of congestion were significantly higher in the peak period because of possible higher penalties for drivers being late and consequently possible higher time pressure. Moreover, results showed that the marginal rate of substitution between travel time reliability and total travel time did not vary across periods and traffic conditions, with the obvious caveat that the absolute values were significantly higher for the peak period. Last, results showed the immense potential of exploiting the growing availability of large amounts of data from cheap and enhanced technology to obtain estimates of the monetary value of different travel time components from the observation of actual behavior, with arguably potential significant impact on the realism of large-scale models.