This paper incorporates both interindividual variation and intra-individual variation into the modeling of car usage time frontiers (CUTFs). A CUTF is defined as the unobserved maximum amount of time that an individual private car user is willing to spend driving and is derived from the concept of a travel time budget. Long-term GPS data collected from private cars in Toyota, Japan, were used. To deal with the panel data, a stochastic frontier model with random parameters was applied as the modeling methodology. The fit of the data for the estimation results demonstrated that models with random coefficients were preferable. Drivers’ CUTFs on workdays were significantly affected by the departure time of the first trip, temperature, home location, gender, age, and occupation. All those explanatory variables except temperature also significantly affected CUTFs on holidays. When the intraindividual variations were ignored, only a few explanatory variables had a significant effect on CUTFs. Predictions made with the estimated parameters showed that the expected CUTFs were about double the corresponding actual times of car usage (expenditures). Therefore, CUTFs are underestimated when intraindividual variations are ignored.
Read full abstract