Temporal Transferability of Long-Distance Trip Rates

The work presented here studies long distance trip rate changes between 1977 and 2001. The performance of trip forecasting models and relationship between trip rates and explanatory factors such as income and education are of primary interest. A long distance trip is defined as traveling one hundred miles or more from home to a domestic (U.S.) location and returning and includes all relevant transportation modes: private vehicle, commercial airline, train, etc. A negative binomial count model structure is found to be appropriate for business and non-business trip rate models. Household income is a consistently strong indicator of both business and non-business purpose trips. The inclusion of additional demographic factors did not significantly improve or degrade the estimation of trip rates. A transferability statistic based on the likelihood ratio test is used to conclude that the long-distance trip models used in this study are not statistically transferable to different years. Aggregate errors are approximately 20% with higher errors observed for business purpose compared to non-business purpose models. Additional intercity mode service quality and destination attractiveness factors are likely required to improve transferability. Future recommended work would be to apply a current model that considers changes in service quality and destination attractiveness to 1977 data to work around the shortcomings of the 1977 long distance travel survey.