Abstract
The level and extent of demand for a transportation service, including the determinants of the demand, can be meaningfully analyzed only by incorporating their evolution over time. Since most travel demand models are based on cross-sectional data, longitudinal analytic methods need to be developed for the study of travel behavior. Heterogeneity and non-stationarity of behavior, lagged effects, and effect of time varying variables are other factors that require using dynamic modeling techniques. A dynamic beta-logistic model using a panel data set of approximately 2,200 Southern California commuters was developed to fulfill this need. Waves 1, 5, and 8 of this panel, which encompasses a period beginning February, 1990 to February, 1993 was used. Seventy five percent of Waves 1 and 5 data were randomly sampled for model development. The remaining 25 percent as well as the data from Wave 8 were used in model validation. The model had a successful prediction rate of about 98.6% for the two two-wave periods between Waves 1 and 5 and between Waves 5 and 8. Policy simulations were carried with Waves 5 and 8 data. For policy simulation, the impact on ride-sharing of reserved parking, cost subsidy, and guaranteed ride home incentives were studied. An increase of over 100% in the usage of shared-ride mode in Waves 5 and 8 was predicted when all respondents were simulated to have perceived a set of three incentives in both waves. This increase in the shared-ride alternative corresponded to a decrease of over 42% in the usage of the drive-alone modes in both waves. There was a decrease of about 35% in the drive-alone alternative when the three incentives were perceived by all commuters only in Wave 5. If the three incentives were perceived by all commuters in Wave 8 only, the drop in solo-driving in the two-wave period was only 7.1%, which demonstrates the existence of lagged and delayed effects in travel behavior. Of the three incentives guaranteed ride home induces the biggest reduction in the use of the drive-alone alternative.