Abstract
Forecasting the demand for alternative-fuel vehicles (AFVs) is quite important for manufacturers, fuel suppliers and environmental planners. AFVs have attributes such as reduced range and limited refueling options that are very different from existing vehicles. Therefore stated preference (SP) data is necessary for demand models. Previous work by Brownstone, Bunch, and Train (1998) shows that there are serious biases in these stated preference data. Another source of households’ vehicle preference, is households’ actual observed transaction behavior (Revealed preference (RP) data). I develop a dynamic stated and revealed preference vehicle transaction model which uses the RP data to control for the biases of using pure SP data in order to better forecast households’ demand for AFVs for California. I implement a “scale factor” to specify the relationship of the different variances of the RP and SP data. Moreover, I examine the nested structure over different fuel-type vehicle choices and estimate both the multinomial logit (MNL) and nested logit (NL) models. In addition, I conduct forecast using the pure SP and joint SP-RP MNL models under the scenario consisting of new vehicle technologies for year 1998. Compared to the new vehicle sales statistics, it is obvious that the joint SP-RP model provides more reasonable forecasts. I also examine the different substitution patterns implied by the pure SP MNL and NL models when new vehicle choices are introduced. The NL model predicts more realistic substitution pattern. I also add the used vehicle choices to the forecast scenario to make the forecast more realistic because the used vehicle market is taken into consideration. Large panel data sets have been collected by the California Alternative-Fuel Vehicle Demand Forecast Project since May 1993. These data contain extensive information on households’ stated and revealed preference vehicle transactions, vehicle utilization and households’ socioeconomic characteristics. This study serves as an example of how to forecast new products or technologies that mark considerable departures from existing products or technologies.