Travel demand forecasting with stated choice data. : Swedish domestic long-distance trips.

Abstract: The travel demand forecasting capability is affected by the model performance and is restricted by the available modes. The Sampers models being used in long distance travel demand forecasting doesn’t deal with nonlinearity and contains no travel service qualitative variables. The RP models can’t forecast the travel demand of the hypothetical mode of the high speed rail. Meanwhile, the value of time which is an important indicator in cost-benefits evaluation needs to be estimated in more specific way. Functional form improving method of Box-Cox transformation is proved to be efficient in dealing with nonlinearity and so does the piecewise function is effective in using discrete variables. Variable related travel service quality is proved to be a significant estimate in enriching model specification. The value of time is re-evaluated by taking into account its distribution. The mean values are sensitive to the model specified and the extent of self-selection has been analyzed by comparing the value of different current mode users with different alternatives Based on the stated choice survey about the high speed rail, the preference and sensitivity is revealed by binary Logit model estimation. The minor difference between X2000 and high speed rail and self-selection effect imply the little attractiveness of high speed rail for both train users and air users.