Predictive validity in stated choice studies: a before and after comparison with revealed preference

Models that explain human behavior within a given domain must be capable of predicting behavioral changes within that domain. Stated response models encompassing stated preference and stated choice models employ hypothetical choice scenarios to generate transportation demand forecasts. This paper presents a validity test for the application of stated choice models to commuter mode choice. A before and after experiment lasting two years was used as the basis of a test for both stated choice and revealed preference forecasts and backcasts. Stated choice was estimated for employees prior to the implementation of two employee trip reduction policies. A revealed preference model was estimated for the same class of employees one year after the programs were initiated. The stated choice model forecasted mode shifts within one percent of the “after” implementation value; the revealed preference model backcasted to within 10 percent of the “before” initiation value