Linear or nonlinear utility functions in logit models? The impact on German high-speed rail demand forecasts

We have shown that nonlinearity of the representative utility functions of the Logit model applied to an intercity mode choice model implied the presence of asymmetry and of thresholds in the response curve, as well as results that had more reasonable properties (diminishing marginal values of time savings, elasticities and values of time that differed among the modes), and yielded very different high-speed rail market shares for Germany than results obtained with the usual linear utility functions.