Diss. Stockholm : Kungliga tekniska högskolan. Skolan för arkitektur och samhällsbyggnad, Transportvetenskap, Väg- och banteknik, 2016

Models that can predict consumer choices are essential technical support for decision makers in many contexts. The focus of this thesis is to address prediction problems in discrete choice models and to develop methods to increase the predictive power of these models with application to car type choice. In this thesis we challenge the common practice of prediction that is using statistical inference to estimate and select the ‘best’ model and project the results to a future situation. We show that while the inference approaches are powerful explanatory tools in validating the existing theories, their restrictive theory-driven assumptions make them not tailor made for predictions. We further explore how modeling considerations for inference and prediction are different. Different papers of this thesis present various aspects of the prediction problem and suggest approaches and solutions to each of them. In paper 1, the problem of aggregation over alternatives, and its effects on both estimation and prediction, is discussed. The focus of paper 2 is the model selection for the purpose of improving the predictive power of discrete choice models. In paper 3, the problem of consistency when using disaggregate logit models for an aggregate prediction question is discussed, and a model combination is proposed as tool. In paper 4, an updated version of the Swedish car fleet model is applied to assess a Bonus-Malus policy package. Finally, in the last paper, we present the real world applications of the Swedish car fleet model where the sensitivity of logit models to the specification of choice set affects prediction accuracy.