Bias and sources of error in discrete choice models: a critical reflection

Discrete choice models have attracted a lot of attention since decades as an alternative to traditional aggregate transport demand models. Since discrete choice models assume that the total utility of an alternative is not exactly known to analysists, researchers have commonly assumed that the utility consists of two components: deterministic utility and error term. The error term is assumed to be composed of different sources. This paper reviews studies on the effect of the different sources of uncertainty (Manski, 1973) such as measurement error, omitted relevant variables, and taste variation in discrete choice models. All this research is based on different assumptions about the source of uncertainty. In this study, we will compare the assumptions and discuss future research directions. In addition, we will discuss the possible effects of the different sources of uncertainty in a new type of discrete choice model: the random regret model.