Consideration of alternatives: development of hybrid modelling approaches and applications to transport mode choice

Transport mode choice models traditionally assume that individuals consider all available alternatives. However, this might not be reasonable, even when the number of alternatives is limited. The biggest challenge for analysts is, however, that consideration sets - defined as the sets of alternatives relevant to the individuals - are unobservable. When only choices are observed, it is impossible to identify whether this is the result of not considering the non-chosen alternatives or preferring the chosen alternative over the other alternatives in the consideration set. Hence, only reduced form models can be estimated to understand the driving factors behind the observed choice. This thesis contributes to the literature on consideration by investigating how direct and indirect indicators of consideration, collected during stated choice surveys, can be used to measure and better understand the role of consideration sets. It aims at answering three research questions. The first is how to measure consideration. By comparing the suitability of a series of indicators of consideration, it emerges that thresholds for attributes are the most informative indicators of consideration. A second research question concerns the driving factors of consideration. Results suggest that thresholds for attributes have a primary role in explaining consideration, and that thresholds themselves are a function of individuals’ socio-economic profiles. A third research question relates to the role that consideration effects play in the estimation of the state-of-the-art RUM-based discrete choice models. Compared to previous studies, it emerges that the identified impact of controlling for consideration effects reduces as soon as the analyst also accounts for unobserved preference heterogeneity. Although calibrated to specific transport choices (i.e. choice of the transport mode for longdistance and airport access trips), the methodology proposed (and, to a limited extent, the findings of this research) can be generalised to other contexts with similar characteristics, e.g. number of alternatives or degree of differentiation amongst the alternatives. These could be either in the transport sector (e.g. mode choice on either systematic or non-systematic urban trips, car purchasing choices, choice of a flight on long-distance trips) or even in other contexts such as purchasing decisions on durable household goods (e.g. TV) or consumer packaged goods (e.g. breakfast cereals or pasta, wine, cosmetics/fragrances)