How to better represent preferences in choice models: The contributions to preference heterogeneity attributable to the presence of process heterogeneity

Discrete choice studies, with rare exception, commonly assume that agents act as if sources of observed utility are captured through a linear in parameters and additive in attributes (LPAA) form, with some interactions. A growing number of transport (and other) choice studies have investigated one or more alternative processing rules adopted by agents in arriving at a choice, raising interest in how best to represent the utility expressions in a joint process and outcome choice model. Given the popular and appealing random parameter treatment of LPAA in mixed logit as a way of identifying non-systematic preference heterogeneity in a sample, this paper considers the possibility that we might be able to interact specific process heuristics with LPAA to uncover sources of systematic preference heterogeneity hidden in the standard LPAA form, and hence establish a link between the LPAA form and candidate process heuristics, offering a way to embellish and hence clarify the contributions to preference heterogeneity attributable to the presence of process heterogeneity. Specifically, we are interested in the extent to which there is a systematic relationship between the simple LPAA form and the more complex (albeit behaviourally realistic) process heuristics emerging in the transport literature which we call conditioning by random process heterogeneity (CRPH). In this paper, in addition to LPAA, we consider two process heuristics - Value Learning, and Relative Advantage Maximisation - with an overlay to account for risk attitudes, perceptual conditioning, and overt experience. The findings, using two data sets, suggest that empirically there exists a significant attribute-specific relationship between preference heterogeneity identified through specific process heuristics and through the LPAA assumption.