Interrogation of responses to stated choice experiments: Is there sense in what respondents tell us?

Abstract

There is an extensive and growing literature on the design and use of stated choice experiments. Such experiments are generally regarded as the preferred framework within which to collect data that is able to reveal the preferences of individuals in a particular application context. Analysts have developed sophisticated ways of analysing such data, typically using a form of discrete choice model that identifies the marginal (dis)utility associated with each observed attribute linked to an alternative, as well as accounting for sources of preference and scale heterogeneity. There is also a growing literature studying the attribute processing rules (or heuristics) that respondents use as a way of simplifying the task of choosing, for all manner of reason. We find that there is relatively less effort placed on looking at the data defining each choice situation for each individual, as represented by the stated choice experiment and the choice response, and seeing if there are ‘evidential’ rules that support in a rational way, the choice responses. This is a complex issue since individuals are complex in their decision making, but it is possible to build up a number of ‘evidential’ rules that can be used to establish confidence in the stated choices. This paper explores the evidence in building confidence in the data used to infer the many behavioural outputs offered from the estimation of choice models using stated choice data. We find a high level of confidence in the evidence, and identify at least two features of choice set processing, namely value learning and majority of confirming dimensions, that are worthy of inclusion in the estimation of choice models. The evidence suggests that there is a great deal of behavioural sense in stated choice responses, for all manner of possible reason.