Latent variables as a proxy for inherent preferences: A test of antecedent volition ☆

We present a test of antecedent volition under conditions of varying choice complexity. Using a stated preference design, decision makers make hypothetical transport choices under three decision scenarios. The choices are made under conditions of three, five, and seven alternatives per choice set. Further, we analyze the data from these three experiments simultaneously using factor-analytic structural choice models. We fit these models because they allow us to specify a latent structure representing a behavioral process of antecedent volition. Past research emphasizes differences in aggregate preferences due to context effects, but is incomplete in some important respects. Model parameters of the most theoretically appealing specification include estimates of means (i.e., aggregate preferences for the attributes studied) and regression coefficients for the effects of latent variables on the taste sensitivities common to the decision scenarios (i.e., structure of preferences in relation to the attributes). Extending past research, we find evidence of latent variables and structures that are common to the taste sensitivities of specific attributes across the decision scenarios (i.e., sources of preference heterogeneity and a structure to the heterogeneity). This pattern of results is suggestive of a behavioral process consistent with literature and theory on antecedent volition. Our model forms and results have implications for policymakers and researchers. They are especially salient to recent literature contrasting context effects with the notion of stable and inherent preferences.