A discrete choice modeling approach to measure susceptibility and subjective valuation of the decoy effect, with an application to route choice

Abstract

The decoy effect reveals a potential violation of the regularity assumption, which is a building block of canonical discrete choice models. This effect has been detected in various choice contexts, but the susceptibility to it and its subjective valuation have been scarcely studied before. This paper proposes, illustrates and assesses two methodologies aimed to fill this gap: systematic taste variations and latent classes. The first approach uses a logit model that accounts for the decoy effect by an emergent value term which coefficient varies among groups of individuals defined by exogenous variables. The second method uses a similar scheme but accounts for heterogeneity, instead, by using three variations of the latent classes approach. This article illustrates and assesses the proposed methods using a case study build from a stated preferences route choice problem that depicts a hypothetical Saturday afternoon car shopping trip. Results show first that both methods are feasible for analyzing the decoy effect and that only respondent's age and response time seem to play a clear role in decoy's susceptibility in the case study. On the contrary, no noticeable impact on susceptibility is detected for gender, imputed income, motorization rate, household size and whether the survey was conducted in person or online. Finally, regarding decoy's subjective valuation, evidence from this case study suggests that presenting the decoy was, on average, as large as reducing about one fourth of the travel time, or between one fourth and one half of the travel cost displayed to the interviewees.