Artículo de revista
Modeling the decoy effect with context-RUM Models: Diagrammatic analysis and empirical evidence from route choice SP and mode choice RP case studies
Fecha
2016Registro en:
Transportation Research Part B 93 (2016) 318–337
10.1016/j.trb.2016.07.012
Autor
Guevara Cue, Cristian
Fukushi, Mitsuyoshi
Institución
Resumen
Evidence outside transportation has suggested that the introduction of a decoy to the choice-set could increase the share of other alternatives. This evidence breaks the regularity assumption, which is at the root of the classical Random Utility Maximization (RUM) model with utilities that ignore the choice context. This article assesses the suitability of various context-RUM choice models that could overcome this limitation. For this we use a diagrammatic analysis, as well as Stated Preference (SP) and Revealed Preference (RP) transportation choice evidence. We begin confirming that the reported decoy outcomes cannot be replicated with the classical RUM models and that such a goal could be achieved instead using a set of five context-RUM models. We then show, for the first time, that the Asymmetrically Dominated (AD) and Compromise (CP) decoy effects were present in an SP route choice setting. We also show that, for a subset of individuals, the relative strength of the different decoy types was coherent with a Data Generation Process (DGP) defined by the Random Regret Minimization (RRM) or by the Regret by Aspects (RBA) parsimonious models. Then, we use cross-validation analysis where we found that RRM and RBA were superior to a classical Logit for all decoy types. Nevertheless, the ad-hoc Emergent Value (EV) model was consistently superior to all models suggesting that, although the parsimonious models may in theory replicate all decoy types, they seem to still make an incomplete representation of the DGP behind the overall decoy effect. We finally consider an RP mode choice experiment with which we detect, for the first time, an AD decoy effect in this choice setting. We also use this experiment to illustrate how to handle the decoy phenomena in a real context with various alternatives and variables. The article concludes summarizing the main contributions of this research and suggesting future lines of investigation for it.