Abstract
This paper shows the explanations of the compromise and attraction effects that are inconsistent with utility maximization via the Generalized Nested Logit (GNL) model that is one of the Generalized Extreme Value (GEV) Family. First, we provide redefinition of the compromise effect confused in Simonson (1989), Simonson and Tversky (1992) and Roe et al. (2001), and the attraction effect defined differently by Huber et al. (1982) and Rooderkerk et al. (2011). This paper provides definitions of these two effects based on relative and absolute probabilities. We also give proofs that definitions based on relative probability definitions connote definitions based on absolute probability definitions. Then, this paper shows the possibility of explanation of these effects via the GNL model through numerical examples and interpretations of these results. In the compromise effect, this paper discloses that 1) the compromise effect based on both relative and absolute probabilities definitions can be explained via the GNL model, and 2) the possible strength of the compromise effect via the GNL model has its limitation but is enough compared with existing research. In the attraction effect, this paper reveals that 3) the attraction effect based on only relative probability definition can be explained via the GNL model, and 4) the possible strength of the attraction effect via the GNL model has its limitation but is enough compared with existing research.
