Double Decoys and a Possible Parameterization: Empirical Analyses of Pairwise Normalization
11 Pages Posted: 17 May 2019
Date Written: April 15, 2019
In this article, we bring the pairwise normalization model to a stochastic setting. Our aim is two-fold. First, we examine novel predictions of the pairwise normalization model when multiple decoys are introduced into the choice set. We contrast these predictions with models which hypothesize that the range of attributes influences choice and conduct an experimental analysis to test for the presence of a “double decoy” effect. Second, we offer a parameterization of the pairwise normalization model for empirical settings. This extension includes weights on attribute dimensions to address subjective preferences and/or unit definitions, and allows for hypothesis testing for the presence of normalization in a choice dataset. We then analyze the pairwise normalization model on data from a choice experiment and find it outperforms the Logit model.
JEL Classification: D87
Suggested Citation: Suggested Citation