Scale Heterogeneity and Its Implications for Discrete Choice Analysis

This paper shows how researchers can make incorrect interpretations of models that include heterogeneity in error variance (scale) if they do not recognize the implications of parameter estimates being conflated with scale. The research contribution is to define best practice management of scale heterogeneity in discrete choice experiments and to empirically demonstrate issues that arise when best practice is ignored. We illustrate the problem using data from two discrete choice experiments, showing that reported parameters vary due to arbitrary normalization decisions. We report examples of papers where authors fall into this trap, potentially leading to erroneous evaluation of models.