Embarrassing Variety of Choice: Modeling Mexican Return Migration Decisions

Claudia Masferrer, McGill University

This work investigates the utility of applying various models for return migration decisions. I study alternative models that fall in a broad class of polytomous choice models widely used when outcomes consist of categories of choices. The three conceptually plausible models considered are the classical multinomial logit, the nested logit and the sequential logit models. The statistical concepts underlying these models are described and investigated, with a special focus on the assumption of independent irrelevant alternatives, and then an application to data of Mexican return U.S. migration is provided. I use the complete set of individual records of the 2005 Mexican Population Count. I find that these models are dependent on how researchers think of these decision processes and that, for this specific type of application, the sequential logit model offers more flexibility in terms of defining the decision structure and in terms of comparison and interpretation.