A Bayesian Mixed Multinomial Logit Model for choice-sets and decision-makers’ heterogeneity

Abstract

ABSTRACT We propose a Bayesian Mixed Multinomial Logit Model (MMLM) to deal with the critical issue of choice-set heterogeneity often present in policy evaluation studies enriched with microsimulated data. We also exploit the comparison of three clustering methods to capture decision-makers’ heterogeneity through a specific random effect. A case study, which aims to describe the determinants of labour choices of females in couples with microsimulated fiscal variables, is the test-bed for our methodological proposal. By virtue of this very flexible specification of the random components, the Bayesian MMLM proves to be more accurate, parsimonious and consistent in terms of point estimates with the research field than other models.