CMStatistics 2018: Start Registration
View Submission - CMStatistics
B1653
Title: A Bayesian mixed multinomial logit model for partially microsimulated data on labor supply Authors:  Consuelo Nava - University of Turin (Italy) [presenting]
Cinzia Carota - University of Turin (Italy)
Abstract: The determinants of labor choices of females within couples are studied in the presence of partially microsimulated data which induce choice sets not identical for all agents under examination. Using traditional random utility models, two main assumptions are violated: the independence of irrelevant alternative and the homogeneity of available choice sets across decision makers. As an alternative, a Bayesian mixed multinomial logit model suitably adapted to describe labor choices made by females within couples is proposed. The constructed specification of random effects allows us to deal with both non-homogeneous choice sets and individual heterogeneity, also taking advantages of different (parametric and nonparametric, frequentist and Bayesian) clustering techniques. By comparing the results obtained under the proposed approach to the ones obtained considering a model without random effects, the relative explanatory ability of these models in the above-described scenario is discussed.