We focus on the determinants of labor choices in the presence of partially microsimulated data and discrete choice sets not identical for all agents under examination. The independence of irrelevant alternative assumption is thus discussed and the variability of the available choice set is taken into account. By comparing a Bayesian mixed multinomial logit model to a model without random effects, we show how the above described scenario affects labor choices made by single females and females within couples when the same discrete choice set is assigned to both individuals in each couple and the partner’s choice is known.