Based on data from the 1988 Alcohol Supplement of the National Health Interview Survey, Mullahy and Sindelar (1996) (M&S) find, for both men and women, that alcohol abuse results in reduced employment and increased unemployment. The estimates from which they drew these inferences were obtained via the instrumental variables (IV) method, which was implemented in order to account for the potential endogeneity of problem drinking. Though these IV estimates qualitatively supported the prior expectation that problem drinking damages individuals' labour market prospects, they were not found to be statistically significant. The present paper revisits this research and offers a new estimation method which, in addition to accounting for endogeneity, explicitly allows for the inherent non-linearity of the underlying regression structure. The new method is applied to the same data and variable specifications as those used by M&S for the male subpopulation. Consistent with their results, problem drinking is found to have a positive effect on the probability of unemployment and negative effect on the likelihood of being employed. Unlike their result, however, the latter estimate is statistically significant. An appealing feature of the new method is that it accommodates the likely possibility that alcohol abuse effects are heterogeneous with respect to the observed and unobserved characteristics of individuals in the population. To illustrate this fact, abuse effects are computed for two widely differing subgroups of the population. The large differential between the estimated effects for these two subpopulations demonstrates the potential importance of accounting for heterogeneity.
