The misuse of alcoholic beverages ('problem drinking') has been demonstrated to result in enormous economic costs; most of these costs have been shown to be reduced productivity in the labor market. The purpose of this paper is to present sound structural estimates of the relationship between various measures of problem drinking and of employment and unemployment. The sample of approximately 15,000 observations is drawn from the 1988 Alcohol Survey of the National Health Interview Survey, the first dataset that enables nationally- representative estimates of alcohol abuse and dependence consistent with generally accepted medical criteria. The structural estimates of the effects of problem drinking on employment and labor market participation are obtained using methods proposed by Amemiya and by Heckman and MaCurdy. For our sample of males ages 25 to 59, we find that using the instrumental variable approach suggests that the negative impact of problem drinking on employment is even greater than that estimated using the OLS approach. Interestingly, the IV estimates on the samples of females change the sign from a positive impact of problem drinking on employment to a negative impact. Thus although the conclusions drawn from raw data comparisons and OLS regressions differ by gender, the IV estimates are very similar for men and women. For women, the unobserved heterogeneity masks the negative impact of problem drinking on employment when using OLS estimation methods.
