Smoking, self-selection and absenteeism

How much, if any, does smoking contribute to absenteeism? Separate samples of employed men and women are drawn from the 1986 wave of the Panel Study of Income Dynamics (PSID) to answer this question. The 1986 wave is the most recent one with information on smoking. In the first analysis, single equation Tobit regressions are run explaining the absence rate. Independent variables include a binary smoking variable together with others for age, race, marital status, and so on. In the single equation models, smoking appeared to raise absence rates by 42 and 232 percent for women and men, respectively. In the second analysis, decomposition techniques, which also adjust for smoking and employment hazard rates, are applied to four separate samples of smokers, non-smokers, men, and women. The decomposition techniques lowered these differentials. Smoking appears to make a moderate contribution to absenteeism for men, but only a slight contribution for women. The decomposition suggests that it is the observed and unobserved personal characteristics of female smokers that account for most of the simple positive correlations between absenteeism and smoking among women found in the single equation models.