We used longitudinal data and instrumental variables (IVs) in a prospective design to test for the causal effects of wages on smoking prevalence among current and past smokers. Nationally representative U.S. data were drawn from the 1999ー2009 waves of the Panel Study of Income Dynamics. Our overall sample was restricted to full time employed persons, aged 21e65 years. We excluded part time workers and youths because smoking and wage correlations would be complicated by labor supply decisions. We excluded adult never smokers because people rarely begin smoking after the age of 20 years. IVs were created with state-level minimum wages and unionization rates. We analyzed subsamples of men, women, the less educated, the more educated, quitters, and backsliders.
