Smoke-free laws and smoking prevalence in the United States, 1998-2010

Background: Previous research indicates that comprehensive smoke-free policies – those restricting smoking in all public places – have a stronger effect on smoking prevalence than partial policies – those exempting certain venues. However, previous research invariably treats smoking policies as exogenous, even though such policies exist endogenous to prevalence. This research treats smoke-free policies as endogenous and investigates how they affect smoking prevalence. Methods: A state-level panel data set was created utilizing smoking prevalence from the Behavioral Risk Factors Surveillance Survey, smoke-free policies from the Robert Wood Johnson Foundation interactive tobacco database, cigarette taxes from the CDC's STATE tracking system, state poverty rates from the U.S. Census, and state unemployment rates from the Bureau of Labor Statistics from 1998-2010. Two statistical models – one controlling for endogeneity (Arellano-Bond dynamic panel data model) and one assuming exogeneity (linear fixed effects model) – were estimated using STATA. Results: The Arellano-Bond model found a strong, partially significant effect of comprehensive smoke-free policies in reducing smoking prevalence, but partial smoke-free policies had no effect on prevalence. While the exogenous fixed effects model also found decreases in smoking prevalence given comprehensive smoke-free laws, the parameter estimates were excessively lower in magnitude than Arellano-bond model estimates. Treating smoke-free policies as exogenous drastically obscures their impact. Discussion: Comprehensive smoke-free policies have a greater effect on smoking prevalence than partial policies, and ignoring the endogenous nature of such policies masks this relationship. Tobacco control efforts should promote comprehensive policies and evaluation research should treat such policies as endogenous.