Abstract
Many public health policies are rooted in findings from medical and epidemiological studies that fail to consider behavioral influences. Using nearly 50 years of data from Framingham Heart Study male participants, we evaluate the longevity consequences of different lifetime smoking patterns by jointly estimating smoking behavior and health outcomes over the life cycle, by richly including smoking and health histories, and by flexibly incorporating correlated unobserved heterogeneity. Unconditional difference-in-mean calculations that treat smoking behaviors as random indicate a 9.3 year difference in age of death between lifelong smokers and nonsmokers; our findings suggest the bias-corrected difference is 4.3 years.