| The three essays in this dissertation focus on cigarette smoking and its impact on adult mortality in the United States. The first chapter investigates sex mortality differentials in the United States between 1948 and 2003 and reveals that these differentials have a cohort structure. The result from an Age-Period-Cohort model shows that cohort smoking histories can partly explain this cohort mortality imprints. Chapter 2 utilizes the relationship between adult mortality and cohort smoking histories established in the first chapter to inform mortality projection. A cohort variable, average years spent as current smoker within each birth cohort, is incorporated into the most widely used Lee-Carter Model for mortality projection. This enables me to introduce a common temporal trend of mortality change for both sexes instead of a separate one for each sex. Morality projections with incorporated smoking histories show faster improvement in adult mortality than suggested by other projections. The final chapter explores the current smoking prevalence between 1970 and 2005. Drift, which is the linear parts of both period and cohort effects, is found to be the dominating force behind current smoking behavior during this period for both sexes. In addition, all three factors, age, period, and cohort, are found to be significant. A Bayesian Age-Period-Cohort model is then used to forecast age-specific current smoking prevalence for each sex separately. It predicts that prevalence of smoking will continue to fall in the next 30 years. |
