Figure 1 tax in states map
AbstractThis study examined smoking and electronic cigarette use associated with geographic location, cigarette tax rate, and personal factors in the United States. The objectives were to assess a) smoking ratio and e-cigarette use ratio against tax in different states b) the association between smoking rate and cigarette tax c) the demographic and personal factors that can predict e-cigarette use status. To do so the geographic mapping, Linear Regression and Random Forest methods were used. In conclusion, the higher cigarette tax rate indicated fewer cigarette consumptions to some degree; while the cigarette tax rate seemed to have no impact on the e-cigarette use. Although some factors such as income, education, exercise habits were expected to influence the e-smoking behaviour, they actually didn’t show statistical significance; while current smoking status, age and some specific diseases can be powerful predictors. The results of this study may provide some valuable directions for policymakers and community smoking educators. IntroductionCigarette smoking is always associated with cancer death and chronic diseases, which not only damages personal life quality but also leads to a heavy burden on the health-care system. Rising cigarette tax to reduce the consumption of cigarette is a general consensus among policymakers. However, some studies indicated that there is no strong inverse relationship between tax and consumptions \citep{Abadie_2007}. Therefore, it is interesting to test whether higher tax rate associated with lower smoking ratio. On the other hand, the use of electronic cigarettes has risen substantially in recent years. Since only a few states exacted the tax on e-cigarette, the indirect effects from cigarette tax on the use of e-cigarette are also valuable to explore. In addition, it is suggested that most people start to use the e-cigarette because of health reasons, and respondents who are older, with low levels of education and with lower income were more likely to use e-cigarettes instead of conventional cigarettes \citep{Schoren_2017}. Therefore, the random forest model can be used to evaluate which factors can best predict whether a person is an e-cigarette user. This paper was constructed by addressing above questions, with the objective to help policy maker better understand their tax policy’s effect as well as understand the characteristics of the e-smokers.  DataThere were three datasets used in this study. First, the cartographic boundary shapefile of the United States was acquired from the Census Bureau’s MAF/TIGER geographic database, which was used to map the contour of the continental United States as background. Second, the amount of cigarette tax in each state was gathered from the State Tobacco Activities Tracking and Evaluation (STATE) System records from the Centers for Disease Control and Prevention (CDC). Raw data was cleaned by using keywords “cigarette tax”, “2017”, and sorted by the states name. Figure 1 provides a quick look at the level of tax over different states.