EXAMINING THE EFFECTS OF PUBLIC POLICIES AND ADDICTION ON PURCHASE OF TOBACCO PRODUCTS WITH CAUSAL INFERENCE AND MACHINE LEARNING METHODS

My three essays explore the effects of tobacco policies and addiction on the consumption of e-cigarettes and other tobacco products. Recently, jurisdictions imposed taxes and other regulations on e-cigarettes, with the hope to raise tax revenues and address health concerns regarding e-cigarette use, especially youth addiction. My first essay in Chapter 1 focuses on the effects of e-cigarette taxes on sales of e-cigarettes. It compares the two types of tax policies on sales of e-cigarettes, cigarettes, and smoking-cessation products. This comparison provides information for lawmakers on decisions of taxes regarding the perspectives of revenue generation and tobacco control. Second, after exploring the tax effects on sales of these tobacco products at an aggregated level, it would be interesting to examine if the tax policy has diverse effects on different groups of consumers of the three tobacco products. My second essay in Chapter 2 provides answers to this question. Since most e-cigarettes contain nicotine, my third essay in Chapter 3 tests if e-cigarettes are addictive from the perspective of economics. The results provide policy implications for future regulations of e-cigarette use. The first essay investigates the effects of two types of e-cigarette tax policies on sales of e-cigarettes using time-series cross-sectional weekly purchases from Nielsen Retail Scanner Data between 2011 and 2017. The first type of taxes is ad valorem taxes, which tax e-cigarettes on a percentage of the wholesale or retail price. The second type of tax is specific excise taxes, which collect taxes according to tax rates per milliliter (ml) of the consumable liquid. With a generalized synthetic control (GSC) identification strategy, my first essay measures the average tax effects of multiple treated regions in various treated periods against the untreated control regions and the untreated periods. The results show an estimate of -3.567 for the own-price elasticity of e-cigarette demand and an estimate of 0.433 for the pass-through rate of e-cigarette taxes to price in all treated regions. With these estimates, one can infer the tax revenues generated and the sales effects of the two types of tax policies. The inference indicates that ad valorem taxes are more effective on tobacco control. In comparison, the specific volumetric excise taxes are more for tax revenue generation. Additionally, the ad valorem taxes show larger average effects but smaller marginal effects than the specific volumetric excise taxes. The results in this essay also indicate that cigarettes and SCP are economic substitutes for e-cigarettes; however, increasing e-cigarette taxes would not raise sales of cigarettes but would increase sales of SCP, when the tax-to-price ratio of e-cigarettes is below or similar to that of cigarettes after tax increases. The second essay examines the policy responses of different groups of consumers of the three tobacco products with the Nielsen Consumer Panel Data between 2012 and 2018. This essay identifies seven groups of consumers according to the data. Some of them consume exclusively on one product while others consume two products or all three products in the meantime. With a generalized difference-in-differences identification strategy, results do not find that ad valorem taxes bring health concerns to any of the seven investigated subgroups of purchasers. In comparison, specific excise taxes increase e-cigarette purchases by subgroups of e-cigarette purchasers except for triple purchasers of the three products. Additionally, the average results for all these groups are that e-cigarette taxes decrease their SCP purchases but do not influence cigarette purchases. Comparing the effects of two types of e-cigarette taxes, though both negatively influence purchases of SCP, specific excise taxes increase purchases of cigarettes and e-cigarettes while ad valorem taxes do not have such effects. The third essay examines the role of addiction in influencing the demand for e-cigarettes using the Nielsen Retail Scanner Data between 2012 and 2017. With a comparison of a myopic addiction model, a forward-looking model, and a rational addiction model, this essay tests whether consumption of e-cigarettes is addictive and rational. Results provide evidence that consumers are rationally addicted to e-cigarettes. The long-run price elasticity estimates are larger than the estimates of the short-run price elasticity. Estimates of both long-run and short-run elasticities are greater than one, -1.50 and -1.05, indicating e-cigarette demand is elastic in both the long-run and short-run. Additionally, the data show evidence that e-cigarettes are less addictive than cigarettes on average because the addictiveness coefficient estimate for e-cigarettes is smaller. These three essays aim to provide insights to policymakers on the effectiveness of tax policies regarding revenue generations and tobacco control. The general findings of the three essays are e-cigarette taxes influence purchases, consumers are rationally addicted to the product but they respond differently to the tax policy. Different responses of the seven groups of consumers provide information for lawmakers to weigh the cons and pros of the two types of policies regarding which groups are the policies mainly targeted for tobacco control. Moreover, addiction could influence tax effects. In the long run, e-cigarette taxes have larger cumulative effects on tobacco control than in the short run. Findings from these essays provide evidence to facilitate future regulations on e-cigarettes. Additionally, they also contribute the literature on applying methods of causal inference and machine learning in exploring e-cigarettes policy regulations