The Effects of Smoke-Free Policies on Cigarette Demand: Empirical Evidence from China’s Cigarette Industry

Smoking and second-hand smoke exposure pose a major health threat to people in China, where comprehensive national-level smoke-free legislation is absent. Smoke-free legislation is a powerful tobacco control measure that reduces not only non-smokers’ risk of exposure to second-hand smoke, but also smokers’ cigarette consumption. Several provinces and municipalities in China have implemented province-level smoke-free policies, while the remaining populations are still at risk of being exposed to second-hand smoke in public places. Although the regions that adopted smoke-free policies are expected to be role models for other regions, the effects of these policies in reducing cigarette consumption of smokers remain unknown. Moreover, different regions tend to have different smoke-free policies, and some policies are stricter than others. Without empirical examinations, it is difficult to judge how effective the existing smoke-free policies are and what kind of policy should other regions implement.

This thesis estimates the effects of smoke-free policies of Shanghai, Sichuan, Tianjin, and Jiangsu on cigarette consumption using the synthetic control method (Abadie, Diamond, and Hainmueller, 2010), a data-driven procedure to generate a suitable control group that resembles the treated group and to estimate treatment effects. The empirical results show that province-level smoke-free policies reduced annual per capita cigarette consumption by about 7.042 packs for Shanghai and 5.648 packs for Sichuan in 2014. The placebo tests demonstrate that the estimated effects of Shanghai’s and Sichuan’s policies are large in magnitude and robust. These results have important policy implications for the potential effects of province/national-level smoke-free policies in China, suggesting that smoke-free policies that are compliant with Article 8 of the World Health Organization Framework Convention on Tobacco Control (WHO FCTC) are more effective in reducing smoking.

The analysis of Jiangsu’s and Tianjin’s smoke-free policies demonstrates the limitations of the synthetic control method. First, the synthetic control approach requires a sizable number of pre-intervention and post-intervention periods. The results of Jiangsu indicate that a one-year post-intervention period is too short for policy effects to manifest. Second, it is not always possible to obtain a synthetic control unit. The synthetic control analysis of Tianjin’s policy does not have a good fit, demonstrating that when too few regions in the donor pool are similar to the treated region, or when untreated regions have significant heterogeneity in characteristics and pre-intervention outcomes, it is not feasible to obtain a good synthetic match.

The current study also uses the difference-in-differences (DID) approach to better understand the marginal contributions of the synthetic control method. The results indicate that the synthetic control method indeed returns a better control unit, which is a convex combination of untreated regions, reducing human discretion required in choosing control units. The synthetic control approach can also reveal the regions that are useful yet might be easily ignored by our subjective judgment. Additionally, the time-varying treatment effects obtained can depict the manifestation of policy effects, providing more information than an average treatment effect. However, the advantages of the difference-in-differences method are that it does not require a long sample period and it allows researchers to pool multiple treated units. When data are not sufficient for a synthetic control analysis, the difference-in-differences approach can serve as an important alternative, as long as control units meet the parallel trends assumption.