A duration analysis of smoking initiation and cessation in China

China has among the highest smoking rates in the world. The health impact of smoking is large; approximately a million deaths from tobacco-related diseases are estimated to occur annually. In recent years, Chinese policy has shifted attention toward implementing certain tobacco control measures, such as implementing media advertising bans and limited legislation on smoke-free areas.  Though nominally tobacco tax policy is also being used, this has been complicated by the fact that tobacco production in China continues to be a government monopoly.  Historical change in the price of cigarettes, therefore, has been driven more by market circumstances than tax change. However, information on how change in cigarette prices can affect smoking behavior in China can be valuable for informing future tax policy.  This paper is an empirical evaluation of the effect of price on smoking initiation and cessation in China. We use pseudo-longitudinal data from the 2010 Global Adult Tobacco Survey (GATS) in China and employ duration analysis on a set of approximately 10,000 individuals whose smoking behavior is tracked from the age of 15 (for the analysis of initiation) or from the moment of initiation (for the analysis of cessation among smokers).

The rate of smoking in a population is a function of the inflow of new smokers and the outflow of current smokers, which in turn is affected by the addictive nature of the product. Duration modeling of initiation and cessation behavior has advantages over traditional static models of cigarette demand because it accounts for addictiveness (by conditioning current behavior on past behavior), and because it sheds light on the underlying process through which prices affect smoking prevalence (whether by preventing takeup among nonsmokers or by encouraging quitting among smokers). In this study, we use the variation of prices over time within several China regions to identify the effect of price.  The GATS provides information on respondents’ broad geographic location using 6 regions (North, Northeast, East, Midsouth, Southwest, Northwest). We use this information to match respondents’ observations to prices dating back to 1990 from a main city in each of 4 of these regions.  Due to lack of region-specific price data for the remaining 2 regions, we assign to them an average price based on the 4 regions with available price data. We control for unobserved region characteristics that are fixed or vary over time using two-way region fixed effects. Preliminary results show that price is considerably effective in reducing initiation but less so in promoting cessation.

The analysis is subject to measurement error in prices. We use prices from a popular local brand in major Chinese cities, but these prices are not representative of the wide cross-brand and within-region price variability of cigarettes in the country. However, they are useful under the assumption that they correctly represent the direction of change in prices over time.  This data limitation is likely to have an attenuating effect on the price estimates.