A (nearly) global look at the dynamics of youth smoking initiation and cessation: the role of cigarette prices

Abstract

This study investigates the role of cigarette prices on smoking initiation and cessation among youth in a sample of eight primarily non-Western low, middle, and high income countries, with a particular focus on a subsample of 40 low- and middle-income countries (LMICs). Using split-population duration models on longitudinally-transformed individual data from the Global Youth Tobacco Survey (GYTS), the average impact of cigarette prices in the presence of unobserved country heterogeneity and shifting cultural norms within countries is identified by the variation of cigarette prices within countries over time. Price increases are found to effectively reduce initiation in early youth, and girls are considerably more responsive than boys. The price elasticity of initiation in LMICs is −0.82 for the combined gender analysis, −0.46 for boys only and −1.5 for girls only. There is some indication that youths in developing countries may be slightly less responsive to price changes than in high-income countries. No evidence is found that cigarette prices increase quitting rates in youth, which may be due to the difficulty of defining true quitting among smokers in early life.

Keywords:

• cigarette prices
• youth smoking
• smoking initiation
• smoking cessation

JEL Classification::

• I12
• I18

Notes

¹ For those who indicate having quit 3 or more years earlier, I assume quitting occurred 3 years earlier. Although this would produce a measurement error for those quitting 4 or more years earlier, I assume that this error is negligible. The average age of the survey participants is fairly young, so it is safe to assume that not many would have had time to start, form a habit, and quit more than 3 years before they were interviewed.

² The answers to this question do not provide the exact age of starting smoking but provide the following ranges: 7 years old or younger, 8–9 years old, 10–11 years old, 12–13 years old, 14–15 years old and 16 years old or older. I take the midpoint of each range as the age of starting smoking which effectively means that initiations in this sample are assumed to have occurred only at the following ages: 7, 9, 11, 13, 15 or 16. This causes measurement error in the year of initiation. Recall bias can also add to the measurement error. However, since year of initiation and year of cessation are dependent variables in this analysis, this measurement error would be reflected in the error term.

³ ‘PPPs are currency conversion rates that both convert to a common currency and equalize the purchasing power of different currencies. In other words, they eliminate the differences in price levels between countries in the process of conversion’ (OECD Statistics Directorate, www.oecd.org/std/ppp). EIU collects nominal price data in local currency and converts them into US dollars using exchange rates at the time of data collection. To standardize the variation of price across countries and make prices comparable across countries with different standards of living, we apply the PPP conversion factor to exchange rate ratio. Prices deflated by the PPP are real prices expressed in US dollars at a moment in time. To standardize the fluctuation of prices over time, we apply the 2000 US Gross Domestic Product (GDP) deflator.

⁴ Results using foreign-brand prices are very similar, and are available from the author.

⁵ This age was chosen because it is young enough to capture the earliest chances of initiation for the majority of people, while keeping the individual sample size as large as possible. Although it is possible to start observation at age 0, this would require losing a large portion of the individuals in the original sample; this is because tracing back cigarette prices to the year of birth is unfeasible for many GYTS respondents due to unavailability of price data before 1990.

⁶ Analysis time t is the time since the beginning of risk. In the initiation model, t = 1 is when age = 8, which occurs in different calendar years for different individuals. In the cessation model, t = 1 is when initiation occurred.

⁷ Note that the initiation hazard model in Equation 1 does not contain Age as a covariate. This is because for each individual, analysis time t is determined solely in reference to age (initiation risk begins at age 8, so t is simply the number of years since age 8). Since the hazard rate h_t ^I is the distribution of initiation times t, the covariate Age_t would be collinear with the outcome. This is not the case in the cessation hazard model, where analysis time t is determined in reference to the year of smoking initiation. Age is therefore included in the model of cessation.

⁸ The coefficients can also be transformed into odds ratios by exponentiation.