Does cigarette price influence adolescent experimentation?

Journal of Health Economics 20 (2001) 261 270 Does cigarette price influence adolescent experimentation? Sherry Emery, Martha M. White, John P. Pierce Department of Family and Preventive Medicine and the Cancer Prevention and Control Program, University of California, San Diego, 9500 Gilman Dr. 0645, La Jolla, CA 92093-0645, USA Received 1 July 2000; received in revised form 1 October 2000; accepted 31 October 2000 Abstract The economics literature generally agrees that state and federal excise taxes can play an important role in deterring adolescent smoking. Teens apparent responsiveness to cigarette prices is puzzling, since the majority of adolescent smokers do not buy their cigarettes. Teens typically do not begin to purchase cigarettes until they have developed an established pattern of smoking. Previous studies have not had adequate measures of smoking experience to explore whether adolescents price responsiveness may vary by smoking experience. This paper uses data from a 1993 national survey of youth smoking to explore this hypothesis. 2001 Elsevier Science B.V. All rights reserved. JEL classification: I18 Public health Keywords: Cigarette excise taxes; Adolescent smoking; Tobacco control 1. Introduction There is general agreement in the economics literature that by raising the price of cigarettes, state and federal excise taxes can play an important role in deterring adolescent smoking (Lewit et al., 1981; National Cancer Institute, 1993; Chaloupka and Grossman, 1996; Evans and Huang, 1998; Gruber, 2000). However, a number of studies have not found the expected relationship between cigarette prices and adolescent smoking (Chaloupka, 1991; Wasserman et al., 1991; Douglas and Hariharan, 1994; DeCicca et al., 1998). One reason for this lack of consensus might be that adolescents are not a homogeneous group with respect to price sensitivity. Gruber and Zinman s (2000) recent findings support this hypothesis. They showed that younger adolescents (8th and 10th grade) were not price sensitive, whereas high school seniors (12th grade) were quite sensitive to cigarette prices; the Corresponding author. Tel.: +1-858-822-2386; fax: +1-858-822-2399. E-mail address: slemery@ucsd.edu (S. Emery). 0167-6296/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0167-6296(00)00081-3

262 S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 estimated smoking participation elasticity for 12th graders ( 0.67) was significant and very similar to other s estimates (Lewit et al., 1981; Chaloupka and Grossman, 1996; Gruber and Zinman, 2000). Studies of the natural history of smoking also support the heterogeneity hypothesis. Younger smokers tend to be in the very early stages of smoking uptake, when most report obtaining cigarettes from friends. In order to be sensitive to cigarette prices, it is necessary to purchase or at least contemplate purchasing cigarettes. Therefore, it is not surprising that younger adolescents may not be price sensitive. Previous research suggests that it is typically not until the later adolescent years (e.g. 16 18 years old), when they start to smoke on a daily basis, that adolescent smokers also start to purchase their own cigarettes. The critical smoking level at which adolescents begin to purchase cigarettes- and therefore may become sensitive to cigarette prices is approximately one cigarette per day on average (Emery et al., 1999). The theory of rational addiction also supports the hypothesis that experimenters may be different from other smokers, since they have not yet built up sufficient addictive capital to make future cigarette consumption and cigarette prices a serious consideration (Chaloupka, 1991). Although experimenters may not yet have developed an addiction to nicotine, they are still an important group to study. Many adolescents who experiment with cigarettes may never go on to battle a lifelong addiction to nicotine. However, there is substantial evidence that for some, the first few cigarettes are enough to trigger a vulnerability to nicotine dependence (Pomerleau, 1995; Pomerleau et al., 1999). Previous research showed that 30 70% of experimenters eventually progress to dependence (US Department of Health and Human Services, 1994; Gilpin et al., 1999). Moreover, adolescents consistently underestimate the risk of becoming addicted to smoking (US Department of Health and Human Services, 1994; Gruber and Zinman, 2000), and therefore may see no danger in experimenting. There appear to be many societal and policy influences that effect the probability that an individual adolescent will proceed from experimentation to dependent smoking, and whether price is one of these influences is an important question. Most previous studies of the effect of price on adolescent smoking, however, have measured smoking participation with a dichotomous variable indicating whether or not the adolescent smoked in the past 30 days. Such a measure is limited because experimenters are categorized with heavier smokers, along with those who may report having recently quit smoking. Each group might respond differently to factors, such as cigarette prices, that could influence their smoking behaviors. The inconsistencies among the adolescent price elasticity estimates in the literature may at least partly reflect differences in the smoking experience or measurement of smoking in the study samples. This paper specifically examines the relationship between experience with smoking and adolescent price sensitivity in a 1993 national cross-section of adolescents. Section 2 describes the datasets and statistical methods used. Section 3 summarizes our results, and Section 4 discusses our findings in the context of other studies and public policy. 2. Data and methods This study employed data from the second wave (1993) of the longitudinal teenage attitudes and practices survey (TAPS). The first wave of TAPS was coordinated by one of

S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 263 the authors (J.P.P.), and interviewed adolescents who were enumerated the 1988 national health interview survey (NHIS). The NHIS are representative annual household interview surveys of the civilian, non-institutionalized population. The National Center for Health Statistics provided state-specific codes for the NHIS surveys as part of a 1990 agreement to evaluate the California tobacco control program. The first wave of TAPS was conducted in 1989, and the follow-up was conducted in 1993. The 1993 wave included a total of 12,952 adolescents (10 22 years of age) from 48 states and the District of Columbia (there were no respondents from South Dakota or Nebraska), who were interviewed by telephone in their homes; of these, 7960 were also interviewed in 1989. 2.1. Measures of smoking behavior We defined never smokers as adolescents who provided a negative response to the following two questions: Have you ever smoked a cigarette? and Have you ever tried or experimented with cigarette smoking, even a few puffs? Those who had smoked in the past 30 days were classified as current smokers. Established smokers were defined as adolescents who had smoked in the past 30 days and who had smoked at least 100 cigarettes in their lifetime. Experimenters answered positively to one of the following two questions, Have you ever smoked a cigarette? or Have you ever tried or experimented with cigarette smoking, even a few puffs?, but had not yet smoked 100 cigarettes in their lifetime. To measure the level of consumption among those who were current or established smokers, we used the average of the number of cigarettes/day reported for the each of the 7 days prior to the survey. Only those who had smoked in the past 7 days were asked about their level of cigarette consumption. 2.2. Cigarette prices The Tobacco Institute (TI) provided a historical compilation of the average price/pack of cigarettes in each state each year, reported on 1 November. We adjusted the TI prices by the US consumer price index (CPI) to obtain the real price/pack. We used codes supplied by the NHIS to identify respondent s state of residence and match it to the appropriate state-level cigarette price. The 1993 TAPS interviews took place between January and April of 1993, with the vast majority of interviews occurring during January and February. This timing is important because in April of 1993, the dominant cigarette manufacturers in the US reduced the retail price of cigarettes by approximately US$ 0.20/pack or 10% of the total price (Shapiro, 1993). The 1993 TI price data reflected this price reduction, but the 1993 TAPS respondents had not yet experienced this decrease at the time of their interview. Therefore, we used the 1 November 1992 real price/pack of cigarettes in our models. 2.3. Model specification We estimated a two-part model of demand, based on the model developed by Cragg (Cragg, 1971). The first part is a model of smoking participation, where the dependent variable is a dichotomous indicator of smoking status. The second part is the model of

264 S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 conditional demand- or the amount smoked among those who are smokers. We analyzed the conditional demand for cigarettes among current and established smokers separately, but did not estimate a conditional demand model for experimenters because the majority of experimenters were not asked about the quantity of cigarettes they smoked/day, due to the skip pattern of the interview. We suspect that, given the sporadic nature of experimentation, the amount-smoked variable would likely be unreliable anyway. The models are summarized as SMOKE i =α + βprice i + δtobacon i + γ SOCDEM i + λpsychsoc i + ε i (1) SMOKE is a measure of smoking (participation or log of quantity smoked conditional on participation). PRICE is the real price/pack of cigarettes. TOBCON is an additive index of indicators of state-level tobacco control activity. SOCDEM is a set of socio-demographic variables (gender, race, rural residence, lives with a single parent, not living with parents, religiousness, works at a job, weekly disposable income >US$ 20, parental education level, and household income), which matches as closely as possible those variables most frequently included in economic models of adolescent smoking in the literature (Lewit et al., 1981; Wasserman et al., 1991; US Department of Health and Human Services, 1994; Chaloupka and Grossman, 1996). PSYCHSOC is a set of psycho-social variables (school performance, depression, rebelliousness, sports participation, parental bond, family smoking, and belief that it is easy to get cigarettes), which appear consistently in the medical and psychology literature as predictors or correlates of adolescent smoking (Chassin et al., 1984; Chassin et al., 1991; Conrad et al., 1992; US Department of Health and Human Services, 1994; Choi et al., 1997; Flay et al., 1998; Pierce et al., 1998). 2.4. Statistical analysis Respondents to the 1993 TAPS interviews ranged in age from 10 to 22 years of age. Among the youngest of this group (the pre-adolescents, 10 13 years of age), experimentation with cigarettes is not uncommon (14%, n = 526), but very few reported smoking within the last 30 days (1.4%, n = 49), and even fewer reported having smoked >100 cigarettes in their lifetime (0.3%, n = 13). Therefore, we restricted our analyses of current and established smokers to include only adolescents who were 14 years old. We present results for models of experimentation for 10 13-year-olds and adolescents 14 years of age and older separately. Because the NHIS used a multistage sample design to represent the civilian noninstitutionalized population of the US, we used the SUDAAN program, which can account for the probability of household selection and post stratification by race, gender and age, and the NHIS weighting procedures (Shah et al., 1997). 3. Results Table 1 provides the means and standard deviations of the independent variables used in our models, by smoking status. Table 2 presents the results of the six models we estimated. Table 3 presents the implied elasticity estimates for each model at the sample means.

S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 265

266 S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 Table 2 Models adolescent smoking behavior a a Parameter estimate (S.E.). Darker shaded boxes indicate significance at the P<0.05 level; lighter shaded boxes indicate marginal significance (0.05 <P 0.10). In the models of experimentation, price was not an important variable, and therefore, we do not report the elasticity estimates for experimentation among either the 10 13-year-olds or the 14-year-olds. As expected, price was negatively associated with current and established smoking participation and conditional demand. Current smoking participation elasticity was 0.83. Among current smokers, conditional demand elasticity was 0.87, and total elasticity was 1.70. Among established smokers, the participation elasticity was 1.56, conditional demand 0.68, and total elasticity was 2.24.

S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 267 Table 3 Elasticity estimates at means Elasticity estimates Experimenters (10 13 years of age) Experimenters (14 + years old) Current smoking (14 + years old) Participation NS a NS 0.83 1.56 Conditional Demand NA b NA 0.87 0.68 Total 1.70 2.24 a Non-significant. b Not applicable. Established smoking (14 + years old) Our measures of tobacco control activity were insignificant across all models. To explore the possible effects of individual tobacco control policies, such as restaurant smoking bans or retail establishment smoking bans, we ran separate models that substituted a dummy variable for one of each such variables in place of the tobacco control index. No single tobacco control policy variable was associated with adolescent smoking across any of the models of participation or conditional demand. Among the socio-demographic variables (SOCDEM), the signs and significance of the coefficients were consistent with other s findings. Across all models, non-white race was negatively associated, and age and living with a single parent were positively associated with smoking behavior. For all but the youngest experimenters, religiosity was negatively associated with smoking. Living with a single parent was positively associated with current smoking and established smoking participation, and marginally positively associated with experimentation and conditional demand among established smokers. Working at a job was positively associated with current and established smoking participation. Conversely, having disposable income of >US$ 20 per week was negatively associated with smoking in each of the current and established smoking models, except for current smoking participation. Although it might be expected that working and having disposable income would be related, a chi-square test revealed no significant association. Household income appeared to be unrelated to teenage smoking, but parental education (another measure of socio-economic status) was negatively associated with experimentation among the youngest teens and with the amount smoked for current and established smokers. As expected, there was a significant positive association between disposable income and household income. To eliminate the potential multi-colinearity effect, we re-ran the models without the disposable income variable, but the results were not substantially different. Fairly consistently across all the models, the psycho-social variables were strongly associated with smoking, in the expected directions. Thus, poor school performance, rebelliousness, and exposure to family smoking were positively associated with adolescent smoking across all the models. Nearly as consistently, depression and the belief that cigarettes are easy to get were also positively associated with smoking behavior, except for conditional demand among established smokers. Participating in sports was negatively associated with current and established smoking, but not with experimentation. Having a strong parental bond was negatively associated with the decision to smoke across experimenters, current and established smokers, but was unrelated to the amount smoked among either current or established smokers.

268 S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 4. Conclusions Our research represents one of the first attempts to explore whether and how adolescents level of experience with smoking may interact with policy variables, such as cigarette excise taxes. Our results showed that price was not significantly associated with experimentation. Rather, other factors, which are less amenable to public policy intervention, appeared to be important in the experimentation phase. Because experimenters typically consume very few cigarettes, and do so sporadically for at least a period of time, it is not surprising that they do not consider the price of cigarettes in their early smoking behavior. At this stage, they do not personally experience cigarette prices because they most often get their cigarettes from friends (Emery et al., 1999), and do not yet pose a burden on their friends generosity. Only after they start smoking enough to justify buying their own cigarettes or asking others to buy the cigarettes for them would price become a relevant issue for these teens. Our results confirm other s findings that price is an important factor in more advanced smoking behavior among adolescents. 1 In addition, in both the participation and conditional demand models of current smoking, several of the psycho-social variables were significant, confirming the well-established findings of many longitudinal studies of adolescent smoking that did not include policy variables, like price. The relative proportion of the total elasticity accounted for by the participation component was lower in our models of current smoking than seen in other studies (Lewit et al., 1981; Chaloupka and Grossman, 1996). Our estimates for conditional demand elasticity among current smokers are relatively high, compared to those seen elsewhere, and suggest that price may play an important role in moderating the amount smoked and therefore, possibly slowing progression toward addiction for some adolescents. Our elasticity estimates for established smoking were slightly higher than elasticity estimates for current smokers seen in the literature. In particular, the participation elasticity for established smokers was higher than the total elasticity estimates reported in most studies. Because we used only the cross-sectional component of the TAPS dataset, we were unable to account for the potential endogeneity of cigarette prices with other state-level variables that may influence adolescent smoking. Therefore, our estimates of the price effect may be biased upwards. Although, this is a salient issue for the interpretation of our models of current and established smoking, it does not threaten the interpretation of our models of experimentation in which price is insignificant. 2 Finally, another important limitation of this research is the omission of measurements of exposure to cigarette advertising and promotional items. Recent research has shown that receptivity to cigarette company advertising and promotional activities independently 1 Others have suggested that the considerable differences in cigarette prices across states may provide incentive for smokers in higher price states to purchase cigarettes in lower price states (Lewit and Coate, 1982; Wasserman et al., 1991). We did not have sufficient detail in our geographic identifiers to restrict our sample to adolescents who did not live near their state s border. For experimenters in particular, we believe it is highly unlikely that they would incur the costs of travelling across state lines to obtain cigarettes, and for the others, this effect would imply that our estimates are conservative. 2 Evans and Huang (1998) and Gruber (2000) addressed the endogeneity issue by using repeated cross-sections, with state fixed-effect models (Evans and Huang, 1998; Gruber, 2000). Both studies showed that price was still important to current smoking participation decisions, apparently resolving some of the controversy about the role of price in adolescent smoking.

S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 269 increases the likelihood of experimentation among adolescents (Pierce et al., 1998). There were no measures of receptivity or exposure to industry advertising in the TAPS data. Thus, our findings may suffer from inflated variances, which could result in the erroneous conclusions about the significance of the parameter estimates for the included variables (Gujarati, 1988). Despite these limitations, our findings represent a unique contribution to the literature on the role of prices in deterring adolescent smoking. Not just younger smokers, but all experimenters appear unaffected by cigarette prices. Therefore, our findings suggest that it is necessary to find alternative public policy approaches that specifically address experimentation with smoking. At the same time, our relatively large conditional demand elasticity estimates for current smokers and the relatively large participation elasticity estimates for established smokers suggest that higher cigarette prices may indeed slow down progression from higher levels of experimentation to established smoking patterns that are likely to persist as a lifelong addiction. Therefore, far from suggesting that price is not important, our research supports the conclusion that cigarette prices are a critically important policy tool in reducing adolescent smoking beyond experimentation. Acknowledgements Preparation of this article was supported by Grant 7KT-0193, funded by the University of California Tobacco Related Disease Research Program, and Grant CA72092, funded by the National Cancer Institute, National Institutes of Health. We would also like to gratefully acknowledge the constructive feedback of Dick Clayton and Scott Novak in preparing this manuscript, as well as the guidance of Frank Chaloupka in the early stages of the analyses. References Chaloupka, F., 1991. Rational addictive behavior and cigarette smoking. J. Political Econ. 99, 722 742. Chaloupka, F., Grossman, M., 1996. Price, tobacco control policies and youth smoking. National Bureau of Economic Research, Cambridge, MA. Chassin, L., et al., 1984. Predicting the onset of smoking in adolescents: a longitudinal study. J. Appl. Social Psych. 14 (3), 224 243. Chassin, L., et al., 1991. Four pathways to young-adult smoking status: adolescent social-psychological antecedents in a mid-western community sample. Health Psych. 10 (6), 409 418. Choi, W.S., et al., 1997. Which adolescent experimenters progress to established smoking in the United States? Am. J. Prev. Med. 13 (5), 385 391. Conrad, K.M., et al., 1992. Why children start smoking cigarettes: predictors of onset. Brit. J. Addiction 87, 1711 1724. Cragg, J.G., 1971. Some statistical models for limited dependent variables with applications to demand for durable goods. Econometrica 39 5. DeCicca, P., et al., 1998. Putting out the fires: will higher cigarette taxes reduce youth smoking? In: Proceedings of the Annual Meetings of the American Economic Association, New York. Douglas, S.M., Hariharan, G., 1994. The hazard of starting smoking: estimates from split sample duration model. J. Health Econ. 13, 213 230. Emery, S.L., et al., 1999. How adolescents get their cigarettes: implications for policies on access and price. J. Nat. Cancer Inst. 91 (2), 184 186.

270 S. Emery et al. / Journal of Health Economics 20 (2001) 261 270 Evans, W.N., Huang, L., 1998. Cigarette taxes and teen smoking: new evidence from panels of repeated cross-sections. University of Maryland Working Paper, College Park, MD. Flay, B.R., et al., 1998. Psychosocial predictors of different stages of cigarette smoking among high school students. Prev. Med. 27, A9 A18. Gilpin, E.A., et al., 1999. How many adolescents start smoking each day in the United States? J. Adolescent Health 25, 248 255. Gruber, J., 2000. Youth smoking in the US: prices and policies. NBER Working Paper, Vol. W7506, Cambridge, MA. Gruber, J., Zinman, J., 2000. Youth smoking in the US: evidence and implications. NBER Working Paper, Vol. W7780, Cambridge, MA. Gujarati, D.N., 1988. Basic Econometrics. McGraw-Hill, New York, NY. Lewit, E.M., Coate, D. 1982. The potential for using excise taxes to reduce smoking. J. Health Econ. 1, 121 145. Lewit, E.M., et al., 1981. The effects of government regulations on teenage smoking. J. Law Econ. 24, 545 569. National Cancer Institute, 1993. The Impact of Cigarette Excise Taxes on Smoking Among Children and Adults. Summary Report of a National Cancer Institute Expert Panel, National Cancer Institute, Division of Cancer Prevention and Control, Cancer Control Science Program, Bethesda, MD. Pierce, J.P., et al., 1998. Tobacco industry promotion of cigarettes and adolescent smoking. JAMA 279 (7), 511 515. Pomerleau, C.S., et al., 1999. Initial exposure to nicotine in college-age women smokers and never-smokers: a replication and extension. J. Addict. Dis. 18 (3), 13 19. Pomerleau, O.F., 1995. Individual differences in sensitivity to nicotine: implications for genetic research on nicotine dependence. Behav. Genet. 25 (2), 161 177. Shah, B.V., et al., 1997. SUDAAN User s Manual, Vol. 1. Research Triangle Institute, Research Triangle Park, NC. Shapiro, E., 1993. Price cut on Marlboro upsets rosy notions about tobacco profits. Wall Str. J., National Edition A1 A10. US Department of Health and Human Services, 1994. Preventing Tobacco Use Among Young People. A Report of the Surgeon General, US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office of Smoking and Health, Atlanta, GA. Wasserman, J., et al., 1991. The effects of excise taxes and regulations on cigarette smoking. J. Health Econ. 10 (1), 43 64.