Youth smoking and addiction: evaluating the wisdom and efficacy of government intervention

According to the best-known economic model of addiction, potential smokers are rational and forward looking [1]. That is, smokers purposefully begin down the path of addiction with full information and their eyes open, weighing the benefits from current and future cigarette consumption against the monetary and health costs, and then making an optimal decision they never come to regret. If smoking is viewed in this light, then government intervention in the form of bans or higher taxes will actually hurt smokers (or in the words of economists, ‘reduce their utility’).

However, even if smokers are rational, there are still economic arguments to be made for government intervention. Below, we discuss and evaluate three of these rationales: negative externalities, imperfect information about health risks and time-inconsistent preferences.

Externalities

The first argument for government intervention is based on the premise that smoking causes negative externalities, and therefore government intervention will benefit nonsmokers sufficiently to make up for the decrease in utility suffered by smokers. A negative externality is a cost imposed on a third party without compensation. So, for instance, the increased risk of death borne by somebody with a two-pack-a-day habit is not a negative externality. Sloan et al. estimated that over 80% of smoking costs are private, totaling US$33 per pack of cigarettes for the average 24-year-old smoker [2]. This cost, although eye-popping, is private and therefore does not provide a rationale for government intervention.

By contrast, any annoyance or harm that the smoker’s habit imposes on friends, family and passersby is a negative externality and could be grounds for government intervention. Sloan et al. estimated the net external costs of smoking to be approximately $6.88 per pack [2]. However, nearly 80% of this estimate was due to mortality and disability risks imposed on spouses. To the extent that private household bargaining and mate selection internalizes these costs, they are not true externalities.

The existence of adverse health effects for unborn or young children represents the strongest case to be made for household-level externalities. Evans et al. estimated the health costs imposed by women who smoke during pregnancy on newborn children to be $0.42–0.72 per pack [3], and recent work by Sabia suggests that health costs may be even higher if smoking persists during the child’s first 3 years of life [4].

Other external costs include medical care costs not borne by smokers, work losses and nonsmokers’ subsidization of smokers’ life insurance policies. However, many of these costs are more than offset by reductions in Social Security and private pension outlays, and life insurers are increasingly imposing surcharges on smokers [2].

Are negative externalities large enough to warrant further government intervention? Sloan et al. argued that if household-level spillovers are internalized and cross-subsidization is minimized, “[current] cigarette taxes are about equal to the external cost per pack” [2]. However, if household spillovers are considered externalities, and current estimates of the cost imposed on unborn and young children rise as we learn more, then cigarette taxes may be too low. Moreover, Gruber noted there might be nontrivial annoyance costs: if each US nonsmoker would value the cost of not dealing with smokers at $10 per year, this would amount to more than ten cents per pack of cigarettes sold [5].

In summary, the existing evidence suggests that the magnitude of externality costs is a matter of some conjecture and that externalities may not, by themselves, justify government intervention aimed at preventing or curbing smoking in general, or youth smoking in particular.

Imperfect information about health risks

A second argument for government action is the notion that individuals may be imperfectly informed about the future health costs of smoking. This concern has prompted changes in federally imposed warning labels on tobacco products. For example, from 1969 to 1984, warning labels on cigarette packages stated simply that smoking was “dangerous” to one’s health. Following the passage of new legislation in the mid-1980s, a specific new set of Surgeon General’s Warnings were added to inform consumers that smoking “causes lung cancer, heart disease, emphysema, and may complicate pregnancy” (PL 98–474). Specific warnings of this type were subsequently extended to tobacco advertisements, cigar packages and smokeless tobacco products, with the purpose of better informing consumers of the health risks of tobacco.

Yet interestingly, Viscusi and Hakes [6] and Viscusi [7,8] have found that both smokers and nonsmokers actually overestimate the risks of smoking-related illness. Specifically, this work suggests that individuals overestimate, by a factor of two to three, the probability of dying “from lung cancer, heart disease, throat cancer, and all other illnesses” because of smoking. A provocative implication from Viscusi and Hakes’ work is that anti-smoking campaigns have actually been, in some sense, too effective in convincing the public that smoking comes with substantial health risks.

Time-inconsistent preferences

If smokers are rational, forward looking and well informed, government intervention is clearly not in their best interest [1]. However, economists have begun to explore models of addiction that admit the possibility that government intervention can actually improve the wellbeing of an addict (or potential addict). For instance, if individuals have what are called ‘time-inconsistent preferences’, then the case for intervention is much stronger [9].

Underlying the rational addiction model is the assumption that the degree to which an individual discounts future costs and benefits is constant. If, however, an individual discounts the future more when faced with an immediate tradeoff than when thinking about tradeoffs down the road, then economists would say that he has time-inconsistent preferences. In other words, an individual might know that he will be better off in the long run if he chooses the path of abstention, but in the short run he must fight the temptation to indulge his desire for a cigarette.

There is a fair amount of evidence to support the hypothesis that preferences are time inconsistent [10], and, in particular, it seems appropriate to ascribe time-inconsistent preferences to teens. A typical teen might know that he should avoid smoking and, in general, do a good job of staying away from cigarettes. However, if for some reason he takes a puff, he may find himself developing an addiction despite recognizing that he would have been better off in the long run had he abstained from that initial puff.

Aside from differences in how future trade-offs are discounted, why might teens be more likely than adults to succumb to a short-term urge to smoke? O’Donoghue and Rabin suggested that teens could inherently derive more utility from smoking and other risky behaviors than do adults [10]. Although difficult to test empirically, this is a straightforward explanation for why teens might be especially susceptible to addiction. Complicating matters, teens could “falsely project their current consumption preferences onto their future preferences,” and as a consequence, make choices that they eventually come to regret [10]. To the extent that teens are especially prone to engaging in this type of ‘projection bias’, there is an argument to be made for government policies specifically aimed at discouraging youth from experimenting with tobacco. Finally, it is possible that teens face unique pressures that push them down the path of tobacco addiction.

Peer effects have received a great deal of attention from economists and other social scientists, and seem to be a logical explanation for why teens might be more susceptible to engaging in activities with short-term payoffs, but long-term net costs. However, pinning down the existence of peer effects empirically has proven to be extremely difficult.

A researcher interested in estimating the influence of peers on adolescent smoking decisions is faced with at least two major hurdles [11]. The first hurdle has to do with the issue of selection. If, for instance, parents who are most opposed to smoking both expend more effort supervising their children, and choose to send their children to schools attended by nonsmokers, then estimates that treat peer behavior as exogenous (that is, randomly assigned) will be biased.

In addition to dealing with the selection issue, this hypothetical researcher must distinguish peer effects from the influence of a shared environment. For instance, a correlation between the probability that an adolescent smokes and the percentage of her peers who smoke may be evidence of a peer effect, but it is also possible that this correlation is driven by shared neighborhood- or school-level attitudes towards smoking.

Recent estimates based on natural experiments provide some, albeit tentative, evidence that peer substance use is contagious [12,13]. However, the peer-effect research that most effectively dealt with the problems of selection and a shared environment focused on the behavior of college students. It seems reasonable to assume that teens are more susceptible to peer influences than college students but, as of yet, there is not enough evidence available to make a definitive assessment.

Effectiveness of US antismoking policies

Having touched upon the arguments for government intervention, we turn now to an evaluation of policy effectiveness. Specifically, given that the case for government action seems strongest for youths, we will briefly review the empirical evidence on whether higher cigarette taxes or school-level anti-smoking policies in the USA have been effective in cutting adolescent cigarette consumption.

Evidence on the effectiveness of cigarette taxes in reducing youth smoking is mixed. Differences in findings have been attributed to differences in empirical strategies, samples used and time periods examined [14,15].

Early research in the USA used cross-state variation in cigarette prices or excise taxes to identify behavioral responses to price changes [16–18]. Results from these studies consistently produced large negative effects of cigarette taxes on youth smoking participation.

However, the identification strategy used in these papers has been open to criticism, most notably by DeCicca, Kenkel and Mathios, who argued that the negative correlation between state cigarette taxes and youth smoking is best explained by policy endogeneity [19]. As state cigarette excise taxes are chosen via political processes in which state legislators and governors choose policies in response to voters’ preferences, lobbying efforts and partisan political pressures, unmeasured factors associated with state policy environments may be associated with both cigarette taxes and youth smoking. If, for example, states with higher cigarette taxes attach the strongest social stigma to youth smoking, then cross-section estimates will overstate the impact of raising the cigarette tax on the proportion of teens who smoke.

DeCicca, Kenkel and Mathios offered this critique in a series of studies using data from the National Educational Longitudinal Study (NELS) [19–21]. When the authors included state-fixed effects in their specifications (and therefore identified the effect of cigarette taxes on smoking using within-state variation over time), they found no relationship between youth smoking participation and changes in cigarette taxes. The authors interpreted this finding as evidence that state-specific unobserved antismoking sentiment creates an upward bias when cross-state variation is used as a source of identification.

In fact, a large number of studies have found that relying on within-state variation reduces the estimated effect of cigarette taxes on youth smoking. Nevertheless, there is evidence of negative and significant tax effects even when state-fixed effects are included as controls [14,22–26]. However, it should be pointed out that in many of these studies, the results are fairly sensitive to the sample employed. For instance, Dee estimated the effects of state cigarette taxes from 1977 to 1992 using data from Monitoring the Future, but only found negative tax effects in the 1985–1992 period [22]; when all 16 years were pooled, the cigarette tax effects were insignificant and of the wrong sign. Gruber [22] and Gruber and Zinman [25] found that cigarette taxes affected smoking participation for high-school seniors, but not younger adolescents.

While the inclusion of state-fixed effects may ameliorate concerns about omitted variable bias, this comes at the cost of reducing potentially important identifying variation, and thus limiting the power of the research design. Two strategies in the literature have attempted to deal with this concern.

First, rather than rely on the ‘black box’ of state-fixed effects, DeCicca et al. attempted to measure a potentially important variable omitted from past cross-sectional regressions: state-specific smoking sentiment [27]. The authors used information from the Current Population Survey’s (CPS) Tobacco Use Supplements. Respondents in the CPS were asked about their views on the promotion and advertising of tobacco products, policies that restrict smoking on public or private property, and whether they permit smoking in their homes. DeCicca et al. used this information to create a state-level antismoking sentiment measure that when included as an additional control, substantially reduced cross-sectional estimates of the effect of cigarette taxes [27].

A second strategy has been to continue to include state fixed effects, but to exploit the fact that in recent years there has been more within-state variation in state cigarette taxes. Using data from national, state and local versions of the Youth Risk Behavior Survey from 1991 to 2005 and controlling for state effects, year effects, clean indoor air laws and demographic characteristics, Carpenter and Cook found that a $1 increase in state cigarette taxes reduced youth smoking participation by 10–20%, with implied own-price elasticities of -0.24 to -0.56. Unlike prior studies, the estimated effect appeared to be robust across a number of specifications [14].

While the results of the Carpenter and Cook study support the view that higher cigarette taxes may discourage smoking, there remains some question regarding how one should interpret estimated tax effects in the youth smoking participation models [14]. That is, do higher taxes cause existing teen smokers to quit or provide disincentives for nonsmokers to start? As DeCicca et al. noted, it is only appropriate to pool samples of smokers and nonsmokers if the latent variables underlying participation initiation and cessation are identical [20]. In their analysis of a longitudinal sample from the 1992 and 2000 waves of the NELS, DeCicca et al. rejected the hypothesis that it is appropriate to pool smokers and nonsmokers [20]. After controlling for antismoking sentiment, they found no evidence that higher cigarette taxes prevent youth smoking initiation, and found only limited evidence that cigarette taxes are positively associated with youth smoking cessation (implied price elasticity of 0.47). However, even the cessation effect was not statistically distinguishable from zero. The authors concluded that, in contrast to conventional wisdom, “adult smoking behavior is more, not less, tax- or price-responsive than adolescent smoking” [20].

Taken together, evidence on the effectiveness of cigarette taxes in reducing youth smoking is mixed. There does appear to be some evidence that state cigarette tax increases from 2000 to 2005 had the effect of discouraging youth smoking, but it is not clear whether this effect operated at the initiation or cessation margin.

Finally, while there is a good case to be made that cigarette taxes are effective, the evidence on school-level tobacco prevention programs is much weaker. A number of studies have found that school-level antitobacco programs may have short-run effects in deterring smoking [28], but a review of school-based randomized control trials found no evidence that such programs have significant long-run effects [29]. Thus, it is not clear that public expenditures on typical school-based tobacco prevention programs will produce important long-run behavioral effects, although there is some evidence that smoking restrictions in public places are associated with lower rates of youth smoking [31,32].

The role of peers

In summary, the most common public policy strategies designed to combat adolescent smoking in the US – cigarette taxes and school-based tobacco programs – appear to be only moderately successful at best. In designing future policies that may be more effective, we believe that greater attention is owed to the influence of peers in triggering youth smoking behavior. DiCicca et al. raised this point in their seminal study: “the monetary price of cigarettes may play a relatively small role in youth smoking decisions … the Surgeon General’s report on preventing youth smoking concluded that the influence of peers plays a “powerful role,” because “smoking initiation appears to be a component of peer associations and peer bonding in adolescence, as peer groups establish shared behaviors to differentiate themselves from other adolescents and adults” [19,30].

However, simply estimating the relationship between the probability that an adolescent smokes (or takes up smoking in the future) and the percentage of his or her peers who smoke is not enough. Due respect must be paid to the problems of selection and shared, but difficult to measure, environmental factors. Specifically, more studies that take advantage of natural experiments that, in effect, randomly assign adolescents to different peer groups are needed. Without more research based on natural experiments, the role of peers in the decision of youths to take up smoking will remain a puzzle.

Financial disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.