The complementarity of teen smoking and drinking

Abstract

Teen drinkers are over twice as likely as abstainers to smoke cigarettes. This empirical study provides evidence of a robust complementarity between these health behaviors by exploiting the “cross-price” effects. The results indicate that the movement away from minimum legal drinking ages of 18 reduced teen smoking participation by 3 to 5%. The corresponding instrumental variable estimates suggest that teen drinking roughly doubles the mean probability of smoking participation. Similarly, higher cigarette taxes and reductions in teen smoking are associated with a lower prevalence of teen drinking. However, the results which rely on cigarette taxes for identification are estimated imprecisely.

Introduction

Over the past several decades there has been a broad variety of aggressive public health initiatives aimed at curbing the prevalence of abusive drinking and tobacco use as well as their related consequences. There has been a particular interest in reducing alcohol and cigarette use among young adults.¹ This interest has motivated a large number of econometric studies that examined the policy responsiveness of teen smoking and drinking.² Most of these econometric evaluations have considered these consumption decisions in isolation. More specifically, relatively little attention has been paid to the substitutability or complementarity of such “sin” goods among teens. Furthermore, the empirical research that has addressed such issues has largely emphasized the substitutability of various drugs and how this might attenuate the desirability of policies that reduce the availability of a particular substance. For example, DiNardo and Lemieux (1996)report that the movement to higher minimum legal drinking ages (MLDA) led teens to substitute marijuana for alcohol.³

However, the consumption of alcohol and cigarettes by teens might constitute an important case where the goods are complements instead of substitutes. In particular, to the extent these goods are economic complements, the considerable public health efforts aimed at reducing the prevalence of teen smoking and drinking may generate important and unintended benefits. This empirical study provides evidence on this policy-relevant question by relying on the “cross-price” effects in models of teen alcohol and cigarette use. More specifically, this study evaluates the complementarity between teen smoking and drinking by exploiting the exogenous variation in the full teen prices of alcohol and tobacco generated by changes in cigarette taxes and state minimum legal drinking ages (MLDA).⁴ These econometric evaluations are based on pooled cross-sections from the 1977–1992 Monitoring the Future (MTF) surveys of high school seniors. An important and unique feature of these data is that because they contain both time-series and cross-sectional variation, the empirical results presented here do not need to rely exclusively on the conventional cross-sectional identification strategies which can confound the effects of state alcohol and tobacco policies with unobserved, state-specific determinants of these teen behaviors.⁵

The MTF data are first used to estimate the magnitudes of the stylized links between teen drinking and smoking. The results demonstrate that there are dramatically large and positive partial correlations between these consumption behaviors. For example, teens who drink are more than twice as likely as abstainers to smoke cigarettes. To the extent such a relationship is robust, it suggests that the initiation of a smoking habit can be understood as a major and largely overlooked welfare consequence of teen drinking. However, these naive empirical models may be highly misleading. The unobserved determinants that increase the likelihood of consuming alcohol are also likely to increase the probability of smoking participation. This study presents less ambiguous evidence on the relationships between teen alcohol and cigarette use by relying on the exogenous variation in these health behaviors generated by their policy determinants (i.e., cigarette taxes and minimum legal drinking ages). These identification strategies are implemented by evaluating reduced-form models which provide direct evidence of “cross-price” effects as well as by evaluating instrumental variables (IV) estimates which provide direct evidence on the structural relationship between these behaviors.

The results of these varied estimations provide consistent evidence that there is a strong complementarity between teen smoking and drinking. For example, reduced-form models indicate that the movement away from a minimum legal drinking age of 18 reduced teen smoking participation by 3 to 5%. The related instrumental variables (IV) estimates indicate that teen drinking roughly doubles the mean probability of smoking participation. These marginal effects are statistically significant and robust to a variety of specification changes. Furthermore, the magnitudes of these marginal effects are consistent with the empirical benchmarks established by the stylized partial correlations between teen drinking and smoking.⁶ However, empirical models that estimate the effects of cigarette taxes on teen drinking participation provide less reliable evidence of this complementarity. In part, this may be due to important structural changes in teen smoking behavior over this period. Evans and Huang (1998)find that the tax responsiveness of teen smoking participation was dramatically higher during the 1985–1992 period than during the 1977–1992 period. Consistent with these results, teen drinking models based on only the 1985–1992 data uniformly indicate that higher cigarette taxes are associated with reductions in teen drinking. However, these marginal effects are statistically imprecise and, in some specifications, implausibly large.⁷ In an important sense, the relative sensitivity of the results based on cigarette taxes is not surprising. Though there is considerable evidence that teen smoking participation is significantly tax-responsive, the magnitude of this responsiveness is relatively small and imprecise. In contrast, the nationwide increases in minimum legal drinking ages generated considerable variation in teen alcohol use and, thus, allow the complementarity between teen smoking and drinking to be evaluated more precisely.