Persistence of cigarette advertising across media and smoking rates

Abstract

This article uses state-level US panel data to examine the effects of cigarette advertising on smoking. We disaggregate cigarette advertising into (1) print, (2) outdoor, (3) entertainment and (4) store. How does advertising persist over time across advertising media? Are there differences in effectiveness of advertising across different smoking rates? Results show that the price elasticity of cigarette demand is negative and within the range of recent estimates, whereas border price elasticities and income elasticities are positive. All elasticities, however, are uniquely sensitive to smoking rates, as is the effectiveness of the Master Settlement Agreement. Current aggregate advertising increases cigarette demand, whereas accompanying negative advertising takes hold in the third year. Upon disaggregation, current store advertising increases smoking in most cases. However, store advertising has a negative effect by the third year, especially in states with smoking rates at or above the median. Other advertising media have weak or no effects.

Acknowledgements

I thank Michael Nelson for comments, Christopher Baum for help with econometric estimation and Michael Naretta for research assistance. Remaining errors are mine.

Notes

¹The study by Goel and Ram (2004) is the only other study examining cigarette demand responses across smoking rates. However, the authors use US data from only 1993 to 1999 and, more importantly, do not control for advertising effects.

²Our sample data begin in 1975. Hence we are not replicating the effects of the 1971 broadcast advertising ban that has garnered attention of a majority of the advertising restrictions literature focusing on the United States (Baltagi and Levin, 1986; Mitchell and Mulherin, 1988; Eckard, 1991; Goel and Morey, 1995).

³We also had information on Internet advertising. However, such data were available for only a few years (limiting the consideration of lag effects) and their magnitude was relatively small. Hence, whereas Internet advertising is included in total cigarette advertising, a separate Internet advertising category is not considered here. Evans (2009) provides a nice overview of the issues related to online advertising.

⁴In the absence of data on individual cigarette consumption, we follow related studies by taking cigarette sales to proxy for consumption (see, e.g. Baltagi and Levin, 1986).

⁵Alaska and Hawaii have no immediate bordering states and are dropped from the analysis.

⁶The Hausman test's asymptotic assumption was not satisfied; thus, the default FE model results are reported.

⁷One exception is the positive, marginally statistically significant, effect of EntrtnADV ₋₂ in the median regression.

⁸This may be viewed as the effectiveness of indirect anti-smoking initiatives. Given appropriate data, one could also include government spending on anti-smoking advertising and compare the efficacy of direct and indirect anti-smoking initiatives (see, e.g. Pechmann and Ratneshwar, 1994).