Panel Paper: An Application of Unconditional Quantile Regression to Cigarette Taxes

Saturday, November 10, 2012 : 1:45 PM
International D (Sheraton Baltimore City Center Hotel)

*Names in bold indicate Presenter

Johanna Catherine Maclean, University of Pennsylvania, Douglas Webber, Temple University and Joachim Marti, University of Leeds

Despite numerous anti-smoking policy initiatives, smoking remains the leading preventable cause of death and disease in the United States, and 20.6% of American adults smoke. Each year, smoking is associated with 443,000 deaths, 5.1 million years of potential life lost, and $96.8 billion (2005 dollars) in productivity costs. There is consensus among policymakers and public health advocates that cigarette taxation is one of the most effective policy measures to curb tobacco consumption, and both Federal and state governments use taxes to reduce smoking levels. Although the majority of empirical evidence on cigarette tax effectiveness suggests that tax increases lead to reductions in smoking behaviors, a small set of studies calls this relationship to question. Despite some exceptions, recent work has mainly focused on the extensive margins (participation, initiation, cessation) or mean effects. Because economic theory hints at heterogeneity in tax elasticity among smokers, the literature has potentially passed over a significant portion of policy relevant information if smokers at different levels of consumption respond differentially to tax increases. This study investigates heterogenous response to state cigarette taxes across the smoking distribution using the recently developed unconditional quantile regression (UQR) technique. We first motivate the use of UQR over conventional conditional quantile regression (CQR). Although we are not the first study to document the appealing features of UQR over CQR, we argue that UQR provides more policy-relevant information and particularly in state policy evaluations. Although we focus on state policies, the central concept of our argument extends to analysis of group level data in general. Second, we document cigarette tax elasticity across a sample of smokers drawn from the Current Population Survey Tobacco Use Supplements between 1992 and 2010. Our OLS regressions show a negative relationship between cigarette tax and smoking volume, but the effect is imprecisely estimated. Results from UQR reveal a U-shaped relationship: light and heavy smokers do not respond to tax increases while moderate smokers (those between the 25th and 45th quantiles of the unconditional distribution) may reduce their smoking behaviors following a tax increase. However, the magnitudes of all estimated effects are small. Our preferred specifications imply that a $1.00 increase in cigarette tax leads to a 5.0% reduction in cigarettes smoked in the past 30 days among the most responsive smokers (those at the 40th quantile). Our findings suggest that focusing on mean effects masks important features of the tax-smoking relationship and that the majority of smokers do not substantially alter their smoking patterns when faced with higher tax. Thus, future tax increases may not lead to large reductions in smoking outcomes. These findings have important implications for effective public health policy development.