Session: Tobacco Control Programs
Room: Hollister McManus Lounge
Time: Tue 10:15-11:45

Presenter: Johanna Catherine Maclean (Cornell University. Economics)

Discussant: Christopher Carpenter (University of California, Irvine)

Abstract

Despite numerous anti-smoking policy initiatives, smoking remains the leading preventable cause of death in the US. The prevalence of adult smoking fell from 25 percent in the mid-1990s to 21 percent in 2004, but since then the decline appears to have stalled and current levels are above the national objective of 12 percent. In the 1990s states began to establish tobacco control programs (TCPs) that fund activities to reduce smoking. By 2002, TCP expenditures across all states totaled $900 million, but by 2007 TCP expenditures had dropped to a $660 million. In this study, we explore the empirical relationships between TCP expenditures and TCP activities and between TCP activities and smoking outcomes.

We extend previous research by conducting a more complete investigation into how TCP expenditures might translate into reductions in smoking. Previous health economics studies take a reduced-form “black box” approach and regress a smoking outcome on TCP expenditures and a set of control variables. The consensus from black box studies is that higher TCP expenditures reduce smoking, but these studies shed little light on the mechanisms inside the box. The outputs of state TCPs are counter-tobacco activities. In black box studies expenditures proxy for activities. In this study we use data on specific TCP activities to go inside the black box.

In our first set of empirical models we estimate cost functions that show how TCP costs vary with program activities. We focus on two activities recommended by the CDC: anti-smoking advertising campaigns and smoking cessation counseling via quitlines. To measure TCP advertising activities, we use data from TNS/ Media Intelligence on over 300,000 televised airings of state-sponsored anti-smoking advertisements between 1996 and 2004. To measure TCP quitline activities, we use data from the North American Quitline Consortium Survey on quitline services between 2004 and 2008. We use residuals to identify high and low cost states conditional on activity level. Stochastic frontier analysis is used to test whether TCPs are operating on the production frontier. Findings suggest that although TCP costs vary with program activities in the expected directions, there are a substantial number of outliers and TCPs may not be operating on the production frontier.

In our second set of empirical models, we estimate cigarette demand equations to test the effect of TCP activities on smoking outcomes. We use three nationally representative individual-level data sets (YRBS, BRFSS, and CPS) and historical cigarette sales data. Unlike many black box studies, we seriously address the econometric challenges in state policy evaluation: ill-conditioned data, correlated policies, unobservable between-state differences, and reverse causality. To address these challenges we estimate a series of increasingly sophisticated econometric models. Naïve cigarette demand models suggest that TCP activities are economically and statistically associated with smoking outcomes in the expected directions. Once more sophisticated econometric techniques are employed to model the demand for cigarettes, many of these relationships vanish. An interpretation of our findings is TCPs have limited effect on smoking outcomes and previous findings to the contrary may be artifacts of model mis-specification and/or data limitations.

Key Terms

Tobacco control program, Smoking initiation and cessation, Program evaluation

Authors:

Donald Kenkel (Cornell University. Policy Analysis and Management) , Johanna Catherine Maclean (Cornell University. Economics) and Hua Wang (Cornell University. Policy Analysis and Management)