Read this article
- Access statistics
- 2,905 article downloads
- 1,909 complete issue downloads
- Total: 4,814
Graphic warning labels are gruesome images on cigarette packages depicting diseased body parts, suffering, and death. An increasing number of countries have mandated them. The scholarly literature unequivocally concludes that they are effective, as measured by elicited cognitive and emotional reactions, increased motivation to quit smoking, and greater awareness of the risks and warnings. But studies investigating actual changes in smoking behavior are more limited. Focusing on a widely cited difference-in-difference study in Tobacco Control by Jidong Huang, Frank Chaloupka, and Geoffrey Fong (2014), we discuss some of the challenges faced in attempting to estimate the impact of cigarette graphic warning labels on population smoking rates in Canada. We demonstrate that spurious correlation may arise when the key underlying assumptions of the difference-in-difference methodology do not hold, or when the specification is plagued by such issues as serial correlation or omitted variables. Our findings suggest that there remains uncertainty in the causal relationship between Canada’s policy that mandates graphic warning labels and smoking rates. Estimates could be improved with additional data, better-specified instruments for problematic data, or alternative methods that would address the limitations identified in the present study.