The Self-Medication Hypothesis: Evidence from 9/11

We use simultaneous instrumental variable models to explore if smoking self-medicates short-term stress. Short-term stress is considered a powerful motivator of smoking, but the decision to smoke could trigger biological feedback that immediately reduces short-term stress. This feedback confounds estimates of the relationship between stress and smoking. Omitted variables, such as genetic or social factors, could also suggest a spurious correlation.

To disentangle these effects, we use a simultaneous instrumental variable model (Roodman, 2011) and 541,983 individual-level observations from the Behavioral Risk Factor Surveillance System for the years 1999-2001. The cross-sectional data contains questions about smoking status and a proxy for stress, which is the number of days over the past 30 days that an individual has had problems with stress, depression, or other emotional problems. We jointly estimate the effect of stress on smoking and smoking on stress. We instrument stress with characteristics of the terrorist attacks occurring on September 11, 2001, measured as both temporal distance (using date of interview) and spatial distance (using distance between county of residence and the terrorist attack epicenters). We instrument smoking with state-level cigarette prices and the prevalence and strength of state-level indoor smoking laws. We find causal evidence that stress increases smoking and that smoking reduces stress, suggesting that in the short run smokers successfully self-medicate stress. However, the long-term effects of smoking on stress are unanswered by this research. The design of tobacco control measures should take into account that smokers are influenced to smoke during periods of high stress, and are successful in the short term in reducing this stress.