Exposure to new smoking environments and individual-level cigarette smoking behavior: Insights from exogenous assignment of military personnel

Abstract

Despite overall declines in cigarette smoking prevalence in the United States (U.S.) in the past several decades, smoking rates remain highly variable across geographic areas. Past work suggests that smoking norms and exposure to other smokers in one's social environment may correlate with smoking risk and cessation, but little is known about how exposure to other smokers in one's community is causally linked to smoking behavior - in part due to endogeneity and inability to randomly assign individuals to different 'smoking environments.' The goal of this study was to evaluate how exposure to localities with high population-level smoking prevalence affects individual-level cigarette smoking behaviors, including quitting. The study addresses key limitations in the literature by leveraging a unique natural experiment: the plausibly exogenous compulsory assignment of military personnel to installations. Logistic and multivariate regressions estimated cross-sectional associations between smoking/quitting behaviors and our proxy for social environments for smoking, county-level smoking prevalence (CSP). Across 563 U.S. counties, CSP ranged from 3.8 to 37.9%. Among the full sample, a 10 percentage point increase in CSP was associated with an 11% greater likelihood of smoking. In subgroup analyses, young adults, women, those without children in the household, and risk/sensation-seekers were more likely to smoke and less likely to quit when exposed to counties with higher CSP. Relocation to areas with high population-level smoking prevalence may increase likelihood of smoking and impede quitting, and may disparately affect some population subgroups. Findings provide novel evidence that community smoking environments affect adult smoking risk and underscore a need for sustained, targeted efforts to reduce smoking in areas where prevalence remains high.

Keywords: Cigarette smoking; Smoking environments; Smoking norms; Social contagion.

Figure 1.. Association between county smoking prevalence and probability of current cigarette smoking in the 2011 Health Related Behaviors Survey

This figure shows the model-based probability of current cigarette smoking status (dependent variable) in relation to county smoking prevalence (independent variable) from the fully-adjusted logistic regression model. Plotted values are model-based probabilities of current smoking at the mean values for all covariates in the model; upper and lower bounds reflect 95% confidence intervals for predicted probability at a given level of county smoking prevalence. The logistic regression model included covariates for state tobacco control policies (cigarette taxation rate; presence of smoke-free air laws in restaurants or bars), county unemployment rate, and individual-level sociodemographic and service characteristics (age, sex, race/ethnicity, education, children in household, marital status, service branch, pay grade, time at installation, combat exposure, and past-year deployment).