The Geography of Smoking

One in five Americans continue to smoke cigarettes, according to a new survey by the Centers for Disease Control and Prevention. The smoking rate varies from low of 9.2 percent in Utah to a high of 26.6 percent in West Virginia. The map below, from the Wall Street Journal, shows the smoking rate by state.

The data are interesting and they allow us to look at the extent to which smoking is associated with all sorts of things, from more obvious ones like cancer and heart disease to the economic and demographic characteristics of states with higher or lower levels of smoking and even the relationship between smoking and happiness. With a helpful analytical assist from my colleague Charlotta Mellander, we decided to take a quick look. We ran some simple correlations and scatter-plots between state smoking rates and these factors. As usual, we point out that correlation does not imply causality, but simply points to associations between variables. Still, a number of interesting things stand out.

It will come as little surprise that states with higher levels of smoking have significantly higher rates of death from cancer, heart disease, and cerebrovascular diseases like hypertension.  There is a significant correlation between state smoking rates and death rates from cancer (.75), heart disease (.7), and cerebrovascular disease (.6).

It might be, however, that states with greater percentages of smokers are those where people pay less attention to their health generally or are more likely to engage in risky behavior. Consider the relationship between state smoking rates and their levels of obesity, where we find significant association both for obesity among adults (.7) and children (.6).

Might smoking be related to states' broader social and psychological climates? To get at this, we looked at the relationship between smoking and a commonly used measure of subjective well-being or happiness developed by the Gallup Organization. Smoking is negatively associated with state happiness (with a correlation of -.7). Since these correlations only reflect associations between variables and not causality, it's hard to say whether this reflects the fact that happier people smoke less or unhappier ones smoke more, or that both smoking and happiness levels reflect something else. To get at this, we look at the associations between state smoking levels and social and demographic factors below.

Common sense would suggest that more affluent people would smoke less and poorer ones would smoke more, but that's not what the data indicate - at least when comparing states. State smoking levels are not related to state income levels or to Gross State Product per capita; the correlations for both are not statistically significant.

One would think that more highly educated people smoke less. And that is borne out by our analysis. Smoking is highly associated with education levels, measured as the percentage of adults with a college degree (with a negative correlation of -.8).

To what extent does smoking reflect the kind of work people do? We examine the relationships between smoking levels and three classes of jobs - creative/professional/ knowledge jobs, blue-collar working class jobs, and standardized service class jobs like those in food processing and home health care. The strongest association is with working class jobs, with a correlation of .5: Smoking is higher in states with a greater concentration of these blue-collar jobs. Smoking is also associated with service class jobs. But here the correlation is negative (-.6). Smoking does not appear to be associated with knowledge-professional-creative jobs, the correlation here is not statistically significant.

That said, smoking rate is associated with concentrations of artists, musicians, and entertainers. Contrary to the stereotypical image of cigarette-puffing bohemians or hipsters, smoking is less prevalent in states with more of these artistic types: The correlation is negative (-.5).

Lastly, smoking is negatively correlated with larger concentrations of gays and lesbians, as well as immigrants (both with correlations of roughly -.45). This likely reflects broader structural characteristics of those states, as more highly educated states also tend to be more tolerant and open to diversity.