Blacks Smoke Less (and More) than Whites: Simpson's Paradox in U.S. Smoking Rates, 2008 to 2012.

In what is known as Simpson's Paradox, a statistical association present within different groups becomes reversed when the groups are aggregated. Because race is confounded with socioeconomic status (SES) in the U.S., the overall smoking rate among Blacks may exceed that among Whites, even while the reverse is true within SES strata (or when adjusting for SES). Indeed, in the most recent five iterations of the Behavioral Risk Factor Surveillance System (2008-2012), a nationwide dataset, unadjusted models found that non-Hispanic Blacks were more likely than non-Hispanic Whites to report current smoking (e.g., in 2012: OR=1.14, p<.0001). However, Blacks were less likely than Whites to report smoking across all five years when models adjusted for income (e.g., in 2012: OR=0.85, p<.0001) and in 2008-2010 when models adjusted for education. This reversal of association reflects racial disparities in SES, which cause Black Americans to be disproportionately affected by smoking risk factors associated with low income and education.