This paper illuminates the demographic and socioeconomic factors assoc
iated with smoking statuses. It employs the 1990 National Health Inter
view Survey's Health Promotion and Disease Prevention Supplement and l
ogistic regression to examine the covariates of smoking status among t
he U.S. adult population. Polychotomous logistic regression, which pro
vides an alternate way to examine smoking practices, simultaneously co
nsiders multiple variables and their interactions, controls for other
important covariates, produces predicted values and patterns, and allo
ws multiple comparisons. By examining interactions, we have found that
age displays distinct, often curvilinear, patterns with smoking; that
compared to females, males have higher rates of cigarette consumption
except at the youngest ages; that Anglos, especially Angle males, exh
ibit high probabilities of cigarette consumption but also high probabi
lities of being former smokers; that Black males exhibit high probabil
ities of light smoking, but only at the older ages, and that they also
exhibit high probabilities of being former light smokers; and that Me
xican-American females are the least likely to currently smoke or to h
ave ever smoked. By clearly elucidating the demographic and socioecono
mic;variations in smoking, arguably the most salient marker of health
behavior, we can better target programs and policies geared toward the
reduction of smoking-related diseases and deaths.