We present a study of demographic factors associated with self-reported back pain prevalence in a population-based study. Study variables were age, gender, occupation, physical demand of occupation, income, education and marital status. Crude associations of the independent variables and reported back pain were in general agreement with other cross-sectional studies. Multiway contingency tables and multivariate models were employed for further analyses. When we controlled each of the other variables, only gender, education and marital status remained strongly associated with reported back pain. Using these three variables plus age, we developed regression models for the prediction of back pain and identified a steep gradient of prevalence: the highest prevalence (44-46%) was in no-longer-married women aged 50-64, regardless of education, while the lowest prevalence (9-11%) was in men who were married and had greater than a high school education. Because the models employ only census-type variables, they should be easy to validate or use to predict back pain prevalence in other populations.