The strong associations between education level and prevalence and severity of arthritis have recently been investigated, with some suggestion that the relationships might be causal. Data from 4 national probability samples were analyzed, in which occupation and income were measured currently and prior to the development of arthritis. These data indicate that previous studies overestimated the strength of the association between schooling and arthritis because income and, especially, occupation were not analyzed as separate covariates. The overestimate appears to be especially high for persons currently employed, for men, and for persons employed in dangerous jobs. Health policy strategies directed toward reducing arthritis rates require not simply a focus on education, but the additional socioeconomic status dimensions of income and occupational safety and health.