Despite the long-standing observation that tuberculosis (TB) case rates are higher among racial and ethnic minorities than whites in the United States (U.S.), the proportion of this increased risk attributable to socioeconomic status (SES) has not been determined. Values for six SES indicators (crowding, income, poverty, public assistance, unemployment, and education) were assigned to U.S. TB cases reported from 1987-1993 by ZIP code- and demographic-specific matching to 1990 U.S. Census data. TB risk between racial/ethnic groups was then evaluated by quartile for each SES indicator utilizing univariate and Poisson multivariate analyses. Relative risk (RR) of TB increased with lower SES quartile for all six SES indicators on univariate analysis (RRs 2.6-5.6 in the lowest versus highest quartiles). The same trend was observed in multivariate models containing individual SES indicators (RRs 1.8-2.5) and for three SES indicators (crowding, poverty, and education) in the model containing all six indicators. Tuberculosis risk increased uniformly between SES quartile for each indicator except crowding, where risk was concentrated in the lowest quartile. Adjusting for SES accounted for approximately half of the increased risk of TB associated with race/ethnicity among U.S.-born blacks, Hispanics, and Native Americans. Even more of this increased risk was accounted for in the final model, which also adjusted for interaction between crowding and race/ethnicity. SES impacts TB incidence via both a strong direct effect of crowding, manifested predominantly in overcrowded settings, and a TB-SES health gradient, manifested at all SES levels. SES accounts for much of the increased risk of TB previously associated with race/ethnicity.