SETTING AND OBJECTIVE: The reversal of the decline in United Kingdom tuberculosis rates has sparked a resurgence of interest in the epidemiology and prevention of tuberculosis in the UK. In this paper we quantify the primary factors explaining the variability in the electoral ward level relative risk of tuberculosis in Manchester, Liverpool, Birmingham and Cardiff. DESIGN: Ecological analysis of the incidence of tuberculosis in 397 wards using hospital admissions data as a proxy for tuberculosis incidence. Admissions were evaluated from the financial years 1991/1992 to 1994/1995. Ward level covariates included measures of country of birth, ethnicity and various socio-economic measures. RESULTS: Separate analyses were carried out for pulmonary and non-pulmonary tuberculosis. For pulmonary tuberculosis the final model included measures of the ward population born in India and Pakistan, overcrowded housing and not-owner-occupied housing. For non-pulmonary tuberculosis the covariates were ward population born in India and Pakistan, overcrowded housing and the proportion of households with no car. CONCLUSIONS: The country of birth of the ward population is the single most influential explanatory factor in the variability of the ward rates for both pulmonary and non-pulmonary tuberculosis in these four cities. Measures of poverty are of secondary importance.