We studied the association between socioeconomic status and survival in a prospective study of 364 HIV-infected homosexual men who were recruited during 1982-84. The participants were divided by annual income; those earning above Canadian $10,000 (high-income; n = 274) and those below $10,000 (low-income; n = 90) at recruitment. The latter threshold closely approximated to the poverty level for this population. Low income men were significantly younger than high income men but the groups were similar with respect to baseline CD4 counts, subsequent use of anti-retrovirals and prophylaxis against Pneumocystis carinii pneumonia (PCP), and number of visits attended during follow-up. Subjects were followed for a median of 9.5 years (range 1.8-13.1). By Dec 31, 1993, there were 135 deaths yielding a cumulative mortality rate of mean 45% (SD 4.0) at 11.5 years. Men aged 30 or more at infection had poorer survival than those under 30 (mortality risk ratio 1.56; 95% CI 1.09-2.24; p = 0.015), and longer survival was significantly associated with a higher CD4 count at the earliest seropositive visit. The age-adjusted mortality risk ratio for low income men compared with high income men was significantly increased at 1.63 (95% CI 1.11-2.40; p = 0.013). The significant risk of death for low income men persisted despite adjustment for age at infection, CD4 count, use of zidovudine, dideoxyinosine, and dideoxycytidine, use of PCP prophylaxis, and year of infection. We cannot attribute our findings to income loss as a result of more rapid HIV progression because the same effect was present in people who provided income data before seroconversion. Similarly, our findings are not due to differential access to care because the study was done within the context of a universal health care system, and the two income groups received treatments equally. This finding is consistent with the association of lower socioeconomic status with increased morbidity and mortality observed within large populations and in other diseases.