The relationship between income and survival rates for chronic dialysis patients in Michigan was examined. To evaluate the relative risk (RR) of dying, by income, a Cox survival regression model was used to adjust for age, race (black versus white), gender, dialytic treatment modality, year of first end-stage renal disease (ESRD) therapy, and primary cause of ESRD. The average household income reported from census data for the ZIP Code of residence for each patient was analyzed as a socioeconomic indicator. Treatment modality on day 120 of ESRD was classified as either center hemodialysis (HD) or CAPD. All new patients, aged 20-59 and registered at the Michigan Kidney Registry between 1/1/80 and 12/31/87 were included in the study. Patients were followed from day 180 of ESRD until death, censoring at transplant, or 12/31/87. The adjusted relative risk of dying decreased for black patients by 3.3% per $1,000 increase in income (p less than 0.01), while the trend by income for white patients was negligible (p greater than 0.10). The difference in trends for the two groups was statistically significant (p less than 0.01). This is a surprising result, since white patients have higher death rates, overall, than do black patients, particularly among subgroups with diabetes and hypertension whose RR was 1.77 and 1.82, respectively. Poor socioeconomic status of the area of residence has a strongly negative effect on survival for black patients, but not for white patients. The reasons for the relationship of death rates with income, especially for black patients, need to be examined in greater detail.