Survival data from eight Cancer and Leukemia Group B (CALGB) protocols were examined for patients with lung cancer (N = 961), multiple myeloma (N = 577), gastric cancer (N = 231), pancreatic cancer (N = 174), breast cancer (N = 87), and Hodgkin's disease (N = 58). After accounting for differences in survival rate attributable to type of cancer, initial performance status, age, and 14 other protocol-specific prognostic indicators, the additional predictive value of socioeconomic status (SES) was evaluated. Race (white v non-white) was not a significant predictor of survival time, but income and education were. People with lower annual incomes (below $5,000 per year in the years 1977 to 1981) and those with lower educational level (grade school only) showed survival times significantly shorter than those with higher income or education, respectively. These survival differences were associated with, but could not be fully explained by, severity of disease at initial presentation. SES continued to exert a small but significant impact on cancer survival, even after controlling for all known prognostic variables. Economically and educationally disadvantaged cancer patients may require treatment programs that include education about treatment and compliance, even after an initial diagnosis is made and treatment is initiated. Because SES is related to survival independent of all known prognostic variables, it should be included in the data bases of clinical trial groups to provide a more accurate test of the effectiveness of new therapies.