We have analyzed mortality data by socioeconomic status for 232 million employed persons aged 35-64 from 1984-1997 in the U.S. Numerators came from death certificates coded for occupation in approximately 20 states. Denominators came from U.S. Census estimates for specific occupations. Four socioeconomic groups were formed based on assigning Nam-Powers scores (which reflect income) to each occupation. For heart disease there was a strong gradient from poorest to richest (rate ratios 2.11, 1.63, 1.40, 1.00). Heart disease mortality for men dropped more quickly over the 14 year period for the wealthier (annual % decrease, poorest to richest, 1.4, 2.4, 3.2, 3.1), but this was not true for women (2.3, 1.6, 2.8, 2.2). For cancer, there was a strong gradient for men (RRs poorest to richest, 1.86, 1.54, 1.35, 1.00) but not for women (1.04, 0.90, 1.01, 1.00). Cancer showed a 0.7% annual decrease over the 14 year period, which did not differ by social class. Analysis of specific sites revealed a strong gradient for lung cancer (RRs poorest to richest, 2.16, 1.77, 1.52, 1.00), a very slight gradient for colorectal cancer (1.16, 1.06, 1.09, 1.00), and an inverse gradient for breast cancer (0.77, 0.80, 0.91, 1.00). All three cancers decreased slightly over 14 years; lung and colorectal cancer dropped more quickly for the wealthier. NHANES 1 and NHANES 3 data suggested that change in risk factors would not fully account for heart disease mortality trends over time, nor for the more rapid decline among the wealthier; it is likely that better treatment for the wealthier is an important factor.