Life Course Earnings in Health and Middle Age

David Rehkopf, University of California, San Francisco and University of California, Berkeley
M. Maria Glymour, Columbia University

Research on income and health frequently relies on self-reported income at a single time point. We assessed how these data limitations modified observed associations between earnings and health in 5,951 Health and Retirement Study participants with Social Security Administration recorded earnings (SSE) available for 1951-1991. Self-reported 1992 health outcomes assessed disability, depressive symptoms, and diabetes, cancer, heart disease, and stroke diagnoses. We compared: 1991 self-reported income; 1961-1991 SSE; multiple-year average SSE; and year-to-year SSE declines (adjusting for average earnings). For most outcomes, findings indicate similar results using administratively reported (SSE) data and self-reported earnings. Earnings averaged across 30 years predicts after controlling for most recent year of earnings. Declines in earnings ages 25-50 predict health independently of average earnings level. Comparing results including and excluding zero earnings years suggests that coding zero earnings years as “low earnings” introduces substantial bias, possibly due to sick individuals leaving the labor force.