Josh will discuss early analysis of a newly created, very large public data set on old-age, individual mortality in the United States. By linking individual records from the 1940 census and Social Security Administration deaths, Josh and colleagues have built a new data set that allows the study of detailed covariates on a sample of millions of people. Josh will explain how the data set was created, some special methods needed to estimate mortality, and preliminary results on trends in educational, income, and racial disparities in mortality at ages 65+.