There is widespread interest in disparities in health status across income groups and other classifications of socioeconomic status. In Canada, as in many other countries, there is considerable evidence showing such disparities. This study reports an analysis of male mortality at ages 65 to 74 in relation to socioeconomic characteristics, specifically employment and self-employment earnings histories during the 10 to 20 years prior to age 65, marital status, disability, and age at retirement. The analysis is based on administrative data from the Canada Pension Plan covering more than 500,000 individuals. Significant mortality gradients are found throughout the earnings spectrum. These gradients are also clearly evident in a multivariate context. The results illustrate the major potential of administrative data for research. Substantively, the results cast doubt on the primacy of causal explanations such as "reverse causality" and "health selection" and raise important questions regarding pension and health policy.