Testing for changes in the SES-mortality gradient when the distribution of education changes too.

We develop a flexible test for changes in the SES-mortality gradient that accounts for changes in the distribution of education, the most commonly used marker of SES. We implement the test for the period between 1984 and 2006 in the United States using microdata from the Census and other surveys linked to death records. Using our flexible test, we find that the evidence for a change in the SES-mortality gradient is not as strong as previous research has suggested. Our results indicate that the gradient increased for females during this time period, but we cannot rule out that the gradient among males has not changed. Informally, the results suggest that the changes for females are mainly driven by the bottom of the education distribution.