Paul Ginsparg and I were discussing that mortality rate adjustment example. I pointed him to this old tutorial that laid out the age adjustment step by step, and he sent along this:

For mortality rate junkies, here’s another example [by Steven Martin and Laudan Aron] of bundled stats lending to misinterpretation, in this case not correcting for the black cohort having a slightly younger average age plus a higher percentage of women.

As Martin and Aron point out:

Why were these differences by race not apparent in the CDC figure? Because for adults ages 65 and older, blacks and whites looked very different in 2015. The average age of blacks was 73.7 compared with 74.6 for whites. Also, a higher share of blacks ages 65 and older were women: 61 percent compared with 57 percent for whites. Because younger people have lower death rates than older people and women have lower death rates than men, comparing a younger and more female population of blacks with an older and more male population of whites offsets the underlying race differences in death rates.