What life expectancy isn’t, and why

If you are using the following phrases in your writing on life expectancy, you are wrong. Don’t feel bad; almost no one uses this concept correctly, including seasoned science journalists and actual scientists who aren’t trained in demography. What’s worse, there are almost no resources to help you figure it out. Maybe I can be that resource?

Wrong: “the life expectancy for the average [woman/man/person] is…”

Why: because life expectancy is a composite measure that takes current age-specific mortality rates and summarizes them in one handy number. So what it tells us is how long a hypothetical person living their entire remaining life under this year’s mortality conditions would live, not the actual lifespan of any real or “average” person. Both real people, and non-real statistically average people, live each year of their life under a different year’s mortality conditions.

Better: avoid using “the average person” in any form.

Wrong: “A person aged X today (or: born today) can expect to live Y more years”

Why: because life expectancy contingent upon reaching age X is a composite measure that tells how long a hypothetical person who is age X this year would live, if they lived the rest of their life under this year’s mortality conditions. (This is true even if X is 0 – that’s life expectancy at birth.) However, that hypothetical person will actually live the rest of their life under the mortality conditions of next year, then the year after that, then the year after that, etc., until they die. So goes the march of time.

Better: although the very term “life expectancy” sets you up to do this, don’t use any variant of “expect to live.” It is never right.

Not wrong: “life expectancy has improved, which means people are living longer on average” or “life expectancy in country X is higher than in country Y, which means people in country X live longer on average.”

Sorry. At least in popular press, this is pretty much all life expectancy is good for: comparisons of mortality conditions in different times or places. Another bummer is that it obscures the components that make up these differences – for example, you can’t tell just by looking whether an increase in life expectancy is driven by a drop in infant mortality or old-age mortality, fewer infectious diseases or better cardiovascular care.

Important caveat: cohort life expectancy is a measure of how long the average person in a given cohort lived, and almost nothing I’ve said here applies to it. But you can’t calculate it until the whole cohort is dead, so it’s not a measure you encounter very frequently.

And, an N.B.: I wrote this post because I took another look at my previous post on life expectancy and it didn’t seem that clear. But if you don’t find this one very clear, maybe you’ll like the other one better – or you can just bug me on Twitter.