The challenge of basing things off of Annuity 2000 tables

In my last post I mentioned that most retirement-related investigations should probably be using Annuity 2000 tables.

It is a good default option but they are probably underestimating longevity risk for relatively youngish people. The problem is that when we do that we are using the chance of someone who is 65-years old today living until they are 95 (or whenever). Then we turn around tell someone who is 40 that, in 25 years, when they are 65, they will have the same probability of living to 95.

Except that’s not how it works.

Our longevity is constantly improving. And the Annuity 2000 tables are now 20 years out of date (they were constructed in 1996).

Someone who was 65 years old in the year 2000 was born in 1935. (That’s approximately, but not quite, the yellow bars above.) We can see that someone in that cohort had a less than 10% chance of living to age 90.

Someone born in the year 2010, however, has an 18% chance of living to age 90: that’s double.

That’s not really a surprise: we all know that life expectancies are going up.

But it does mean that it is a challenge to interpret results generated by using current life tables unless you are already close to age 65, since they rarely take longevity improvements into account.

(I’ve never tried to do it, in my own simulations!)

(For those who are curious: from the 2004 life table to the 2011 life table the chance of an 80-year old male dying dropped from 0.066557 to 0.059565. That’s nearly an extra 1% chance of living, in just 7 years.)