When Bengen published his seminal paper on withdrawal rates he seems to have started the common usage of a 30 year distribution period for retirement studies.
Assuming they have normal life expectancies, they should live at least 25–30 years.
His paper was published in 1994…22 years ago. Given how long it takes papers to be written and published, it seems plausible to say that the concept is a quarter-century old.
Life expectancies have gone up, on average, about 1 year per decade. In the year 1900 a 60-year old man could be expected to live about another 14 years. Now they can expect to live another 21 years.
Based on that alone, you’d expect to need to add 2 or 3 years to Bengen’s 30 year distribution period.
Except that the life expectancy gains aren’t evenly distributed among the population. The gains disproportionately go to people who are white and well-off. If you’re in that demographic, then your life expectancy has gone up by 5 years in just two decades.
So does that mean we need to add five years to Bengen’s distribution period?
A number of authors do exactly that, arguing that a 30-year distribution period is not conservative. A 40-year distribution period is a common choice. Guyton and McClung are two authors who seem to prefer a 40-year distribution period.
If you look at just a single male, aged 65, they have a 5% chance of living another 31.61 years, according to the Social Security Administration. But that’s an average across all Americans.
The Society of Actuaries says that, for the kind of people doing retirement planning, there is a 5% chance of living another 34.57 years.
But few retirees are single. Most are in a couple. For a couple that are both 65-years old, the Society of Actuaries says there is a 5% chance they will live another 36.98 years.
These are the numbers that cause some authors to say a 40-year retirement distribution is a better choice nowadays.
There is a somewhat hidden assumption above. Actually, there are two hidden assumptions above.
That the couple is a man and a woman.
And that they are exactly the same age.
Obviously neither one is a great assumption but the one about them being the same age is especially bad.
Age differences in couples
The Census conducts a Current Population Survey which tells us that only 13% of couples are within 12-months of one another’s age.
The average age difference is 2.3 years. (And two-thirds of the time, it is the man who is older, which is “worse” when it comes to retirement planning.)
So instead of a 65-year old male and a 65-year old female, we should be using a 65-year old male and a 63-year old female, right?
That couple has a 5% chance of living 38.38 years. We’re still under the 40 year distribution period that some authors use; but we’re pretty far from the 30 year period that is most commonly used.
Again, though, we are misled by averages. Younger couples are more likely to be closer in age. A couple that is aged 20 usually only has about a year’s age difference. But once the couple gets up around 50, the age difference is nearly 5 years.
For a 65-year old male and 60-year old female, there is a 5% chance they will live 41.05 years.
Now we’re exceeding even the conservative 40-year distribution period some authors use.
Same sex couples
As the chart above indicates, the age gap tends be to even larger for same sex couples. This has an especially strong effect for female-female couples, since they’re already likely to live longer.
A male/female couple (both aged 65) has a 5% chance of living 36.98 years. A female/female couple (both aged 65) has a 5% chance of living 37.50 years.
But a female-female couple near retirement age is likely to have an age difference of over 7 years. A 65 year old female and 58 year old female have a 5% chance of living 43.06 years.
Now we’re seeing a spread of 11 years from the base case of a single 65-year old man and a same-sex couple with age differences. That’s obviously a pretty big change in planning needs.
A realistic couple for analysis
It seems clear that assuming the couple is the same age makes very little sense. Anything between 2–5 years of age difference (with the woman being younger) is a better default, though even that papers over the many complexities of the real world.
In any case, the evidence seems overwhelming that real world couples should be planning on more than a 30-year distribution period…
…Of course, using a fixed-distribution period like that is almost certainly the wrong thing.
Joint Mortality again
In 2008, Blanchett and Blanchett published “Joint Life Expectancy and the Retirement Distribution Period”. The main thrust of their paper is that basing “probability of failure” on a fixed period of time overstates the actual chance of failure in the real world.
(They were not the first to raise this objection but they were among the first to have a detailed analysis.)
It is implicitly assuming that you have a 100% chance of living until you are 95 years old. To assess your retirement plan you need to know the chance that your portfolio will run out and the chance that you are still alive when that happens.
As their table shows, once you take life expectancy into account, even with a 5% withdrawal rate, there is only a 4.8% that a 65-year old couple will be alive when the portfolio expires.
Following the argument I made above, it would be better if they had used a couple that weren’t the same age. This creates an interesting tug of war. On the one hand, assuming a fixed distribution period of 30 or 40 years is overly conservative because people usually die sooner than that. On the other hand, using modern life expectancy numbers for realistic couples we see that large numbers of them can live even longer than the “conservative” 40 year period.
The net result is that using a 40-year period without any mortality weighting is certainly too conservative on average but not quite as conservative as Blanchett & Blanchett suggest.
But let’s try it for ourselves and see…
30 year fixed span: 92.62% chance of success
40 year fixed span: 85.26%
Couple 65/65 : 93.77%
Couple 65/60 : 91.56%