I'll keep this short. The following chart is a representation of planning horizons by age using the full term-structure of longevity conditional on achieving a given age[1], all of it framed with respect to age 65 and 30 years.
Blue is the horizon for "mean longevity" calculated in years by collapsing the conditional survival probability for age x. I calculate it like this:
Orange
Orange is the horizon in years based on the median or 50th percentile expectancy. This is determined by calculating age(t) - x, where age(t) is the year representing 50th percentile of the probability mass implied by tPx rounded to the nearest year. The rounding creates the discontinuities.
Grey
Grey is the same as orange except that we use the 95th percentile expectancy.
Commentary
- 30 years is a decent-ish conservative estimate for age 65 but we haven't considered gender.
- Early retirements and more conservatism might (and do) need long horizons.
- Late retirement, or mid-course retirement in the context of a continuous approach, can have pretty short horizons. This may possibly influence portfolio choice and optimization for both assets and products (e.g., annuities) as the horizon shortens...and also if one does not have a set-and-forget program in place.
- Personal health and risk-aversion or -taking behavior would also influence the choice of horizon. Infinite MPT horizons and single period approaches seem to make little sense here, though.
-------------------------------
[1] SOA IAM table with G2 longevity adjustments to 2018.
I think one of the biggest differences between reality and "ret-fin" is the assumption that either
ReplyDeletea) As you said, we use the lifetime of a single, male, 65-years old; or,
b) We assume a couple...but they are both 65-years old.
Yet we know from the actual real world evidence that couples are rarely the same age. We know from the Current Population Survey that the average age difference in a couple is 2.3 years and 2/3rds of the time it is the woman that is younger. From a retirement portfolio sustainability perspective that's terrible; the person who is already predisposed to live longer is younger.
And we know from other data that the age gap increases as you get older. That is, couples in their early 20s have an age gap less than that but couples in their 40s are higher than that.
Let's stick with 2.3 years as the difference and see how it affects life expectancy. All numbers are the 95th percentile based on SOA tables
Single, male, 65: 34.74
couple, both 65: 35.64
couple, 2.3 year gap, woman younger: 37.47
Yes. I tend to ignore couples because I mostly selfishly think about my own single situation. That's also because I live in FL where the state flag has the motto "self interest above all else" printed in bold letters ;-) Kidding aside, when I've (rarely) thought of couples I have often thought of them as a single person with the lifespan of the younger and gendered person with perhaps a jump-rate in spending at the death of the other. Is this close to the approach in your comment? Btw, I was also going to risk a joke about our world now being gender-fluid and he or she is really xe or ze and age and gender no longer matter but thought maybe I'd be shredded for that kind of comment which I guess I've now made. You know, I wasn't really following the "age gap increases as you get older" thing. Is that a generational "cohort" difference in how people have married closer in age since the 60s or do you mean something else? Your implicit point, if I may, looks like it is one of having a bias for being more conservative than not. Fwiw, this last Saturday, I had coffee with Dirk Cotton up in Chapel hill and had an interesting conversation between old guys with weird hobbies in ret-fin. His last comment was eminently quotable: "I think *THE biggest risk* in retirement is overconfidence." Assuming long horizons goes a long way towards hedging overconfidence and staying conservative given our long, and getting longer, single or joint lifespans.
ReplyDeleteYeah, my implicit point was that I think "30 year retirements" are the wrong number to pick for most people....for a whole host of reasons. (I think David Blanchett has some research showing that most people actually retire at 62...not 65, which throws another wrench into "assume the retiree is 65 years old and the plan runs for 30 years").
ReplyDeleteWhat I mean by "growing age gap" is that when you look at 20-year olds who are married, the partner is generally within 1-2 years of their age. But when you look at 40- or 50-year olds who are married, the partner is more like within 5-7 years of their age.
I think it is just a combination of simple math -- when you are 23 it is illegal to have a wife who is 7 years younger than you! -- and a reflection of later-in-life remarriages due to death & divorce.
I've never seen the numbers for "average difference for a couple at retirement" but it seems plausible that it is higher than the overall average due to that. Anecdotally, it seems like 3-5 years is the most common, at least among my friends & acquaintances.
My wife is a fair bit younger than me, so yeah I essentially just plan based around her age and ignore mine.