Longevity risk is one of THE big planning risks along with investing risk, spending risk, inflation, spend shock risk, and some others. It is so important that when I built a simulator I made longevity randomizing an important part of it. In fact I did it twice, once with a distribution using the SS 2013 Life Table and a second time with code for an option to use a Gompertz distribution so that I could play around with the shape of the PDF. This kind of thing is great for research and fun to play around with[1] but it is a little abstract and I realized that my own planning is not abstract it's very real. It's not just that I don't plan to die at 97 and 82 and 88 or that I don't plan to die at 97 or 82 or 88 it's that I expect to do it just once. And that one age, for planning purposes, I want to be a conservative number and I want to know as much as I can about what will happen to "the plan" if I get to that age.... or beyond.
I got to the point of view in the last sentence like this: while I was patting myself on the back for creating my very own own sim and for being genius enough to add random longevity I got tipped off to the trade-offs first by David Cantor at PwC who asked me "why?" The second was Joe Tomlinson, an eminent practitioner-researcher who suggested that I might be underestimating my risk by just a little bit. I got tipped a third time in an article by John Curtis called "Monte Carlo Mania" in Retirement Income Redesigned (Evensky and Katz, 2006) where along with a number of critiques of Monte Carlo simulation, most of which I had seen before and most of which I agree with, he ding-ed longevity risk modeling in simulators in particular. "...this factor is statistically accurate but not very useful to real advisers and real clients...[and] can produce misleading results... How does a person die some of the time. This is another example of sound mathematics conflicting with sound planning."
So given the nudges I went back and, since it's easy enough to do, coded the sim to use either fixed or random longevity (why not keep it around, right? It makes some great data visuals for thinking about and analyzing certain things) and then I compared the two approaches using my own data, once with random and once (actually trial and error) with fixed. What I found was that for a given fail rate (using some personal assumptions with random longevity) I would have to adjust the fixed longevity to around age 84 to get the same fail rate. That makes a lot of sense -- since 84 is somewhere between the mean and median for a 58 year old -- but it really does conflict with me wanting to use 95 as a bad-conservative estimate to shake out weaknesses in my plan. It is, of course, easy enough to get around the problem of the random longevity approach by extracting the right age/death cohorts from the data pile after doing enough sims but it even easier and faster to just do a fixed age run. And more comprehensible to boot.
So now? I will, as always, endeavor forever to remember that Monte Carlo simulation can be a weak and easily misinterpreted and distracting kind of retirement analysis in the first place but I will also: a) lean towards fixed longevity for my own planning, and b) keep both types of sim around as an early-warning-insight-window into the risks that always threaten to arise right around the next corner. Because the issue is not really longevity-simulation it's about adaptation when things look bad and they (we) are about to hit the wall.
----------------------------------
[1] I also now know more than I ever wanted to about longevity: it's probability "shape"...at different survival ages, it's movement up over the years, how it is "squaring off" of the curve as society and medicine advances, how to generate longevity PDFs in R, the way it interacts with financial plans, etc. etc. On the other hand, I also think that financial planners could do a way better job of clue-ing clients into the fact that longevity is random and a "distribution" and does have an outsize impact on a retirement better than has been done. Why am I 58 -- after spending freakishly large sums on fees and planning over some 30 years! -- and just figuring this out now...and on my own? I couldn't have been that blind, could I?
No comments:
Post a Comment