I hate to even take up the space to define this, I suspect you know the drill as well as I do. Numerous studies (Trinity, Bengen, etc.) determined that if you take the invested assets in a diversified portfolio you can withdraw 4% of the total value your first year of retirement. Each year after that you can withdraw the same amount plus any extra needed to offset inflation. It makes it easy to determine what you need to have invested to pull the trigger on retirement. You need 1/(4%) or 25 times your annual expenses.
Now there have been all kinds of recent reviews of this concept by Vanguard, Fidelity, Schwab and even Bengen himself as to whether 4% is a little too high or a little too low or even way too high. It all depends on your assumptions about the future. What will inflation look like, it’s pretty high right now. What about the performance of stocks, that’s pretty dismal this year so far. And how about that old standby bonds, not looking too great if you ask me. I have that classic diversified portfolio, nearly exactly like the one in the original studies and it’s been quite battered.
Since you are reading this I’m going to make the intuitive leap that you want to know what my opinion is regarding the 4% rule. And I will tell you, but first I need to give you some back story. I’m a chemical engineer and was always on the technical side of my profession. A lot of engineers get into management or sales and don’t do a lot of heavy lifting when it comes to computer modeling of complex processes, but I did exactly that. I was using artificial intelligence software to model chemical plant reactions back when nobody had heard of it. This is like thirty-five years ago, long before iPhones and self driving cars.
My task was to model a complex reactor system that produced a high octane gasoline component that was used for aviation gasoline and high performance car engines. A chemical additive had been created that the manufacturer claimed would improve the yield and quality of our gasoline significantly. If we purchased this chemical for a few thousand dollars a week we’d see a million dollar improvement over the course of a year. But there was a problem. While a million dollars sounds like a big improvement it was actually only a fraction of an octane number and maybe 0.1% of an increase in production. And our product octane varied by a full number every day and production varied by several percent due to constant unpredictable variations in feed stocks amount and quality. In short the very improvements claimed by the additive manufacturer were hidden by the “noise” of the natural variations in our reactors performance.
So what to do? I needed to be able to come up with a model so accurate that it could tell me what the octane and production should be based on feed quality and amount. Then I could look at the actual performance versus the model and any differences would be due to the additive. I tried a whole lot of conventional computer modeling methods, none came close to working. Then I came across something brand new and exotic. It was called artificial intelligence modeling software. I could feed it enormous amounts of past data and it would “learn” what impact things like reactor temperature, feed quality, amount of production and a hundred other data points had on the variables I was trying to predict. And it was great at predicting the past. I could feed it the data from any past date and I’d get results that matched up with what really happened. So problem solved, right? I wish.
There was a huge problem, in that we live in a dynamic changing world. Our facility was constantly being expanded to meet market demand and so each year we ran more feedstock with different components in it than at any time in the past. And my model that was so talented at predicting anything that was within its past experience, lost its mind when it encountered a future that was outside the range of what it had learned from in the past. I suppose it was like a self driving car finding a charging elephant in its path, it’s going to be quite confused and it might not do the right thing, if there even is a right thing.
So what does that have to do with the 4% rule. Everything, in my opinion. The 4% rule applied today is going to base its assumptions on what the next thirty years look like, based on the past. And the past it knows about started in 1925 and ended in 1995 in the original studies. The next thirty years for you and me will be comprised of 2022 through 2052. It’s kind of obvious isn’t it? Does anyone really think that the 2052 economy can be accurately represented by the economy of 1925? We do not have a clue where the next thirty years will take us, but assuming it will be just like some thirty year past set of years is quite a stretch, in my opinion. If we knew, if we could see into the future, we might find the 4% rule should really be the 14% rule, or maybe the 0.4% rule. I’ll let you know in thirty years, if I’m still around.
I hate to attack a sacred cow, because I have used this rule to comfort myself since my withdrawal rate is very low. I tell myself that being well under 4% is surely safe. But then I think back on how poorly my history based artificial intelligence models did when applied to the future and it gives me pause. I realize I do not have any reliable basis for feeling like 4% will mean much of anything heading toward 2052.
So what to do? Are we helplessly adrift with no reliable guides to the future? Should we just make do with 4% because it’s all we have?
Do you have faith in a 4% initial withdrawal rate? If you do, why?
As usual if you’d like to comment and don’t see a place to do so, just click on the title of the post.