An earlier blog entry described the background of my recent correspondence with Michael Kitces on safe withdrawal rates (SWRs). Set forth below is the text of an e-mail that I sent to Michael on August 24.
Michael:
Thanks again for being willing to engage in some helpful (I hope on your end as well as mine) back-and-forth discussion.
I do not believe that the Old School SWR studies look at the worst-case scenario. This was so even before we got to the crazy valuation levels that applied in the late 1990s. I of course understand that they purport to look at the worst case. The error was in failing to take valuations into account. You get a different worst case when you consider valuations than you get when you fail to do so.
Imagine a study that examines whether it is safe to drive drunk or not. The researchers look at the result of 100 cases in which a driver drives a car for 20 miles. In 98 of the cases the driver is sober, in 2 cases he is drunk. The researchers note that in each case the driver makes it to his destination alive. In two cases, however, he is involved in accidents and is hospitalized. These are the two cases in which he was drunk. Would it be a reasonable conclusion that it is safe to drive drunk?
This is the procedure that was followed in the Old School studies. The historical data shows that a 4 percent withdrawal is more than safe for retirements that begin at times of low or moderation valuations; withdrawals a good bit higher than 4 percent survive in all such cases. But 4 percent comes close to failing in the two cases in which valuations went to sky-high levels. The fact that 4 percent barely survived two tests does not show that it is a safe withdrawal in such circumstances. It shows that it is a high-risk withdrawal, just as it is a high-risk driving practice to drive drunk. There is no case in the historical record in which retirees took a 4 percent withdrawal for retirements beginning at times of high valuations and did not come close to suffering busted retirements. It is not safe to follow a practice that has always caused bad results in the past.
The core problem with the Old School studies is that they use the word “safe” when they should use the word “survive.” It is certainly true that a 4 percent withdrawal has always survived. But the fact that risky behavior produces acceptable results on two occasions does not transform that risky behavior into safe behavior. All that you can say is that it is possible for those engaging in risky behavior to sometimes get lucky and achieve an acceptable result. The risk is real and the risk remains.
John Walter Russell’s research shows that those who retired in the mid-1960s with a high stock allocation and a 4 percent withdrawal rate had about a one-in-three chance of seeing their retirements fail within 30 years. It turned out that those retirements survived. It turned out that those people got lucky. But the fact that the retirements survived did not transform them into safe retirements. Those retirements were always high-risk retirements. The retirements that began in the late 1990s were of course even higher-risk retirements. But you don’t need to get to those crazy valuation levels for the analytical errors of the Old School studies to generate highly misleading findings.
I agree with you that there is an “underlying implicit assumption” in the Old School studies that the worst return pattern that we had seen was the worst-case scenario. I don’t agree that this assumption “may” be proven false. I say that those of us who do not believe in the Passive Investing model know that it is false today. It is false because it does not take valuations into account. Valuations affect long-term returns and it is not possible to determine safe withdrawal rates accurately without taking valuations into consideration. There is not one worst-case scenario. Different worst-case scenarios apply at different valuation levels. At some valuation levels, 4 percent is indeed safe. At others, 4 percent is more than safe (that is, withdrawals a good bit higher than 4 percent are safe). At still others, 4 percent is dangerous, not safe.
I have devoted much thought to this and I am not able to come up with any explanation of the way in which the Old School studies were set up other than a misplaced belief in the Passive Investing model. If it were true that valuations did not tell us anything about long-term returns, the Old School approach would make sense. If valuations did not tell us anything, it would be reasonable to not identify different worst-case scenarios for different valuation levels, to just throw all of the data in one big bowl of data soup. But valuations do tell us something! (there’s that explanation point again!). Valuations tell us the range of long-term returns that are possible in a PARTICULAR investing environment. Knowing that, we are able to say what the worst-case scenario is not in general but for the particular case under examination. Knowing that, we are able to report the safe withdrawal rate ACCURATELY. The safe withdrawal rate cannot be reported accurately without taking valuations into account (this is only my opinion, of course, but it is one that is based on a good bit of study of this topic and one that I hold strongly).
I hope that you do not take any of these comments as a disparagement of the work you have done. I believe strongly that your work represents a significant advance. You are quite right that most of the material in the existing literature fails to take into account valuations. Your study does take this factor into account, and that is a big deal.
My personal belief, though, is that you need to go the next step. I urge you to do a follow-up study that shows that valuations also can bring the SWR well below 4 percent. There are millions of retirees who are at grave risk of suffering busted retirements in years to come as a result of having placed their faith in the Old School studies. We can help those people by making them aware of the errors in these studies while there is still time for them to make adjustments.
There are all sorts of other good things that would follow from taking this path. I’ll refrain from listing all of those good things in this e-mail to keep it to a reasonable length, but please let me know if you would like me to provide more detail in a follow-up.
I hope you’ll give some thought to the idea. Or, if you have other ideas for how I can get the word out about our findings of recent years, I would like to explore those. For example, if you can connect me with some people in the field with an interest in exploring the side of things that I have focused on, I would be most grateful. Or, if you could connect me with someone in the media that might want to publicize this issue, that would be a big help. Or perhaps you know of a conference where I could give a talk to financial planners with an interest in learning more about these questions. I’m open to all possibilities.
Please let me know what you think. John and I have done lots and lots of work (and hundreds of other community members have made significant contributions). If you have questions, please just ask — the odds are that your question is one that we have struggled with at an earlier time. You may not fully appreciate it at the moment, but I believe that you have grabbed a tiger by the tail. The implications of this reach out in many directions.
Let’s pray that he turns out to be a friendly tiger!
Rob
My personal belief, though, is that you need to go the next step. I urge you to do a follow-up study that shows that valuations also can bring the SWR well below 4 percent.
One way to do this is to scale from the P/E10 value that corresponds to a 4% safe withdrawal rate. This is not the most accurate approach. But it is easy to understand.
If, for example, one believes that the 1929 P/E10=29 corresponded to a 4% SWR, then P/E10=44 would correspond to 4%*(29/44)=2.6% SWR.
This is far from the best approximation, but it starts telling the right story.
Have fun.
John Walter Russell
NOTE: P/E10=44 occurred in Year 2000.
This is far from the best approximation, but it starts telling the right story.
Here’s my simple way of thinking about it.
We never know in advance what withdrawal rate is going to work because we don’t know what sort of return pattern is gong to turn up. But we know that the range of possibilities covers three percentage points of withdrawal rate. That is, the withdrawal rate that will be the highest one that works is always going to be something between “x” (this is the withdrawal rate that works if a worst-case return pattern comes up) and “x plus three” (this is the withdrawal rate that works if a best-case return pattern comes up).
“X” is not a constant number. “X” varies, depending on the valuation level that applies at the starting date of the retirement. So, when the P/E10 level is 14, the highest withdrawal rate that works is going to be a number between 5.41 and 8.41. And, when the P/E10 level is 26, the highest withdrawal rate that works is going to be a number between 3.12 and 6.12.
To say that the safe withdrawal rate is 3.12 is silly. To say that the safe withdrawal rate is 5.41 is silly. To say that the safe withdrawal rate is 4 is silly. There never can be one safe withdrawal rate. To know the safe withdrawal rate that applies at any one particular point in time, you need to know the valuation level that applies at that point in time. All that you can say without knowing the valuation level is that the range of possibilities will be three percentage points of withdrawal rate.
What the Old School studies tell us is the withdrawal rate that happened to work in the two cases in which we went to very high valuation levels. The fact that 4 percent happened to work in two tests does not tell us 4 percent is safe. All that it tells us is that the SWR at high valuations cannot be more than 3 percentage points below 4 percent (that would be so if the best-case return pattern happened to turn up in those two cases).
Rob