The Great SWR Investigation-Part 2

Financial Independence/Retire Early -- Learn How!
bpp
** Regular
Posts: 98
Joined: Tue Nov 26, 2002 6:46 am
Location: Japan

Post by bpp »

Any tool used for assessment of the safety of a retirement portfolio that does not take valuation levels at the start of retirement into account is fundamentally flawed. This factor always affects the result. A tool that does not take it into account cannot produce the correct result. That sort of study has not taken into account all the inputs that matter. Without taking into account all the inputs, it is not reasonable to expect the analysis to produce the correct output.


How exactly do you propose to take valuations into account? I don't see anything quantitative here, and I don't see how to get anything quantitative. The most suggestive plot in this regard is the PE10 vs SWR plot that BenSolar linked to on another thread. But even that doesn't seem to lend itself to any hard and fast rule, and it is not clear how to extrapolate it out to the extremes of the 2000 bubble peak. I could imagine reasonable functional fits to the data which would give anywhere from 4% down to 0% SWRs out there, but these would all involve purely empirical functions, nothing with any theoretical basis behind them, so there's no way to choose among them. Not to mention statistical confidence problems. (I might play with such fits for fun if I get a chance, but wouldn't believe anything that came out.)

I guess it makes sense to say, as a rule of thumb, that it may be wise to tweak down the withdrawal rate when starting from high valuations, but I just don't see how to quantify this into anything really useful.

Of course I also believe Gummy is right that the whole SWR idea based on absolute fixed withdrawals is only really useful for guessing how big a pot one might have to accumulate in advance. As an actual withdrawal strategy, it is too rigid and fails in very unforgiving ways.

Cheers,
Bpp
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

hocus:
Both plan to take 4 percent withdrawals from their portfolios ...

And how did they arrive at this 4%?
My question is: "How does one generate a SWR? What information is required?"
Yours seems to be: "Having decided upon a withdrawal rate, how does one determine the probability of surviving N years?"

Anyway, we may not be talking about the same problem.

If you think we ARE talking about the same problem, then perhaps you can suggest what inputs one should use to generate a SWR (but avoiding the phrase "it depends upon valuation levels" which our Black Box will NOT accept as input :^)

Perhaps one of your inputs is:
The average annual return for the past m years.
For Mary, with m = 1, that's 100%.

Perhaps one of your inputs is:
Last year's deviation from the Mean Annual Return.
For Mary, that could be more 100%.

If you can identify numbers (not phrases), then I'd be interested in analyzing the consequences of incorporating these numbers.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

How exactly do you propose to take valuations into account? I don't see anything quantitative here, and I don't see how to get anything quantitative.

There are a variety of reasonable ways to do the calculation. One reasonable way is to follow the process used by Bernstein is coming to his conclusion that the SWR in the year 2000 was 2 percent. Another reasonable way is to follow the process used by JWR1945 in coming to his conclusion that the SWR in the year 2000 was 2.3 percent. All we need do is follow one of these two methodologies or some other reasonable variation for all the years other than the year 2000 and we have better SWR assessments for all those years than we possess today.

These two numbers were developed independently. The fact that they are both in the same ballpark lends credibilty to both of them. It suggests that both William Bernstien and JWR1945 were on the right track with their analyses. The fact that the intercst number of 4 percent is so far off from them should undermine confidence in that number. It suggests that intercst is not on the right track.

There's room for reasonable differences of opinion as to exactly how changes in valuation should be reflected in a valid SWR calculation. I don't believe that there is any longer room for reasonable differences of opinion as to whether this factor should be incorporated into the analysis in some manner. Its effects are too pervasive and too significant to ignore.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

I am looking forward seeing how we input guesses (or estimates as some prefer) and get "one correct answer" guess I will have to be patient. I imagine that one day hocus will volunteer to reveal his "tool."
Intercst says that both retirements are 100 percent safe (let's call it 95 percent if you count in the chance of the future being unlike the past). He says the two retirements are equally safe.


It seems like we can't have this discussion without reference to intercst although I have tried. Intercst certainly didn't say this on 5/13/02 (the day that the world supposedly changed.) I think it wrong if he did say it. Bob says "100% safe" refers to the past and isn't (on the rehp) mentioned in connection with any prediction of the future. Bob doesn't know the content of the 10,000 messages intercst has written and isn't sure why it would be of any importance to know about what went on there.
Have fun.

Ataloss
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

If you can identify numbers (not phrases), then I'd be interested in analyzing the consequences of incorporating these numbers.

Thank you! Coming up with some numbers might help make things more concrete for a lot of people. I'm not capable of of running numbers, or I would have already done it myself. I very much appreciate your willingness to help us out with this.

My suggestion would be to start by seeing what you come up with for years other than 2000 if you employ either the methodology that Bernstein used to arrive at an SWR of 2 percent for a retirement beginning in the year, or the methodology that JWR1945 used to arrive at an SWR OF 2.3 percent for a retirement beginning in that year.

JWR1945 had a post just yesterday in which he described the Bernstein methodology, and he also had a recent post in which he described his own methodology at least in part. I'll look for those two posts, and add links to them to this thread.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Bob says "100% safe" refers to the past and isn't (on the rehp) mentioned in connection with any prediction of the future.

The only data that we have access to is data from the past and the present. No one has data from the future, so no one is suggesting that we use data from the future to calculate the SWR. Please pass the word on to Bob for me, OK?
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

ataloss:
I am looking forward seeing how we input guesses (or estimates as some prefer) and get "one correct answer"...

Getting a "correct answer" is easy.
(Gimme some numbers, I'll crunch 'em ... and give y'all a "correct answer".)

Alas, the hard part is defining "correct" and, more important, asking the "correct question" :^)

hocus:
I'll bet that, when we're finished with this "incorporating valuation levels" stuff, we'll wind up agreeing that an appropriate withdrawal rate will depend upon what the market has done recently ... and the withdrawal rate will change as market conditions change.

That's the point of my "sensible withdrawals" strategy which, as I recall, you didn't think was very sensible.
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

I apologize in advance for another Sam & Sally story, but ...

Sam: "In a thousand tosses of a coin, how many heads would one expect?"
Sally: "That's easy! It's five hundred!"
John tosses a thousand time and gets 678 heads.
Sally: "Oops. I guess I was wrong."

Which suggests the question:
How do we determine whether Bernstein (or anybuddy else) is correct with his 2% (or any other) withdrawal rate?

Oh, I forgot to comment on the John & Mary story:
In 2004, John is withdrawing at (about) 2% of his current portfolio whereas Mary is withdrawing at 4%. Any respectable mathematical ritual would conslude that John is safer.
Last edited by gummy on Fri May 30, 2003 9:27 am, edited 1 time in total.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

That's the point of my "sensible withdrawals" strategy which, as I recall, you didn't think was very sensible.

There's been a misunderstanding here.

I don't hold any particular views as to the withdrawal strategies that people should or should not follow. The issue that I am working--whether the effect of changes in valuation levels should be reflected in an SWR calculation or not--is keeping me plenty busy.

I think that the comment to which you are referrring was one I made in response to a post in which you had said that one could retire on an assumption that the SWR was 4 percent and then just "forget about" the issue of the validity of various ways of calculating the SWR. I don't think that is good advice.

Anyone who retired in the year 2000 thinking that the SWR for that year was 4 percent has made a big mistake. People in those circumstances need to revise their plans for what we know now (or at least what some of us think we know now) about SWRs that we did not know at the time those 4 percent figures were being put forward.

The sensible withdrawals strategy may be a great strategy, but I believe that it will work a lot better for those who crafted their plans with an understanding of what the correct SWR was than for those who crafted their plans based on confidence in a SWR that has subsequently been shown not to apply.

I don't object to those people following the sensible withdrawals strategy after making adjustments in their plans (or returning to work and saving up enough to begin a new retirement) to reflect our new understanding of SWRs. But I believe that the prudent thing to do is to take the new understanding of SWRs into account before doing anything else.

The difference between a 4 percent SWR and a 2 percent SWR is huge. People who retired in 2000 or sometime near then need to take the realities into account. Employment of the sensible withdrawals strategy may help in circumstances in which one's understanding of the SWR at the time of retirement is a little off. But in the year 2000, the common understanding of how to calculate SWRs produced numbers that are off by a country mile. I question whether employment of the sensible withdrawals strategy will be enough to address the potential problem in some cases.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

How do we determine whether Bernstein (or anybuddy else) is correct with his 2% (or any other) withdrawal rate?


I can tell you (but not for 30 years)
I would see correct as the result that gave the maximal withdrawal rate as of a given date. Hocus is sure that 4% is wrong but he doesn't (apparently) know what is right :wink:
Have fun.

Ataloss
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

How do we determine whether Bernstein (or anybuddy else) is correct with his 2% (or any other) withdrawal rate?

I think we need to talk it though until we come to some sort of consensus.

My view is that we are the experts in this area. We have a responsibility to do all that we can to get it right. We need to inform ourselves the best we can, we need to enter into debate with our minds open to the other side (I mean all of us, including me, of course), and so on. We have a responsisbility here to the people who read our work product and possibly make money decisions based on what we say here.

The question you ask, Gummy, hints at the concern that I am trying to address with my proposal that we invite experts to comment on a few questions that we come to view as particularly significant and particularly hard to resolve. I think that some of us may feel a little intimidated with the idea that we are rewriting the rules for how SWRs are computed. It is an ambitious undertaking, to be sure. But I think it is an important and exciting undertaking, and I can't think of any group more qualified for the task than the group that congregates here.

A lot of issues have been brought to the table in the course of the past 12 months. We don't need to resolve them all, certainly not real soon. I think that the next step is to pick a few of the issues that seem most fundamental and explore those in depth until we are reasonably sure that we possess a proper understanding of them. Then we should approach the four experts (you get into the hot seat at this stage of the process if you elect to accept the invitation, Gummy) and ask for their counsel on those core issues.

I view that as a responsible and constructive way to proceed. If we succeed at addressing those core issues, I believe that the experience will give us the confidence and understanding of the subject matter needed to press forward and address a whole bunch of other stuff in future days. I see this as a long-term project, not something that is going to be wrapped up in early 2004.

There is no rush on this. We are under no obligation to come up with final answers by any specified time. There is no purpose in stretching things out unnecessarily. But we should proceed at a pace at which we feel comfortable.

We don't want anyone to feel that he or she cannot put his or her question forward. We should encourage questions and criticism because it is in the process of asking and answering questions that we gain confidence that the direction in which we are moving is the right one.

So long as we follow a sensible process, we will arrive at a sensible answer, I believe. The answer will not be perfect. Others will come after us and update our work just as we are giving consideration to the idea of updating the work of some who have come before us. That's OK. It's no more reasonable for others to demand that we get it perfect than it would have been for us to have demanded that those who came before us get it perfect.

People learn things over the course of time. When they learn things, they use what they have learned to make the world a better place. SWR studies are part of the world that people should be trying to make better. SWR studies are the little part of the world over which this particular community can have some influence, if we play our cards right.

The goal is to make a postive contribution. The goal is to take the tool that exists today and make it better in some way. What more important project than enhancing SWR analysis could a community of people with a mutual interest in FIRE possibly work on?
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

In 2004, John is withdrawing at (about) 2% of his current portfolio whereas Mary is withdrawing at 4%. Any respectable mathematical ritual would conslude that John is safer.

Of course!! That is the point!

No respectable mathematician would say that these two retirements are equally safe. But that is what the conventional SWR analysis says. The conventional SWR study says that they are both equally safe and they are not. If the tool doesn't show a difference between the safety of the two retirements, there is something wrong with the tool, and we should be trying to determine just what that something is.

If you do the SWR analysis the way that Bernstein did it in his book, it would show that John's SWR is somewhere around 4 percent and Mary's is somewhere around 2 percent (I haven't done any calculation, I am using very rough numbers just to illustrate the point). Do the analysis the correct way, and the obvious diiference in safety reveals itself.

John retired with a high enough real SWR that he is able to lower his withdrawal rate (while maintaining the same spending level) in times of extreme valuations and thereby preserve the safety of his plan through those times. Mary did not do this. She deceived herself by thinking that her plan was safe just because it meets the test of the conventional SWR studies, and her plan as a result is not safe. She is taking out too much each year to have confidence that her plan will work, given the extreme valuation levels that applied on the day she retired.

Isn't the idea for the plan to be safe? Why are so few troubled that the conventional analysis fails to fulfill the goal that the tool was developed to serve?
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Hocus is sure that 4% is wrong but he doesn't (apparently) know what is right

Well, it's a lot easier to identify that something is wrong than to determine what is needed to set it right.

I knew that the conventional methodology was wrong the moment that I saw that it did not include an adjustment for changes in valuation levels. If you don't consider all the required inputs, you sure as shootin' ain't gonna come up with the correct output.

It's a harder and completely separate step to determine how to incorporate the missing factor into the analysis. I believe that Bernstein took a good stab at it, and I believe that JWR1945 took a good stab at it. But my understanding of statistics is not good enough for me to have too much confidence in my views on the how-to-fix-it question.

I know for sure that the conventional approach does not provide the right answer. If we could reach a consensus on just that much, I think it would be a breakthrough. If we got that far, I can't imagine that we would not eventually be able to reach a consensus on one of the several possibie means of incorporating the missing factor into the analysis.

For me, the second step is harder than the first. My sense is that, for some others, the first step is harder than the first. That's why discussion boards are a good thing, we all get a chance to contribute in the areas in which we have the most to offer.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Gummy:

Here is a link to a JWR1945 post in which he describes what Bernstein did to come up with his 2 percent number.

http://nofeeboards.com/boards/viewtopic ... 6953#p6953

Here is a link to a JWR1945 post in which he describes what he himself did to come up with his 2.3 percent number.

http://www.nofeeboards.com/boards/viewt ... ght=#p6940
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

No respectable mathematician would say that these two retirements are equally safe. But that is what the conventional SWR analysis says. The conventional SWR study says that they are both equally safe and they are not. If the tool doesn't show a difference between the safety of the two retirements, there is something wrong with the tool, and we should be trying to determine just what that something is.


I think this is obviously a misinterpretation
Have fun.

Ataloss
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

I think this is obviously a misinterpretation

It's not obvious to me that it is.
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

hocus:
Thanks for the links. (I've bookmarked them.)
At first blush, it seems you can get Bernstein's 2% simply by making the Expected Annual Return "input" (to our blackbox) something smaller than the normal 10% "input".

Although I once wrote a tutorial on Gordon's Equation (I think), I didn't think much of it (I think). Guess I'll have to find it.

Anyway,here's an interesting spreadsheet:
http://home.golden.net/~pjponzo/SWR-saving.xls

If a young fella starts investing for his eventual retirement assuming an eventual 2% withdrawal rate, he'll be eating beans for a while :^)
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Post by BenSolar »

ataloss wrote:
How do we determine whether Bernstein (or anybuddy else) is correct with his 2% (or any other) withdrawal rate?


I can tell you (but not for 30 years)
I would see correct as the result that gave the maximal withdrawal rate as of a given date.


I disagree with this, ataloss. Everyone agrees that the specifics of future returns that will determine the maximal WR are unknowable. So, to me, the effort should be to determine the best estimate(s) of a withdrawal rate that will very likely (say 90% (though such precise knowledge of probablilities is unavailable)) survive the desired time period. In the absence of a crystal ball we use the best tools available. Different methodologies will produce different best estimates. If a methodology suffers from a serious logical flaw, then it's result should be discarded as incorrect.

I agree with hocus that valuation is so fundamentally linked to long term stock returns (and hence SWR from a stock heavy portfolio) that it should be considered in generating a best estimate. At least with a high, mid, and low range of valuations and associated SWR.

For instance, from average valuations, there are historical start dates that yielded close to 9.5% maximum withdrawal rates. But at that start date it would have been impossible to have said the SWR was 9.5% when we had seen starts from the same valuation provide only 4.5% maximum yield. The 'sensible withdrawal rate' strategy would apply here. Pick the 4 or 4.5% WR which seems safe from all evidence, and then take some of excess gains if you find yourself in a sequence where returns are high.

2 or 2.5% may not prove to be the maximum WR from a 2000 start for a 75/25 port, but given the evidence that valuation affects long term stock market returns, and that the stock market reverts to mean valuation, one can certainly build a fairly strong case that one would need to go with such a low rate to be 'safe'.

The big question is why anyone would choose a 75% allocation to S&P 500 at times of such extreme valuation :?:1.1% dividend yield can't be expected to support a high rate of withdrawals even in the absence of reversion to mean valuation. :shock:

Regards,
Ben
"Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things only hoped for." - Epicurus
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

Bob never said they were equally safe and he isn't sure what intercst said. Both withdrawal rates are safe based on the historical data. This is what safe means.
"100% safe" in terms of the safe withdrawal studies means that the selected withdrawal rate survived all the pay out periods examined from 1871 to 2000. If you are predicting that something worse....World War II, or the Reagan or Clinton presidencies, then the safe withdrawal rates are less than 100% safe. That's all it is saying.

If you find comfort in imagining that something worse than all the things we've experienced over the past 130 years will happen to your retirement portfolio and want to use a lower withdrawal rate that's fine.


Bob is optimistic that the future is bright and a 4% withdrawal rate will survive for 30 years. He doesn't know what the probability of this is. He doubts that intercst ever took a stand on this and he asked me for a link showing intercst indicating a numerical probability of future success. Do you have such a link? The 5/13/02 thread only incresed Bob's favorable impression of intercst. Bob tells me that a 2% withdrawal rate must be safer but based on historical data it is also much mor likely to spiral out of control and result in a very large portfolio. Bob doesn't aspire to extreme wealth so he prefers the 4% withdrawals.
Have fun.

Ataloss
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

Different methodologies will produce different best estimates. If a methodology suffers from a serious logical flaw, then it's result should be discarded as incorrect


Historical swr analysis will fail when long term future returns are lower than the past. Believing this I agree with hocus and intercst that the logical thing to do is plan for lower withdrawals. I disagree with hocus that intercst was "wrong." He has correctly presented a historical method which has limitations. From what he has said, I am not sure that hocus appreciates the uncertainty of the output of monte carlo analysis.
Have fun.

Ataloss
Post Reply