Safe withdrawal ideas, thoughts and provocations

Financial Independence/Retire Early -- Learn How!
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

I accept that you have been treated badly on another board,

I appreciate the comment. I don't believe, however, that anything that happened to me personally at the other board is any of this board's concern. The part that I think is this board's concern is what was done to the FIRE movement and to a goal that we all should share of developing as many resources as possible to help people learn how to achieve financial independence early in life. That's why we are here, to learn about that stuff. So the loss of a valuable learning resource is our business.

You, yourself, appear to have become more and more intolerant of others' comments

I have grown weary of the semantics games, that's so. Some of these things that are being raised have been discussed 20 or 30 times before. Couldn't people just look through the Post Archives if they want to re-live discussions of whether an SWR applies on a going forward basis or not? Things like that bore me at this point. There comes a time when you just have to move on.

And your statements appear to have become wilder.

My views have grown stronger over time. Every bit of data that has been presented argues the same way. We have seen nothing pointing in the other direction. At some point, I think you just have to accept what the data is trying very hard to tell you. Valuation matters.

All, and I repeat all, of the conventional studies on SWRs based on historical data are expressed in moderate and reasonable terms.

It's entirely possible to make claims not in accord with the historical data and to offer them in moderate and reasonable language. The key question is whether these studies consider the data that bears on the question or not. All of it? Or just some of it?

They... go back to the Harvard study in 1973; Bengen 1994; Trinity 1998; Jarret/Stringfellow 1999; Armstrong 1999 and, of course, REHP. Generalising, they all indicate an SWR of about 4%.

Does this not strike you as a problem?

This is one of the aspects of this matter that most alarms me. If one study was saying "the SWR is 11 percent" and another "the SWR is 23 percent" and another "the SWR is 1.4 percent," I think people would catch on that the whole thing is a bunch of nonsense and pay no attention.

The situation we have is a dangerous one because the fact that the studies using the conventional methodology agree with each other tends to give each of them more credibility in people's eyes than any of them standing alone would possess. There were many times when you would hear someone say of the intercst study, "oh, that's based on the Trinity study, so it must be right" Each of the existing studies has gained credibility by following the methodology used in the others.

What if they are all wrong?

That's what I am saying. I am saying they are all wrong. If that's so, it's a problem.

People say from time to time, "It's just a message board, what does it matter if intercst got it all wrong?" It matters because he is not expressing his personal viewpoint, he is reporting on data. And it matters because his claims have added credibility due to the fact that other researchers have put forward similar claims. The combination of factors adds up to a terribly dangerous situation.

Most people don't take this stuff as gospel, but many people think that it must be more or less on the mark. Otherwise, there wouldn't be so many people saying it. And iwe have discovered that there are times when it is way off the mark. For someone planning to live on $60,000 in retirement, $1.5 million off the mark. That's not pocket change.

Within the context of historical studies they are most definitely valid

What the heck does the phrase "within the context of historical studies" mean? Either they tell you what the historical data says re what is safe or they do not. I say they do not. They leave out a critical factor.

Note that "safe" is expressed as "safe".

What are you suggesting that the quote marks signify? I have always taken them as an indication that the future may be worse than the past, and, in that event, all bets would be off.

Are you suggesting that what they signify is "I am just kidding about the claim that this study reports on what the historical data says, I included some data and left other stuff out so that I could produce a number I have a personal fondness for.?" If that is what the quote mark signifies, the whole thing is a sham.

I don't think that is what intercst meant for the quote marks to signify. I think he didn't know that changes in valuation levels affected the result. I think it was an honest mistake at the time he put out the study. Now that it has been revealed that changes in valuation levels affect the question examined, the study should be updated to reflect this reality.

All the actual studies, including those both by respected academics and by REHP, are couched in terms sufficient to cause any reasonable person to implement an SWR strategy with care based on the findings of the studies.

I agree. But how much care is needed? You need to know whether the thing is based on data or not to know how much care to exercise.

If it is based on data, you might use SWR minus one in your plan to be absolutely sure. That would be a 3 percent withdrawal. According to the historical data, a 3 percent withdrawal was not safe in the year 2000. So even those exercising lots of care were burned.

They were burned because they were led to believe that the number produced was generated by looking to the data bearing on the question. That's a perfectly reasonable thing to expect.

Can you give me one example outside of the realm of SWR studies where it is common practice to deliberately ignore data that always bears on the question being examined and then issue a study saying that the data that was examined answers the question satisfactorily? I cannot think of one other area where research following this sort of practice is considered valid. Why are there special rules for SWR analysis?

The fact that we now have an extreme event outside the historical range merits updating of the studies' results

Changes in valuation levels is not an "extreme event." Stock market valuations change all the time. It is true that the bubble valuations were extreme, but they revealed a problem with the conventional methodology that existed before the bubble ever came along. The problem is less dramatic at non-bubble valuation levels, but changes in valuation levels always affect SWRs. At least that's what the data we have looked at during The Great Debate indicates.

Historical studies obviously take into account the changes in valuation levels and the extremes that have occured historically wthin the study period. It is only the extremes of a few years ago that have not been included.

I disagree.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

hocus:Andrew61 is not a cartoon characrter. He is a real person. I believe that we have an obligation to the Andrew61's of the world to try to get this right.

Chips: If you have that obligation, it is because you have taken it upon yourself. That is public spirited since you are under no coercion.

I disagree. I believe that we are under an obligation to do what we can to help others in the FIRE community who are being misled by studies done pursuant to an invalid methodology.

My guess is that I will not convince you of the obligation until I first convince you that the methodology is invalid. There is more I could say about the nature of the obligation that I believe we have towards other in the FIRE community, but I believe that those would be wasted words until we make more progress on the validity issue.

So I think it is best to hold off for now on putting forward those thoughts.
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

This goes to a lot of what I have been working on the last few days.
ataloss wrote: Interesting work in swr could include looking at the value of alternative asset classes in reduction of variability, variable withdrawals, systems of cash buffers. The quest for an unarguable swr is doomed.

Petey
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

I should receive Bernstein's book tomorrow and will start reading it, but I think it pretty unlikely that I will agree with what Bernstein says on an accurate way to predict long term market returns. Even if it were possible to do so, you could not predict their sequence. Perhaps he means what Bogle means, that 7% real has been achieved averaged over decades. There is no way to predict future returns accurately using any methodology. It is an impossibility. That's just plain common sense.

Petey
hocus wrote: William Bernstein says on Page 54 of "The Four Pillars of Investing" that: "The Gordon Equation provides an accurate way to predict long-term stock market returns."
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

I think it pretty unlikely that I will agree with what Bernstein says on an accurate way to predict long term market returns

That's fair enough, Petey. What I am saying is rooted in my acceptance of what Bernstein says in Chapter Two of "The Four Pillars of Investing." If you reject Bernstein's claim, you are rejecting my claim too. I am OK with that.

What I do not follow is how anyone could accept what Bernstein is saying and reject what I am saying. I am just taking Bernstein's general comments and applying them to the specific question of how to calculate a SWR.

There is no way to predict future returns accurately using any methodology. It is an impossibility. That's just plain common sense.

i believe that Bernstein has common sense, and he says different. I think it would be a big help if everyone would read Chapter Two of "The Four PIllars of Investing." Then we all would be talking about the same thing.

If there is a flaw in what Bernstein is saying, I obviously want to know what it is. If he is right in what he is saying, I think it opens up some exciting possibilities for SWR analyis in the future, and I would like to see us get about the business of exploring them.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

I should receive Bernstein's book tomorrow and will start reading it, but I think it pretty unlikely that I will agree with what Bernstein says on an accurate way to predict long term market returns.


my take on Berstein is that he thinks that higher recent returns (as of 2000 mean overvaluation and lower returns going forward.) Like Bogle and Buffet (and raddr) Using the Gordon model as an approximation he gets a real return figure of 3.5% (based on assumptions about inputs) Obviously the swr can't be 4% so he estimates 2% using appropriate language to indicate that this is an estimate (guess.)
Have fun.

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

Post by hocus »

My take on Berstein is...this is an estimate (guess.)

It was a calculation, ataloss. I wish you would read the book.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

It was a calculation, ataloss. I wish you would read the book.


swr calculated by the Gordon equation?
Have fun.

Ataloss
raddr
*** Veteran
Posts: 265
Joined: Mon Nov 25, 2002 3:25 am
Contact:

Post by raddr »

ataloss wrote:
It was a calculation, ataloss. I wish you would read the book.


swr calculated by the Gordon equation?


I'll take a stab at it. The SWR cannot be derived from the Gordon Equation. If hocus or anyone thinks differently then I offer a friendly challenge to them to show me the mathematical proof. I doubt I'll have any takers. :wink:
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

What does a perfectly constructed SWR get you though?

Petey
hocus wrote: If we had perfect data, we could calculate the SWR perfectly.
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

Actually, no. Quite the opposite. If you had a bunch of different smart people looking into this matter in different decades and they were all over the park with their estimates, that would be a big problem. The fact that they're all on their own coming up with similar numbers is a good sign that it is right. Of course it is always possible that they're all wrong, it happens like a lemming effect, but given the subject matter I doubt it.

Possible but not probable.
hocus wrote: They... go back to the Harvard study in 1973; Bengen 1994; Trinity 1998; Jarret/Stringfellow 1999; Armstrong 1999 and, of course, REHP. Generalising, they all indicate an SWR of about 4%.

Does this not strike you as a problem?

This is one of the aspects of this matter that most alarms me. If one study was saying "the SWR is 11 percent" and another "the SWR is 23 percent" and another "the SWR is 1.4 percent," I think people would catch on that the whole thing is a bunch of nonsense and pay no attention.

I always took the quotes to mean it was safe in a theoretical sense, as no one can know if history will repeat with investment returns.
hocus wrote: What are you suggesting that the quote marks signify? I have always taken them as an indication that the future may be worse than the past, and, in that event, all bets would be off.

Petey
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

hocus wrote: i believe that Bernstein has common sense, and he says different. I think it would be a big help if everyone would read Chapter Two of "The Four PIllars of Investing." Then we all would be talking about the same thing.

Hocus,

I agree, he probably has.

Clearly he is using a fancy approach to predict the results over a 30 year period. That's wonderful. Good luck to him.

One question. Does he have a crystal ball?

No. See, there's the flaw in your arguement. No formula, fancy approach or author you seem to take as gospel can know what the returns would be. That would be like looking at 100 years of statistics and thinking you can predict who the next winner of the superbowl will be. Gamblers everywhere swear they are beating the house but the annual reports of the bookmakers show another story..

What Bernstein has done is use a method of calculation that looks like it predicts the future because he's used it at data points in the past and it came up with what actually happened. That's fine, but it is no different than the various safe withdrawal studies which use the past to indicate what a safe withdrawal might be over a certain number of years. It gives you an indication but as anyone knows and Bogle says, " No one knows what the markets are going to do. " It is reasonable to assume we might have sub-par results in the markets over the next few years but we don't know for sure that will happen. A reliable SWR will only be an indicator nothing more. We'll then have to have our own take on whether we want to rely on the study and/or downgrade the future returns so we know we can survive in retirement if the returns are 1% less positive or 2% or whatever margin of safety we want to build in on historic vs. predicted actual. A SWR certainly gives up a starting point but it is only one approach.

A cash buffer system is a different approach that would need a whole different set of scenarios over history to be run to see how well that would protect you from a 15 year downturn in the 70s, high inflation etc. There isn't just one approach to the issue.

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

Post by hocus »

I'll take a stab at it. The SWR cannot be derived from the Gordon Equation. If hocus or anyone thinks differently then I offer a friendly challenge to them to show me the mathematical proof. I doubt I'll have any takers.

You should read the book, raddr. It would help a lot.

Bernstein does not spell out in great detail how he came up with his 2 percent number. He sets forth the number on page 234 and refers the reader back to Chapter Two for an explanation of how he came up with it. Chapter Two deals with the Gordon Equation and related matters.

My understanding is that he used the historical sequence methodology to come up with a number unadjusted for valuation, and then used the Gordon Equation to make an adjustment for valuation. I do not know the details of how this is done. But the approach seems to me to be a sound one conceptually.

Isn't the goal to know what is safe? The conventional approach does fine on all aspects of the question except valuation. So what is wrong with the idea of using the Gordon Equation to make an adjustment for valuation? Isn't that a whole lot better than making no adjustment for valuation at all?

Would you have more confidence in a SWR calculated the Bernstein way or in one with no adjustment for valuation at all?

Do you agree with Ataloss that anyone using the Gordon Equation to assess the effects of valuation changes is engaging in an act of pure guesswork?
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

One question. Does he have a crystal ball?

You need to read the book, Petey.

My recommendation to anyone who sees use of the Gordon Equation as akin to use of a crystal ball is that he or she needs to read the book.
[KenM]
*** Veteran
Posts: 133
Joined: Tue Mar 04, 2003 12:54 am

Post by [KenM] »

Hocus
As usual, you appear incapable of differentiating between what the studies actually say and what was apparently said in some of the wilder, more absurd posts on TMF.

The studies, including REHP, based on historical data, include all relevant data including peaks in valuations and changes in valuation for the period studied. The studies are valid.
The methodology used in the studies is logical and uses all the relevant historical data including peaks in valuations and changes in valuation for the period studied. The methodology is valid.

The studies, including REHP, do not make the wild sort of claims apparently made in internet rants - I'm totally uninterested in digging up or pursuing such nonsense. I suggest you're going to be very lonely on your new board if you can't forget about all that stuff.
What the heck does the phrase "within the context of historical studies" mean? Either they tell you what the historical data says re what is safe or they do not. I say they do not. They leave out a critical factor.
It is very obvious that, for the period studied, the studies tell you what the historical data says - no critical historical data is left out.
What are you suggesting that the quote marks signify? I have always taken them as an indication that the future may be worse than the past, and, in that event, all bets would be off
Very obviuosly there is the risk that the future may be worse than the past as for many things in life - but that is no reason to concude that all bets should be off.
Can you give me one example outside of the realm of SWR studies where it is common practice to deliberately ignore data that always bears on the question being examined and then issue a study saying that the data that was examined answers the question satisfactorily? I cannot think of one other area where research following this sort of practice is considered valid. Why are there special rules for SWR analysis?
For the period studied all historical data was included - there were no special rules.

However that does not mean to say that studies cannot be updated with the most recent data and the conclusions cannot be refined and improved.

.... but, hocus, have you actually read any of the original studies?
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

You appear incapable of differentiating between what the studies actually say and what was apparently said in some of the wilder, more absurd posts on TMF.

I understand the distinction you are making. Some of the intercst claims were absurb, that's true. The words in the study itself are not as bad, that's also true. That doesn't prove that the study methodology is valid.

You might want to ask yourself this question--Given the absurdity of the intercst SWR claims, why were there not more people objecting to them on the TMF board.? Many people there see the absurdity of his claims. Even someone like Arrette says that he posts like "an idiot" and Prometheuss promised to "denounce" him if he continued making some of the more absurb claims. But these people do not find the intercst claims sufficiently damaging to insist that he stop making them (Prometheuss never followed through). Why?

I think it is because they view the intercst claims as mere overstatements of points that are in some way valid. The statement you heard a lot was "Oh, 4 percent may not be exactly right, but it's a good rule of thumb." Not for the allocation recommended in the study. Not at the valuation level of the year 2000. Not according to the historical data.

My sense is that there are many people who believe that the methodology is valid, that intercst overstates things a bit, and that it is no big deal. The intercst overstatements don't help matters at all, but the core problem is the flaw in the conventional methodology. There are lots of people who read SWR studies who have never once been to the TMF board. What of them? Do you see no problem if their retirements go bust because of their confidence in a methodology Bernstein describes as "sometimes misleading." I do.

The studies, including REHP, based on historical data, include all relevant data

This is the core point of contention, KenM If the studies include all relevant data, they are valid. We are in agreement on that. I say they do not include all relevant data. I say that changes in valuation affect what is safe. The studies purport to show what is safe, yet they do not incorporate data dealing with the critical factor of changes in valuation levels.

The methodology used in the studies is logical and uses all the relevant historical data including peaks in valuations and changes in valuation for the period studied.

There is no adjustment made for changes in valuation levels. Reading one of these studies using the conventional methodology, you would think that the SWR was the same for a retirement beginning in the year 2000 as it is for one beginning today. That can't be.

hocus, have you actually read any of the original studies?

I read the intercst study when I first came to the TMF board in 1999. I haven't read any others since.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

The fact that they're all on their own coming up with similar numbers is a good sign that it is right.

The observation you are making here underlines the concern that I was trying to raise in my earlier post. If there were only one study out there with an invalid methodology, we might reasonably conclude "oh, not too many people will be taken in.": When there are a whole bunch of studies saying the same wrong thing, the risk that retirements will go bust as a result of the flaws in the methodology is far greater.

The practical purpose of the invalidity claim is to protect aspiring FIRE retirees from false claims as to what is safe. BenSolar put up a poll in June in which the board agreed 9 to 1 that changes in valuation levels affect what is safe. So we are agreed that this is an important factor. The conventional methodology does not include an adjustment for this factor. You need only move one step down the logic chain to see that the conventional methodology at least sometimes misleads people as to what is safe, just as Bernstein says.

To me that all adds up to "invalid." To the rest of the board, it does not. We disagree on what word to use to describe this flaw in the conventional methodology, but we reached a consensus in June that the failing is real and significant.

I suppose we could try to find some other way of saying it. Is it more acceptable to say "studies using the conventional methodology deliberately mislead readers as to what the historical data says is safe?" I see the word "invalid" as being a short and sweet way of conveying the same basic concept.
[KenM]
*** Veteran
Posts: 133
Joined: Tue Mar 04, 2003 12:54 am

Post by [KenM] »

The conventional approach does fine on all aspects of the question except valuation

hocus, that's where you begin to sound unreasonable. The historical studies use historical data over a particular timeframe on peak valuations, low valuations, changes in valuations, inflation, deflation, dividends, etc, etc. You have this bee in your bonnet only about valuations. If you want to use the findings of historical studies and interpret the results to simulate what-if, worse-case scenarios then that is a different matter and is going beyond the studies themselves. You would then have to make assumptions on extreme situations outside the historical range on valuations, dividends, inflation, deflation, etc, etc. If that's your intention then that's admirable and I would not disagree. But please do not say that historical studies carried out over a particular time frame are invalid - that only leads to ridicule. The studies may be out of date but their methodology is not invalid.
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
peteyperson
**** Heavy Hitter
Posts: 525
Joined: Tue Nov 26, 2002 6:46 am

Post by peteyperson »

I'm officially bowing out of this discussion now. It's just eating too much time and getting people nowhere.

Petey
KenM wrote: But please do not say that historical studies carried out over a particular time frame are invalid - that only leads to ridicule. The studies may be out of date but their methodology is not invalid.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

If you want to use the findings of historical studies and interpret the results to simulate what-if, worse-case scenarios then that is a different matter and is going beyond the studies themselves.

It's ridiculous that investors using these studies should be required to jump through all sorts of hoops and make all sorts of adjustments to the number reported in the studies to determine what the data says is safe. That's exactly what the study is supposed to do, tell you what is safe.

Why should each and every reader of an SWR study need to perform all these calculations for himself or herself? Why not just incorporate the valuation aspect of the question right into the study? The researcher does the work once, and the task is done. They a person could quote the results of an SWR study with a reasonable expectation that he was not going to cause busted retirements by doing so.

Why is it such a controversial idea to ask that people researching what is safe be willing to include in their studies data on all aspects of the question of what is safe?

That only leads to ridicule.

I'm used to it, KenM If you have the data on your side, you keep pounding on people to look at the data. If you don't, I guess you employ ridicule to shift people's attention away from the data. The continuing use of ridicule by those defending the conventional methodology reveals the weakness of their case on the merits, in my view.

The studies may be out of date but their methodology is not invalid.

You keep saying "out of date." My view is that the conventional methodology generates inaccurate results not only in bubble periods, but always. I am saying that changes in valuation levels always affect SWRs.
Post Reply