The Great SWR Investigation - Part 1

Financial Independence/Retire Early -- Learn How!
Post Reply
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

The Great SWR Investigation - Part 1

Post by JWR1945 »

The Great SWR Investigation - Part 1

This is the beginning of a response to a post that hocus has put up on the Alternative Withdrawal Strategy - Part 2 Thread on this FIRE board. I have separated it because of its scope. It is typical of one of hocus's products at an intermediate stage. It is very clear. It is useful. Yet, upon reflection, a comprehensive response covers a tremendous scope.

In its most basic context hocus is asking for a reasonable, yet precise, definition of the term Safe Withdrawal Rate. He has suggested using an intermediate level of sophistication...something between an approximate rule of thumb and something else that covers all conceivable possibilities. He suggests limiting the scope to cover those factors that we can do something about. He suggests excluding political instability, which is entirely reasonable within the United State. It might not be for a portfolio that includes international diversification.

Again, at a most basic level, let me suggest that a Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio based primarily upon a mathematical calculation based on existing, historical information. It should always be placed in context. It should be defined in terms of those factors that are covered by the calculation, those factors that are not covered and those that are only partially covered. The term Safe Withdrawal Rate should always be defined in terms of the reliability of that estimate. That includes known sensitivities to the assumptions inherent in any projection.

There is one more requirement that I would add to the definition of a Safe Withdrawal Rate. It is the application of the estimate. It is also the applicability of the estimate for other purposes.

This last requirement is what expands the scope of this investigation. I have only a limited concept of what it implies. We already think in terms of asset allocations. But there is more. Fortunately, I am able to provide an example to clarify our thoughts.

Just over a year ago, hocus asked a simple question that set off fireworks. He had some money to invest. He had already determined by his own independent investigation that a retirement portfolio could reasonably be expected to support a withdrawal rate of 4% over a long period of time. He was aware that stock market valuations were still at levels well above the historical norm. He wanted to know whether he should invest in stocks (consisting of S&P 500 index funds for purposes of discussion) or, because of valuations, whether he should invest in ibonds at 2%. If he could have bought TIPS at 3.5% as he had done in the past, he would have. Again, this was a matter of valuations. But at 2% the decision was not clear-cut.

In retrospect, the best decision would have been to invest in ibonds because of the stock market valuations at that time. We have subsequently developed an understanding of the effects of valuations on the safety of withdrawal rates that is sufficient to support such a decision. This required non-trivial research. It covered new ground.

Part of this understanding is knowing whether taking a risk is justified or whether it is reckless. In terms of investing right away one year ago, such a risk would have been reckless. The penalties for any downward fluctuation were not offset sufficiently by the prospect of greater returns. Waiting for more reasonable valuations had minimal penalties and a much greater upside potential.

Today, we think of this conclusion in simple terms. Cash is an asset class. Back then it was heresy.

Notice that when we talk about asset allocations, we must also think in terms of penalties and rewards and their impacts upon the investor. This is not simply a matter of one's psychological make up. It includes objective consequences as well.

This is where we begin...settling on a definition and looking for more general applications than a simple rule of thumb.

Have fun.

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

Post by ataloss »

I think this is a good discussion of safe withdrawal rates:

http://www.retireearlyhomepage.com/restud1.html

Intercst discusses what he calls maximum "100% safe" withdrawal rates referring to the historical record (and I think implicitly recognizing the limitations of the method)

Also the 4% rule (by Dory36)
http://boards.fool.com/Message.asp?mid=18870026

It's simply the maximum withdrawal that one can take from a portfolio every year for 40 years, and know that even in the worst times this country has ever seen, the portfolio would not have run dry.

can't quote too much w/o permission :?

This assumes that "the worst of times" are
a. in the past
b. in the 1870-2000 data

Datasnooper has pointed out that all non-zero withdrawal rates are unsafe to some degree. Although undoubtedly true most retirees haven't found this to be very helpful.

Of course we would like to know the maximum future safe withdrawal rate (fSWR in some circles) which is obviously unknowable.

We would like to consider all the factors which would influence SWR but we don't know what those factors were in the past or will be in the future. I assume that those who suggest that this is possible but difficult mean to be amusing in some way that I don't understand. Unknown unknowns, black swan events and so forth.


Bernstein has pointed out that long term equity returns can't really match the 20th century US record:
http://www.efficientfrontier.com/ef/198/returns.htm

perfectly valid tool for the use to which you intend to put it--making allocation decisions consistent with what the historical record says is safe.


I am not sure that the idea of a safe withdrawal rate is primarily of value to determine asset allocation given the limitations of the data, changing correlations and returns of asset classes. Personally I would like to have some guidance on how much I could withdraw from my portfolio.



I may have misunderstood the point so I would be interested in any clarification. :wink:

I think that there is a middle position between asserting that future swr couldn't be worse than the past and the position that if you got enough data you could predict the future. Adjust for valuation and consider the result. Don't pretend that you know the level of precision of what is essentially a prediction of future events.
Last edited by ataloss on Sat May 24, 2003 8:06 am, edited 1 time in total.
Have fun.

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

Post by hocus »

Let me suggest that a Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio based primarily upon a mathematical calculation based on existing, historical information.

Why primarily? Why not entirely?

In retrospect, the best decision would have been to invest in ibonds because of the stock market valuations at that time

I don't agree.

It might have been the best decision for some investors in some circumstances. It was almost certainly not the best decision for others in other circumstances.

Also, I do not believe that the asset class with the highest SWR is necessarily "the best decision."

Part of this understanding is knowing whether taking a risk is justified or whether it is reckless. In terms of investing right away one year ago, such a risk would have been reckless. The penalties for any downward fluctuation were not offset sufficiently by the prospect of greater returns. Waiting for more reasonable valuations had minimal penalties and a much greater upside potential.

This is worded too expansively, in my opinion. There are probably cases in which what you say above is so. But there are also probably cases in which it is not so.

In any event, I don't think that this aspect of the question affects calculation of the SWR. It is probably better classed as an allocation consideration..

Notice that when we talk about asset allocations, we must also think in terms of penalties and rewards and their impacts upon the investor. This is not simply a matter of one's psychological make up. It includes objective consequences as well.

This I very much agree with.

Cash is an asset class

This I very much agree with.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

These are just a few quick comments. More will follow. Especially for ataloss, I know that Gummy has mentioned at his web site that there are many similar but different definitions of the term Safe Withdrawal Rate. He has extracted the essence of what they had in common. They disagree on details.
ataloss
I think that there is a middle position between asserting that future swr couldn't be worse than the past and the position that if you got enough data you could predict the future. Adjust for valuation and consider the result. Don't pretend that you know the level of precision of what is essentially a prediction of future events.

What we can do is to supply the full context behind our predictions. We make predictions where there is randomness all of the time. For example, we expect gambling casinos to win. But we place this in the context of probabilities and statistics.

hocus
Let me suggest that a Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio based primarily upon a mathematical calculation based on existing, historical information.

Why primarily? Why not entirely?

It is because of qualitative factors.

This is what I have in mind: When he developed his Monte Carlo model, Gummy decided that he should introduce the effects of inflation by randomly selecting the first year and then extracting inflation percentages from the historical sequence that began with that year. Gummy excluded most, if not all, of the years when we were still on the gold standard.

I think that he made a good decision. I think that any prediction involving inflation that we make today should exclude the gold standard. There are other cases in which we would want to consider earlier years. The most obvious example is to include stock returns from a longer time period. We need the additional data points.

hocus
In retrospect, the best decision would have been to invest in ibonds because of the stock market valuations at that time.

I don't agree.

I will back away from the strong form of my assertion. It is based upon a very narrow context and it has some hidden assumptions.

hocus
Also, I do not believe that the asset class with the highest SWR is necessarily "the best decision."

It will be fascinating to examine this in greater detail. I know for sure that I don't comprehend this fully. There is much worth learning along these lines.

We need to examine this assertion that I made originally.
There is one more requirement that I would add to the definition of a Safe Withdrawal Rate. It is the application of the estimate. It is also the applicability of the estimate for other purposes.

Do we want to do this? Or should we simply identify such applications when we provide the context that goes with our numbers?

Have fun.

John R.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Intercst discusses what he calls maximum "100% safe" withdrawal rates referring to the historical record (and I think implicitly recognizing the limitations of the method)

I disagree with the parenthetical. The quote marks used by intercst are intended as a recognition that the numbers he reports do not account for the low--probability "asteroid" type events. He is not acknowleding that his study fails to consider factors with a critical effect on calculation of the SWR. If he were using the quote marks to acknowledge that he failed to consider critical factors, he would need to include a discussion of those factors somewhere in the text of the study, and offer an explanation of why he did not incorporate them into his analysis.

It's simply the maximum withdrawal that one can take from a portfolio every year for 40 years, and know that even in the worst times this country has ever seen, the portfolio would not have run dry.[i/]

I think that's a reasonable definition. It's not perfect, but it's not bad. It's a whole lot better than the unwritten and unspoken definitions that many people have been employing in their discussions of the matter.

We would like to consider all the factors which would influence SWR but we don't know what those factors were in the past or will be in the future.

Should we consider those we know had an effect in the past and that we know will have an effect in the future? That's the most important question on the table. Is it proper to exclude a factor from one's analysis that one knows for certain affects the result?

Bernstein has pointed out that long term equity returns can't really match the 20th century US record:

We should consider this statement in relation to the Dory36 definition of SWR stated above. Dory36 says that a SWR
is "the maximum withdrawal that one can take...." Please note the tense. It is the future tense. He is not saying that the SWR reports on something that happened in the past. He is saying that it provides you an indication of the probabilities of future possibilities. It provides insights to be used for taking actions in the future.

If Bernstein is right in what he says above, then the factors driving his statement need to be accounted for in a valid SWR analysis. Otherwise, the number produced will not serve the purpose that the Dory36 definition says it is to serve.

I think that there is a middle position between asserting that future swr couldn't be worse than the past and the position that if you got enough data you could predict the future.

I hope there's a whole lot of middle positions between those two. Both of those two positions are absurdities.

Actually, the whole idea of putting the word "future" in front of the words "safe withdrawal rate" is an absurdity. It makes the concept a self-contradicting one. If you were able to tell the future, you would not be determining a "safe" withdrawal rate, you would be determining a "certain" withdrawal rate. I don't think there can ever be such a thing as a Future Safe Withdrawal Rate. It is an imaginary animal.

The Historical Safe Withdrawal Rate is another imaginary animal. If you calculated the number that people seem to be referring to when they employ this phrase, you would be determining a safe withdrew rate, not a safe withdrawal rate. Adding the word "historical" turns the concept into a self-contradicting one.

Adjust for valuation and consider the result. Don't pretend that you know the level of precision of what is essentially a prediction of future events.

I disagree strongly with the second sentence.SWR analysis has nothing to do with making predictions of future events. I see no langauage referring to the need to make predictions of future events in the Dory36 definition provided above.

If the calculation of SWRs is some sort of variation of astrology, I want no part of it. I had never heard anyone describe it in this fashion prior to May 13, 2002. This idea that you need to make predictions of the future to deternine the SWR is a relatively new phenomenon, as far as I can tell.

I don't agree with the suggestion i hear in the first sentence that adjusting for valuation is an optional exercise. It's not just that adjusting for valuation is a nice thing to do because you can then "consider the result" you get from doing so. Valuation levels affect returns (and therefore SWRs as well) as a matter of "mathematical certainty," according to William Bernstein. If you haven't accounted for valuation levels, you haven't determined the SWR.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

I know that Gummy has mentioned at his web site that there are many similar but different definitions of the term Safe Withdrawal Rate.

In the event that Gummy is reading this, he might be able to help us out with the core dispute (which has been more of a problem at the REHP board than here, but which seems to have caused some confusion here as well). I think it is fair to say that Gummy's site has won near universal praise and that he is as much of an expert on this subject as exists on Planet Earth at this time.

I don't know how much Gummy knows about the background of the dispute. Rather than go into any details that might bore him, I will just ask him to comment on the core question, in the event that he is willing and able. I think that it is only fair to provide Gummy a brief description of the use to which I intend to put any answer he provides, so I have made brief reference to that as well.

The core question is--Do valuation levels affect SWRs?

William Bernstein says yes. He says in his book "The Four Pillars of Investing" (Page 12) that valuation levels affect returns as a matter of "mathematical certitude." Returns obviously affect SWRs. So valuation levels affect SWRs, in the view of William Bernstein.

On Page 234 of the book, Bernstein says that, when you factor in the effect of valuation levels, the SWR at the top of the bull market comes to 2 percent. This is way off from the number that intercst puts forward in his SWR study.

Now, I understand that there are other studies that put forward numbers close to the intercst number. My sense is that the authors of these other studies were not aware that valuation levels affect SWRs and therefore did not incorporate this factor into their analyses. I have no problem with that. The SWR concept is a new one, and those who prepare studies in good faith that do not account for every factor subsequently determined to have an effect are helping advance knowledge by doing so. I offer no cirticism of any study other than the intercst study. Actually, I offer no criticism of the intercst study per se either. I object only to the SWR claims that intercst has put forward on the REHP board. He cites the study as support for the SWR claims he posts to the board, but the data in the study does not support many of the claims.

I put up a post at the REHP board on May 13, 2002, saying that I thought that valuation levels might affect SWRs (I was not aware of the Bernstein quote at this time, so I was tentative in my assertion), and suggesting that it might be a good idea to determine three SWRs: (1) the one that applies at times of high valuation levels; (2) the one that applies at times of moderate valuation levels; and (3) the one that applies at times of low valuation levels.

The board expressed great enthusiasm for the idea. But intercst said that it was "loony" and "irrational" for me to say what I was saying. Subsequently, he said that it showed that I was suffering from "mental illness." He declared that the "board culture" of that board required "ridicule" of any poster putting forward such ideas. I did not back down from my claim, and defended myself from the various smears directed at me, and earlier this week Motley Fool informed me that my posting privileges have been revoked as a result. It is as a practical matter not possible at that board to raise the question of whether valuation levels affect SWRs. Motley Fool has for practical purposes prohibied at its site discussion of the idea that valuations levels affect SWRs.

Do you think that valuation levels affect SWRs, Gummy, or that it is at least possible that they do and that therefore the subject is one that should be discussed at a board on early retirement in a reasoned manner? It is only fair to let you know that my purpose in asking is that I am hoping to have intercst removed as leader of that board, and one of the things that I hope to do in future days is accumulate the testimony of various experts in the field that his position is deceptive and that his SWR claims may cause great financial losses to people who are exposed to them on a discussion board where it is not possible for community members to raise valid questions about the methodology used in the study.

If you would prefer to wait until a later time to offer any views on this matter, that is fine with me. I can understand if you would want to better inform yourself before venturing an opinion. But I hope that you will be able to help us out on the question at some appropriate time. I believe that it is a matter of great importance to the community of people who employ SWR analysis to finance their early retirement plans.

Gummy excluded most, if not all, of the years when we were still on the gold standard. I think that he made a good decision. I think that any prediction involving inflation that we make today should exclude the gold standard.

OK. I see what you are getting at.

My view is that SWR analysis should include only aspects of the question that can be quantified. If there is some way using data to incorporate the factor you are raising here, then I think it is valid to include it in a SWR analysis. If that is not possible, I think the thing to do is not to include it in the calculations of the SWR, but to make reference to the factor in the text of the study describing the methodology that was used to produce the number.

I maintain that the SWR is a mathematical construct. That means that there is no room for subjectivity in regard to some factors. For example, valuation affects the SWR without question. So you simply must incorporate data on this factor or you have not done a valid SWR analysis. (In the event that you have issued a study that does not factor in valuation because you were not aware at the time that it was a factor, you need to update the study when you become aware that it does.)

The factor that you are raising might affect the SWR but it might be hard to gather data on this factor. If there is no data at all, then all you can do is make note of the factor in the text of the study and note that it was not included in the analysis because of a lack of data.

If there is some data, but it is not good data, there might be some room for subjectivity in making a decision as to whether to incorporate that factor into the analyis or not. Even if you elect not to include the factor in the objective mathematical calculations that follow, however, I can see no justification for not making note of the ignored factor in the text of the study. If you believe that it has an effect and your only reason for not including it in your calcuations is a lack of data, you are misleading the readers of the study not to make note of your concerns that failure to include the factor inevitably has diminished your confidence in the results.

If we had access to perfect data on every single factor that affects the SWR, we would have perfect knowledge of the SWR. This does not mean that we would be able to predict the future. SWR analysis does not aim to predict the future. But we would know the SWR perfectly.

Given that we do not possess perfect data on every single factor, we have to do what responsible people do when faced with this sort of obstacle. They include all of the factors that they possibly can and are careful to inform readers of the possible effects of factors that they were not able for practical reasons to incorporate into the analysis.

My bottom line on this is that your decision to include such factors or not does not change the SWR into something other than a mathematical construct. The SWR is a mathematical construct. The only place where subjectivity rears its head is in making decisions as to whether to include poor data when that is all that is available or to leave that factor out of the equation altogether. The subjectivity comes into play in design of the study, but there is no subjectivity to the SWR calculation itself.

Misunderstanding of this point has generated a great deal of confusion, in my view. It is hard to make progress without first coming to a consensus as to whether the SWR is a mathematical construct or not. If it is not a mathematical construct, then you can come up just about any number under the sun and maintain that it is a "valid" SWR. If the SWR is a mathematical construct, then there are things which you simply cannot do. You must include all critical factors; you have not choice in the matter.

My position is that the SWR is a mathematical construct. It is not a nice rule-of-thumb. It is a tool in which you use data, not subjective impressions, to inform your allocation decisions.

Do we want to do this [referring to the idea of considering applications of the SWR concept at the part of the process at which you calculate the SWR]? Or should we simply identify such applications when we provide the context that goes with our numbers?

I favor the latter. I think that the SWR should be viewed as a mathematical construct. Application questions require reference to subjective considerations. So I think it is best to keep these separate.

The only sense in which I see it as appropriate to consider applications up front is in coming up with a definition of "SWR." If someone were to say in his or her proposed definition that the SWR should not account for valuation, I would respond that using that definition defeats the purpose of the exercise.

If you don't factor in valuation, you cannot use the SWR to determine the safe take-out percentage or to inform your allocation decisions. So why bother with the exercise? You need to consider the application to which you are going to employ your results at the time you are coming up with a definition.

But once you have a good definition in place, it is best to leave subjective considerations out of calculation of the SWR. There is a place for subjective considerations in making allocation decisions, but it is not at this stage of the process. The appeal of the SWR concept is that it quantifies considerations that, prior to the development of the SWR concept, had been discussed only in subjective terms. Add subjective considerations to the calculation of the SWR, and you are back to where people were before there even was such a thing as SWR analysis.

We can each of us offer our subjective views as to the "best" take-out percentage and the "best" allocation percentage from now until the end of time, and never really get anywhere. When you are speaking subjectively, 4 percent is a good number and 1 percent is a good number and 11 percent is a good number, and they are all good numbers. Who is to say which of the various possibilities is really "best"?

What I like about SWR analysis is that it gets you out of that quicksand. The SWR is a number generated by data You don't have a right to an opinion different from mine on what the data says. It says what it says, and that's that.

That to me is the benefit of SWR analysis. People who say that they think 4 percent is a good rule of thumb may be entirely correct about that, but they are not engaging in SWR analysis. Once you start employing subjective considerations in the calculation of SWRs, I don't see where you stop. The subjective stuff overtakes the objective stuff. The concept loses its purity and its value once you let subjectivity in the door.

I say, Hash out what are legitimate factors or not at the start. Once you have identified the factors that apply, collect the data and then calculate the number. Do it that way and the number you calculate means something. Do it the other way, and it is all a mish-mash combination of objective data and subjective impressions. Do it that way, and you don't end up with a terribly useful tool.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »


Have fun.

Ataloss
WiseNLucky
** Regular
Posts: 84
Joined: Tue Nov 26, 2002 3:59 am
Location: Florida

Post by WiseNLucky »

Even with a MC result accounting for known factors and successful in 95% of runs, we will have basically no idea if it will be safe in the future. This makes it essentially a well refined rule of thumb.


I agree with this statement. The assertion that we can reduce SWR to a mathematical equation applies only to the past.

The fact that there is a certain safe withdrawal rate for a given portfolio over a given point in time in the future is not useful if we can't estimate it. But estimation is the best we can do.

The concept of SWR has saved a number of peoples' financial lives because the very recent conventional wisdom was withdrawals in the 7-10% range. Studies like the Trininty study and intercst's work showed that to be untrue. But to extend their work to apply it with certainty to the future would be equally unwise. It may not require a catastrophe to have future returns fall far below historical averages.

So I believe that SWR will remain a "rule of thumb" concept. I believe it can be further refined to include starting valuations, but it will remain a rule of thumb. It will need to be "subjectively" reviewed on a periodic basis, and a cash reserve will be needed to smooth out the flow. But I feel that the mathematical work is still very useful.
WiseNLucky

I just wish everyone could step back and get less car and less house then they want, and realize they don't NEED more. -- NeuroFool
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

But I feel that the mathematical work is still very useful.


I agree. The ability to turn down the real returns and look at the result has been very interesting to me. (Increasing the historical er to check the historical swr was something I would never have thought of- glad raddr is here)
Have fun.

Ataloss
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Revised definition

Post by JWR1945 »

Revised Definition

My original definition was:

Again, at a most basic level, let me suggest that a Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio based primarily upon a mathematical calculation based on existing, historical information. It should always be placed in context. It should be defined in terms of those factors that are covered by the calculation, those factors that are not covered and those that are only partially covered. The term Safe Withdrawal Rate should always be defined in terms of the reliability of that estimate. That includes known sensitivities to the assumptions inherent in any projection.

My first revision is:

A Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio that is mathematically calculated based primarily on existing, historical information. It should always be placed in context. It should be defined in terms of those factors that are covered by the calculation, those factors that are not covered and those that are only partially covered. The term Safe Withdrawal Rate must always be defined in terms of the reliability of the estimate. This should include known sensitivities to the assumptions that are inherent in making any projection. It is desirable to include general comments about the applicability of the estimate.

An example of a Safe Withdrawal Rate that does not depend on existing, historical information involves the use of mathematical formulas and theorems. If you started out with $1.0 million, you could withdraw $50K annually for 20 years. You would simply keep the money in cash at zero interest (such as stuffing dollars into a mattress).

Comments:

In this revision I have required that the term Safe Withdrawal Rate is always a mathematical calculation and that it must always include remarks about the reliability of that estimate.

The additional requirements would always be present in a good Safe Withdrawal Rate calculation. I allow for not so good calculations as well.

Have fun.

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

Post by ataloss »

An example of a Safe Withdrawal Rate that does not depend on existing, historical information involves the use of mathematical formulas and theorems. If you started out with $1.0 million, you could withdraw $50K annually for 20 years. You would simply keep the money in cash at zero interest (such as stuffing dollars into a mattress).


What about inflation/hyperinflation? The Weimar Republic?
Have fun.

Ataloss
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

For ataloss:That would be covered in explaining the context.

My example is very simple so as to be easily understood. However, I have posted about the interest rates needed for TIPS to provide truly safe (100% safe) withdrawal rates of various lifetimes. IMHO, that kind of calculation should be included.

Have fun.

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

Post by ataloss »

Although simple, wouldn't the example would have failed in some circumstances (in terms of preservation of spending power)

ataloss
not sure we can be independent of history
Have fun.

Ataloss
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

For ataloss: I think that my illustrations are transparent enough so that none of this is hidden. The context behind the numbers is adequate. In my earlier post I did address the availability of TIPS at various interest rates on today's secondary market. That helped to improve the context.

My revised definition reads: A Safe Withdrawal Rate is an estimate of the safety of a retirement portfolio that is mathematically calculated based primarily on existing, historical information. It should always be placed in context. It should be defined in terms of those factors that are covered by the calculation, those factors that are not covered and those that are only partially covered. The term Safe Withdrawal Rate must always be defined in terms of the reliability of the estimate. This should include known sensitivities to the assumptions that are inherent in making any projection. It is desirable to include general comments about the applicability of the estimate.

I do not believe that these requirements have been violated. There must be a reasonable amount of flexibility in terms of how much detail is presented. Where there has been trouble, it has been because certain factors and sensitivities (including valuations and allocations) were found to be relevant but they were not acknowledged as such. In fact, there was a concerted effort to suppress any investigation into such issues.

Have fun.

John R.
raddr
*** Veteran
Posts: 265
Joined: Mon Nov 25, 2002 3:25 am
Contact:

Post by raddr »

WiseNLucky wrote: The concept of SWR has saved a number of peoples' financial lives because the very recent conventional wisdom was withdrawals in the 7-10% range. Studies like the Trininty study and intercst's work showed that to be untrue. But to extend their work to apply it with certainty to the future would be equally unwise. It may not require a catastrophe to have future returns fall far below historical averages.


Yes, in fact the future returns may not have to be any lower in order for the SWR to be significantly lower. I have done studies in the past that show that the sequence of returns play a huge role in what is a "safe" withdrawal rate. Even switching just 2 years out of the last 130 could've been devastating. Also, a slight uptick in portfolio volatility compared to the past can make a big difference on the negative side. All of this assumes that avg. yearly returns remain the same going forward which is in and of itself a huge leap of faith given current valuations on the TSM indices. :cry:

The next time you hear one of the sages at the REHP say that a 4% (or whatever) SWR assumes that "the future will be no worse than the past" then run the other way. :wink: The same average returns going forward could be devastating given slight changes in sequence, volatility, or covariance among assets - virtually a given IMO.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

The next time you hear one of the sages at the REHP say that a 4% (or whatever) SWR assumes that "the future will be no worse than the past" then run the other way.

I agree, raddr. The problem with the intercst claims is not that they may be proven wrong in the event that the future is worse than the past. The problem is that they certainly will be proven wrong if the future is exactly like the past. The only way that someone who retired in the year 2000 could have reasonable confidence in a 4 percent withdrawal was if he or she possessed a strong confidence that the future would be very different from the past (such as is in the way argued in the book "DOW 36,000" by James Glassman).

Your post noted one reason why the intercst claims don't work even if future returns are as good as past returns. The odds are good that the sequencing of returns will be less favorable. That's an extremely powerful point, in my view.

The valuation concern that I have focused on is an independent factor that also reveals why it is not reasonable to expect that a 4 percent withdrawal will work from the year 2000 forward presuming that the future is like the past. In this case, the concern is that stock market behavior will be like the past and therefore stock market returns will be less than those that were achieved in any of the time-periods covered in the 130 years examined in the study.

The caveat in the intercst study is not that returns may be worse than the past. It is that "the future" (a broader concept) may be worse than the past. Bernstein is saying in no uncertain terms that future returns will be worse than past returns in the event that the future itself (the broader concept) is anything like what the past was.

I have seen you argue that there is a "reversion to the mean" in stock valuation levels, that stock market performance is not random. I agree with you. More importantly, What Bernstein is saying in the most emphatic terms possible indirectly supports what you are saying. He is a careful writer.. But on this question he uses a strong phrase, one that leaves absolutely no room for ambiguity. He says that valuation levels affect returns as a matter of "mathematical certitude." Those who are not clear on why he is saying this would benefit from reading Chapter Two of "The Four Pillars of Investing," in my view. The argument he is making is of great significance to all investors and the support he offers for the argument is very strong, in my assessment.

Bernstein is most definitely not saying that there may be some terrible event (political instability or whatever) that will cause future returns to be worse than past returns. He is not predicting the future. He is stating with no equivocation whatsoever that future returns will be worse than past returns, assuming that the future (broad concept) turns out like the past.

The point is that valuation levels have always affected returns in the past, no exceptions. So why would you develop a safe withdrawal rate presuming that at this point in history for the first time ever the world would turn on its axis and this time valuation levels would not affect returns. That is an extremely far-fetched assumption. Go with the assumption that the future will be more or less like the past, and you are forced to conclude that returns will be less than the past, and you are thereby forced further to conclude that SWRs at this point in history (the year 2000) less than ithey ever have been in the past.

We all accept that the 100 percent safety assessment that intercst assigns to his resutts is an overstatement because of the possibility of asteroid-type events. So let's stop using the 100 percent number. Let's say that a SWR identified under the intercst approach to possess 100 percent safety really possesses 95 percent safety. Bernstein says that it is probably a waste of time to go for more than that, so let's presume that all we demand of this exercise is 95 percent safety.

Bernstein is saying that you cannot get it with a 4 percent withdrawal, not in the year 2000. He is saying that to possess reasonable confidence in a 4 percent withdrawal, you would need to assume return levels close to those that prevailed in the 130 years examined in the study and he is further saying that such an assumption flies in the face of what has happened throughtout the history of the stock market.

There has never before been a time when valuation levels did not affect returns. So you must include this factor in performing any reasonable assessment of what will happen in the future, presuming that you accept the condition that the future is going to turn out something like the past.

If you presume that the future will turn out entirely unlike the past (the Glassman scenario), then SWR analysis as we know it up to this point in time just isn't for you. To get a 4 percent number in the year 2000, you need to presume a future entirely unlike the past, and then you need to perform a different sort of analysis that what has commonly been referred to as safe withdrawal rate analysis. It might well be possible to come up with a number using the Glassman hypothesis, I see nothing wrong with the idea of someone doing this. But it should be clearly labled as the number you come up with when you assume a future unlike the past.

Intercst says that you come up with a 4 percent number when you assume that the future will be like the past, and this is just flat-out wrong. It is not just my opinion that it is wrong. It is wrong as a matter of "mathematical certitude.:" The data does not support this assertion. He should stop saying it. All others should stop saying it too. It is an assertion that at this point in time has been proven false beyond any reasonable doubt
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »


Have fun.

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

Post by ataloss »

Hocus
I disagree strongly with the second sentence.SWR analysis has nothing to do with making predictions of future events. I see no langauage referring to the need to make predictions of future events in the Dory36 definition provided above.

If the calculation of SWRs is some sort of variation of astrology, I want no part of it. I had never heard anyone describe it in this fashion prior to May 13, 2002. This idea that you need to make predictions of the future to deternine the SWR is a relatively new phenomenon, as far as I can tell.


I am still unable to understand this. It appears to me that the swr for the 1870-2000 period is whatever intercst says it is. The swr for 1973 forward (in the 30 year sense) remains to be determined i.e. it depends on future events. Any assertation that the swr will be x% rests on assumptions about future returns, correlations, volatility and other factors. I think this is by definition a prediction of the future. Could you explain this further?
Have fun.

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

Post by raddr »

Quote:
Your post noted one reason why the intercst claims don't work even if future returns are as good as past returns. The odds are good that the sequencing of returns will be less favorable. That's an extremely powerful point, in my view.


Hocus, what are those odds exactly?


Approximately 75-80%:
http://nofeeboards.com/boards/viewtopic ... ence#p2252
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

Approximately 75-80%:


Is there a chance that there is some reason for the "lucky sequence" that isn't captured by your mean reversion technique?

Raddr (from that thread)
I will say that I am confident that the simulation comes very close to modeling past market behavior but none of us know how different the magnitude of returns and volatility will be in the future.


Would this mean that you can't provide 95% confidence intervals for applicabiility of the 3.55% swr to the future?
Have fun.

Ataloss
Post Reply