The Great SWR Investigation - Part 1

Financial Independence/Retire Early -- Learn How!
therealchips
*** Veteran
Posts: 174
Joined: Sat Jan 04, 2003 4:00 am
Location: Henderson, Nevada, USA

Post by therealchips »

Hi, ataloss. Yes indeed, it is easy to take growth in the assets. You said
Bernstein never said that it was a matter of mathematical certitude that the 4% withdrawal rate was unsafe. He suggested that with current market valuations it would be prudent to plan for a lower withdrawal rate- which makes sense to me.
Did he say how much lower? Did he say that the prudent rate currently is 2%? As I recently posted, having just discovered that I could cut from my spreadsheet and post here, my currently planned average withdrawals rates at various ages are these:
63 to 69 3.22%
70 to 79 3.34%
80 to 89 3.69%
90 to 94 4.04%
Does anyone know whether this plan would qualify as prudent according to Bernstein? (Not that I will lose any sleep over the matter, but just to use his work as a sanity check on my own.) I may have to buy his book. I went so far as to price it at amazon just now.
He who has lived obscurely and quietly has lived well. [Latin: Bene qui latuit, bene vixit.]

Chips
[KenM]
*** Veteran
Posts: 133
Joined: Tue Mar 04, 2003 12:54 am

Post by [KenM] »

On a couple of real estate issues........

referring to a 1% or 2% annual appreciation in residential housing values. Hocus, are you overlooking the value of free rent?
I fully agree. If appreciation of owner occupied housing is included in an annual valuation of total assets then the free-rent issue has to be assessed - back to an overall 4% annual increase?????

In looking at long term appreciation in values of real estate I'm not sure whether, in the US, there has been in parallel a long term deterioration in value-for-money. Anecdotally and in very general terms (don't ask me for figures but I'm sure there're out there somewhere) I know of real estate markets in the UK, Singapore, Hong Kong, Malaysia. In all of those, although in the long term the average price paid for a single family home has perhaps not increased dramatically over inflation, that single family home has become smaller and in many cases with a smaller land area and in not such a good area. For example (again anecdotally), in the UK a 40 year old doctor in 1900 could have afforded a very large detached house in large grounds. His son, in 1930 still a very large house but smaller grounds. His grandson in !960 a detached house in a very good area in the suburbs. His great-grandson in 1990 a semi-detached in not quite so good an area. Therefore overall there has been substantial appreciation on a like-for-like basis. As I say, not sure if this applies to the US?
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
[KenM]
*** Veteran
Posts: 133
Joined: Tue Mar 04, 2003 12:54 am

Post by [KenM] »

If it were 1993, we would not be nearly as concerned about this as we are today

I'm not so sure about that. In 1993, the S&P500 was 80% higher than 5 years before; 200% higher than 10 years before. I tend to think that the topic of discussion would still have been about "high" valuations. Not, as today, attributable to the excesses of the late 90's but I'm sure that many would have found a reason somewhere to support a proposal for a 2%SWR :)

The problem I have with the whole high valuation debate is that, looking at the past, the only sensible prediction is that, over 10 to 20 year timeframes, the market is unpredictable, has no commonsense and will always surprise me. In the past we have had long periods of low valuations. Therefore over the timeframe of a retirement starting now, it could well happen that high valuations could continue for the next 15 years or so. If that occured, I would be a very disappointed retiree if I'd kept a low standard of living at 2% for 15 years when I could have enjoyed myself at 4% for the period of retirement when I would probably be at my fittest. IMO whatever the valuation levels, having regard to the unpredictability of the market over 10 to 20 year timespans, the optimal choice must be to start out at 4% and cut back to 2% if necessary.
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
wanderer
*** Veteran
Posts: 363
Joined: Tue Nov 26, 2002 9:33 am
Location: anytown, usa

Post by wanderer »

I fully agree. If appreciation of owner occupied housing is included in an annual valuation of total assets then the free-rent issue has to be assessed - back to an overall 4% annual increase?????

hocus said two things (well, twelve :wink:):

1. that he expected a 4% real return on the value of the property (ain't no way, unless very unusual circumstances, imo - fmo and i use the same data, i think)

2. a 4% take out - (i don't know what that means, but if he includes imputed rent, i have high confidence he can achieve that number).
regards,

wanderer

The field has eyes / the wood has ears / I will see / be silent and hear
wanderer
*** Veteran
Posts: 363
Joined: Tue Nov 26, 2002 9:33 am
Location: anytown, usa

Post by wanderer »

I'm not so sure about that. In 1993, the S&P500 was 80% higher than 5 years before; 200% higher than 10 years before. I tend to think that the topic of discussion would still have been about "high" valuations. Not, as today, attributable to the excesses of the late 90's but I'm sure that many would have found a reason somewhere to support a proposal for a 2%SWR

we stopped adding to that pig in mid 95. but we didn't get out (something hocus did do, iirc). 4500 on the djia. About 1000 pts higher is what raddr and some guy named Gross at PIMCO keep calling for. :shock:
regards,

wanderer

The field has eyes / the wood has ears / I will see / be silent and hear
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

BPP: "Valuation" seems to be a rather vague idea with intuitive appeal, but which seems maddeningly hard to pin down.

Ataloss: I agree.

Whether this is so or not is an extremely important issue that we very much need to resolve. One of the reasons why I keep asking that people buy "The Four Pillars of Investing" and read Chapter Two is that it addresses points like this in detail.

Here is what William Bernstein says on this question (the bolds were added by me):

Page 54: "This formula, which is known as the 'Gordon Equation,' provides an accurate way to predict long-term stock market returns....

"The Gordon Equation is as close to being a physical law, like gravity or planetary motion, as we will ever encounter in finance.....

Page 55: "The Discounted Dividend Model (DDM) model gives us the Gordan Equation, which allows us to estimate stock returns. This raises an important point. Wall Street and the media are constantly obsessed with the questiion of whether the market is overvalued or undervalued (and by implication, whether it is headed up or down). As we have just seen, this is essentially impssible to determine. But in the process we've just acquired a much more valuable bit of knowledge: the long-term expected return of the market. Think about it, which would you rather know: the market return for the next six months, or for the next 30 years? I don't know about you, but I'd much rather know the latter. And, within a reasonable margin of error, you can. .....

Page 56: "Going forward, it looks like stock and bond returns should be in the 6 percent range, not the 10 percent historical reward. Don't shoot me, I'm only the messenger. [Comment: This translates into an SWR of 2 percent]....

"On an intellectual level, most investors have no trouble understanding the notion that high past returns result in high prices, which, in turn, result in lower future returns. But at the same time, most investors find this almost impossible to accept on an emotional level.....

Page 57: "At the end of the day, the Fisher DDM method of discounting interest streams is the only proper way to estimate the value of stocks and bonds. Future long-term returns are quite accurately predicted by the Gordon Equation....

"The rub is, the Gordon Equation is useful only in the long term--it tells us nothing about day-to-day, or even year-to-year, returns"

Bernstein is saying that you cannot predict the short-term effects of differences in valuation
levels, but that you can indeed predict (with good but not perfect accuracy) the long-term effects of differences in valuation levels. SWR analysis concerns itself with the long-term. So the long-term effects of changes in valuation should be reflected in SWR analyses.

You notice that he [Bernstein] doesn't say (as a matter of mathematical certitude) that future returns will be less than the 6-7% long term historical rate

I am saying that he indeed is saying this, with the caveat that the future may turn out different than the past. I'll quote Bernstein again:

Page 58: "The Gordon Equation does not account for changes in the dividend multiple....Over relatively short periods of time--less than a few decades--this change in the dividend or PE multiple accounts for most of the stock market's return, and over periods of less than a few years, almost 100 percent of it....

"On the other hand, the long-term increase in stock market value is entirely the result of the sum of long-term dividend growth and dividend yield calculated from the Gordon Equation."

SWR analysis aims to assess stock investment safety only in the long term. In the long-term, changes in valuation levels always make a difference. It is therefore improper to ignore this factor in analysis of the SWR.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

In 1993, the S&P500 was 80% higher than 5 years before; 200% higher than 10 years before. I tend to think that the topic of discussion would still have been about "high" valuations.

You are right, KenM. Valuation is always a factor. The effect was more pronounced at the top of the recent bull market, that's all.

I'm sure that many would have found a reason somewhere to support a proposal for a 2%SWR

That depends on how the term "safe withdrawal rate" is defined. I am saying that it must be defined as the product of mathematical calculations performed on relevant data. Define it that way, and people will not be able to use SWR analysis to say things that are not so.

If you are excessively loose in the definition, then you are right that people will be able to use SWR analysis to "prove" any personal littlen point that they want to prove. Some will say that the SWR is 2 percent, some will say 4, some will say 8, some will say a negative 6.

It is my view that SWR analysis loses its appeal when it becomes a completely subjective exercise. The apppeal of the tool to me is that it cuts through the subjectivity and tells you what the data says.

I had all my money in stocks until I performed a SWR analysis in 1996. I was surprised to see what the numbers told me but I was too concerned with the safety of my plan not to act of the insights I gained from that analysis.

SWR analysis saved my skin. I would be back in the corporate workforce today were it not for the insights provided by valid SWR analysis. That's why I would like to see the value of that tool opened up to other aspiring early retirees.

It is not my goal to undermine confidence in the tool, to make it just one more thing that people can use to engage in endless arguments about which way to invest is "better" or "OK." The terms "better" and "OK" are so vague that there is no way to get a handle on them and resolves the differences. The beauty of a properly done SWR analysis is that it is an exercise in mathematics.

The problem I have with the whole high valuation debate is that, looking at the past, the only sensible prediction is that, over 10 to 20 year timeframes, the market is unpredictable

Please see my response post to ataloss from earlier this morning, which addresses this question in some detail.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

I may have to buy his book. I went so far as to price it at amazon just now.

That's highly encouraging news. I think that the best way to bring this to resolution is for as many people as possible to read the actual words of Chapter Two rather than my summaries of them.

Does anyone know whether this plan would qualify as prudent according to Bernstein?

Bernstein only gave us a calculation of the SWR which reflects the effect of valuation levels for the year 2000. I don't know that anyone has done this calculation for all the other years.

It was my proposal that we do this calculation for the other years that kicked this thing off. The suggestion that the community of people interested in early retirement find out what such calculations reveal has generated a good bit of controversy, to say the least.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

JWR1945
We can even specify two constraints such as satisfying a minimum balance throughout the portfolio's lifetime and staying above a minimal ending balance.)

therealchips:
Pardon me, but I don't think that is so. In retirement, when the value of my retirement assets falls below any particular number, there is no way I can repair the damage....

I think that Chips has misunderstood me. I am requiring that a hypothetical portfolio stays above some number, always, each and every year throughout the entire period. I demand that it is above a second number at the particular year at the end of the period. (The second number should be higher than the first.)

I am allowing the portfolio to fluctuate in the sense that hocus has mentioned. You can allow it do go down for a year or two just so long as you are certain that its value will be restored.

Does Chips really mean that his portfolio has no growth possibilities ever? I don't think so. That is why I think that he has misunderstood my words. I think that he means that if his portfolio takes a big hit, then he is unable to recover.

Have fun.

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

Post by JWR1945 »

I have finally figured out how William Bernstein came up with his numbers. (Reference: the Four Pillars of Investing by William Bernstein.) He relies entirely on the Gordon Equation and Monte Carlo simulations. He validates his Monte Carlo simulations by using the results from the Trinity Study and other studies that use the historical sequence methodology.

He causes a whole lot of confusion because he likes to come up with plausible rules of thumb. He throws around percentages all over the place. There are some logical errors hidden within the details behind these plausibility arguments.

Now let us take advantage of some quotes that hocus has extracted.
Page 56: "Going forward, it looks like stock and bond returns should be in the 6 percent range, not the 10 percent historical reward. Don't shoot me, I'm only the messenger. [Comment: This translates into an SWR of 2 percent]....

"On an intellectual level, most investors have no trouble understanding the notion that high past returns result in high prices, which, in turn, result in lower future returns. But at the same time, most investors find this almost impossible to accept on an emotional level.....

Page 57: "At the end of the day, the Fisher DDM method of discounting interest streams is the only proper way to estimate the value of stocks and bonds. Future long-term returns are quite accurately predicted by the Gordon Equation....

What Bernstein has done is this: He used the Gordon Model to estimate the long-term growth rate of stocks at recent valuations. His estimate was 3.5% real growth. Then he used his Monte Carlo simulation and calculated a Safe Withdrawal Rate. The answer was 2%.

raddr makes similar calculations assuming 3.5%. I suspect that his Monte Carlo simulator is much more accurate than William Bernstein's.

I (JWR1945) believe that there needs to be an adjustment to the Gordon Equation/Dividend Discount Model. It is because payout ratios are much lower now than they were previously. See the following thread (by JWR1945) dated Tue Feb 11, 2003 at 10:27 pm CST about stock market return projections.
http://nofeeboards.com/boards/viewtopic.php?t=494

With the help of BenSolar and others, my final numbers are in the neighborhood of 4.7% for the real return of the stock market for the next decade or two.

We now have three numbers. The lowest is 3.5% and it is favored by William Bernstein and possibly raddr. The intermediate projection is 4.7% that I favor and possibly BenSolar. The long-term historical real rate of return is 6.5% to 7%.

If I were to use Bernstein's approach, I would have come up with a larger estimate of the Safe Withdrawal Rate for 2000 than he did. If would have been lower than that of the historical sequence method because I would have projected lower real stock market returns. I actually used a different approach. (I scaled the answer back into the historical range of valuations.) My estimate was 2.3%.

Have fun.

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

Post by ataloss »

Bob pointed out to me that real returns of 7% gave a swr of 4% so using my estimate of future returns of the s&p 500 of 3.5% he gets a swr of 2%

3.5 *4/7 = 2%

Bob doesn't have access to MC simulation but he has a slide rule.

Bob asked me why we would limit our investments to US equities if we thought the returns would be so low ( vs international, value, emerging markets, junk)

Of course Bob pointed out that things didn't look so good at other points in history but a 4% withdrawal rate worked out fine.
Have fun.

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

Post by raddr »

hocus wrote:

That depends on how the term "safe withdrawal rate" is defined. I am saying that it must be defined as the product of mathematical calculations performed on relevant data. Define it that way, and people will not be able to use SWR analysis to say things that are not so.

If you are excessively loose in the definition, then you are right that people will be able to use SWR analysis to "prove" any personal littlen point that they want to prove. Some will say that the SWR is 2 percent, some will say 4, some will say 8, some will say a negative 6.

It is my view that SWR analysis loses its appeal when it becomes a completely subjective exercise. The apppeal of the tool to me is that it cuts through the subjectivity and tells you what the data says.


hocus,

What is the "tool" that you refer to? Does it give you a SWR number? With confidence levels?

Please step us through the calculations. I think that it would be very interesting to most of us.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

ataloss
Bob asked me why we would limit our investments to US equities if we thought the returns would be so low ( vs international, value, emerging markets, junk)

Studies based on historical sequences have used S&P 500 index values (and estimates that fill in the early years) because that is all that is available.

Monte Carlo simulations permit you to introduce additional asset classes. You have to be able to provide your own statistical information. Monte Carlo models are usually validated by comparisons with historical sequence calculations. Validation capabilities are quite limited except for the S&P 500.

Have fun.

John R.

P.S. Have Bob look into volatility (variances and covariances). They don't scale. If they were linear, the formulas would be simple. Safe Withdrawal Rate calculations do not scale linearly as a general rule.
P.P.S. Why volatilities? That is what kills retirement portfolios. You sell too many shares in a down market. If you had less volatility, you could have better safety even if your growth rate were lower. It would be nice to know that kind of trade-off. If linearity held, we would have the answers.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

What is the "tool" that you refer to?

My reference to the "tool" was just a reference to SWR analysis in general. I haven't done any calculations. I was just trying to say that I like using SWR analysis as a tool for assessing risk because it provides an objective data-based approach to the question.
raddr
*** Veteran
Posts: 265
Joined: Mon Nov 25, 2002 3:25 am
Contact:

Post by raddr »

hocus wrote: I like using SWR analysis as a tool for assessing risk because it provides an objective data-based approach to the question.


I still don't understand. If the "tool" is an objective function then you must have some specific calculations that spit out a number or numbers. Please share them with us. Inquiring minds want to know. :wink: If you don't have a numerical model that you can describe to us then I don't see how your "tool" is any more objective than all of the other (subjective) opinions out there.
therealchips
*** Veteran
Posts: 174
Joined: Sat Jan 04, 2003 4:00 am
Location: Henderson, Nevada, USA

Post by therealchips »

John R. said
I think that Chips has misunderstood me.

John, you are right. There is no problem in applying the constraints you specified in retrospective analysis of what the market actually did. Thanks for clarifying the matter. I withdraw my objection. I read your post too quickly. :roll:
He who has lived obscurely and quietly has lived well. [Latin: Bene qui latuit, bene vixit.]

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

Post by hocus »

I still don't understand. If the "tool" is an objective function then you must have some specific calculations that spit out a number or numbers.

I am saying that the tool should be set up so that it provides objective results. I am saying that to define what an SWR is so loosely that it can produce anything under the sun as a result is not a good idea because the tool becomes just one more way to engage in subjective evaluations of risk.

If you don't have a numerical model that you can describe to us then I don't see how your "tool" is any more objective than all of the other (subjective) opinions out there.

Prior to The Great Debate, SWR analysis had always been viewed as an objective tool when it was discussed at the Motley Fool board. Prior to May 13, 2002, there were never suggestions put forward that you could define an SWR to be anything you wanted it to be. That idea is something new.

I think that the old understandong of the purpose of SWR analysis was the way to go. I think that the tool should be an objective one. The old approach does not adequately account for changes in valuation levels, so it is not a valid answer to the question that an SWR analysis is intended to answer, in my view. But the old approach was intended to be objective at the time it was developed; there was no deliberate decision made to leave out data that was known to bear on the answer of the question being explored. I suggest that we follow the same approach in redefining the concept to take account of what we have learned about valuation.

I have not proposed any particular method of incorporating valuation into the analysis. I am open to different approaches. But I believe that we should start from the premise that a valid analysis must include all factors bearing on the question of "what is the safe withdrawal rate?" It is not proper to deliberately leave out any factor that we know for certain affects the answer to that question, in my view.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

I am saying that the tool should be set up so that it provides objective results. I am saying that to define what an SWR is so loosely that it can produce anything under the sun as a result is not a good idea because the tool becomes just one more way to engage in subjective evaluations of risk.

I look forward to existence of a tool with these characteristics. :wink:
Have fun.

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

Post by JWR1945 »

raddr
I still don't understand. If the "tool" is an objective function then you must have some specific calculations that spit out a number or numbers. Please share them with us. Inquiring minds want to know. If you don't have a numerical model that you can describe to us then I don't see how your "tool" is any more objective than all of the other (subjective) opinions out there.

Please continue on to the Part 2 thread. I have put up Revision 2 and Alternative A versions of a definition. (I am favoring Alternative A.)

Both definitions require that the SWR be the numerical result of a mathematical calculation. I think that we have a general agreement on that point. (That point was one that hocus advocated strongly.)

The next issue is the matter of which calculation to select. I ended up permitting several different answers in the context of what you consider to be a best estimate. You have to be very specific and explain what you mean by the term best estimate. I have introduced some other restrictions. One of them is that the Safe Withdrawal Rate cannot be defined as being the result of any specific method or model.

(You have been spared seeing posts...[paraphrased here for dramatic effect]...that hocus is wrong because the SWR is defined to be the result of [a particular] historical sequence calculation. Since hocus thinks that the SWR might be some other number, he is stupid and should shut up.)

The reason that I allow for multiple answers is a matter of context. Some methods are very well understood and their strengths and limitations can be identified in a fair amount of detail. Some other methods are also well understood but their strengths and weaknesses are different. Other methods might include factors that make them seem to be superior, but we really don't know enough to know for sure.

Which answer is best depends upon the needs of the individual. I expect most individuals to benefit from using several answers, each for a different purpose.

Have fun.

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

Post by ataloss »

The reason that I allow for multiple answers is a matter of context. Some methods are very well understood and their strengths and limitations can be identified in a fair amount of detail. Some other methods are also well understood but their strengths and weaknesses are different. Other methods might include factors that make them seem to be superior, but we really don't know enough to know for sure.

Which answer is best depends upon the needs of the individual. I expect most individuals to benefit from using several answers, each for a different purpose.

Since we won't be able to identify the "best swr estimate" in terms of the one closest to the maximum withdrawal rate I think you need to state your assumptions, present your method and let people decide. Afterall isn't one fire guru being blamed for being overly positive about a single method?
Have fun.

Ataloss
Post Reply