A Sidebar to the Bernstein's 2% Thread

Research on Safe Withdrawal Rates

Moderator: hocus2004

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

A Sidebar to the Bernstein's 2% Thread

Post by JWR1945 »

This is a side benefit from hocus2004's questions on the Bernstein 2% thread. These observations have come up as I have started to answer his questions.

A Sidebar to the Bernstein's 2% Thread

Here is another way to look at what William Bernstein's 3.5% prediction of future stock market returns tells us about Safe Withdrawal Rates. Our New Tool strips valuations away from the actual calculations and works directly with the real, annualized total return (return0) of a portfolio.

From A New Tool Application: Summary dated Wed, Jun 02, 2004.
http://nofeeboards.com/boards/viewtopic.php?t=2548
New Tool errata dated Wed, Jun 23, 2004.
http://nofeeboards.com/boards/viewtopic.php?t=2655

Applying the equations to HDBR50:
If the real, annualized return0 is 3.5%, the 14-year Calculated Rate is 5.4437279% or 5.44%. The confidence limits are plus and minus 0.64%. The Safe Withdrawal Rate by this method is 4.80%.

Applying the equations to HDBR80:
Using 3.5% for the 14-year real, annualized return, the Calculated Rate is 5.64%. The confidence limits are plus and minus 1.02%. [This has been corrected from the originally reported value of 1.10%.] The Safe Withdrawal Rate by this method is 4.62%.

Calculations using 6-year and 10-year Calculated Rates are similar. Their confidence limits are wider.

We get a much different story when we look at valuations.

One thing that comes to mind is that we do not know the time period that the 3.5% return is supposed to represent. It is supposed to be a long-term number, but exactly how many years that requires is not mentioned.

Another thing that comes to mind is that we have no confidence limits associated with the 3.5% predicted stock market return.

But something else is far more important.

This tells us something about the Historical Sequence method as compared to the Monte Carlo method. The Historical Sequence method produces the kind of earnings yield behavior that we have reported in a natural manner. It is just a matter of looking at historical valuations. The Monte Carlo approach does not. It behaves similar to the New Tool. It depends upon the real, annualized total return (and the standard deviation), but valuations are a separate matter. Once a total return number is identified, it produces the same result regardless of the initial level of valuation.

We can tie in earnings yield only indirectly by placing an upper limit on the total return.

The New Tool and the Monte Carlo approaches accept an assumption that we have reached a permanent, higher level (or plateau) of valuation. To them, it makes no difference. The Historical Sequence method rejects such a notion since it differs radically from the historical record.

Have fun.

John R.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

"The Historical Sequence method produces the kind of earnings yield behavior that we have reported in a natural manner. It is just a matter of looking at historical valuations. The Monte Carlo approach does not."

This post is a bit over my head.

But the point you are making brings to mind the argument that SalaryGuru often puts forward that the various factors that influence SWRs are inter-connected and that to examine them separately renders the results "artificial."

I am largely in sympathy with the SalaryGuru claim. Where we part company is that I believe that valuation is a different sort of factor altogether from the other factors that are affecting the result. I go along with the William Bernstein viewpoint that there are a bunch of factors that go under the heading of "Factors Affecting Future Risk" and then there is a second cateogory with just one factor in it that goes under the heading of "Expected Return."

There is a strange way in which intercst, SalaryGuru, and me are on the same side re this thing. Lots of others are taking the view that SWR analysis doesn't really tell you much, that it is of little consequence. Intercst, SalaryGuru, and me are not taking that view. My differences with intercst and SalaryGuru are just that they say that it is reasonable to make zero adjustment for changes in valuation (as if it were just one more Future Risk factor) while I see it as mandatory to make a separate assessment of Expected Return.
unclemick
*** Veteran
Posts: 231
Joined: Sat Jun 12, 2004 4:00 am
Location: LA till Katrina, now MO

Post by unclemick »

Well I'm back in the bleachers with nay sayers again (Curmudgeon in training). I like the segment analysis where 30-40 yrs is broken into shorter chunks to analyize what is going on and being preprejudiced - the looks at dividend and capital gain fractions - a move in the right direction.

You may end up with a box of tools to deal with returns sequence, valuations, and inflation consequences.

So here's my Bronx cheer - SWR sucks - Yea dividends and age related pie charts.

In a left handed way - that should inspire you to soldier on.

Heh, heh heh - here's a sneaky - ??? ever wonder what the SEC yield was on some of Bernstein's portfolios???
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

I've reconciled this problem in my own mind.

William Bernstein's 3.5% real return estimate excludes the speculative component that John Bogle mentions. The Gordon Equation is dividend yield + dividend growth.

Bogle prefers dividend yield + earnings growth + the speculative component.

Bernstein acknowledges that there can be multiple compression (in the price to earnings ratio) for shorter time periods such as two decades. He even states that this could result in flat earnings: that is, with zero percent real growth over twenty years.

There is a flaw that underlies this thinking. If the Gordon Equation is supposed to be for the very long-term, what happens at 50 or 60 years or even longer? The influence of any speculative term will necessarily diminish. [The annualized return depends upon the Nth root of a gain multiplier, where N is the number of years and the gain multiplier is the final P/E divided by the initial P/E. When N becomes very large, the Nth root gets closer and closer to 1.0.] This brings about an internal contradiction: history shows that the long-term return is around 6.5% to 7.0%. Bernstein's application of the Gordon equation implies that it must be 3.5%.

I know that, when I make projections, I include the speculative term. That is why my own calculations do not show high SWR estimates when I use the New Tool looking forward. I use 0% as my high end estimate for the next decade (ouch?) and minus 3% (or even worse) as my lower estimate.

Looked at from this vantage point, does it really make sense to ignore the speculative adjustment and expect stocks to return 3.5% per year above inflation for the next decade? If P/E10 were to drop in half, we would have to include a speculative adjustment minus 6.7%. If the underlying return without the speculative adjustment were 3.5%, the expected return over the next decade should be minus 3.2%. That's right: there would be an annualized loss of 3.2% per year in real dollars over the next decade.

My own calculations with the New Tool have assumed annualized real returns in the neighborhood of minus 3% and they are consistent with what earnings yield calculations predict.

Have fun.

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

Post by JWR1945 »

unclemick wrote:Well I'm back in the bleachers with nay sayers again (Curmudgeon in training). I like the segment analysis where 30-40 yrs is broken into shorter chunks to analyze what is going on and being prejudiced - the looks at dividend and capital gain fractions - a move in the right direction.
You may end up with a box of tools to deal with returns sequence, valuations, and inflation consequences.

These are the kinds of things that SWR analysis investigates. Or should investigate.

You have been unduly influenced by those who take a very narrow view of Safe Withdrawal Rate investigations. Too many people think in terms of one particular withdrawal strategy and with a final balance always equal to zero.

SWR research focuses on such things as monitoring portfolio safety, improving the flexibility of your decisions (by letting you know what to watch out for and what doesn't matter too much) and supplying reliable backup solutions if things don't work out as expected.

I have been wanting to look into strategies with ending balances other than zero for quite a while. I have been able to learn about the balances. But I have not been able to know what the withdrawal amounts would be (without a whole lot of effort)! Now I can produce tables that make it easy to examine portfolios whose withdrawal amounts and balances are allowed to decline but never below a lower bound (such as 50% or 75% of the portfolio's initial buying power).

I am interested in the influence of interest rates and dividend yields. For example, we have found out that there is no single number that should be used to characterize commercial paper's real interest rate. The commercial paper of the late 1800s behaved differently from that of the Great Depression, which behaved differently from that of the years of Stagflation. Similarly, the great dividend yields of long ago have been gone for decades. Still, the stability of dividend amounts is a key factor affecting the safety of retirement portfolios. I suspect that dividend based strategies can be superior during retirement. I am not at all convinced that Dividend Achievers are (always and necessarily) poor investments. I have noticed some attractive companies in Mergent's book (formerly collected by Moody's). I suspect that the combination of price discipline (buy them when they are cheap, when companies screw up badly and when they are being punished by Wall Street) and quality (as reflected in their dividend history) is a winning combination.

In any event, I consider all of this to fall into the category of SWR Research. The key is that we insist on using a high level of safety when we make our calculations. [What you do with our numbers is a different story. You can be an optimist or a pessimist when it comes to applications. It is up to you, your own temperament and your particular investments.]

Have fun.

John R.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

"These are the kinds of things that SWR analysis investigates. Or should investigate. "

I strongly agree with this statement. There are scores of things that can be done with SWR analysis that have never been done before. The tool is far more powerful than most people realize.

The reason why the potential of the tool has not yet been explored is that the pioneers made the mistake of not including an adjustment for valuation. They achieved a breakthrough in creating the conventional methodology, so they should not be faulted much for making this mistake. But it is a objectively established fact that they made a serious analytical mistake.

The result of that mistake is that the conventional methodology pumps out all sorts of crazy numbers. If you believed that the conventional methodology accurately reported what the historical data says, you would agree with intercst that anyone who does not always have a 74 percent stock allocation is investing in a non-"optimal" way. That obviously is a nutso thing to believe.

The conventional methodology results defy common sense. So the conventional methodology needs to be retired. The sooner that happens the better for all investors seeking to take advantage of the insights offered by looking in an informed way at the historical data. Once you begin calculating SWRs in an informed way, all sorts of powerful insights are opened up to you.

I've known all this since the mid-90s and I can personally vouch for the power of the tool. I say again that we all should be deeply grateful that we have JWR1945 willing to do all the work that is required in looking at the numbers and seeing what they really say. Prior to May 13, 2002, it had been a source of frustration for me that I do not possess the skill set necessary to do that.

I think we are getting to a point now where at least a few others are beginning to see what I have been seeing since 1996. The SWR tool is wonderful, the conventional methodology is holding us back. The people who developed the conventional methodology would not want that to happen. They developed the methodology to advance knowledge of what the historical data says, not to keep it confined to what people thought they knew in the summer of 1999.

At some later point in the discussions, I hope to be inviting some of the pioneers to this board. My guess is that they will be more thrilled than anyone else to see what JWR1945 has come up with already and what JWR1945 and others will be coming up with in coming days.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

"Looked at from this vantage point, does it really make sense to ignore the speculative adjustment and expect stocks to return 3.5% per year above inflation for the next decade?"

I don't think that Bernstein is saying that the return will be 3.5 percent per year above inflation for the next decade. He says that the Gordon Equation only works in predicting long-term returns.

"This brings about an internal contradiction: history shows that the long-term return is around 6.5% to 7.0%. Bernstein's application of the Gordon equation implies that it must be 3.5%. "

I'm certainly no expert on the Gordon Equation. But it seems to me that the resolution of this contradiction is that the return from the date at which Bernstein did his calculations will be lower than the long-term return measured from other starting points.
Mike
*** Veteran
Posts: 278
Joined: Sun Jul 06, 2003 4:00 am

Post by Mike »

That's right: there would be an annualized loss of 3.2% per year in real dollars over the next decade.
The average employee is still faced with the grim choice in his 401k of fixed income that is virtually guaranteed not to provide a positive real return, or equities. They may soon be faced with this same choice in private Social Security accounts, which could pump another 7 trillion into equities in the next couple of decades. Congress seems to have decided that they should tell people that after tax S&P dividends are capable of supporting a majority of the population, despite the math. If the Republican SS plan passes, the S&P dividend yield could go lower, much lower.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

More about A Sidebar to the Bernstein's 2% Thread

I have found the underlying problem with using the Gordon Model for predicting stock market returns. It is based upon present value calculations. No dividends are ever reinvested. There is no compounding.

If you look at the long-term return of the stock market, the numbers that we have are based upon reinvested dividends. This is what is predictable.

This discrepancy helps to explain why one often reads ambiguous words related to average or typical dividend yields and growth rates. The predictability in the stock market is with dividends reinvested.

When you read that the stock market is expected to have a real return going forward of 3.5%, that number is similar to a bond yield. Reinvestment is an issue. The rate of return of the reinvested dividends is not steady.

We can correct our calculations by focusing on the dividend amount. If priced fairly, stocks (overall) will return their predictable 6.5% to 7.0% when reinvested. We can make good predictions even as short as 10 or 20 years, but only if we first scale our prices to match the levels in the long-term history of the market.

Mathematically, the Gordon Model looks like this:
Return = Dividend Yield + Dividend Growth
That is deceptive because growth is really a percentage. The underlying equation has the form
Return = [Dividend Amount/Price]*G, where G is a growth multiplier.
We must have a return close to 6.5% to 7.0% to make projections. The scale factor is [Fair Price/Price] = [Return/(6.5% to 7.0%)].
If we are currently projecting a 3.5% real return based on the Gordon Model, then the Fair Price is roughly one half of the current Price.
Over N years, (1+return adjustment)^N = [Fair Price/Price].
Our projections must include this term. Otherwise, dividends are being reinvested at varying rates. That violates the (simple) mathematical assumptions that underlie the Gordon Model.

We may need a second adjustment at the end of a projection. The original adjustment does not include valuations directly. Rather, it aligns the Gordon Model to match what is known to be predictable about stock returns (or, at least, predictable in the long-term).

Have fun.

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

Post by JWR1945 »

Mike wrote:The average employee is still faced with the grim choice in his 401k of fixed income that is virtually guaranteed not to provide a positive real return, or equities.
I think that this is only a temporary situation.

I agree that things are bad for investors right now. I expect real long-term returns to be under 6.5% to 7.0% for a while because of competitive buying by the baby boomers. We could have an extended period of time with higher than normal price to earnings ratios or, even worse, a sudden fall to very low prices (at which time most people would be afraid to buy any stocks).

I think that the key is that the Federal Reserve is getting away from its extremely low short-term interest rate policy. I expect short-term rates to rise slowly but steadily in a manner typical for Alan Greenspan.

People desperate for yield have been forced to take on more risk, sometimes much too much risk.

I am not sure that investments will become attractive anytime soon. I am confident that they will become less unattractive.

Have fun.

John R.
Mike
*** Veteran
Posts: 278
Joined: Sun Jul 06, 2003 4:00 am

Post by Mike »

I think that this is only a temporary situation.
To a degree this is likely, but there are structural changes in the economy as a whole. Fixed income will likely become less unattractive than it is now, but it is not likely to return a significant positive real return over the long run. The constant inflation policy has destroyed it as an asset class, and taxing interest that does not even keep up with inflation has magnified the problem. Equity yield may increase once the boomers pass their prime, but placiing 7 trillion of Social Security money into the equities market could significantly delay this. Boomer aging may also coincide with less boomer spending, leading to declining corporate profits for a while. Longer term, the population is not likely to grow as fast during the 21st century as it did during the 20th century. Robust population growth enhanced earnings growth during the 20th century. Alternative investments may need to be considered, at least for part of the portfolio.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

Here is a recap about return predictions based upon the Gordon Model. I have included a couple of corrections.

1) The Gordon Model calculates a discount rate R when dividends grow steadily.
2) This discount rate R is calculated as R = dividend yield + dividend growth rate.
3) This discount rate tells us about the present value of the income stream absent any reinvestment of dividends. It is similar to bond coupons.
4) This discount rate equals (the predicted long-term) stock market return if and only if suitable reinvestments can be found. That is, R = the long-term stock market return (with dividends reinvested) if and only if dividends can be reinvested at the same discount rate.
5) The annualized, real, long-term stock market return with dividends reinvested has been amazingly consistent, falling within the range of 6.5% to 7.0%.
6) As long as the discount rate R calculated from the Gordon Model falls between 6.5% and 7.0%, stocks are fairly priced and R = the stock market return.
7) When R is less than 6.5%, stocks are overpriced and suitable reinvestments will not be found. Prices will fall relative to dividends until the discount rate R from Gordon Model (eventually) lies between 6.5% and 7.0%. That is, prices will not grow fast enough for dividend yields to stay the same. Instead, dividend yields will increase until R, which is the sum of the dividend yield and the dividend growth rate, lies between 6.5% and 7.0%.
8 ) When R is greater than 7.0%, stocks are underpriced and suitable reinvestments will not be found. Prices will increase relative to dividends until the discount rate R from the Gordon Model (eventually) lies between 6.5% and 7.0%. That is, prices will grow so fast that dividend yields will not remain the same. Instead, dividend yields will decrease until R, which is the sum of the dividend yield and the dividend growth rate, lies between 6.5% and 7.0%.
9) In all cases the investments available for purchases made from dividends will change until the Gordon Model discount rate falls within the range of long-term stock market returns, between 6.5% and 7.0%.
10) Dividends themselves are well behaved. The sum of a dividend amount and its growth rate is stable enough for us to draw useful conclusions based on the Gordon Model.
11) We can take advantage of the stable, known, long-term, real, annualized return of the stock market and apply it to the dividends themselves. This leads us to scale prices so that the discount rate from the Gordon Model falls between 6.5% and 7.0%.
12) We calculate a Fair Price when we scale the discount rate: [Fair Price/Current Price] = [the discount rate R calculated from the Gordon Model] / [6.5% to 7.0%].
13) Notice that the dividend yield (and, therefore, the Price to Dividend ratio) is not constrained. Instead, the constraint applies to the sum of the dividend yield and its growth rate. This means that the Fair Price adjustment is affected by the growth rate.
14) For the mathematically inclined (who have Gummy's CD), using R for the discount rate and g for the dividend growth rate, the Gordon Model can be written as (R-g) = the dividend yield. Notice that the growth rate term g is not the growth multiplier G = (1+g). When we scale to calculate the Fair Price, a more accurate equation would be [Fair Price/Current Price] = [the dividend yield at Current Prices / dividend yield at a Fair Price] = [R-g] / [6.5% to 7.0% - g] = [the dividend yield] / [6.5% to 7.0% - the dividend growth rate].

Here are a couple of observations.
1) The Gordon Model is helpful for SWR investigations because it focuses on what retirees need: income streams.
2) Applying the Gordon Model directly to come up with a stock market total return input to Monte Carlo models is wrong when calculating Safe Withdrawal Rates. This includes William Bernstein's applications.
3) The effect of errors (related to stock market returns) in calculating Safe Withdrawal Rates diminishes greatly soon after the first decade.
4) The error caused by using the Gordon Model is mitigated because it is the income stream during the first few years of retirement that influences the Safe Withdrawal Rate most heavily.

Have fun.

John R.
unclemick
*** Veteran
Posts: 231
Joined: Sat Jun 12, 2004 4:00 am
Location: LA till Katrina, now MO

Post by unclemick »

Yeah! Good post - Income streams ala dividends plus div. growth. Of course - the other struggle is to understand active rebalancing and tease out the part cap gains (P/E swings) plays in simple(Bogle) and complex(Bernstein slice and dice asset classes with turnover at the edges).

Again given 'my SWR' probably doesn't anybody else's - I've come to point of dropping the term.

NOW - a calculated take out (HSWR or whatever label) constrained within a good data set can provide powerful insight as to HOW TO CONSTRUCT INCOME STREAMS for a contunued happy ER. Put me down as a short end of the stick guy who keeps a $ bill on page 108 of Ben Graham's 4th ed when Mr Market roils portfolio values.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

"Good post - Income streams ala dividends plus div. growth...."

SWR analysis is income stream analysis. I have three income streams: (1) the income stream from the writing work I do; (2) the income stream from my non-volatile investment classes (TIPS and such); and (3) the income stream from any volatile investment classes (Stocks and such) that I invest in (this is a non-existing income stream today but one that I will be building in the future, when the cost attached to building it is more appealing). If I ultimately collect on Social Security (I expect that I will collect something), that will be a fourth income stream.

The purpose of SWR analysis is to "translate" the income stream received from volatile investment classes into terms that make those income streams comparable to the income stream generated by non-volatile investment classes. You have a government guarantee on your TIPS income stream. SWR analysis is providing you a historical data guarantee on your stocks income stream.

The reality is that the income stream from stocks may end up being a good bit bigger than what the SWR suggests. But it is impossible to put together a plan without having a number to plug in for your investment stream from stocks. You can't plug in a range of numbers. You need to plug in one number. You need one number. The SWR is the number that is comparable to the easy-to-compute number for TIPS. SWR analysis tells you the number that you can plug in for stocks that will work if the worst of all returns sequences that have turned up in the past (but nothing worse than that) happpens to turn up in your particular retirement.

This is a valuable number to know. You can't put a plan together without plugging in some number. And where else are you going to get the number from if not from the historical data? You could put numbers in a hat and pull one out. That approach might work from time to time. But I have not been able to come up with any better approach to coming up with a plug-in number than looking at what the historical data says.

To employ SWR analysis, you need to be willing to make an assumption that stocks will perform in the future somwhat in the way that they have performed in the past. If you are OK with that assumption, SWR analysis is a big help. It allows you to make an informed assessment of what sort of income stream you are likely to obtain from stocks.

"given 'my SWR' probably doesn't [equal] anybody else's"

The SWR is determined by looking at historical data. Saying that the SWR that applies for you is not the same as the SWR that applies for others is like saying that two plus two equals four for some of us and two plus two equals six for others. The SWR is a mathematical construct. The calculation of the number is an objective enterprise. If someone is not reporting what the historical data actually says, he is not involved in legitimate SWR research.

What you mean to say here is that your Personal Withdrawal Rate is not the same as other Personal Withdrawal Rates. [I am putting to one side for the time being the issue that I raised in "The Coin Toss" post, the fact that personal differences between various investors' personal circumstances can properly be included in SWR analysis in cases in which there is an objective means of determining the effect of the factor being taken into consideration.]

If you are more risk averse than another investor, you need to take a lower Personal Withdrawal Rate than that investor. That is of course so. But that reality does not transform the calculation of SWRs into a subjective enterprise. The SWR is whatever the historical data says it is.

The way that you adjust for personal circumstances is to make adjustments to the objectively determined SWR in deciding on your Personal Withdrawal Rate. One investor might take the SWR as his PWR. Another might take SWR Plus One Percent. Yet another might take SWR Minus One Percent. All sorts of variations are possible.

If you don't like using the take-out number that is 95 percent safe according to the historical data (that is the SWR under the JWR1945 methodology), you could use the number that his methodology identifies as 85 percent safe or 75 percent safe. These other numbers are not techncially SWRS as that term is defined in the studies, but they are the product of SWR analysis. There is no rule that says that anyone must start their analysis with use of the 95 percent safe number. We use that starting point in most discussions just because that is the number that is comparable to the number generated by the conventional SWR methodology (the number that is purported to be "100 percent safe").

"I've come to point of dropping the term. "

I see no purpose being served by dropping use of the term "Safe Withdrawal Rate." SWR analysis has a rich history. People have discussed how to use the concept for years, and there are insights to be harvested by making reference to those discussions. If you change the terminology, you lose access to those insights. It is confusing to change the words by which we refer to things. Aspiring early retirees have long made use of SWR analysis in putting together their plans. I see no reason to ask them to stop doing so.

Also, changing the label put to the concept would not change the practical reality. People need to know what the historical data says re what take-out number is safe to craft their plans. You can call this number "The X Factor" and you still have the same questions that have dominated the SWR discussions to contend with.

How is The X Factor to be calculated? Are we going to look at all of the historical data that comes into play in an informed calculation of The X Factor or only some of it? Are we going to look only at the historical data that deals with the calculation of the "Future Risk" factor or are we also going to take into consideration the historical data that deals with the "Expected Return" factor?

You are going to get different answers if you do it different ways. If William Bernstein is correct that changes in valuation levels affect long-term returns as a matter of "mathematical certainty," then there is no way to determine The X Factor accurately without taking changes in valuation levels into account.

So you circle back to the same old problem. You have some people saying that The X Factor is 4 percent and some people saying that the X Factor is 2 percent. Changing the labels that you put to things doesn't change the underlying realities. It is the underlying realtiies that matter. There is a reason why aspiring early retirees have long wanted to know what the SWR is and changing our name for the SWR concept will not change the reality that people will not be able to put together sound plans unless we report accurately what the historical data says on the question being examined.

"a calculated take out (HSWR or whatever label) constrained within a good data set can provide powerful insight as to HOW TO CONSTRUCT INCOME STREAMS for a contunued happy ER."

The historical surviving withdrawal rate (HSWR) does not provide powerful insight for the purpose of putting together a plan. The number you need for putting together a plan is the SWR.

The HSWR concept does have value. Determining the HSWR is an important step in the process of determining the SWR. But it is dangerous for anyone to use the HSWR as the number informing them as to the likely size of the income stream they will receive from stocks. The HSWR always gives the wrong answer to this question. Why would anyone want to use a number that always is wrong? What constructive purpose is served by doing that?

More than that, the HSWR is always wrong by unpredictable amounts. Sometimes it is a little wrong on the low side, sometimes a little wrong on the high side. Sometimes it is a lot wrong on the low side, sometimes a lot wrong on the high side. Without taking valuation into account, the aspiring early retiree has no way of knowing in what direction the HSWR is wrong or by how much. If the aspiring early retiree takes valuation into account so that he has a rough idea of how much the HSWR is wrong by, he is calculating an SWR rather than a HSWR. The difference between the two concepts is that calculation of the SWR requires consideration of all of the factors with a critical bearing on the question of what is safe while the HSWR does not.

Doesn't it make more sense to just tell the aspiring early retiree the number he needs to know to put together his plan? The number he needs to know is the SWR, not the HSWR. There might be some researchers that for various purposes would want to know the HSWR. But what direct practical value does this number have to someone putting together a Retire Early plan? The only thing he knows for sure about this number without making an adjustment for valuation is that it is certaintly not the SWR, and the SWR (or some variation of it for which a different safety percentage applies) is the number he needs to know to construct an effective plan.

"Put me down as a short end of the stick guy who keeps a $ bill on page 108 of Ben Graham's 4th ed when Mr Market roils portfolio values."


I am not sure what this reference is to. My guess is that it might be a reference to the advice he gives to take one's stock allocation as high as 80 percent at times of low valuation and as low as 20 percent at times of high valuation.

I agree with the idea of changing one's allocation as the intrinsic value of a stock purchases waxes and wanes. I think that a good argument can be made for using a percentage other than 80 and 20. My personal view is that the average investor might want to move his allocation only as low as 35 percent and only as high as 65 percent, depending on valuation. But there are circumstances that call for exceptions to the general rule. I need to be completely out of stocks at today's valuations because of my particular circumstances, and there are others for whom a 50 percent stock allocation makes sense even at these valuation levels.

I do think there is value in pointing to the 80-20 rule of thumb, though. That rule of thumb shows that there was a time when many informed stock investors understood that changes in valuation levels affect the value proposition offered by stocks, and suggests that informed investors of today should be taking the changes in the value proposition being offered into account in determining their strategies.

Bernstein says that it is a repeating phenomenon for people to get caught up in the emotion of bull markets and to throw the time-tested analytical approaches to the winds at times when emotion-based analytical approaches come to dominate reason-based ones. The other side of the coin is that, when stock prices go down to levels where the value proposition for stock purchases is very high, many will be saying that middle-class investors should not be investing in stocks at all.

The benefit of the Data-Based SWR Tool is that it provides the aspiring early retiree a means of cutting through all the emotion-laced rhetoric that is put forward both in Bull markets and Bear markets and allows him to simply assess the historical facts. It is a fact that the data says that the SWR today for an 80 percent stock portfolio s roughly 2.5 percent. That's not a Bull claim nor a Bear claim. It is an historical data claim.

The value of the tool is that it tells investors what the historical data says re the value proposition offered by stocks at a particular time. When you start mixing the personal viewpoints of the researcher into the mix, the number no longer serves this purpose. Leave out some of the data to push the number higher than the number you get from looking at the historical data, and you are getting the subjective viewpoint of a Bull reseacher. Leave out other data to pull the number lower than the number you get by looking at the historical data, and you are getting the subjective viewpoint of a Bear researcher.

There is lots of stuff out there giving the views of Bulls and Bears. My purpose in developing the Data-Based Tool was to provide objective insights, not more subjective Bull and Bear stuff. I see it as a critical first principle that we report what the data says, not what we would like it to say or wish it to say or at one point perhaps thought it would say. It is a fundamental principle of SWR analysis for me that the SWR is whatever the historical data says it is.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

"Put me down as a short end of the stick guy who keeps a $ bill on page 108 of Ben Graham's 4th ed when Mr Market roils portfolio values."
I am not sure what this reference is to. My guess is that it might be a reference to the advice he gives to take one's stock allocation as high as 80 percent at times of low valuation and as low as 20 percent at times of high valuation.

I have the Fourth Revised Edition of The Intelligent Investor by Benjamin Graham, which was the fifth version of his book. Page 108 is where he introduces us to Mr. Market.

Benjamin Graham recommends that you form your own estimate of what a company is worth instead relying on current price quotes. You may decide to sell to Mr. Market when he offers a ridiculously high price or buy when he offers a low price. But at all times, you use your own estimate of what a company is worth, not what Mr. Market has to say.

Have fun.

John R.
Post Reply