The Great SWR Investigation-Part 2

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 2

Post by JWR1945 »

The Great SWR Investigation-Part 2

I am introducing this new thread to address the definition of the term Safe Withdrawal Rate. The Part 1 thread has picked up a fascinating discussion that I want to see continue. I had originally thought that I would copy a couple of those posts to start a Part 2 thread and then continue to discuss definitions on the original thread. I did not act fast enough.

Let the Part 1 thread continue along its current lines of development. Let us look at definitions on this thread.

Have fun.

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

Post by JWR1945 »

The Great SWR Investigation-Part 2

This is where I left off on the Part 1 thread. I will also post an alternative definition that I think is better.

Definition

The Safe Withdrawal Rate is the best estimate of the maximum initial percentage of a retirement portfolio that, on an inflation-adjusted basis in future years, can be withdrawn without completely depleting the portfolio. It is mathematically calculated based primarily on existing, historical information. The sense in which it is the best estimate must always be identified. There must be an assessment of the degree of safety. This may be stated in terms of probabilities. The phrase Safe Withdrawal Rate 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.

The Safe Withdrawal Rate must be described in terms independent of any methodology. The Safe Withdrawal Rate must be based on calculations that are independent of the actual occurrence of future events. The Safe Withdrawal Rate must be described in terms that acknowledge our inability to conduct carefully controlled experiments.

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). Another example is the introduction of a new asset classes on a theoretical basis instead of a historical basis such as including Treasury Inflation Protected Securities (TIPS). Even though they have existed only recently, it is possible to define a meaningful Safe Withdrawal Rate calculation that includes TIPS. Another example would be to include the effects of a new law.

Have fun.

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

Post by JWR1945 »

Here is what I promised. I think that this is better.

Alternative Definition A. This is a brand new definition.

The Safe Withdrawal Rate is the best estimate of the maximum withdrawal amount for a retirement portfolio based on both a well-defined procedure and a set of well-defined constraints. It is stated as a percentage of the initial portfolio balance. It must be a mathematical calculation that is based primarily on existing, historical information. The sense in which it is the best estimate must be identified. The algorithm or procedure must be stated clearly. The constraints must be stated clearly. There must be an assessment of the degree of safety. This may be stated in terms of probabilities.

The phrase Safe Withdrawal Rate 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 context of a Safe Withdrawal Rate calculation must always address 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.

The Safe Withdrawal Rate must be described in terms independent of any methodology. The Safe Withdrawal Rate must be based on calculations that are independent of the actual occurrence of future events. The Safe Withdrawal Rate must be described in terms that acknowledge our inability to conduct carefully controlled experiments.

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). Another example is the introduction of a new asset class on a theoretical basis instead of a historical basis such as including Treasury Inflation Protected Securities (TIPS). Even though they have existed only recently, it is possible to define a meaningful Safe Withdrawal Rate calculation that includes TIPS. Another example would be to include the effects of a new law.

Several example of well-defined procedures include:
1) Withdrawing a fixed dollar amount every year of the portfolio's lifetime.
2) Withdrawing an initial amount and then increasing (or decreasing) that amount to adjust for inflation (or deflation).
3) Withdrawing a fixed percentage of the current balance of the retirement portfolio every year.
4) Withdrawing a minimal amount every year and taking out an additional amount that depends upon how much the portfolio size has increased. If the balance has decreased, the additional amount is zero. If the balance has increased greater than a specified amount (or percentage), take a specified percentage of the increase up to a maximum (amount or percentage).
5) Withdrawing an initial amount based on a specified percentage of the initial balance of the retirement portfolio. At the end of each year two amounts are calculated. One is to increase (or decrease) that amount to adjust for inflation (or deflation). The other is to use the originally specified percentage, but this time to apply it to the current portfolio balance. Withdraw whichever amount is larger.

Several examples of well-defined constraints include:
1) The portfolio balance must remain above zero for a specified number of years.
2) The portfolio balance must be greater than or equal to the original balance after a specified number of years.
3) The portfolio balance must be greater than or equal to the original balance as adjusted for inflation after a specified number of years.
4) The portfolio balance must remain between a minimum amount and a maximum amount for a specified number of years. All amounts are adjusted for inflation.

Have fun.

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

Post by ataloss »

Bob tells me that he likes the new definition. He is interested in how we will pick a favorite from competing "best estimates of swr" and how the safety factor will be expressed.
Have fun.

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

Post by JWR1945 »

ataloss
Bob tells me that he likes the new definition. He is interested in how we will pick a favorite from competing "best estimates of swr" and how the safety factor will be expressed.

My intention is that we do not pick a single favorite. My intention is that there be many calculations of Safe Withdrawal Rates, each of which applies to its own case. Then every individual chooses whatever comes closest to meeting his own, unique needs.

I also anticipate that there will be several competing calculations for the same SWR (i.e., same definition of what is best). Each calculation would have its good features and its limitations. We have that to some extent when we compare the historical sequence calculations to Monte Carlo calculations. Historical sequences are highly credible when they suggest that something could fail in the future...because something similar has failed in the past. But their statistics are lousy. Monte Carlo models have great statistics but some of their predictions correspond to things that have never happened previously. You hate to depend too much on those particular results.

A safety factor will be described in the best way that we know how. Sometimes we will know a lot. Sometimes we will know almost nothing. Over time, I suspect that we will learn more and more and add that to our description. That is the way it should be.

Valuations provide an example. It has been only recently that we have known the importance of valuations and how to quantify their effects. Even now, there is a lot of uncertainty in the details. We are in the process of adjusting our previous answers to today's conditions. We expect to improve our adjustments continually in the future. The description of the safety factor will continue to improve.

Have fun.

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

Post by hocus »

My intention is that we do not pick a single favorite. My intention is that there be many calculations of Safe Withdrawal Rates, each of which applies to its own case. Then every individual chooses whatever comes closest to meeting his own, unique needs.

I agree with this thought up to a point, but I think you are taking it too far. There are some questions on which our knowledge is foggy. In those cases, it makes sense to me to "let a thousand flowers bloom." Let people calculate the number in different ways so long as they are careful to note the limitations of the approach chosen,. and the users of the studies can decide which studies to make use of.

That said, there are some issues which are so clear that the realities do not permit for reasonable differennces of opinion. I am saying that the question of whether valuation levels affect SWRs or not is in this class. Bernstein is saying that as a matter of "mathematical certitude" this factor always makes a difference. So I cannot see a justification for ignoring it.

If we go down the road you are proposing, we will so cheapen the common understanding of what SWR analysis is that we will do damage to the perceived usefulness of the tool. Using your definition, someone who is a follower of Warren Buffett could declare that "using the Buffett investing techniques, I am sure that I can attain an average return of 15 percent" and construct his SWR analysis to refflect this presumption.

But does an analysis crafted along these lines really tell you what is safe? I say no. There is no harm in doing the analysis, but it should not be called SWR analysis. SWR analysis should always be aimed at telling you what is safe, according to what the data says. The data says that most people are not capable of investing like Warren Buffet, so a study rooted in the presumption that one can invest like Buffet violates the premise of the SWR calculation project.

It is no more reasonable to presume that the valuation level in effect at the time of your retirement will not affect your long-term return than it is to presume that you can invest like Warren Buffet. These are far-fetched assumptions. To say that studies rooted in these studies are legitimate undermines confidence in the tool.

We can't pass a law to stop people from preparing such studies, but we should not encourage it in any way. Such studies do not help people learn what they need to learn to put together effective FIRE plans.
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

Bob and I are entirely satisfied with the current definition.
Have fun.

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

Post by JWR1945 »

hocus
It is no more reasonable to presume that the valuation level in effect at the time of your retirement will not affect your long-term return than it is to presume that you can invest like Warren Buffet. These are far-fetched assumptions. To say that studies rooted in these studies are legitimate undermines confidence in the tool.

Let me make this clear. All other things being equal, the best calculation always takes valuation levels into account.

The reason that I permit additional answers to be best in some other sense is that the context associated with each calculation is different. We know a lot about those calculations which have not included valuations. We can discern many of their good points and bad points. When we introduce valuations, we are not nearly as confident that we know how good our new estimates are. We know that the old estimates are bad, but we have a good idea of how they are bad and why they are bad. We cannot say the same thing when we introduce valuations. There may be flaws that we don't know about, at least, not yet.

I hope that this resolves the issue.

Have fun.

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

Post by [KenM] »

The definition gets looonger and loooooonger :) Is anybody going to read it apart from us obsessives (is it a reasonable assumption I'm not the only obsessive?)

Preparing the definition seems to have started with the primary focus on the "SWR" itself and then adding all the qualifying bits on as an afterthought. My alternative view is that the definition should include in order of importance:-
1. retirement income objectives (how much/how safe/more now/more later/preserve capital/etc)
2. strategy to achieve these objectives (optimistic/pessimistic/fallback plan/etc)
3. personal choice of a suitable strategy (important IMO - don't blindly follow what someone tells you on the internet :)- several strategy options may suit the objectives)
4. an SWR applicable to the selected strategy.

So my definition might be something similar to:-

The Safe Withdrawal Rate applied to the value of a portfolio in order to suit a retiree's chosen strategy for achieving his/her retirement income objectives from his/her portfolio.

....or something like that??????
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] »

As usual, I'm not sure I entirely agree with some things:-

It is stated as a percentage of the initial portfolio balance
I suggest that an overall definition of SWR should not be limited to the initial portfolio value. This is OK for the original "100%% safe" concept where starting with the initial $withdrawal would always be safe until the end of the period however low the portfolio value reached during the period. However for, say, 90% safe there must come a point where the value gets so low that there is a very high probability that it is one of the 10% that fails. An SWR applicable throughout the period would be useful in order for a retiree to assess whether he needs to reduce his current $withdrawal in order to avoid failure in those circumstances.

All other things being equal, the best calculation always takes valuation levels into account.
Even starting at high levels, I doubt that I would personally set an SWR by guessing at or trying to predict the future direction of valuation levels for the limited timeframe of the withdrawal period of my retirement. In the past we have had low levels for long periods; why not high levels for long periods? Therefore I don' think a best estimate SWR always includes valuations. However within my withdrawal strategy I would allow for adjusting withdrawals downwards if and/or when valuations dropped. And it would be imprudent not to do that even starting at valuations within the historical range. There is considerable emphasis on the board that we cannot predict the future except in the very long term. Why we would want to make an exception of valuations I'm not sure.The "irrational exhuberance" statement made no difference to the direction of valuations in the short term. And I don't believe in "rational pessimism" either :)
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Post by BenSolar »

Greetings :)

KenM wrote:
In the past we have had low levels for long periods; why not high levels for long periods?


I agree that valuations could feasibly stay in the high range for a long time. But, starting from and staying at high valuations still means reduced return, because at high valuations the return from dividends is much lower: 1.7% (or so from the S&P now) vs. the average 4.5%. That unavoidable reduction in return will lower the SWR unless you assume valuations continue to climb indefinitely to offset the reduced dividend. Which assumption, I think, is not a good one.

Ben
"Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things only hoped for." - Epicurus
raddr
*** Veteran
Posts: 265
Joined: Mon Nov 25, 2002 3:25 am
Contact:

Post by raddr »

BenSolar wrote:

I agree that valuations could feasibly stay in the high range for a long time. But, starting from and staying at high valuations still means reduced return, because at high valuations the return from dividends is much lower: 1.7% (or so from the S&P now) vs. the average 4.5%. That unavoidable reduction in return will lower the SWR unless you assume valuations continue to climb indefinitely to offset the reduced dividend. Which assumption, I think, is not a good one.

Ben


Ben,

You're exactly right. The only other way we can get the same returns in the future at current valuation levels is for the sustainable real GDP growth rate to nearly double from what it has been since the 1800's - not very likely anytime soon. It appears to me that is only way that you'd get enough profit growth to make up for the dividend shortfall. Remember too that about 1.5% of the nearly 7% yearly real return from stocks since 1926 has been from a quadrupling of PE ratios. This isn't likely to happen again unless you envision a sustainable PE of about 100 later in this century. :wink:
[KenM]
*** Veteran
Posts: 133
Joined: Tue Mar 04, 2003 12:54 am

Post by [KenM] »

But, starting from and staying at high valuations still means reduced return, because at high valuations the return from dividends is much lower: 1.7% (or so from the S&P now) vs. the average 4.5%
You're exactly right

Ben, raddr
Thanks. Instead of taking the pessimistic view of the market falling from now, I'd imagined it perhaps going sideways as 1966-82. But had not thought it through clearly that returns would not be able to sustain normal SWR's. Some months ago before I came across all this SWR stuff I'd made a "subjective judgement" (i.e guess) of about a 2% return after inflation when I retire next year. Looks as though my guess might have been right at this point in time.
KenM
Never try to teach a pig to sing. It wastes your time and annoys the pig.
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

Aha! Another thread, eh?
(I bookmarked the earlier one and got lost there :^)

Anyway, I think SWR is like ice cream.
Everybuddy has their favourite flavour (favorite flavor).
I don't think you can convince a chocolate-lover that strawberry is better.

Personally, as a method for generating an SWR estimate
(for purposes of planning your investments, in anticipation of eventual retirement)
I like the MonteCarlo/SWR
It appeals to me because it's well-defined, like MACD or Efficient Frontiers or Dividend Discount Models or whatever.
(But NOT like P/E ratios which have a jillion definitions ^$%#@*??)

I don't claim that it's "best" (whatever that means), only that one can calculate it with some precision.
(Tho' it's fun to argue about what is best!)


1: Pick a return distribution D (from historical data or whatever)
2: Pick an inflation rate I (or a distribution of rates)
3: Pick a probability p (like 95%)
4: Calculate SWR/mc(D,I,p)

Example: SWR/mc(D,3%,95%) = 4%
where D is D(lognormal,Mean=10%,SD=15%)

The nice thing about SWR/mc is that there are lots of parameters to play with ... and it gives an SWR to which all the others may be compared - a kind of benchmark.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

KenM to Ben and raddr
Thanks. Instead of taking the pessimistic view of the market falling from now, I'd imagined it perhaps going sideways as 1966-82. But had not thought it through clearly that returns would not be able to sustain normal SWR's. Some months ago before I came across all this SWR stuff I'd made a "subjective judgment" (i.e guess) of about a 2% return after inflation when I retire next year. Looks as though my guess might have been right at this point in time.

I will start a new thread this afternoon (possibly more than one) relating to your latest posts. (They have ended up being posted all over the place and understandably so). I see a way to brighten your day. So let me do that right now.

When I calculated a true Safe Withdrawal Rate of 2.3%, that corresponded to an S&P 500 index level of 1400, not 950.

When William Bernstein estimated a true Safe Withdrawal Rate of 2%, it was based on the market in 2000, not 2003.

Although raddr has shown convincingly (at least, convincing to me) that the 4% historical sequence number was a lucky number and that it corresponds to a true level of safety around 50%, you do not have to drop all of the way down to a 2% withdrawal rate to return to safe levels.

My personal assumption is that earnings (as approximated by P/E10) support stock market returns and safe withdrawal rates. Stated another way, if the S&P 500 were 900 today, I think that something close to 4% would be safe. (It would have to be reduced slightly to overcome raddr's concerns about lucky sequences. In addition, I have not made any corrections to my safe number for changes in the P/E10 during the last few months.)

Therefore, I conclude that your 2% withdrawal rate (plus adjustments for inflation) is an extraordinarily conservative number. You can expect to have a lot of upside potential. By adopting gummy's Sensible Withdrawal Rate strategy or by coming up with your own blend of several good strategies to meet your own needs, your financial outlook is exceedingly bright.

Have fun.

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

Post by gummy »

I've been playing with a MonteCarlo/SWR spreadsheet
and here's something that I found interesting:

I follow 10,000 investors for 40 years, all of whom start with a 4% withdrawal rate.
Withdrawals increase at an annual (inflation) rate of 2%.
Annual returns are drawn (at random) from a lognormal distribution with Mean = 9%, SD = 15%.
When a portfolio drops to $0 (or less!), it's ignored.

At years 10, 20, 30 and 40, the current withdrawal rates are calculated for those investors that have survived this far.

In the final year (that's year 40), what do you think the average withdrawal rate is (for those investors that have survived)?

Before I clicked on the "calculate" button (on the spreadsheet) I thought it would be large. After all, investors can withdraw everything that last year, eh?

Then I realized that we're not talking about a 1-year timespan but those investors that have survived 40 years
... and their average portolio is probably large (about $14M if investors started with $1M), so the average withdrawal would be tiny.

Anyway, it looks like so (which, for some reason, surprised me):
User avatar
ataloss
**** Heavy Hitter
Posts: 559
Joined: Mon Nov 25, 2002 3:00 am

Post by ataloss »

so the average withdrawal would be tiny.

with the $14 million dolllar portfolio I think I would be doing more than playing canasta :wink:
Have fun.

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

Post by JWR1945 »

ataloss
with the $14 million dollar portfolio I think I would be doing more than playing canasta.

You haven't played canasta with Gummy's wife.

You might really have fun.

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

Post by gummy »

Here's the distribution of final 40-year portolios (for our 10,000 investors):



Pick the one you like!

P.S. My wife cheats at canasta ...
User avatar
gummy
*** Veteran
Posts: 107
Joined: Wed May 14, 2003 4:00 am
Location: Hogsville, ON
Contact:

Post by gummy »

PS (again):

Anybuddy recognize the curve (outlining the final portfolio distribution)?

It must (surely) be defined by the parameters: lognormal and Mean = 9% and Standard Deviation = 15% and Inflation = 2% and years = 40 ...
Post Reply