Driving Drunk & Other "100% Safe" Actions

Research on Safe Withdrawal Rates

Moderator: hocus2004

Post Reply
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Driving Drunk & Other "100% Safe" Actions

Post by hocus2004 »

JWR1945 is the numbers guy. I am over my head talking about that stuff. So I often try to explain the realities of SWRs by making analogies that bring home the point of why there are some ways of reporting what the historical data says re what is safe that are analytically valid and other ways that are not.

I will use this thread as a place to collect various SWR analogies that I have put forward during the first 26 months of the Great SWR Debate. The Drunk Driving analogy set forth below is my favorite because it makes two important points through the use of one story. The two points are that: (1) the conventional methodology was analytically invalid before we entered bubble-level valuations; and (2) the problem got a lot worse in the late 1990s.

This post comes from the "SWR of 6.21 Percent for 26 Years"￾Â￾ thread at the Early Retirement Forum. That thread is the second most popular thread in the history of that site; I recommend that those with more than a casual interest in the SWR matter read it all. The Drunk Driving analogy appears as Post Five on Page Four of the thread. Here is a link:

http://early-retirement.org/cgi-bin/yab ... 8;start=45

Here is the full text of the Drunk Driving post:

Say that you have a friend who has been to your house 130 times over the course of the time you have known him. On 128 of those occasions, he drove home without incident. Twice he got drunk at parties you were throwing. On the first of those occasions, he banged into another car at a red light. On the second, he smashed up a neighbor's fender while parking. In neither case did he get himself killed as a result of his drunk driving.

This weekend, he gets twice as drunk as you have ever seen him before. You see him heading to the car and suggest that he sleep over rather than taking a risk of getting himself killed.

He says "I was drunk on two earlier occasions, and I didn't die either of those times. If I bang up another car again this time, I'll just somehow come up with the money needed to pay the bill. At least I can be certain that I won't get myself killed."

You say: "But two times isn't much to go on, and you're a lot more drunk this time. I think there's a real chance you actually will get killed this time."

He says "you're forgetting that I have driven home from your place 130 times. That's an awful lot of times that I drove home from here and didn't get killed. Considering that I've done this 130 times before, I'm 100 percent confident that I can pull it off again this time."

I am not persuaded by this logic. What matters is what happened the two times your friend was drunk. On both occasions when retirees tried to get away with a 4 percent withdrawal at 1929-type valuation levels, their retirements barely survived 30 years. And during the bubble, valuation levels went a lot higher than they were in 1929. I do not think that aspiring early retirees should be placing their confidence in claims that, because on two earlier occasions retirees just barely mananaged to survive on 4 percent withdrawals, it is safe for new ones to try test their luck in hoping this same withdrawal rate will barely squeak by yet a third time. This is especially so when the valuation level at which they are starting their retirements is far higher than it was on the two earlier occasions on which this withdrawal rate squeaked by.

Your friend might survive the ride home no matter how drunk he is this time. Anything is possible. But I don't view it as a safe practice for you to hand him back his car keys.

I believe that the historical data is trying to tell us something important. I think that we should be soberly trying to come to terms with the message it is trying very hard to communicate to us.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

Here's a link to a speech in which a comparison is drawn between investing and poker.

http://www.leggmason.com/billmiller/con ... nelson.asp

Juicy Quote #1: "People want to mix it up. They want to be in the hands and get the action and see what the next card will bring. So they will not fold to the extent that you need to because it's no fun and it requires discipline and patience. You can see where I'm starting to head in comparing investing with poker. Successful investing is boring. It takes place over years of time and involves accumulations of wealth in periods of years and decades as opposed to minutes, hours and weeks."

Juicy Quote #2: "Both are games of incomplete information. The tough thing about investing is the amount of information that you don't know and don't control. Lisa talked about scenario analysis and probability analysis. You don't know what's going to happen a lot of the time, but you do enough analysis so that the probabilities are in your favor when you make a particular investment. In poker, you weigh the odds in your favor by only playing those hands where you either have the best hand or have the best draw."

Juicy Quote #3: "In poker, they say, "have the best hand, the best draw, or get out." In other words, know when you have the 60/40 end of the proposition. Know when the odds are in your favor and bet, or know when the odds are not in your favor and get out of the way."

Juicy Quote #4: "Control your emotions. Losing your cool in poker-regret, anger, self-pity-is called "going on tilt." If you get angry and you start feeling sorry for yourself, you better get up and walk away because you're getting ready to give somebody a lot of money. The most important decision in poker and in investing is always "what is the right thing to do next?""
MacDuff
* Rookie
Posts: 31
Joined: Wed Jun 18, 2003 8:20 pm
Location: Centralia, WA

Post by MacDuff »

Hi Hocus, JWR, et.al.

I read this Legg Mason speech about probability, as well as some of the other presentations. Very good! How did you find this?

Anyway, in his speech David Nelson poses the question whetner one should call a $10 bet with $50 in the pot, and having either of the following hands: a four flush ( ie need one more card of same suit to complete) or an open straight ( eg 4,5,6,7) which could be completed at either end. He concludes that one should call the bet.

I went thorough the following reasoning and concluded that one should fold. Could someone criticize for me?

First assumption that if you complete the straight or flush you will win the pot. This is likely, but some unknown probability under 1.

First the case of the flush. In a 52 card deck, there are 13 cards of each suit. you have 4 in your hand. That leaves 13-4=9 cards that will complete the flush. As I see it, you then have 9/47 probabity of taking the pot. (47 cards still unknown after the 4 in your own hand) The pot is $50, so 9/47*$50=$9.57. To my way of thinking, to send $10 ( the call) looking for $9.47 is a long term loser.

The case of the straight is even more clear. Here we have 8 cards that will complete-any of 4 threes or 4 eights. So expectation =8/47*$50=$8.51. Another loser, since you must risk $10.

Did he make a mistake, or did I?

Thanks,

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

Post by JWR1945 »

MacDuff wrote:Did he make a mistake, or did I?
You got it right.

While we are at it, let us look at the first part of the speech.

David Nelson talked about lower returns (or higher losses) with certainty versus higher returns (or lower losses) with uncertainty. He has a point as long as the same decision is repeated many times. The effect of randomness [almost always] decreases by taking many samples.

But the Sharpe ratio favors certainty.

Implicit in using the Sharpe ratio is an assumption that you can leverage your investment choices, magnifying gains, until all of their volatilities are the same. You make comparisons at that point, after matching volatilities.

We have to be very careful about implicit assumptions. This issue came up in the What is the complaint? thread on the FIRE board.

It was also relevant to Hocus2004's initial choice of investments. Some gloss over the fact that his decisions were uniquely tied to his own circumstances. There is no one-size-fits-all solution for investment choices.

Have fun.

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

Post by JWR1945 »

JWR1945 wrote:
MacDuff wrote:Did he make a mistake, or did I?
You got it right.
On second thought, maybe you didn't.

The question is whether to "call a $10 bet with $50 in the pot." If you call the bet, the pot becomes $60. Based on $60, the expected returns become [9/47]*$60 = $11.49 and [8/47]*$60 = $10.12.

Have fun.

John R.
MacDuff
* Rookie
Posts: 31
Joined: Wed Jun 18, 2003 8:20 pm
Location: Centralia, WA

Post by MacDuff »

JWR1945 wrote:
JWR1945 wrote:
MacDuff wrote:Did he make a mistake, or did I?
You got it right.
On second thought, maybe you didn't.

The question is whether to "call a $10 bet with $50 in the pot." If you call the bet, the pot becomes $60. Based on $60, the expected returns become [9/47]*$60 = $11.49 and [8/47]*$60 = $10.12.

Have fun.

John R.
Thanks for your comments, JWR. I woke up about 4 AM with this same thought- I am not sending $10 after $50, but rather $10 after $60, the gross pot that I will take given success.

No doubt about this. But then I went back to Nelson's article. He assumes that in a 5 card draw game the set of unknown cards is 47. But it isn't, at least it doesn't appear so to me. The bettor is looking at 4 cards, not five. So the universe of possiblities of unknown cards appears to be 52-4=48. He only sees 5 cards once he has drawn.

So case one becomes 8/48*$60=$10
And case 2 becomes 9/48*$60=$11.25

Case one is a no-go. Case 2 IMO is marginal, since there is a small possibility that the hand if made could be beaten. With more knowledge of poker, that could be quantified in a probabilistic way.

There is more to this also. For example, the right choice depends on a correct assessment of current chances, and has nothing to do with one's sunk costs. This is a hard thing to get into one's head in the investing arena! :)

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

Post by JWR1945 »

Where is Ben Solar when we need him?

Although I am not a poker player, I do know enough about draw poker to know that Nelson got this part right. All players are dealt five cards initially (in their hands, not on the table). Later, you are allowed to replace any number (from zero to five ) of cards in your original hand. In these examples, you would be holding on to four cards and discarding the fifth. You would draw [actually, be dealt] one card to complete your new hand.

You may have been thinking about five card stud or some other game that is played one card at a time on the table.

Have fun.

John R.
MacDuff
* Rookie
Posts: 31
Joined: Wed Jun 18, 2003 8:20 pm
Location: Centralia, WA

Post by MacDuff »

JWR1945 wrote:You may have been thinking about five card stud or some other game that is played one card at a time on the table.John R.
You are right. I finally have this understood. Thanks.

Mac
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

Hyperborea put forward a Bridge Analogy last night at the Early Retirement Forum. Please scroll down to Response #39 at the link below.

http://early-retirement.org/cgi-bin/yab ... 1;start=30

The flaw in the argument is that hyperborea is assuming that there is only one factor that affects bridge safety. He mentions two factors--weather conditions and the amount of traffic being carried on the bridge. But he does not integrate these two factors in an analytically valid way when he makes the comparison to SWR analysis.

In order to make a fair comparison to SWRs, you need to have two factors affecting the result, as there are two factors known to affect SWRs. The two factors are: (1) the returns sequence that comes up during your retirement (this factor is examined in the study published at RetireEarlyHomePage.com) and (2) the valuation level that applies at the beginning of the retirement (this factor is ignored in the REHP study). William Bernstein discusses the two factors in Chapter Two of his book "The Four Pillars of Investing." It is because the conventional methodology ignores the second factor that Bernstein terms the results of conventional methodology studies "highly misleading" at times of high valuation (such as those we have been living through since formation of the Motley Fool board in May 1999).

Juicy Quote #1: "That might only be 130 years of weather observations and some of the early decades might only have spotty data. That would be much better than going with just normal traffic conditions on a sunny day.

Juicy Quote #2: "In the case of a retirement plan the numbers that can retire with double what "normal traffic" would allow is a small fraction of those who would normally be able to."

What hyperborea is saying is that, if the worst weather you have experienced is Storm Level D, and the bridge stood up to Storm Level D on the six times in the past 130 years in which weather conditions of Storm Level D were experienced, then the bridge may reasonably be declared "safe." I say no.

If you look at the data with a little bit of care, you see that the storm level is a significant factor in bridge safety but not the only significant factor. The returns sequences experienced by 1929 and 1965 retirees were not necessariily the worst on record. My guess is that they were not the worse, although I do not know for sure. What made the results following from 1929 and 1965 so dismal was the COMBINATION of valuation levels that applied in those years and the effect of returns sequences which were not so great. It is when you get both a bad valuation level (which is something you can avoid by informed planning) and a bad returns sequence that you get a busted retirement.

Say that we look at the weather history for the past 130 years and we find that there were 6 times when Storm Level D conditions were experienced. In two of these cases, low traffic levels were experienced, in two of these cases moderate traffic levels were experienced, and in two of these cases high traffic levels were experienced. Say that in the low-traffic and moderate-traffic cases, the bridge held up just fine, but that in the two cases where there was a combination of high-traffic and Storm Level D weather conditions, the bridge buckled and came within a whisper of collapsing altogether. What does this tell us? Does it tell us that Storm Level D presents no problem whatsoever re bridge safety, that so long as we do not experience a Storm Level E (a returns sequences worse than any we have seen in the past) that we can be certain that the bridge will not collapse. Not in my book.

It may be that Storm Level D will cause a collapse the next time one is experienced. It came close to causing a collapse the last time it was experienced and traffic levels are higher now than they were then. What if there is a Storm Level D (something no worse than what we have seen in the past) COMBINED with a higher traffic level (a worse starting-point valuation level)? What happens then? Given the fact that the bridge came within a whisper of collapse at lower traffic levels, my common sense (and the data too) tells me that there is a significant risk of collapse if we see another Storm Level D combined with a higher traffic level than applied in the earlier experiences of Storm Level D.

If all that mattered were weather conditions, you would not need to factor in the effect of traffic levels. You could just look at weather conditions and know whether the bridge was safe or not. But if you have two separate factors affecting safety, then you must analyze both or you have not done an analytically valid safety analysis. It's the same with long-term (or intermediate-term) stock returns. The returns sequence is one factor and the starting-point valuation level is another. Look at only one of them and there is no possible way that you can hope to ascertain what take-out number is safe. The only way to do a valid analysis is to look at both of the factors that have had a critical effect on safety in the past.

Is it possible that the bridge will survive even though traffic levels are now higher than they ever were in the past? It is. If the worst weather we have is a Storm Level B, the bridge is likely to remain standing regardless of the higher traffic levels. JWR1945's research shows that a 4 percent take-out has a 50 percent chance of working out even for 80-percent-S&P-allocation retirements beginning at today's valuation levels. So we are not making "predictions" that retirees following the advice set forth in the REHP study will experience busted retirements,. We are saying that there is a 50 percent chance that they will. If they are lucky re weather conditions (returns sequences), they will make it regardless of the valuation levels that applied on their retirement date. But if a Storm Level D pops up as their retirement sequence, their retirements will go bust. Even a Storm Level C will probably cause busted retirements for followers of the REHP study. It is only the 50 percent of sunniest weather conditions that permit these poorly informed retirement strategies to result in something other than severe life setbacks for those following them.

There is a 50 percent chance that our collective unwillingness to permit honest and informed posting on SWRs is going to cause a lot of busted retirement in our community. If you measure from the top of the bubble, the odds are a good bit worse than that. We have experienced the easy stuff over the past 30 months. It's when people's retirements go bust that we are going to see the truly bad kind of fireworks. People will then be asking questions as to why they were denied access to information that could have saved them from losing hundreds of thousands of dollars of accumulated capital. They will not be happy with the board administration decisions that were made at a number of Retire Early/FIRE/Passion Saving board communities. They will be as angry as hornets. Will any of us be able to offer a word of doubt as to whether they will be justified in feeling that these boards let them down is a serious way? I say no.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

I am not posting at the Early Retirement Forum at this time. The environment has become too hostile. I expect things to settle down as Dory36 gradually regains control of his own board. I have been following its content, however.

I addressed the original question in the referenced thread over here at the SWR Research Group. Read this post and the related posts with withdrawal amounts and portfolio balances: Tapping into Portfolio Gains dated Sat Oct 23, 2004.
http://nofeeboards.com/boards/viewtopic.php?t=3021
My post quantifies the penalty of removing portfolio gains in terms of Historical Surviving Withdrawal Rates. [I did not make any adjustments for today's market outlook and today's valuations.]

The original poster was under the erroneous impression that he could strip off portfolio gains (whenever they occur, but not making extra withdrawals in down years) without penalty.

I have also noticed mention of strategies similar to withdrawing a constant percentage of a portfolio's current balance instead of withdrawing a percentage of its initial balance plus inflation. I have already explained how to investigate such problems on FIRECALC at that site. We can do that as well with our own calculators. A withdrawal rate of 4% of the initial balance plus inflation corresponds roughly to a withdrawal rate of 5% of a portfolio's current balance.

We have consistently included confidence limits with our calculations, especially when we calculate Safe Withdrawal Rates versus the percentage earnings yield (100E10/P). This provides one way of separating out the effects of the sequence of returns from the overall return of a portfolio.

In addition, we have separated the effects of the sequence of returns from the overall return of a portfolio using a second approach, one that does not rely directly on valuations. Read these posts for details:
A New Tool: Overview from Wed Apr 28, 2004
http://nofeeboards.com/boards/viewtopic.php?t=2426
A New Tool from Wed Apr 28, 2004
http://nofeeboards.com/boards/viewtopic.php?t=2427

We have provided the numbers that go along with the bridge construction analogy.

We have made the kind of adjustments over here that hocus2004 has pointed out in his critique of the bridge building problem. It makes a tremendous difference.

Have fun.

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

Post by JWR1945 »

hocus2004 wrote:The returns sequences experienced by 1929 and 1965 retirees were not necessarily the worst on record. My guess is that they were not the worse, although I do not know for sure. What made the results following from 1929 and 1965 so dismal was the COMBINATION of valuation levels that applied in those years and the effect of returns sequences which were not so great. It is when you get both a bad valuation level (which is something you can avoid by informed planning) and a bad returns sequence that you get a busted retirement.
hocus2004 has guessed right. Those two worst case sequences were typical for their valuations. In fact, they were a little bit lucky.

Although our approach increases the sensitivity of our statistics (i.e., it increases the effective number of degrees of freedom), we cannot eliminate the effect of overlapping sequences entirely. When we plot the Historical Surviving Withdrawal Rates versus the Overall Return (or Return0) of a portfolio, the randomness caused by year-to-year price fluctuations is reduced so much that the effect of overlapped sequences is visible, especially when we look at data related to year 14. The data are clustered just above the regression line when portfolio returns are very low. This makes those two worst case sequences very easy to recognize.

The fact that the Historical Surviving Withdrawal Rates are slightly above the regression line (i.e., what I call the Calculated Rate) means that those particular sequences were just a little bit lucky, but not dramatically so. Of greater interest, however, is that those were far from being unlucky sequences in terms of their valuations.

Have fun.

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

Post by hocus2004 »

"The original poster was under the erroneous impression that he could strip off portfolio gains (whenever they occur, but not making extra withdrawals in down years) without penalty. "

My impression is that most community members think they know more or less what the REHP study claims, but that only perhaps 20 percent have taken a serious look at the study itself. The other 80 percent depend on community leaders to alert them to any serious problems.

Unfortunately, the 20 percent who claim expertise on what the study says is largely made up of people who made extensive use of the study in putting together their plans. So this group is loathe to acknowledge the flaws of the study. I believe that in most cases they understand that there are problems, but they have an emotional committment both to large stock allocations and to a 4 percent take-out. They so much want the study to "work" that they are uncomfortable with any discussion of reasons why it may not, and are certainly not inclined to add their own voices to explanations of the flaws.

The result is that those who know ain't about to tell and those who need to know have no practical means of gaining access to the explanations they need to make sense of things. The raft picks up speed in its race to the waterfall and all aboard confirm each other's observations that things would be proceeding about as smoothly as could be hoped if only it weren't for those few annoying individuals shouting from the shoreline to "Watch Out!" and "Stop!" and "Jump!" Such nuisances!
Post Reply