SWR of 6.21% for 26 years

[cute fuzzy bunny]

Re: Self Correction

I disagree that it's necessarily equally risky, so I can still argue it's not exactly the same thing, and I can argue that FIREcalc doesn't say 6% in 2004 is 100% safe. : (Although I think I've exhausted all ways of saying it.)

Sort of. In 2000 we still thought it might be "all different" and it wasnt clear what the impending risk might be. Sort of like riding a dirt bike full tilt on a big flat open space towards an invisible precipice. The perceived risk might have been lower before you went 3 miles down the face of the precipice. But the point is right on. 4% in 2000 is no more or less likely to work than 6% in 2004. Although I'd feel less good about the 6% than the 4% because I know there be precipices nearby...but I might feel better about 4% in 2004 than I did in 2000 with the same sized portfolios.

As far as two sequential slumps? Oh yes.

From July of 1929 to June of 1932, the Dow fell from 380 to 42. From then to August it ran up to 78. Then from August to February of 1933 it fell back to 53. A huge drop, rise and drop, percentage-wise. Substantial drops in 1937, 1938, and 1940 and 1942 after run ups. In fact there are a number of large percentage up and down cycles throughout.

 

[wabmester]

Give me two variables related to the stock markets, and show me the correlation formula that linearly relates them.  SHOW ME.

Next, you'll ask SG to prove that you have free will.

Two generally accepted correlations are short-term price momentum and longer-term mean reversion. Here's a random paper I googled that attempts to model mean reversion:

http://www.stat.fi/isi99/proceedings/arkisto/varasto/kim_0290.pdf

 

[cute fuzzy bunny]

I need no proof of that

I understand the rtm concept as an (unintended) corrollary to the gordon equation. Stocks move in the long term as a function of earnings and dividends. The net of both is a rise, the market rises to that tune. The net of both is a fall, the market falls at that rate. Thats the long term market return.

That is a fair correlation, in 20-30+ year periods.

However, in terms of less than 20 years, and especially in terms less than 10, market movements are significantly away from the mean or the "line" drawn from the gordon equation.

But my confusion in this thread exists because SG says long term return determinations are not easily gotten, while extensive data shows they are. And he believes that short term determinations are better gotten by some as yet unexplained correlations, when extensive data shows they cannot be. Although he notes that those short term determinations that lead to correlations are hard to identify and quantify.

I hope this all doesnt come across as adversarial or angry/upset. I'm more confused and perplexed than grouchy. Thats why I've asked for the specific points to be drawn from these thousands of words, and something I can read or poke with a stick that provides background or proof points to that.

Always looking to add more arrows to my quiver...

 

[sgeeeee]

. . . Give me two variables related to the stock markets, and show me the correlation formula that linearly relates them.  SHOW ME.

If they did exist, they would be incorporated into a sophisticated monte carlo program.  To my knowledge, they dont exist, and therefore dont find themselves in a monte carlo.  Disagree?  SHOW ME.

TH, I'm not going to teach a course on mathematics on this board. You can choose to believe that a correlation coefficient cannot be calculated for any two data sets if you wish. That is not an informed position. You are simply wrong about that. If you seriously want to find the correlation in some data sets, I suggest you purchase or borrow a statistical analysis package. All of the stat package software I've ever seen includes a built-in function that will compute correlation of data for you. Even Excel includes built-in functions to compute correlation coefficient for any two sets of data. Formula to compute that coefficient are also provided via Excel help.

I could take on this same kind of argument with you. . . You believe that no correlation between data sets exist. SHOW ME.

Clearly, we have gone past the point of productive discussion with that kind of response. That is unfortunate.

. . . I unconcede the correlations bit, because you are alternately saying there are clear correlations that cause market behavior, then saying they cant be (easily) measured or explained.  If they are there, they are measurable or explicable.  If they arent measurable or explicable, they arent there.  PROVE ME WRONG.

I hope I never said that the correlations are not measurable. If I did, I misspoke. Correlation coefficient is always measureable. The correlation may be positive or negative and may be week or strong or zero. The root cause of the correlation may not be known or understood, but the correlation can be quantified.

. . . You haven't responded to any of my invitations to show me examples of correlations that most professional mathematicians who have performed extensive analysis of the market say do not exist.  You havent shown me examples of the correlation formulas that are incorporated into monte carlo simulations, or a monte carlo simulation that uses formulae to calculate where the market can or should go based on these correlative factors.

I think you misread my post about monte carlo analysis. The only example of correlation I have ever seen included in a SWR monte carlo simulator was the one I mentioned that was discussed on the nofeeboards discussion. But here's how it works.

For the typical non-correlated monte carlo simulation, the simulation of each year begins by using a random number generator along with a modeled distribution to produce a number to represent stock return. Two additional random number generations and distributions also produce a number to represent bond return and inflation. These numbers are used to advance the state of the portfolio 1 year. The process is repeated for each additional year.

To include correlation between stock return and bond return (for example), you would modify the above so that the distribution of possible bond returns is made to be a function of the stock return that is generated each year. One way to establish the function you use can be to first plot all annual stock return and bond return data, then fit that data to a well-chosen curve. There are other more sophisticated ways.

I spent several years of my life developing monte carlo simulators to predict the flow and scatter of electrons within compound semiconductors. In that analysis, many of the variables were assumed to be correlated and the inclusion of that correlation was routine. Monte carlo simulations for many other types of physical and social problems is common. These types of simulations are significantly more complex than the SWR simulators -- primarily because those simulation developments are highly funded. It was not unusual to have multiple graduate students work for a year or more to develop an appropriate correlation technique for a single variable.

I think if you read some of the other posts here by ***** and gummy and a few others, you will see that several of them have produced correlated data as a key element of the analysis they describe.

. . . My failure to do so successfully indicates you are a bad explainer or I'm a bad understander.  I'll accept fault, so please tell me what your point(s) are exactly again?  Maybe someone else can explain them to me?

. . .

Help me help you.

I'm in no need of help, thank-you.

 

[sgeeeee]

Hi SG,

As usual you make a lot of good points.   I will nitpick this one though: . . .

 

Of course you are correct, raddr. But you will note that I chose my words carefully. I did not say that this was of no concern to me, but rather that it was of little concern to me.

There is also the problem that there is some correlation of financial results from year to year while there is no correlation of the coin toss results.

Clearly, independent series of years would be preferred over the overlapping data we have. But, as I pointed out, if we had more non-overlapping data I would worry about the relevance of data earlier than 1871.

What we need is a reliable crystal ball -- not a historical simulator.

 

[sgeeeee]

. . . It is possible to determine from looking at the historical database what effect changes in valuation levels have had in the past and then make the necessary adjustment to reflect that factor. But the bubble level valuation levels have never been experienced before. It is not possible to "find" them by looking at the data in the database without making an adjustment.

Well . . . the problem is that an empirically determined dependance cannot necessarily be extrapoltaed outside of the range of available data. When I look at the data you posted, for example, it looks to me like the data is best described by an assymptotic function that is nearly totally saturated at any PE10 above 18 or 20. In fact, I fit your data to a very simple exponential function that assymptotically approaches a HDBR50 value of 4.15%. If I use that function to extrapolate beyond the existing data, I get a very different result.

The conventional methodology does not call for any adjustment for this critical effect. That's why Bernstein says that the results of the conventional methodology studies are "at times misleading." Do you think that Bernstein is wrong about this?

I think he may be right and he may be wrong.

 

[wabmester]

But my confusion in this thread exists because SG says long term return determinations are not easily gotten, while extensive data shows they are.

Sort of. If you include a fudge factor called the Equity Risk Premium. And when that doesn't work, throw in another fudge factor called the rate of P/E growth. And when that doesn't work, then throw up your hands and insist that mean reversion will make things right some day.

And he believes that short term determinations are better gotten by some as yet unexplained correlations, when extensive data shows they cannot be.

Huh? Did you notice the market dive when bombs went off in Spain? There are lots of correlations that drive short-term moves. What economists would like you to believe is that you can't exploit any of them to your advantage. I don't believe them.

 

[cute fuzzy bunny]

TH, I'm not going to teach a course on mathematics on this board.[...]Formula to compute that coefficient are also provided via Excel help.

You didnt specify your points (again). And you are misspeaking my point(s). I did not say that correlation cannot be determined between two data sets. You get specific, and then overgeneralize. I said that the short to medium movements of the stock market can not be determined mathematically. You disagreed and said that it could, and that there were factors in the market that demonstrated correlation, but that they were hard to express. I disagreed that there were any correlations that could be used to determine future market direction next year, five years later, or any time in between. I dont need a statistical package to show me what a thousand mathematicians have already proven. There are no short term correlative data that you can use to predict market movements. Not even a hint of one.

I could take on this same kind of argument with you. . . You believe that no correlation between data sets exist. SHOW ME.

Ok.

Markets and individual stocks have moved both up and down or stayed the same, repeatedly when:

Inflation is low or high or moving in one direction or the other.
Interest rates are low or high or moving in one direction or the other.
Corporate earnings are low or high or moving in one direction or the other.
Dividends are low or high or moving in one direction or the other.
Yesterdays price was up or down.
Todays was trending higher or lower.
A company's stock has gone higher and lower and stayed the same when announcing better, worse, or spot on earnings or any other result or expressed report, news or number.

Do I need to go on? Do I need to bring up specifics? Wheres the correlation?!?

Clearly, we have gone past the point of productive discussion with that kind of response. That is unfortunate.

I didnt think so, and I'm sorry you do. I said I didnt understand your position and asked you to re-express it so I could understand it. So far a lot of the opinions you expressed are contrary to what I have seen, and I wanted to see the background and data that supports your opinions. Instead you've said I didnt understand your points, and havent provided me with any backup. I can either give up (a frequently considered option), or ask again. Is your intention to "win", to have me "give up" or to express an opinion and have a reasonable person understand it?

I hope I never said that the correlations are not measurable. If I did, I misspoke. Correlation coefficient is always measureable. The correlation may be positive or negative and may be week or strong or zero. The root cause of the correlation may not be known or understood, but the correlation can be quantified.

Yes you did.

"That does not imply that the correlation is 1 (ie completely predictive) or even that I can quantify it. "

"The correlation is not simple and it is not easily captured mathematically, but it is real."

You then went on to say that whether or not it is quantifiable, you can still draw a correlation from it. I'm not a mathematician, but correlation DOES say that you can quantify it. In every textbook definition I can find. IS there a correlation for short to medium term market movements? WHAT is it? HOW do you quantify it?

 

[cute fuzzy bunny]

I think you misread my post about monte carlo analysis. [...]

I do not doubt you have created non stock market monte carlo simulations. I am far from a stranger to them. However your original comment on these said:

"[1]Financial returns and inflation are correlated to each other and to the results of previous years. [2]The correlation is not simple and it is not easily captured mathematically, but it is real. [3]This is precisely the reason why Monte Carlo simulations tend to predict less optimistic withdrawal rates than historical simulators. [4]Monte Carlo simulations fail to capture the correlations between the returns and inflation data. "

Your statement [1] as discussed above is unsupported by ANY data that I have ever seen and essentially says that there should be a way to determine current and future gains based on the past, hence a formula, hence actively managed funds that follow the formula should beat indexes. Your [2] is discussed above. Your [3] I debated that the reason why monte carlo situations return less optimistic results is NOT due to a failure to capture this immeasurable correlation, but because they string random years returns together to produce worse than historical data. You argued that NO, this is not the reason, its because monte carlo models DO incorporate this. Which is it? Then you use the word "random" several times, without inclusion of any magical correlation formulas in them, in describing how the stock market/portfolio monte carlo situations work.

I think if you read some of the other posts here by ***** and gummy and a few others, you will see that several of them have produced correlated data as a key element of the analysis they describe.

Please save me a hundred days of reading their posts and show me a demonstrable correlation that you feel is relevant to calculating market returns and SWR's of a given portfolio. If you're talking about the ones included within this thread, I see no obvious correlations.

I'm in no need of help, thank-you.

I am, in fact, I need a glass of wine after THIS discourse.

As I said, I suspect if you simply state your opinions, one sentence to each, on this total topic, we may find we completely agree. It just seems to me that an abundance of your posts contradict each other, are unnecessary argumentative (with me), and are not based in any well accepted or explained science that I'm aware of.

Tell me what your positions are, please be objective, and please give me a balance of data that I can peruse to add this understanding to my own. I try to do the same when I offer opinions and will certainly do so when challenged.

Or I'll offer "I got nothin'".

 

[John Galt]

I had a Monte Carlo once. Never ran worth a damn.

John Galt

 

[gummy]

About using correlation in Monte Carlo simulation, there's a spreadsheet described here which does that:
http://home.golden.net/~pjponzo/Monte-Carlo.htm

About correlation between various assets, there's a table (somewhere near the bottom) here:
http://home.golden.net/~pjponzo/Covariance.htm

And a question which has always interested me:

For those on this board who suggest using SWR when you're retired (and withdrawing from your portfolio),
do you imagine doing this each year on (say) January 1 and using the result to determine what you'll withdraw
... for the remainder of that year ?

I'm retired (and withdrawing) but I've never considered doing any SWR calculation to determine my withdrawals.

As has been noted by arrete, after a bad year me & the missus don't go nowhere or spend much ... but play canasta instead

 

[gummy]

Oops ... another question that interests me**:

If somebuddy asked you to suggest a "Safe" withdrawal rate for the next 27 years, would you suggest something around 4%?

If the person then mentioned that they had started withdrawing three years ago at 4% and now their rate happened to be 6%, would you change your suggested "Safe" withdrawal rate to something around 6%?

** Sorry, but it seems in old age almost everything interests me

 

[BigMoneyJim]

Oops ... another question that interests me**:

If somebuddy asked you to suggest a "Safe" withdrawal rate for the next 27 years, would you suggest something around 4%?

If the person then mentioned that they had started withdrawing three years ago at 4% and now their rate happened to be 6%, would you change your suggested "Safe" withdrawal rate to something around 6%?

** Sorry, but it seems in old age almost everything interests me

My using FIREcalc is like a kid with a new toy, and I have no near-term stake in determining a "safe" SWR, but I'll offer the opinion I'm forming based on FIREcalc's analysis, my confidence in the statistical support and my complete ignorance/disregard for market movements or valuations which I haven't studied yet.

I would now go for less than 4.19%. I'm not sure how much less--maybe a few tenths of a percent, maybe a half to 2% less--but 4.19% is clearly pushing the limits of historical 30-year survivability, and even if one were to believe no future 30-year period could be harsher on a portfolio than the past periods--in which I don't have high confidence--the granularity of the data and overlapping samples alone indicate to me the need for a buffer zone / lower SWR.

With the above judgement qualifications, I would not adjust a SWR upward based on today's 6% being the result of a past 4% or "safe" rate.

To my eyes, the biggest problem with SWR is you have to decide what it is when you're still in prime earning years but only find if it succeeds after you've been out of the workforce for many years and have a higher probability of health problems, general crumudgeonness towards potential coworkers and the complete loss of restraint for telling a boss he's stupid and can go to hell.

What caught my interest and got me involved in this discussion is the apparent paradox of "100% safe" 4.19% in 2000 and the proposed conclusion that 6.21% in 2004 is equally safe.

 

[gummy]

BigMoney:
I always found it curious that people who calculate a 30-year SWR based upon the worst 30 years (in the last 75 or 100 years) assume (I suppose?) that they're considering calendar years: Jan thru Dec.

Has anybuddy considered years starting in Feb or Mar or whatever?

Suppose that one DID determine the "worst-of-the-worst". (Would it be even worser?)

And what's the effect of interchanging some of the years in that "worst-of-the-worst" 30-year period?

For example, in the Jan, 1930 - Dec, 1959 example I considered earlier, the "Safe" rate was 4.08%.
If, however, I interchange just two returns and inflation rates, switching 1933 with 1937, that SWR becomes 3.13%.
(You'll note that the two sets of 30 annual returns/inflation have exactly the same Mean,Volatility and Distribution.)

Although I (normally) hate to give advice, I do make suggestions for my kids ... like so:

Figure out what income you think you'll need from your portfolio, at retirement, by using that notorious 4% rule.

(That is, determine what annual income you'll need and multiply by 25 to get an estimate of your required portfolio at retirement).

Then, when you retire, completely forget SWR and withdraw (each year) according to what happened to your portfolio during the previous year

... and learn to play canasta

 

[sgeeeee]

. . . However your original comment on these said:

"[1]Financial returns and inflation are correlated to each other and to the results of previous years.  [2]The correlation is not simple and it is not easily captured mathematically, but it is real.  [3]This is precisely the reason why Monte Carlo simulations tend to predict less optimistic withdrawal rates than historical simulators.  [4]Monte Carlo simulations fail to capture the correlations between the returns and inflation data. "

I'll stand by that statement completely. I am quite certain of the truth of that statement and believe that most people who develop and use such techniques are well aware of the truth of that statement. The fact that you interpret what I've said to be meaningless or false is just something that I will live with.

The reason that the correlation I am speaking about in this particular statement is difficult to capture mathematically is because one of the data sets (ie. future returns) is not yet known and the cause and effect relationship of the correlation is too complex to describe with simple formula -- or even to capture in advance. Once two data sets are known (as in the case of historical data) the correlation can always be captured.

 

[sgeeeee]

. . . So far a lot of the opinions you expressed are contrary to what I have seen, and I wanted to see the background and data that supports your opinions.  

At this point, I don't even understand what you think I said. And I certainly don't understand what you are expecting from me? I do value your inputs on this board and respect you, however, so lets see if we can find some common ground.

This whole discussion started because I pointed out that amt's conclusion (that the RE of today could retire with a 6.2% initial withdrawal rate and be as safe as the RE2000 who started with a FIRECALC approved 100% safe 4.1% withdrawal) was correct, logical and mathematically sound. I went on to point out that if you don't believe Mr. RE2004 will be safe with a 6.2% initial withdrawal rate, then you should question the safety of the 4.1% rate.

Do we agree on that? Because that's the key point here.

The reason that the statements above are true is not related to some post-FIRECALC analysis that is inconsistent with the underlying principles within FIRECALC. It is not because we used a 100% safe rate as opposed to a 90% or 80% safe rate. It is because FIRECALC calculates the safe rate based on the worst case. So as long as you can know that a prior year will provide a retirement series that is worse than the retirement year you are currently in, you can start your calculations from that point.

Do we agree on that?

I don't believe that I can develop a simulator that will predict future returns -- long, mid- or short.

Do we agree on that?

And finally, just because I cannot predict the future does not mean that the future is not causally connected to the present and to the past.

Do we agree on that?

 

[BigMoneyJim]

BigMoney:
I always found it curious that people who calculate a 30-year SWR based upon the worst 30 years (in the last 75 or 100 years) assume (I suppose?) that they're considering calendar years: Jan thru Dec.

I have assumed the reason is that freely-available market return data are for end-of-year only, but I may be mistaken. That's part of what I mean when I mention FIREcalc's data having granularity concerns, although I may be gobbeldygooking the terminology. It would certainly feel a little more comfortable to have a simulator run data sets beginning in a given month or even a given day of the year or model multiple intra-year withdrawals or even randomly scattered withdrawals throughout each year and see how that affects historical SWR. It's not immediately clear to me if it would reduce the statistical error, but that's nitpicking the process again.

If there is a finer-grained freely useable data set, somebody point me to it and maybe I'll attempt Yet Another SWR Caculator, but my enthusiasm is already waning. But sometime in the next 15-30 years I'll probably get interested again.

For example, in the Jan, 1930 - Dec, 1959 example I considered earlier, the "Safe" rate was 4.08%.
If, however, I interchange just two returns and inflation rates, switching 1933 with 1937, that SWR becomes 3.13%.
(You'll note that the two sets of 30 annual returns/inflation have exactly the same Mean,Volatility and Distribution.)

That kind of talk can spawn a whole new 8-page thread on historical analysis versus Monte Carlo versus market motion versus fluctuation. Intuitively I feel like short-term losses can easily be greater than historical happenings but long term performance will on average be positive. I have no evidence to support that; it's just my gut. (Of course it's such a vague proclomation that it's hard to be wrong.) Switching historical yearly returns like that doesn't sound statistically reasonable, but it's an interesting exercise nonetheless.

Your advice is among the best I've seen. But I don't know how to play Canasta; can other card games be substituted without disturbing statistical analysis, correlation or valuation?

 

[wabmester]

And finally, just because I cannot predict the future does not mean that the future is not causally connected to the present and to the past.

Yeah, I don't believe in free will either. We're really deterministic meat machines, but I still enjoy the illusion

 

[BigMoneyJim]

This whole discussion started because I pointed out that amt's conclusion (that the RE of today could retire with a 6.2% initial withdrawal rate and be as safe as the RE2000 who started with a FIRECALC approved 100% safe 4.1% withdrawal) was correct, logical and mathematically sound. I went on to point out that if you don't believe Mr. RE2004 will be safe with a 6.2% initial withdrawal rate, then you should question the safety of the 4.1% rate.

That wasn't aimed at me (at least not directly), but worded that way I agree completely. My earlier arguments were directed at a percieved paradox and then a perceived perversion of statistics. But as quoted I agree 100%. (and 100% safely)

I think the catch is the first time I heard the argument I took ". . . and be as safe as . . ." to mean that 6% was supported by FIREcalc as 100% safe. But now I read it that if 6% doesn't sound good now then why did 4% sound so good 4 years ago?

I suppose the easy answer is that we didn't really expect 2000 to be remotely comparable to 1929, and 4 years later we're not as willing to let the bet ride. The more interesting answers we are now getting to is how 2000-2004 affects our conclusions from our analyses and whether we should have seen it in 2000.

EDIT: I wouldn't have seen it; in 2000 I was just getting into personal finance and TMF, and I was a happy follower of TMF's stock cheering. That DOW 40,000 book sounded like a reasonable possibility, although I didn't read the book and now forget when we were supposed to hit 40,000. I'm just thankful I didn't try getting out of index funds to pick my own stocks.

 

[Bob_Smith]

Although I (normally) hate to give advice, I do make suggestions for my kids ... like so:

Figure out what income you think you'll need from your portfolio, at retirement, by using that notorious 4% rule.

(That is, determine what annual income you'll need and multiply by 25 to get an estimate of your required portfolio at retirement).

Then, when you retire, completely forget SWR and withdraw (each year) according to what happened to your portfolio during the previous year

Excellent advice! I plan to follow it. And thanks for the excellent tools you provide on your site. They have been very helpful.

 

[gummy]

BigMoney:

But I don't know how to play Canasta; can other card games be substituted without disturbing statistical analysis, correlation or valuation?

For Canadians, it has to be Canasta:
Canadian Alternate Norm After Stockmarket Tanks, A?

:

 

[gummy]

Bob_Smith:
My personal favourite "Safe" withdrawal calculation is (sorta) like this:
http://home.golden.net/~pjponzo/sensible_withdrawals.htm

 

[ESRBob]

Here's a post from the SWR Research Group board where JWR1945 puts forward an allocation strategy that the historical data indicates would support a 4 percent withdrawal at today's valuation levels.

http://nofeeboards.com/boards/viewtopic.php?t=2158

.

Is JWR1945 somehow business-related to Zvie Body and Mike Clowes? They are also advocating a 100% TIPS strategy for ER, but I am having trouble buying it...

 

[hocus]

Is JWR1945 somehow business-related to Zvie Body and Mike Clowes?

No.

I bear full responsibility for JWR's move to the dark side. I think he had about two or three messages to the Motley Fool board in his entire posting career when I put up the now famous (infamous?) May 13, 2002, post asking whether SWR analysis should include an adjustment for changes in valuation levels. One of the funny historical facts to know and tell about all this is that I was fed up with all the heat being directed at me after three days, and I bowed out of the debate on May 16, 2002. Then JWR pulled me back in. He really is the one to blame for all that followed!

After I bowed out, JWR1945 did a statistical analysis checking whether there was any validity to my claims. He found that there was was, and put up two posts--"***** is Really Onto Something!" and "***** Started It All (and I'm Having a Ball)" that each received about 50 or 60 recs from other members of the board community. That strong expression of a desire to further explore the realities of SWRs prompted me to put up the "Coin Toss" and "My Plan" posts, which became two of the most popular posts in the history of the board. The board had gotten stuck in a rut of off-topic posting, and there were scores of posters that came forward then thanking me for getting things back on the right track.

Lots of complications followed. There have been tens of thousands of posts put forward at the four boards that have participated in the Great SWR Debate. I think it is fair to say that the road has been a rocky one. The other side of this, however, is that those who are interested in learning about how to put together an effective plan for early retirement have learned an awful lot from the best of those tens of thousands of posts. I learned enough that I now have an outline for a second book which will bascially be my in-depth exploration of the most important concepts explored during the course of the Great SWR Debate.

I'm too close to this today to say whether this all has been a good thing or a bad thing. It's sucked up an awful lot of my time, so much that it's painful for me even to think about that aspect of it. But asking me not to participate where there's learning about the subject of early retirement going on is like asking a guard-dog not to bark when someone breaks in through the window at night; learning about this stuff is what I live for.

My view is that this debate has revealed to us both the best of the new discussion-board communications medium and the worst of the new discussion-board communications medium. The bad stuff has been real bad. The good stuff has been too wonderful for words.

To get back to your actual question (I didn't forget!), JWR1945 has revealed himself to be an absolute giant of the Retire Early movement. He has devoted himself on a full-time basis to preparing research for us on these important questions for close to two full years now without charging us one shiny dime! It amazes me and astounds me every time I think about it.

That sort of thing happens on internet discussion boards all the time. I tell people from time to time about the trouble I have experienced during this thing, and the usual reaction I get is something to the effect of "Why do you waste your time on such nonsense?" My answer is that it is not nonsense. There are things that you can learn on an internet discussion board that you cannot learn anywhere else. I know because back in the mid-90s I searched everything available in book stores and libraries. During the Golden Age of the Motley Fool board, I regularly gained access to insights that were not available anywhere else on Planet Earth. So I see real value in well-run discussion board communities.

When they are working right, these boards to me are a wonder. People come to these boards and talk about their dreams, and share details of their budgets, and put their hopes for a more free and fulfilling future in our hands. We have an obligation to talk straight with them. They can ban me from six boards if they like, and I will never let the people who read my stuff down in that regard. I feel a moral obligation to shoot stright with the people who come to these boards expecting to hear my story of how to retire early the best that I am able to tell it.

Anyway, the point I am trying to get across is that JWR1945 has been the brightest of the bright spots of the past two years for our little movement. He is the model that all other posters should aim to emulate, in my view. He is positively selfless, so far as I can tell. He works hard and he works smart and he works free. Where else do you get a deal like that nowadays?

There is hardly anyone who ever talks to him when he posts over at the SWR Research Group board because the leaders of the community there have imposed a boycott of that board. I think it would be great if a few posters from here would head over there and provide him a little company from time to time.

You will learn from tapping into his research and he will learn from the feedback you provide. That's what community is all about. It's each of us learning from the others, and thereby all achieving our life goals sooner. JWR1945 has an awful lot to offer those who tap into his statistical expertise, so I hope that you will give it a little bit of thought.

 

[cute fuzzy bunny]

At this point, I don't even understand what you think I said. And I certainly don't understand what you are expecting from me? I do value your inputs on this board and respect you, however, so lets see if we can find some common ground.

Thats why I asked you several times to simply restate your points. And I agree with the sentiment.

This whole discussion started because I pointed out that amt's conclusion (that the RE of today could retire with a 6.2% initial withdrawal rate and be as safe as the RE2000 who started with a FIRECALC approved 100% safe 4.1% withdrawal) was correct, logical and mathematically sound. I went on to point out that if you don't believe Mr. RE2004 will be safe with a 6.2% initial withdrawal rate, then you should question the safety of the 4.1% rate.

Do we agree on that? Because that's the key point here.

Yep, we agree completely. Shocked? I'm not. And I think that I added something that may be useful: if you are calculating SWR for a given starting point, it may be useful to eyeball investment valuations and short term impacts to your portfolio and if the valuations are largely considered "high" by some measures, or there are significant short term impacts (ie, imminent interest risk), you might consider toning down the SWR for the short haul or delaying retirement a few more years until things settle down. I think over long hauls the calculators truth is good, but starting out on a high note may induce some indigestion.

The reason that the statements above are true is not related to some post-FIRECALC analysis that is inconsistent with the underlying principles within FIRECALC. It is not because we used a 100% safe rate as opposed to a 90% or 80% safe rate. It is because FIRECALC calculates the safe rate based on the worst case. So as long as you can know that a prior year will provide a retirement series that is worse than the retirement year you are currently in, you can start your calculations from that point.

Do we agree on that?

Yep. One hundred percent.

I don't believe that I can develop a simulator that will predict future returns -- long, mid- or short.

Do we agree on that?

Yep, one hundred percent. But at one point you said that you felt the market was somewhat predictable in the short term, and unpredictable in the long term. We disagree on that.

And finally, just because I cannot predict the future does not mean that the future is not causally connected to the present and to the past.

Do we agree on that?

We agree on that 100%. Where we diverged is that I stated that if a causal connection demonstrating correlation <>0 existed, that it should be identifiable, and measurable. In the absence of that identifiability and measurability, it has no benefit to us in this time to determine future market movements on that basis. You disagreed with that, saying it was beneficial. Hence my confusion. Tying that to your previously stated belief that shorter term market movements are predictable, I determined that you felt there were measurable correlations that would allow such prediction in a beneficial manner. Hence my frustration. Are there correlations. What are they. Are they measurable and actionable. What measurements and actions would I take if they are.

So as far as I can tell, we're in complete violent agreement on these topics.