Further withdrawal study results from Darrow Kirkpatrick of Can I Retire Yet?

[Ed_The_Gypsy]

I am not sure this update has been reported here yet.

As reported here earlier, on Dec 28 of last year, in his blog Can I Retire Yet New Research: The Best Retirement Withdrawal Strategies - Can I Retire Yet?, Mr. Kirkpatrick posted the results of his search for the best retirement withdrawal strategies. He found that for an initial 50/50 AA, an annual 4%+inflation withdrawn in equal amounts from stocks and bonds for 30 years had superior survivability and terminal value to all but one of his schemes. It was so simple and the relative success was so striking to me.

Mr. K has just posted the results of a further study in his blog of March 8 The Best Retirement Withdrawal Strategies: Digging Deeper - Can I Retire Yet? (also reported in Yahoo Finance/Money Magazine, but the post in his blog is more informative). I found the new results even more interesting. He discarded some of the earlier schemes as not worth further investigation and corrected some small errors. He then examined several withdrawal strategies with and without annual rebalancing for different AA’s--80/20, 60/40, 50/50, 40/60 and 20/80—using the usual data from 1928 to 2014.

Now some interesting stuff:

He found that all of his withdrawal strategies had at least 90% survivability until the AA dropped to 50% stocks and at least 80% survivability until stocks dropped to 20%.

Again, his CAPE withdrawal strategy came out on top but he acknowledged a comment that his use of CAPE required advance knowledge. In other words, he was applying CAPE averages from 1928 to 2014 to all tests, not looking backwards only. He says he may test backwards-looking-only CAPE averages. His method requires knowledge of the current value of CAPE, which may not always be easy to find. I am thinking of my surviving spouse or myself when the gears get rusty. He also states that market timing has been thoroughly debunked, but his CAPE strategy seems to me to be a kind of market timing.

The second runner-up again is what he now calls Proportional, whether or not the AA is rebalanced annually. Assets are harvested in proportion to the original AA. This is again wonderfully simple and I find it extremely attractive.

He found that annual rebalancing reduced volatility and also reduced the ending portfolio value. No surprises there. Take your choice. I prefer less volatility.

He observed that having a cash bucket is really part of AA and does not add anything. He noted earlier that he sells assets and takes withdrawals as he needs them (presumably monthly?), but he has also said that he personally is comfortable with a years’ cash in the bank. For his studies, he took distributions at the first of the year and rebalanced to the original AA at the end of the year when rebalancing was the option. (Note that he has a withdrawal strategy he calls Rebalancing which is not exactly rebalancing to the original AA. Beware. What he means is that he sells assets to try to bring the AA closer to the original AA but does not sell any more than the annual distribution requires so the post-withdrawal AA may not wind up equal to the original AA. This is to minimize transactions in your account. Clever! Aside from this, it does not appear to have any advantage over Proportional.)

This study looks worthwhile putting in FAQs as a sticky. Does anyone agree?

I have turned most of my AA to cash for the purpose of converting my TIRA to a Roth before MRD hits me so I have not committed to a future AA yet. I have in mind a 2:1 ratio of VRB/TLT (SCV/LTB), or 66%/34% because it is simple, does not require reference to an external number and dodges the market-timing issue. Paul Merriman has recently opined that one may as well choose SCV for the only equity class. I do plan to take a maximum 3.5% or 4% flexibly with no adjustment for inflation (another external number—still trying to keep it simple). A few months ago, I used data back to 1974 (IIRC) and it looks like annually rebalanced SCV/TLT (50/50 or 60/40) has better performance and less volatility than VWELX (Wellington, approx 60/40), my current standard. Wow. For myself, I have been taking distributions only as needed, a month at a time.

[Svensk Anga]

Odd that you state you prefer low volatility but intend to use 50-60% SCV and LTB to build your portfolio, possibly the most volatile corners of the basic stock and bond categories. (okay, maybe SCG, emerging markets stocks, and junk bonds are worse.) Perhaps the small and value factors being discovered in this period and most of it being a bull market for bonds goosed returns more than can be expected going forward.

On the question of which asset to liquidate, I am already set. I am using the Bill Bernstein Liability Matching Portfolio, which means I will be selling "safe" assets to fund most needs, that is CD's and individual TIPS mostly. Also will be spending dividends. The consequence of this is a rising equity allocation and potentially a large residual, assuming equities outperform over the long haul. This balances being safe for us and doing well by our heirs.

[LOL!]

I read this and came away with "Well, duh!"

If one rebalances back to 60/40 every year AND one makes withdrawals that gets one closer to target allocation when making withdrawals throughout the year, then isn't that what Kirkpatrick says works?

For instance, if equities are overweighted, then one doesn't reinvest the quarterly dividends, but just spends them. If bonds are overweighted, then one doesn't reinvest the monthly bond fund dividends, but just spends them.

And are not rebalancing bands a proxy for CAPE tracking?

I wonder if withdrawing in a tax-aware way has more effect than any of the different methods that he looked at.

[Markola]

I subscribe to Kirkpatrick's blog and am glad he and his large brain work so hard on this sort of retirement-for-regular-people stuff. He has another recent and persuasive post, IMHO, about the probable pointlessness of Roth Conversions. Anyway, I liked his chart in the post above with the different strategies according to AA. For comparison to others here, my aim in FIRE is to have 90% of our assets in a single balanced fund that auto-balances (either the Vanguard Life Strategy Growth [80/20] or the Moderate Growth one[60/40] with maybe a year or two of cash for spending in downturns. So, we will necessarily do a Proportional withdrawal, which appears pretty darn successful in Kirkpatrick's model, happily.

[nun]

I rebalance whenever my AA deviates by +/-5%. If I need income I'll sell some winners to get closer to my idea AA.....if I'm already there then I just sell the proportional amounts of stocks and fixed interest. Do we really need to think about this so much? There seem to be plenty of ways to successfully withdraw money and my intuition tells me that differences of a few percent in success rates are meaningless as there are
too many unknown variables in our future. So pick an AA with at least 50% stocks and use any withdrawal strategy that takes your fancy.

[Dog]

I just found his blog earlier this evening and find it very informative and timely too!


Sent from my iPhone using Early Retirement Forum

[CaliforniaMan]

This quote stuck out to me, the bolding is mine.

"Note that I used the long-term median currently being reported for CAPE. One reader pointed out that this requires future knowledge. It’s a valid concern. I’ve put on my list to investigate a simulation that looks only backwards. But how much it matters may boil down to whether we consider the CAPE median to be constant or variable over retirement-length time spans. I don’t know. There are distinguished experts on both sides of that question."

It seems to me this pretty much invalidates the CAPE method since it uses future certain knowledge of an unknown value. The question of whether it is a constant or variable seems to be pointless, since we can never know if it is a constant, all we have is past measurement data. It is a statistical measurement, not a known physical principle. We must assume it is a variable.

It would be much better to use a running median, since that is all that is really known at the time a decision has to be made.

It is never permissible and meaningless to use knowledge about the future when making a model.

[Totoro]

Quote:

Originally Posted by Ed_The_Gypsy His method requires knowledge of the current value of CAPE, which may not always be easy to find.

Just to say that the CAPE (as per Shiller's definition) is easy to find, for example:

Shiller PE Ratio

[Ed_The_Gypsy]

Today, it is. For me, with a clear mind. (Link was provided by the author as well.)
I am thinking about DW on her own or me in my dotage. (With luck, by that time everything will be converted to VWELX and on autopilot, but s*it happens.)

Sent from my SM-G900V using Early Retirement Forum mobile app

[Ed_The_Gypsy]

Quote:

Originally Posted by CaliforniaMan This quote stuck out to me, the bolding is mine. "Note that I used the long-term median currently being reported for CAPE. One reader pointed out that this requires future knowledge. It’s a valid concern. I’ve put on my list to investigate a simulation that looks only backwards. But how much it matters may boil down to whether we consider the CAPE median to be constant or variable over retirement-length time spans. I don’t know. There are distinguished experts on both sides of that question." It seems to me this pretty much invalidates the CAPE method since it uses future certain knowledge of an unknown value. The question of whether it is a constant or variable seems to be pointless, since we can never know if it is a constant, all we have is past measurement data. ... It is never permissible and meaningless to use knowledge about the future when making a model.

Agree.

Sent from my SM-G900V using Early Retirement Forum mobile app

[Ed_The_Gypsy]

Quote:

Originally Posted by nun I rebalance whenever my AA deviates by +/-5%. If I need income I'll sell some winners to get closer to my idea AA.....if I'm already there then I just sell the proportional amounts of stocks and fixed interest. Do we really need to think about this so much? There seem to be plenty of ways to successfully withdraw money and my intuition tells me that differences of a few percent in success rates are meaningless as there are too many unknown variables in our future. So pick an AA with at least 50% stocks and use any withdrawal strategy that takes your fancy.

Excellent points. It is nice, though, that someone did look at the issue, ran the numbers and compared methods. Some methods are worse than others.

Questions about withdrawal methods have been raised by newcomers here several times and this work brings a lot of clarity to the subject. I know how to withdraw baffled me when I was starting out. The answer is that it does not need to be complicated. Keep it simple and try to not do something stupid. I have done stupid.



Sent from my SM-G900V using Early Retirement Forum mobile app

[Ed_The_Gypsy]

Svensk, good points, but I ran the numbers. They are anti-correlated, so if my numbers are correct (still checking) when suitably rebalanced, this combo has better returns and holds up better in downturns than VWELX, my current standard.

I will post my findings here when I have a few minutes.

Sent from my SM-G900V using Early Retirement Forum mobile app

[harley]

So, when is one of these big brained economist types going to do a study on a SWR for a 50 year retirement? Or at least 40 years. I think they don't take us early retirement sorts into consideration.

[GTFan]

OP thanks very much for this post, I have checked out Darren's blog from time to time and finally subscribed to email updates based on this thorough analysis. I have his programmer/engineer's mind as well but certainly not his intellect or diligence to create these models.

I found it very interesting to see that his simple CAPE WD model outperformed any model with rebalancing included (at the expense of more volatility, check out in particular the 85% ending stock allocation for a starting 60/40 in his most recent article). And in particular, that CAPE combined with annual rebalancing ends up much worse off than just CAPE WD itself.

As others have said this model is highly dependent on the CAPE median value being valid over time. Having said that, it looks like the simple rebalancing WD model does well enough for me to stick with it.

[nun]

Quote:

Originally Posted by harley So, when is one of these big brained economist types going to do a study on a SWR for a 50 year retirement? Or at least 40 years. I think they don't take us early retirement sorts into consideration.

You can test 40 or 50 year retirements in FIRECalc......with 50/50 AA you'll find that the SWR drops to 3.6% and 3.4%....ish...respectively

[Fedup]

Tsp does EqualTarget, my husband is 20/80, I'm glad to see from this chart he doesn't run out of money.


Sent from my iPad using Early Retirement Forum

[walkinwood]

Quote:

Originally Posted by Ed_The_Gypsy Questions about withdrawal methods have been raised by newcomers here several times and this work brings a lot of clarity to the subject. I know how to withdraw baffled me when I was starting out....p

Me too! I would have been thrilled to have seen this article 8 years ago. With the real-world knowledge gained reading this forum, I have stumbled on a way that works for me - take all distributions & then if more is needed, take it when re-balancing. If that is not optimal - so be it.

[walkinwood]

Quote:

Originally Posted by harley So, when is one of these big brained economist types going to do a study on a SWR for a 50 year retirement? Or at least 40 years. I think they don't take us early retirement sorts into consideration.

Guyton considers a 40 year retirement in his paper on Decision Rules
http://www.bobneiman.com/NWM_Pages/D...ton%202006.pdf

A point to remember is that people like Bengen, Dirk, Guyton and kitces are not economists. They are financial planners or engineers who have used historic data to reach their conclusions.

[samclem]

Kirkpatrick ignored rebalancing in his first blast, which significantly reduced the utility of his findings (unless, for some reason, a person would want to depend entirely on withdrawals to keep their allocation in line. That makes sense only if a person is consumed with worry over cap gains taxes or doesn't have access to a tIRA, Roth IRA, or a 401K).

His new effort still gives too little attention to rebalancing, as far as I can tell. If you are going to bring your portfolio back into balance, then which "withdrawal strategy" (his term for deciding which asset class to draw from) used makes almost no difference. He's only "discovered" that, without rebalancing, the high-return asset class (stocks) will grow to a larger share of the overall portfolio over time, and that this will improve returns (at the risk of higher overall volatility).

Are there a lot of people who don't rebalance their portfolio as an independent step from taking their withdrawals?

[Ed_The_Gypsy]

SCV actually was safer in the 2008 meltdown than VWELX was (a balanced fund and presumably less volatile--but it wasn't). See first figure.

And, going back to 1972, during a period of fast-rising interest rates, combinations of SCV and LTB performed better than VWELX and were less volatile. See other figures. (Portfolio P2 was adjusted to give the same standard deviation as VWELX and it gave more than 1% better long term results. Wow again.)

In fact, going back that far, 50/50 looks better than my 66/34 idea and just as simple.

(Note: VBR only goes back to 2004, but VISVX is a good surrogate. And VUSTX is a good surrogate for TLT, but TLT performed a touch better. I do not know about style drift before 2004.)

Who woulda thunk it?

Comments are welcome.