Evaluating various withdrawal strategies

[omni550]

Darrow Kirkpatrick ("Can I Retire Yet" blog author) has a new blog post where he compares the historic results of 6 different withdrawal strategies. His results show up to a 12% difference in success rate, depending on which strategy is used.

New Research: The Best Retirement Withdrawal Strategies - Can I Retire Yet?

"These results show that the order in which you liquidate asset classes can make a substantial difference in your overall retirement income. Two withdrawal strategies stand out: Equal Withdrawals for its simplicity, and the CAPE Median for its performance. Equal Withdrawals is dead simple, and the CAPE Median isn’t much harder. To use it, when you need cash in retirement, you look up CAPE on the web. If it’s greater than the long-term median (currently about 16), you sell some stocks. If it’s less, you sell some bonds.
According to my research, if you do this consistently, you’ll come out ahead, possibly way ahead, of other withdrawal strategies. So that’s what I’ll be doing, until I learn something better."

I'm wondering if anyone here is using the CAPE strategy or something similar?

omni

 

[GrayHare]

At first glance I expected similar results from the 7-year Moving Average Strategy and the CAPE (10-year) Strategy, but they are acutally quite different, apparently due to CAPE's incorporation of earnings. That was a good read. The CAPE Strategy appears worthy of serious consideration.

 

[tmm99]

Is it surprising that the runner-up is equal withdrawals? Well, it is very surprising to me.
Another surprise to me is rebalancing WR,s success rate being low.

Sent from my iPad using Early Retirement Forum

 

[Ed_The_Gypsy]

I am gob-smacked! My surviving spouse can do Equal Withdrawals. In my dotage, I can do it. For a while.

This is vital information.

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

 

[audreyh1]

Is it surprising that the runner-up is equal withdrawals? Well, it is very surprising to me.
Another surprise to me is rebalancing WR,s success rate being low.

Sent from my iPad using Early Retirement Forum

The success rate wasn't low - it was close to equal withdrawals and higher than 3 of the six methods. It was the ending portfolio value that was low.

I guess the other strategies let the stocks run - on average your ending stock allocation will be higher - so you end up with a higher median ending portfolio value. Portfolio volatility increases as you age, but you ignore it taking out your inflation adjusted amount each year regardless.

These types of models can vary greatly depending on the assets used. For example - intermediate treasuries instead of long might have given quite different results. Same with a broader stock allocation than the S&P 500.

 

[big-papa]

Yet another method!
There is one thing that puzzles me about using PE10 in any withdrawal strategy whether its this one or the one from a while back that dynamically changes the asset allocation based on PE10. That is, PE10 is reasonably correlated to stock performance over the coming 10 years. It's specifically stated to have poor correlation for any single year. Yet these methods make changes as often as once per year. If you're going from a high stock allocation to a low stock allocation (as shown in other studies) or changing your withdrawal from stocks to bonds (as in this study) and doing so in a single year based on PE10 then aren't you basically predicting something about a single year's performance out in the future? Perhaps these methods add some value because PE10 doesn't often change quickly.

I actually did an analysis a while back using PE10. Suppose on year 0, PE10 was predicting a certain return over the next 10 years. Now let time go by and let the returns accumulate for 9 years. Based on those 9 years of genuine returns + the original PE10 you started with what is the most likely return for year 10 so that the original PE10 prediction holds? As you might expect the predicted results weren't all that highly correlated to what the actual results were. But maybe high correlation isn't needed for these methods to add a little value..

 

[DEC-1982]

I enjoyed reading that article. It is something to think about. The problem is, it's a matter of time before another article comes out suggesting that a different strategy is better.

 

[photoguy]

I think it's incredibly dangerous to follow the "CAPE withdrawal" strategy as presented given that we know CAPE is non-stationary (the mean varies over time) due to accounting changes. In fact, CAPE has only been below it's historical median once in the past 25 years (last time was 2009 and before that 1988) so going forward this will probably mean you will draw from equities even when they have a down year.

Also note that the "rebalancing strategy" is not rebalancing as typically discussed.

 

[REWahoo]

I enjoyed reading that article. It is something to think about. The problem is, it's a matter of time before another article comes out suggesting that a different strategy is better.

+1

I dropped my Withdrawal Strategy of the Month Club membership years ago.

 

[marko]

I dropped my Withdrawal Strategy of the Month Club membership years ago.

I haven't run the numbers too heavily, but I get the feeling that, provided one of the 20 or so conventional wisdom strategies are used, the differences are minimal.

I tend to shy away from things that require a lot of work to deliver a .01% advantage. And if you're shaving it so close that .01% makes a difference, maybe the real problem is that you're not ready to RE.

 

[Lsbcal]

...
"These results show that the order in which you liquidate asset classes can make a substantial difference in your overall retirement income. Two withdrawal strategies stand out: Equal Withdrawals for its simplicity, and the CAPE Median for its performance. Equal Withdrawals is dead simple, and the CAPE Median isn’t much harder. To use it, when you need cash in retirement, you look up CAPE on the web. If it’s greater than the long-term median (currently about 16), you sell some stocks. If it’s less, you sell some bonds.
According to my research, if you do this consistently, you’ll come out ahead, possibly way ahead, of other withdrawal strategies. So that’s what I’ll be doing, until I learn something better."

I'm wondering if anyone here is using the CAPE strategy or something similar?

omni

Relying on a fixed number is dangerous. Two authors that I know of (Siegal and more recently Swedroe) have noted difference in accounting and other things that might change CAPE numbers upwards.

I'm reminded of a long ago market newsletter (before the web) which came out with a strategy that used the dividend yield of stocks i.e. one measure that was not normalized to anything like interest rates. The backtesting looked good but that was before the 1990's bull market.

 

[big-papa]

I enjoyed reading that article. It is something to think about. The problem is, it's a matter of time before another article comes out suggesting that a different strategy is better.

I actually love to play with spreadsheets. Like the author, I'm an engineer. I have yet to find a method that I would call "better", though there are definitely some that are "worse" . There literally are a many ideas as there are people to think them up and since different people have different goals, it's doubtful if there will ever be a one-size-fits-all method.

 

[audreyh1]

I think it's incredibly dangerous to follow the "CAPE withdrawal" strategy as presented given that we know CAPE is non-stationary (the mean varies over time) due to accounting changes. In fact, CAPE has only been below it's historical median once in the past 25 years (last time was 2009 and before that 1988) so going forward this will probably mean you will draw from equities even when they have a down year.

Also note that the "rebalancing strategy" is not rebalancing as typically discussed.

What was wrong with their rebalancing technique? They indicated they withdrew from the two assets classes such that the fund would be balanced after the withdrawal. I didn't see anything wrong with that.

The CAPE thing bothered me too - the outcomes in recent history would look super different than then ones in the last 3 decades of the 20th century.

 

[audreyh1]

I enjoyed reading that article. It is something to think about. The problem is, it's a matter of time before another article comes out suggesting that a different strategy is better.

You also have to trust that the author implemented the models completely correctly to get the published results. It helps if someone else checks their work.

I would have liked to see the spread/volatility of the different methods, not just the median values.

 

[W2R]

+1

I dropped my Withdrawal Strategy of the Month Club membership years ago.

+2

I think that for me, the method described in the article:

(1) is more complicated than a withdrawal scheme needs to be; and
(2) is too reminiscent of market timing.

I am doing fine by just withdrawing the coming year's spending money from my money market account at Vanguard during the first week in January each year, and moving it to my bank.

For example, this weekend I will download my 2016 spending money from Vanguard to my bricks and mortar bank. Then I will rebalance and leave everything alone until next year.

My dividends and capital gains distributions are dumped into my Vanguard money market account throughout the year, so it is more than replenished each year. The extra cash will be used to buy funds as needed during rebalancing.

The easier and the simpler the withdrawal procedure, the more foolproof it is and the less likely mistakes would be, it seems to me.

 

[CaliforniaMan]

+2

I think that for me, the method described in the article:

(1) is more complicated than a withdrawal scheme needs to be; and
(2) is too reminiscent of market timing.

I am doing fine by just withdrawing the coming year's spending money from my money market account at Vanguard during the first week in January each year, and moving it to my bank.

For example, this weekend I will download my 2016 spending money from Vanguard to my bricks and mortar bank. Then I will rebalance and leave everything alone until next year.

My dividends and capital gains distributions are dumped into my Vanguard money market account throughout the year, so it is more than replenished each year. The extra cash will be used to buy funds as needed during rebalancing.

The easier and the simpler the withdrawal procedure, the more foolproof it is and the less likely mistakes would be, it seems to me.

With such a limited, and as has been pointed out non-stationary, data set it is very easy to over fit, that is tune the model to fit the noise. It is why I have avoided so far making any models to fit various withdrawal and balancing schemes. IMHO we really don't have enough information to fine tune our withdrawal or balancing like this and it is a waste of good time and energy that could be better spent worrying about the weather or when the roast should be taken out of the oven.

 

[Lsbcal]

If the OP would look at the VPW thread here and on Bogleheads, he will have at least a historical spreadsheet tool to investigate. A good compliment to FIRECalc. There is even a place to select a fixed starting withdrawal percentage (see the Table tab in VPW).

I think this is an excellent backtest tool and also a way to think about things. See start retirement years 1968 and maybe 1929 for some worst case scenarios.

 

[photoguy]

What was wrong with their rebalancing technique? They indicated they withdrew from the two assets classes such that the fund would be balanced after the withdrawal. I didn't see anything wrong with that.

If I understand the article correctly, if the withdrawal (4%) is not sufficient to reach the target allocation, the portfolio will be left out-of-whack.

I guess his rebalancing approach will fail to maintain the target allocation whenever stocks have a greater than 8% gain/loss (he uses 50-50 portfolio).

 

[Svensk Anga]

I'm reminded of a long ago market newsletter (before the web) which came out with a strategy that used the dividend yield of stocks i.e. one measure that was not normalized to anything like interest rates. The backtesting looked good but that was before the 1990's bull market.

Sounds like you are referring to "Growth Fund Guide". My father subscribed in the 1980's and passed the newsletters on to me. It probably set my FIRE back a couple years by keeping me light on stocks. It might have done serious damage except we were just starting out and there wasn't enough in the stash to make a material difference. Stocks just kept going up despite the low dividend yields, so I eventually figured the guy's system was broken. Maybe there was value then in seeing that the newsletters gurus did not have the answers.

 

[Lsbcal]

Sounds like you are referring to "Growth Fund Guide". My father subscribed in the 1980's and passed the newsletters on to me. It probably set my FIRE back a couple years by keeping me light on stocks. It might have done serious damage except we were just starting out and there wasn't enough in the stash to make a material difference. Stocks just kept going up despite the low dividend yields, so I eventually figured the guy's system was broken. Maybe there was value then in seeing that the newsletters gurus did not have the answers.

Yep, you win the prize for coming up with that one.

As I recall I was impressed with their graphics but I don't think it affected my investing much. I think I stopped subscribing in the early 1980's. Just was shredding some stuff a few days ago and came across some old notes on this.

Those were great years in the markets, 1982 through to 2001. People actually began to think that 20% up was the norm! Rude awakening in the new millennium.

I wonder how our current thinking will evolve in the next few decades.

 

[audreyh1]

If I understand the article correctly, if the withdrawal (4%) is not sufficient to reach the target allocation, the portfolio will be left out-of-whack.

I guess his rebalancing approach will fail to maintain the target allocation whenever stocks have a greater than 8% gain/loss (he uses 50-50 portfolio).

OK - good point. That is not complete rebalancing. Then the model is just confusing.

 

[photoguy]

Yeah I hate it when people don't use an appropriate default when testing their new fangled algorithm. Sometimes I think it's intentional (instead of cluelessness) but either way doesn't instill confidence in his work.

Sent from my Nexus 5 using Early Retirement Forum mobile app

 

[donheff]

I read an article years ago that suggested a starting WR % calculated as 1/Schiller's PE/10 (CAPE) at the start of ER. That would amount to 3.8% today. And someone else suggested reducing that a bit ~.83/PE/10 (3.2% today). Does anyone know where I could find a reference that discusses using the PE/10 to set a starting WR?