Using Shiller PE to Time the Market

[RobLJ]

Using Shiller PE as an investment timer is a bad idea--see Hussman funds.
Using it to tweak investment allocations (from over to comparatively undervalued) is not a bad idea--I did a version of that in 1999-2002 and in 2006-8 and am now rebalancing to foreign stocks, but YMMV.
I see Dallas made my point above, which was to consider pulling some or a lot of gains from overvalued markets to undervalued, which would be Emerging Market, Europe, perhaps China, etc, perhaps on a slow and steady periodic reallocation. If the US continues to soar, you will continue to benefit on your prior allocation but also have side bets from the reallocations. I've done this over the last 14 years, which smoothed out both the high years and the low years since there generally was an undervalued market jumping--other than late 2008/9.

The market can stay irrational longer than you remain solvent or sane. (Or earnings can accelerate to justify the PE--the essence of growth investing.) And the 2008 crash was a once or twice in a lifetime anomaly, which raises some caution in the Schiller 10 data--one would think. Will Texas score <10 points against OU every year and get beat by 50 points?

Using Schiller PE10 to tweak a safe withdrawal % guide is an interesting thought that makes intuitive sense, although I'll think about it more.

 

[samclem]

Resurrecting this thread, sorry.
It turns out that Kitces and especially Pfau have looked at using PE10 as a means to adjust portfolios for reduced risk/greater return, and there appears to be something useful here.

As many people know, this idea is far from new. Benjamin Graham suggested such a weighting scheme. In a 2011 paper, Dr Pfau did some analysis of Graham's system (he used PE10 as the valuation metric, of course it didn't exist when Graham and Dodd did their work)

From this paper:

As a case study, I will compare a fixed 50/50 asset allocation strategy against a strategy introduced by Graham and Dodd (1940), in which investors maintain a 50/50 asset allocation when valuations fall within a range between 2/3 and 4/3 of their historical average value; the stock allocation is 75% when valuations are less than 2/3 of their average and 25% when valuations are more than 4/3 of their average. These numerical bounds correspond to evolving PE10 values of approximately 10 and 21 over time

The results were impressive. Compared to a static 50/50 rebalanced portfolio, on average retirees using the valuation-shifting strategy could have afforded SAFEMAX withdrawal rates about 1/2% higher, and it was 2% higher for some retirement years. That might not sound like a lot, but if we were planning to take 4% per year, that's 12% to 50% increase in allowable spending. The best part is that it is achieved with less portfolio volatility.

Pfau didn't try to datamine and reverse-engineer the signal points or weightings, he just used what Graham proposed.

In this paper (Be sure to pull up the second available dowload--"35006"--it has been updated), Pfau takes a look at a previous popular study that had found negative results from PE10 timing. Pfau found the opposite--jumping from 0% to 100% stocks (to short-term commercial paper) based on 4 possible timing signals each produced average geometric returns better than 100% stocks. More significantly for most of us, he looks at other AA "spreads" centered on 50/50 (10:90, 20:80. 30:70, 40:60) and how an investor would have done (see Table 3). In general, the volatility of the "timing" portfolio was higher than the fixed 50/50 portfolio, but the average geometric return was from .3 to 1.5% better per year. The Sharpe ratio (a measure of how much risk is rewarded) was better for all the valuation-based portfolios than for the fixed 50-50 portfolio.

Pfau and Kitces collaborated on this piece. Similar positive results for varying the portfolio weightings based on valuation. As we have recently discussed elsewhere, they also found that using govt bills rather than commercial bonds proved to be much more effective during times that stocks were overvalued by historic standards (as they are today). Rising equity allocations later in retirement are also indicated when a person retires at a time of high stock valuations (i.e. it's best to start with a higher % of bills at first, to get past the point where sequence of returns risk could scuttle your plans if the stock market crumps--and crumping is more likely if you retire when valuations are high.

I think there's something to this. And it is intuitive--own less stocks when they are more likely to take a plunge. The historical record seems pretty solid, and it isn't highly dependent on picking just the right trigger points or AAs. Pfau even showed it was effective in the Japanese market.

DMT? Maybe. But hardly day trading or reacting to the 200 day moving average. Most of these schemes would trigger a change in allocation on average intervals of 4-5 years, but the wait can be a lot longer than that. PE10 has very little predictive ability on a year-to-year basis, and (as noted previously here) that may shield these techniques from being used by many advisors/MFs, and even institutional investors--you'll be "wrong" for years at a time, and that's not good for a career. Even if the technique works well over the long term.

 

[samclem]

+1
The big problem with all of these in my opinion is that there is no way to get any meaningful historical statistics not to mention reasonable projections of future probabilities. No way to do a FIRECalc type measurement for example on Greece or Russia or even China. Might be some great values here but at this point in my life they are not for me.

+1. As an alternative to US bills/bonds as a place to move the money when US equities are overvalued, I don't think foreign equities are a good idea. We're trying to get ahead of the future US market with these AA shifts in response to today's (US) PE10--what's cheap or expensive in other markets today might even be negatively correlated to what happens when the US market dives--maybe they dive worse, even if their valuations today are favorable. I'd be more content to park money in USG bills and wait.

But what to do about shifting the allocations of non-US equities in our portfolio? Shift them into bonds/bills (US? Foreign?) when we reduce US equities? Keep them constant at a fixed % of portfolio and just use the shifting-with-PE10 strategy with US equities? As a first guess: Based on the apparent increased correlations between US and foreign equities, I'd reduce my foreign holdings along with US ones when US PE10s are high. And I'd probably stick with USG bills rather than foreing ones--if there's a "flight to quality", the money will probably come to the US again (I hope!).

 

[Lsbcal]

....
But what to do about shifting the allocations of non-US equities in our portfolio? Shift them into bonds/bills (US? Foreign?) when we reduce US equities? Keep them constant at a fixed % of portfolio and just use the shifting-with-PE10 strategy with US equities? As a first guess: Based on the apparent increased correlations between US and foreign equities, I'd reduce my foreign holdings along with US ones when US PE10s are high. And I'd probably stick with USG bills rather than foreing ones--if there's a "flight to quality", the money will probably come to the US again (I hope!).

I tend to keep US/international ratios constant when adjusting the equity/bond mix. From my studies equity returns are much more correlated to the yield curve then to PE10. I too would stick to US bonds as the bonds are there for capital preservation with only modest hoped for real returns.

 

[big-papa]

In the paper : Investing for Retirement: The Defined Contribution Challenge -- Ben Inker and Martin Tarlie , currently at GMO LLC - Home

They talk about the correlation between the Shiller P/E value and future returns.

At 1 year, it's about a 20% correlation,
At 5, almost 45%,
at 10 years, 60%,
At 20, about 70%.

See chart 10 on pg. 7.

So, like others have said, not a perfect system. But, overall, this paper addresses a lot of the variables of asset allocation, and a thought provoking discussion of valuation, timing/timeframes, returns simulations, etc.

-CC

Good article - thanks for the link! There are those of us who don't use the SP500 and I have yet to find PE10/CAPE for other US based market cap/valuation based indices . Yes, there is correlation from them back to the SP500), but I'd bet that given that other indexes are twice removed, I suspect that the correlation to future returns is likewise reduced as well. Wonder if Fama-French have the ability to extract it from their work?

 

[urn2bfree]

My understanding is that the CAPE10 has a negative correlation with returns of no better than -0.5 for durations of less than 8 years and peaks at -0.65 at 18 years, with poor correlation for shorter time frames like 1 year. This seems to make it a moderately poor tool for timing markets. Is it the best of a lot of bad tools? Or just another bad tool?


Sent from my iPad using Early Retirement Forum

 

[samclem]

My understanding is that the CAPE10 has a negative correlation with returns of no better than -0.5 for durations of less than 8 years and peaks at -0.65 at 18 years, with poor correlation for shorter time frames like 1 year. This seems to make it a moderately poor tool for timing markets. Is it the best of a lot of bad tools? Or just another bad tool?

By what standards do we judge a tool?
-- Ease of use? The PE10 is easy to calculate (for US stocks, anyway), and easy to apply. Shifting assets (with PE10 or anything else) can have tax implications, but the availability of IRAs and 401Ks reduces that somewhat, particularly if used to support a moderate shift (i.e. from an allocation of 30% stocks (high PE10), 50% stocks ("average" PE10) to 70% stocks (low PE10). Most people have got more than 20-40% of their assets in accounts that don't cause tax problems if traded.
-- The results the tool produces? Using the Graham-Dodd criteria, Pfau looked at how $1 invested in 1871 would have grown by the beginning of of 2010. A fixed 50/50 allocation produced an ending value of $13,426, switching between 30 and 70% stocks produced an ending value of $33,211, so about 247% higher. And the switching portfolio had less downside variability (measured by Sortino ratio). Better returns, less risk of the important kind--that's good.
-- Is it a versatile tool? Depends. It doesn't work for short-term market timing, so it won't help if we need to amass a fortune by next Thursday. But it is a robust and flexible tool--it appears to be insensitive to the allocations chosen or the trigger points.

It's not a perfect tool by any means, but I don't think I'd call it "bad".

 

[big-papa]

Good article - thanks for the link! There are those of us who don't use the SP500 and I have yet to find PE10/CAPE for other US based market cap/valuation based indices . Yes, there is correlation from them back to the SP500), but I'd bet that given that other indexes are twice removed, I suspect that the correlation to future returns is likewise reduced as well. Wonder if Fama-French have the ability to extract it from their work?

Well, a quick internet search revealed that somebody did this already with small cap value, but using PE10 back in 2005. Kinda crazy results, but it also shows that the last time period ended in 1980 when interest rates were pretty astronomical......
Edited: Switching Allocations with Small Cap Value

 

[rbmrtn]

Well, a quick internet search revealed that somebody did this already with small cap value, but using PE10 back in 2005. Kinda crazy results, but it also shows that the last time period ended in 1980 when interest rates were pretty astronomical......
Edited: Switching Allocations with Small Cap Value

I believe your author used to post here as "JWR1945"

I noticed in the credits from Pfaus paper ...

I am also grateful to Rob Bennett for motivating this investigation

Is this not the same AKA "*****" that was banned from this forum, bogleheads, RetireEarlyhompage, and maybe others. I recall he advocated PE10 switching as VII ( Value Informed Indexing )

 

[ERD50]

By what standards do we judge a tool?
-- Ease of use? ...
-- The results the tool produces? ... 30 and 70% stocks produced an ending value of $33,211, so about 247% higher. And the switching portfolio had less downside variability (measured by Sortino ratio). Better returns, less risk of the important kind--that's good.

-- Is it a versatile tool? Depends. It doesn't work for short-term market timing, so it won't help if we need to amass a fortune by next Thursday. But it is a robust and flexible tool--it appears to be insensitive to the allocations chosen or the trigger points.

It's not a perfect tool by any means, but I don't think I'd call it "bad".

I need to read through it in detail, but what concerns me about any tool like this is, it may provide superior results in back-testing, but did it 'generally work' in all the cycles of the back-test? And by 'generally work', I mean were there any cycles where the results were significantly worse than Buy & Hold?

The trouble with the historical reports like FIRECalc is that we only have a handful of cycles in all that data (many years are just a basic repeat of a cycle in that period), and the same applies to a tool like this. We may really only see one, two, maybe three economic cycles in our retirement? So if those superior returns are based on averages, but the system works against us on one of our future cycles, then it may not be so great, despite averages.

I guess I'm just chicken. Presented with unknown paths, I feel like sticking with the path I'm on. Right or wrong, who knows? Or is this like the Door # 3 paradox?


-ERD50

 

[haha]

It is very hard for me to imagine how buying stocks only when they are at average price or better could possibly be worse over time than buying them at all times irrespective of valuation. (Leaving aside any taxation)

Ha

 

[samclem]

And by 'generally work', I mean were there any cycles where the results were significantly worse than Buy & Hold?

IIRC, one of the papers had stats like you're looking for, it showed that shifting the allocation performed worse than a fixed strategy on about 10 of the 100+ cycles examined. I don't recall how big the lag was.

 

[ERD50]

It is very hard for me to imagine how buying stocks only when they are at average price or better could possibly be worse over time than buying them at all times irrespective of valuation. (Leaving aside any taxation)

Ha

Agreed, but that is only one part of the equation, isn't it?

Timing like this also involves selling. And it looks to me like there were times when the trigger would have you selling and missing out on the remainder of a long bull market. And you may never see an entry point below your exit point.

That's the problem I see. Of course it's unreasonable to expect any tool to get it right every time. But I keep asking myself, can I risk the potential reward against the risk that I get out and can't know when to get back in? You can look at a chart in hind-sight, and say "look, you could get in here!", but as it is happening, you might miss a 'near entry' point, while you wait for a better one that never comes. I just don't know.

If I was changing my AA, I would wait for above average valuations to get out, as I don't plan on getting back in. Heck, I'd probably still DCA out.

-ERD50

 

[samclem]

It is very hard for me to imagine how buying stocks only when they are at average price or better could possibly be worse over time than buying them at all times irrespective of valuation. (Leaving aside any taxation)

Ha

But it's not just "buying" stocks, right? It's actively selling stocks you already own when valuations get too high. During the time you don't own them, they pay dividends (maybe at a higher rate than the bonds/bills you replaced them with), etc, so there's the possibility to lose out. But, on average, it seems to work.

 

[samclem]

Of course it's unreasonable to expect any tool to get it right every time. But I keep asking myself, can I risk the potential reward against the risk that I get out and can't know when to get back in? You can look at a chart in hind-sight, and say "look, you could get in here!", but as it is happening, you might miss a 'near entry' point, while you wait for a better one that never comes. I just don't know.

It's got this going for it:
-- It's mechanical, not based on emotions, hunches, etc.
-- It doesn't appear to be very sensitive to the allocation levels chosen or the trigger points. If it "worked" at PE10 triggers of 10 and 20, but failed at 12 and 22, I'd be very suspicious. If it worked with two allocations (e.g. 40 and 60% stocks), but not with three graduations (30, 50, and 70), then I'd be more leery. I don't really know if PE10 is going to drift higher or lower overall for structural reasons over the next decades, so it's good that it appears not necessary to hit things right on back-tested perfect spots.

 

[ERD50]

IIRC, one of the papers had stats like you're looking for, it showed that shifting the allocation performed worse than a fixed strategy on about 10 of the 100+ cycles examined. I don't recall how big the lag was.

Thanks, I'll try to dig that out - I didn't see it in my first pass of one of the papers.

~ 10/100+ is sounding pretty good, especially if the lag was small. Taking a small risk is not painful if I 'lose', the large ones are the worry.

It's got this going for it:
-- It's mechanical, not based on emotions, hunches, etc.
-- It doesn't appear to be very sensitive to the allocation levels chosen or the trigger points. If it "worked" at PE10 triggers of 10 and 20, but failed at 12 and 22, I'd be very suspicious. If it worked with two allocations (e.g. 40 and 60% stocks), but not with three graduations (30, 50, and 70), then I'd be more leery. I don't really know if PE10 is going to drift higher or lower overall for structural reasons over the next decades, so it's good that it appears not necessary to hit things right on back-tested perfect spots.

Yes, I've done some models that ended up looking just like that - they 'worked' with certain inputs, but as you move up/down from there, the returns jump around rather than a smooth progression - which tells me the 'working' points were a bit of data-mining (not the good kind).

I may be seeing the effect of some 'behavioral economics' here - taking an active roll in the decision and possibly being wrong might be more 'painful' than doing nothing (staying the course), and being wrong - but staying the course is a decision as well. It just doesn't 'feel' like it!

-ERD50

 

[karluk]

It is very hard for me to imagine how buying stocks only when they are at average price or better could possibly be worse over time than buying them at all times irrespective of valuation. (Leaving aside any taxation)

Ha

I don't find this hard to imagine at all. Buying stocks when they are expensive isn't competing with buying stocks when they are cheap. Rather, buying stocks when they are expensive is in competition with all of the other choices to invest one's money at that given point in time. If you don't put your money in stocks, it doesn't simply disappear. It's sitting in some other investment vehicle such as bonds, cash, real estate, gold, or any other choices you can think of.

So the question for someone contemplating PE10 market timing is not whether stocks are expensive right now, but whether there is another investment available that will offer better investment performance in the time one needs to wait in order for stocks to become cheap again. Maybe there is, maybe there isn't. It all depends on how big the gains are before the correction occurs vs. how big the correction is when it finally happens. If you indulged in PE10 market timing in, say, 2012, you would have missed out on some really large gains in equities that most likely will greatly exceed whatever losses happen in the next correction. If you had been clairvoyant enough to hold your stocks until now, in spite of sky high valuations, you will have enjoyed more of the gains and may very well fare better. It all depends on what happens in the future, which unfortunately is not something I am privy to.

 

[samclem]

I may be seeing the effect of some 'behavioral economics' here

Well, I'll admit I'm worried about the same thing in my case. It's quite a coincidence that I'm falling out of love with strict fixed rebalancing and starting to like the idea of going to a lower stock allocation just as markets are at high PE10. I'm telling myself it is evidence-based, but . . .

 

[NW-Bound]

... the question for someone contemplating PE10 market timing is not whether stocks are expensive right now, but whether there is another investment available that will offer better investment performance in the time one needs to wait in order for stocks to become cheap again. Maybe there is, maybe there isn't...

+1

Equities have been expensive in terms of P/E, but so are bonds in terms of yield. Foreign stocks are a bit less expensive, but I have enough of that, and they have been trounced by the S&P. All I can do now is to wait with my 25% cash, and see how things develop.

 

[RetireBy90]

So I like a good "system" as much as the next guy esp when it allows me to improve my financial situation and doesn't require much effort on my part. However, the problem I see is that with few exceptions, once a "system" is popularized then market traders will trade ahead of it and it won't work any more. These are interesting to study and discuss, but seems to me like spinning gold from staw. Lots of examples where a system can work, but many more where they don't work. So how do you know which to trust and how do you know when they are loosing the effectiveness?

Just asking

 

[REWahoo]

So how do you know which to trust and how do you know when they are loosing the effectiveness?

Just asking

Simple: only trust a system who has you buy stocks that will go up and sell any that will go down. Once it stops doing that, look for another system.

 

[Lsbcal]

Well in the end one must choose a method. For me, I like having my own plan which is an amalgam of reasonably well researched and respected ideas (includes PE10 in there). But they are just ideas supported by existing data and there will always be new ones cropping up.

It hard to stay on plan. I don't like tracking error which is why I select an alternate portfolio to benchmark against. But how much and how long does one accept "tracking error" before giving up and going with the alternate portfolio?

 

[Lsbcal]

Well, I'll admit I'm worried about the same thing in my case. It's quite a coincidence that I'm falling out of love with strict fixed rebalancing and starting to like the idea of going to a lower stock allocation just as markets are at high PE10. I'm telling myself it is evidence-based, but . . .

At the start of this year, I reduced our equity percentage. I think the main reason was to reduce risk because we could support our spending needs with that new AA. The main tool I used to convince myself was VPW.

 

[ERD50]

So I like a good "system" as much as the next guy esp when it allows me to improve my financial situation and doesn't require much effort on my part. However, the problem I see is that with few exceptions, once a "system" is popularized then market traders will trade ahead of it and it won't work any more. These are interesting to study and discuss, but seems to me like spinning gold from staw. Lots of examples where a system can work, but many more where they don't work. So how do you know which to trust and how do you know when they are loosing the effectiveness?

Just asking

I agree in general, but this is a long term method. Not likely to attract the 'get rich quick' or hedge fund types. Patience is required.

-ERD50

 

[haha]

It hard to stay on plan. I don't like tracking error which is why I select an alternate portfolio to benchmark against. But how much and how long does one accept "tracking error" before giving up and going with the alternate portfolio?

I understand a public money manager worrying about tracking error. But why should an individual investor become concerned about tracking error? Nobody can fire him, no clients can leave. IMO what counts for this person is absolute return over time.

Ha