CAPE As A Risk Management Tool

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Park
Posts: 467
Joined: Sat Nov 06, 2010 4:56 pm

CAPE As A Risk Management Tool

Post by Park » Wed Oct 04, 2017 8:05 am

http://advisoranalyst.com/glablog/2014/ ... tml/print/

The above article looks at the predictive power of Q ratio, PE10, PE1 and market cap/GNP ratio among others. See Matrix 1. There is some predictive value at 5 years, which is increased at 10 years. At 15 years, they tend to peak, with declines by 20 and 30 years. At 15 years, they have about the same predictive power, except PE1 has less than the other three.

http://www.starcapital.de/files/publika ... imling.pdf

Look at Figure 6 in the above link. It looks at the correlation of CAPE (PE10) with average real stock returns over the next 10-15 years. Once CAPE is less than 10, return has been a minimum of 5%. But at a CAPE of about 14, return has been as low as 0%. Once CAPE is greater than 60, returns are usually negative, with the best being around 2%. But at a CAPE of 58, there has been a return of 5%.

CAPE might be the best predictor, and its greatest ability to predict may be over the next 10-15 years. But even under those circumstances, its ability to predict is limited to the extremes to some extent. Between a PE10 of 14 and 58, the variability of returns is considerable.

Look at Figure 8. It divides maximum drawdown according to CAPE. The maximum historical drawdown over the next 3 years tends to increase, as CAPE increases. For CAPE 0-10, it is -51%. For CAPE 30 or greater, it is -76.8%. However, the average (all countries) maximum drawdown over the next 3 or 15 years is interesting. For CAPE 0-10, it is -5.7% and -5.2% respectively. For CAPE 30 or greater, it is -28.8% and -39.5% respectively. There is a steady increase in drawdown as CAPE increases. With a CAPE of less than 20, the average maximum drawdown over the next 3 or 15 years is IMO somewhat modest.

This illustrates why market timing using valuation is a long term tool, that works best at the extremes. It's another reason why market timing is tends to be unprofitable.

However, the use of valuation as a risk management tool is another question entirely. Although far from perfect, CAPE to minimize drawdown risk has worked reasonably well historically.

https://papers.ssrn.com/sol3/papers.cfm ... id=2129474

About CAPE being less useful for short time horizons, the above paper may be a dissenting view. Fig 3A is relevant. The relationship of CAPE to real average US stock returns over the next 1, 3, 5, 7 and 10 years is given. If anything, the relationship progressively weakens as the time interval increases. But that progressive weakening is basically at the extremes, PE less than 10 or greater than 40. Between PE10-40, there's not much relationship between return and time interval. Fig 7 gives similar data, except for non US markets. With a PE of less than 15, the relationship between PE10 and return weakens, as the time interval is increased from 1 year to 10 years. Otherwise, I don't see a relationship. It should be noted that only averages are given, with there being no mention of how variable the results are.

My conclusion is that it might be possible to use low valuations (PE10 of less than 10 or 15) to invest over the next 1-5 years. As to how to profit from bubbles by shorting, that would be at best difficult.

asif408
Posts: 1097
Joined: Sun Mar 02, 2014 8:34 am
Location: Florida

Re: CAPE As A Risk Management Tool

Post by asif408 » Wed Oct 04, 2017 10:04 am

My takeaway about CAPE has been that it is useful along with some of the other valuation predictors (e.g., P/B) as a blunt tool to use for determine where to allocate to stocks for intermediate term horizons. I don't think it has much use for going back and forth between stocks and bonds as a lot of folks try to use it for, expect in extremes like 2000, when most countries had CAPE ratios in the 40-80 range, with the exception of Australia and some of the Asian countries (e.g., Singapore, Hong Kong).

This is how I use CAPE:

1) Along with other valuation factors such as P/B, first determine which group of investable countries offer the lowest CAPE
2) Find the cheapest way to invest in those countries and tilt portfolio to those regions, ensuring I buy a basket of countries and not just a few (since you could risk putting all your money into a stock market that goes to zero, like China or Russia has in the past)
3) The more years of underperformance of the low CAPE stocks relative to higher CAPE stocks, the more tilt I will consider.
3) Sit tight for 5-10 years and reassess

For example, in the late 90s Singapore, Hong Kong, and Australia were at low CAPE levels. Of course, they had drastically underperformed everything else in the years leading up to 1998, so their low CAPEs were deserved. If you had invested in those countries going forward they still dropped along with the US during the dot.com bust, but that recovered quicker and performed much better going forward. So it worked, you just had to wait 4 or 5 years to see results, which some aren't willing to do.

User avatar
TheTimeLord
Posts: 4883
Joined: Fri Jul 26, 2013 2:05 pm

Re: CAPE As A Risk Management Tool

Post by TheTimeLord » Wed Oct 04, 2017 10:17 am

I look at CAPE as a fever, symptom not a cause. If the patient has the flu they will likely have a fever but just because the patient has a fever doesn't mean he has the flu.
Run, You Clever Boy! 8723

onthecusp
Posts: 302
Joined: Mon Aug 29, 2016 3:25 pm

Re: CAPE As A Risk Management Tool

Post by onthecusp » Wed Oct 04, 2017 10:28 am

Kitces put together some good analysis on this. For example:

https://www.kitces.com/blog/should-equi ... valuation/

As indicated in his reply in the comments at the bottom of that page his 'big conclusion' is not so much that 4% SWR is in danger at high PE10 values as that even higher SWR are reasonable if you retire in a period of low PE10. Alas I am most likely to pull the plug during a high period.

lazyday
Posts: 3015
Joined: Wed Mar 14, 2007 10:27 pm

Re: CAPE As A Risk Management Tool

Post by lazyday » Thu Oct 05, 2017 12:30 pm

Anyone seen P/B or CAPE that's adjusted for industry weighting?

For example, by reweighting each country to global industry weighting.

Especially for PB this would make sense to me. P/B really should vary a lot by industry. So for example if UK is more energy weighted and US more tech, the overall P/B should be higher for US.

Park
Posts: 467
Joined: Sat Nov 06, 2010 4:56 pm

Re: CAPE As A Risk Management Tool

Post by Park » Sat Oct 07, 2017 5:02 pm

About valuations predicting stock returns best over longer periods of time, the Japanese stock market illustrates how long bubbles can take to pop.

http://investingforaliving.us/2017/10/0 ... 0-edition/

The above link gives a figure showing the relationship of CAPE to the Japanese stock market in the 1980s. Around 1980, CAPE increased above 20. For the next 47 months, it was between 20-39. It then spent 19 months between 40-59. This was followed by 9 months between 60-79. Finally, about the last 33 months were 80 or greater. The peak was reached at the end of those 33 months with a CAPE of 94 in December 1989.

Although the link above from Mebane Faber would suggest that cheap stock markets can revert to the mean over periods considerably shorter than 5-10 years, that doesn't always hold true. The Russian stock market has been cheap for years. One reason for that is the concern that the Russian government won't respect property rights. I have read that the Hong Kong stock market was cheap for years also, for somewhat similar reasons. GREK, the Greek ETF, presently has a price/book of 0.61 as per the ETF.com website. That low P/B ratio is despite the fact that the Greek stock market index is at a similar level to where it was January 2012, about 5.5 years ago. Even if cheap stock market prices don't mean revert, you can still profit from them though, because dividends may be higher.

Why do valuation predict individual stock returns over a shorter time horizon than they predict stock market returns? One reason may be that individual stocks go up and down for cyclical reasons often. This also happens in stock markets, with the nadir of March 2009 being an example. However, there may be a greater tendency for stocks markets, relative to individual stocks, to go up or down for structural reasons. Those structural reasons may be more long lasting, resulting in valuations being slower to predict stock market returns than individual stock returns.

User avatar
CyclingDuo
Posts: 926
Joined: Fri Jan 06, 2017 9:07 am

Re: CAPE As A Risk Management Tool

Post by CyclingDuo » Sat Oct 07, 2017 5:21 pm

Can anyone provide any recent examples (say the past 20 years) where CAPE indicated one should be buying equities?

Valuations:

Yes, they're high, but since 2009 we've become enchanted by what everyone calls Shiller's Cape. We went back to the 1980s and found that in an entire 10-year span, there was only one mention. This isn't a law that's been in existence ever since the New York Stock Exchange has been around. We weren't even aware of it in the last cycle. My argument against the whole thing is: It's never gotten you in the market. We can catalog a bunch of articles that show no one's ever said to buy. It's always been "The market's overpriced," "The market's expensive," "The market's high," but no one's ever said buy. In July 2009, there were some articles saying that according to some valuation measures, the market was fully valued. In July 2009! To me, something that's never told me to buy is not something I'm going to listen to when deciding when to sell.


Laszlo Birinyi: http://www.businessinsider.com/laszlo-b ... age-2017-9

https://www.birinyi.com/

Park
Posts: 467
Joined: Sat Nov 06, 2010 4:56 pm

Re: CAPE As A Risk Management Tool

Post by Park » Sat Oct 07, 2017 6:13 pm

About the relevant importance of cyclical versus structural reasons for cheap stock markets, I think it may be important to differentiate between incidence and prevalence.

When you look at the incidence of cheap stock markets, both cyclical and structural reasons will play an important role. However, when one looks at the prevalence of cheap stock markets, structural reasons may play a more important role. The reasoning is as follows.

A stock market will have bear market perhaps once every 4 years. So prices will be cheap for perhaps 1-1.5 years every approximately 4 years in many (virtually all?) stock markets. That's a cyclical reason for a cheap stock market. Since world markets tend to cycle together, especially in the bear phase, there may be many stock markets that are cheap at around the same time. If one looks back at the incidence of cheap stock markets, they will be frequent. But those periods of cheapness won't last that long and they may occur at around the same time for different stock markets.

Structural reasons for a cheap market may be less common. However, structural reasons for a cheap market can persist. Especially if you look at what markets are cheap outside of periods, such as 2008-2009, those that are cheap on a long term basis may be more prevalent.

So if you're investing in cheap stock markets outside of the cyclical bear time frame, you will likely need a longer time horizon.

About expensive markets, it doesn't surprise me that they can persist, with the Japanese market of the 1980s being an example. Stock markets have correcting mechanisms, when they're too cheap or too expensive. When too cheap, people buy. When too expensive, people short. But shorting is more complicated and risky than buying, so expensive markets may last longer than one would think.


Edited to include the following.


Valuations can be used to predict individual stock returns and individual stock market returns.


As mentioned above, valuations predict return for individual stock markets at 5 years. However, the signal is stronger at 10 years and stronger yet at 15 years. The following link from Vanguard is relevant; see Fig. 2.

Fig.2 gives the proportion of variance of future real US stock returns that is explained by various metrics from 1926 to 2011. For regular P/E ratios and PE10 (Shiller PE), the numbers at 10 years are 0.38 and 0.43 respectively. The corresponding numbers at 1 year are about 0.02 and 0.07 respectively. PE10 has had a poor ability to predict the return of the US stock market one year from now. And this is the US stock market, where structural reasons to cause the market to be cheap would be less of an issue compared to other markets, and cyclical reason would play a greater role.

For individual stocks, the time period for valuation prediction is shorter.

https://www.investingdaily.com/10913/th ... ing-period

"To truly assess the performance of growth stocks, a long-term horizon assuming a buy-and-hold strategy must be examined. The results indicate that over the first five years or so after portfolio formation, the performance of the growth stocks lags behind that of the value stocks. However, after the seventh or eighth year, the value indexes for the growth stock portfolios, in general, pass up that for the value stocks."



Value investing is a bet on mean reversion. However, with cheap stocks, you can make a bet on reversion of hundreds of stocks. With cheap stock markets, it's a bet on markets in the two digits. So for cheap stock markets, the signal will be weaker.



So the valuation signal for individual stock markets is weaker and more long term, compared to individual stocks. That has consequences, when it comes to investing.

One can own value ETFs, but there will be turnover.

https://screener.fidelity.com/ftgw/etf/ ... ymbols=RZV

A more extreme example might be RZV, a small cap value ETF with an annual turnover of 51%; Fidelity gives the asset class median at 27%. So over a 4 year period, the asset class median will have about 100% turnover.

But when it comes to turnover between individual stock markets, those that I read don't recommend high turnover. Many recommend rebalancing alone. And those who do recommend more than rebalancing recommend modest shifts in portfolio allocation.

Another consequence may be how frequently you modify your portfolio.

https://alphaarchitect.com/2016/05/13/h ... gs.HUIZctc
https://alphaarchitect.com/2015/03/19/h ... ing-funds/

The above links found that at the individual stock level, rebalancing monthly gave the optimal precost return. As the rebalancing frequency increased up to 5 years, the return decreased. So for individual stocks, frequent is good, on a precost basis.

But rebalancing between individual stock markets should be done every few years.

The one caveat here is Mebane Faber's paper, where he found that when individual stock markets are very cheap, the return over the next year is excellent.

https://poseidon01.ssrn.com/delivery.ph ... 66&EXT=pdf

If you look at Fig. 3A, when US stock markets are very cheap (PE10 less than 5), returns over the next year are excellent. But the returns over the next 7 years are similar to those over the next year. And such markets constitute 0.8% of US stock market history from 1881-2011, and will be less than 0.8% for the last 30 years. When the US stock market PE10 is between 5-10, real return over the next year is 14.5%, but the real return over the next 5 years is 12.6%. If I had to hazard a guess, the frequency of the US stock market having a CAPE less than 10 will be small in the future.

If you look at Fig. 7, the data for world stock markets is more persuasive. When CAPE is less than 15, the return over the next year is impressive and declines relatively steeply after that. Also CAPE less than 15 occurs 30.7% of the time in these stock markets between 1980-2011. He indicates that the data comes from about 10 countries in 1980, 20 in 1990 and 30 in 2000. My guess is that the data represents equal weighting, rather than market cap weighting. He also gives data for world stock markets when PE10 was less than 5. The returns are excellent, but it occurs about 1% of the time.

Why the discrepancy between the US stock market and world stock markets? Some of the included world stock markets were small. An extreme example is the Greek stock market which has a market cap of $27 billion. The US stock market is around $26 trillion.

http://money.visualcapitalist.com/all-o ... s-by-size/

For purposes of comparison, the largest stock by market cap in the world is Apple, at $725 billion.

What's been written so far assumes a dichotomy between individual stocks and individual stock markets. But if a stock market is small enough, will it start to behave more like a stock? Stocks are more volatile than stock markets, and smaller markets tend to be more volatile than larger markets. Perhaps that's a reason why Mebane Faber found that cheap nonUS stock markets do best after one year, and then decline thereafter.


Overall, I think the data for using valuations on individual stock markets is more persuasive as a risk management tool than as a way to increase returns.

Second edit:

http://www.businessinsider.com/sp-500-c ... rns-2015-2
http://www.hartland.com/wp-content/uplo ... 05/May.pdf
https://www.kitces.com/blog/shiller-cap ... -planning/

I found more data linking PE10 to S&P500 returns over the next year. The R2 was 0.06 in the first two analyses; it predicts return poorly. The one possible exception is when PE10 is less than 12. The are a few outliers of excellent returns over the next year, when PE10 is less than 12. At a PE10 of around 6, there have been 2 years where return was around 120% each year. And when PE10 is less than 12, the risk of a bad bear market goes down a lot. However, when PE10 has been around 15, there have been several years where returns have been very bad, with one year being around minus 65% return.

I wonder how much of Mebane Faber's international data linking PE10 to excellent one year returns in cheap markets is due to a few outliers increasing the average, along with a decreased risk of bad returns in cheap markets. As mentioned previously, his nonAmerican data has some small stock markets. The smaller a stock market is, the more volatile it tends to be. I wouldn't be surprised if there are some outliers in his data on cheap international stock markets.

The following are some links giving 3 year American return data and its relationship to PE10.

http://awealthofcommonsense.com/2016/12/expected-risk/
http://www.businessinsider.com/shiller- ... rns-2015-6
http://mebfaber.com/2016/12/02/missed-7 ... ood-thing/

When one looks at average returns, PE 10 might predict 3 year return at least as well as it predicts 5 or 10 year return But when you look at the variability of return at 3 and 5 and 10 years, that's where the problem arises. The variability at 3 years, and to a less extent at 5 years, is such that it considerably limits the usefulness of PE10. However, when PE10 is less than 15, the maximum drawdown over the next 3 or 5 years is quite modest IMO. Once again, PE 10 may be a useful risk management tool over such time spans. The data to support PE10 as a risk management tool over 1 year periods is not nearly as good IMO.

Post Reply