Tactical Asset Allocation with Market Valuations: Magic of Myth?

Tactical Asset Allocation with Market Valuations: Magic of Myth?

April 14, 2015 Research Insights, Value Investing Research, Tactical Asset Allocation Research, $SPY, $vlue
Print Friendly
(Last Updated On: June 29, 2015)

Executive Summary

Although it has been very difficult to overcome our initial skepticism, we’ve finally accepted the notion that simple technical analysis may serve as an effective way to manage risk and to time markets. As John Adams said many years ago, “Facts are stubborn things.” However, our stance on the hypothesis that “market valuations can be an effective tactical asset allocation mechanism” is still a long way from full acceptance–some tests suggest it works; others suggest it doesn’t. These tests are also fragile and often lack robustness: For example, in a piece we highlighted a few weeks ago, we found that identifying whether or not the market is “overvalued” can be highly dependent on framing.

But what is, “Tactical asset allocation with market valuations?”

Simply put, a tactical asset allocation model that incorporates market valuations typically suggests that we sell when prices are high relative to some historical benchmark, and we buy when prices are cheap relative to some historical benchmark. Buy low, sell high–an intuitive concept. We also know that over long horizons valuations are directly related to realized returns, however, just because we know valuations relate to long-term returns doesn’t necessarily imply we can beat a buy-and-hold strategy by timing the market based on valuations. So how should we proceed?

As evidence-based decision-makers we decided to test the various stories told in the market when it comes to tactical timing using valuations:

  • We’ve already tested the “extreme valuation” hypothesis–sell when the market hits 90-percentile+ valuation–doesn’t work.
  • Here we test the “relative valuation” hypothesis–sell when market is above average valuation; buy when below average–doesn’t seem to work.
  • Gestaltu tests “absolute value” hypothesis–buy when valuation yield minus realized inflation is high; sell when it is low–this seems to work, but we need to do our own testing.

Bottom line: We are still on the fence when it comes to tactical asset allocation using valuations.

Tactical Asset Allocation and Intuition

Often, “intuitive” concepts don’t actually work in practice. In our experience, the more intuitive the concept, and the “easier” it is to implement, the less likely it is that it actually works. A few common examples (at least recently) come to mind:

  • Buying high quality dividend stocks
  • Avoiding treasury bonds
  • Avoiding European stock markets
  • Buying hedged currency products

Who knows how any of these trades will end up, but for many investors they are both intuitive and easy to digest, which is to us a potential sign these “obvious” trades probably won’t end up being that great on a relative basis.

Another intuitive trade is to tactically allocate a portfolio’s market exposure based on valuation metrics (e.g., P/E, P/B, CAPE, etc.). As hard-core value-investors, with a strong belief in applying value-investing concepts to stock selection, we would love nothing more than to identify a way to apply valuation to market timing! Talk about easy to digest. Yet it’s a tough row to hoe — market timing has traditionally been considered to be a fool’s errand. Although we have been taught that attempts at market timing usually end badly in the long run, there is evidence that some approaches have merit.

Consider simple technical analysis: I thought technical analysis was a combination of voodoo and blasphemy before finally accepting the notion that trend-following rules are one potential way to avoid large drawdowns and maintain reasonable returns in the market.

Here is a chart from a simulation study we did a while ago on the subject of simple technical analysis in the form of moving-average rules:

Tactical Asset Allocation with Market Valuations

2014-07-08 11_14_32-Microsoft Excel - MA_simulation_v2 ex5 sp.xlsm

Sharpe ratios certainly seem interesting relative to a buy-and-hold investor.

But my acceptance of trend-following rules took over 4 years!

My acceptance of a timing rule that was based on valuations would take a lot less evidence, because I suffer from confirmation bias and I’d love to see something actually work in practice. But it hasn’t happened…at least not yet…

Where we sit on Valuation as a Timing Tool

The “99% percentile,” or the broader timing rule that suggests we dump stocks when they are at a historical high, doesn’t seem to work.

A more detailed study on dumping stocks at extreme valuations dug a little deeper. We looked at the following measures and compared/contrasted them with trend-following:

  • 1/CAPE = Inverse of Shiller’s Cyclically Adjusted PE ratio.
  • Dividend Yield = Total Dividends for SP500 over the past 12 months divided by the SP500 closing price.
  • Default Yield Spread = BAA Yield – AAA Yield.
  • GNP/Marketcap = GNP of U.S. divided by the total market capitalization of U.S. Equity Markets.

Again, no dice.

Finally, we are investigating an interesting post by Adam and his team at Gestaltu.com. Prelims look promising, but there are potential robustness issues. More to come…

And finally, our friend Meb Faber posted a tweet recently:

People often say you can’t use CAPE (or value) to time the market.  That isn’t true. Simple example: Long when CAPE < average rolling CAPE.  Otherwise bonds. Pair that with simple trend system and you have a beauty value/trend system.

We certainly agree that adding a simple trend system will add serious risk-management mojo to most historical return streams. But this post isn’t focused on trend-following or the blended valuation/trend model suggested by Meb.

Let’s focus at the valuation component of the concept.

Testing the simple CAPE < Avg (CAPE) Rule

Let’s run a quick test on this simple valuation-timing rule.

  • Strategy: Long the S&P 500 when last month’s CAPE < Average Rolling 10-year CAPE (when market is relatively cheap); Otherwise, long the 10-yr bonds (when the market is relatively expensive).
    • We also test CAPE < Average CAPE using an expanding window (using cumulative-to-date data)
  • Data Period: 1/1937-1/2015 (For the rolling analysis we burn 10 years of data, so our analysis starts in 1937, not 1927).

Summary Statistics:

The simple CAPE < Avg 10-year Rolling CAPE strategy (rolling window) and CAPE < Ave CAPE to date (expanding window) lowers volatility a bit. But in terms of CAGR and Sharpe ratio, it is worse than B&H. By comparison, the 12-month simple moving average strategy performs much better. We also examine different rolling lookback periods: 1-year, 3-year, and 5-year…

None of them beats B&H; all lose to simple trend-following.

summary stats
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Maybe CAPE is busted? How about other metrics?

Here we test the market-timing performances of 5 other value metrics that we have examined in prior research. In prior research we focus on value investing metrics to identify cross-sectional predictions (i.e., stock selection ability), which seems to work, however, we need to remember that our focus now is on time-series prediction (i.e., timing market exposures over time).

The value investing metrics are as follows:

  • Earnings to Market Capitalization (E/M).
  • Book Value of Equity to Market Capitalization (B/M)
  • EBITDA to Total Enterprise Value (EBITDA/TEV).
  • Free Cash Flow to Total Enterprise Value (FCF/TEV).
  • Gross-Profits to Total Enterprise Value (GP/TEV).

Note, that all these metrics are in “yield” format, so higher is cheaper and lower is more expensive. This is in contrast to the CAPE ratio where higher is more expensive and cheaper is less expensive. A little mental gymnastics, but that is healthy for our brains 🙂

Rolling-Window Performance: 

The rolling technique looks back over a fixed time period (e.g., 10 years) to calculate an “average” value that will serve as the anchor for our valuation-based trading signal.

  • Strategy: Long the S&P 500 when last month’s value metric > Average Rolling 10-year value metric (when market is relatively cheap); Otherwise, long 10-yr bonds (when market is relatively expensive).
  • Data Period: 7/1970 to 12/2013 (we burn 10 years of data, so our analysis starts in 1960).

All of the valuation-based timers are roughly equivalent to B&H, and all of these systems substantially underperform simple trend-following rules. (Click chart to enlarge).

Value metrics market timing performance 01
Click to enlarge
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Expanding Window Performance: 

Here we investigate a different lookback technique: the expanding-window method.  In contrast to the rolling lookback, a expanding lookback method uses a ever growing lookback period that contains the entire history available in the past. So if we are standing in 1940 and our data starts on 1930, we use 10 years of data to determine our “average.” However, if we are standing in 1960 and our data starts on 1930, we would use 30 years of data to determine our “average.”

  • Strategy: Long the S&P 500 when last month’s value metric > Average value metric to date (when market is relatively cheap); Otherwise, long 10-yr bonds (when market is relatively expensive).
  • Data Period: 7/1970 to 12/2013

All of the valuation-based timers are roughly equivalent to B&H, and all of these systems substantially underperform simple trend-following rules. (Click chart to enlarge).

value metrics market timing performance 02
Click to enlarge
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Is there a Valuation-Timing Holy Grail?

Maybe it exists, but thus far, we haven’t found it.

We will continue to look, but as our search widens and our techniques get more and more complex, we must consider that we may be susceptible to optimization and data-mining.

Or perhaps valuation-timing simply isn’t effective on the S&P 500? That is possible.

But that doesn’t mean it may not work across countries or on other asset classes. We simply can’t identify a reliable methodology in the context of US stocks.

Where to from here?

Well, we are working on a promising technique we’ve reviewed from Gestaltu and our results will be shared with our community. Until then, if you have any valuation-based market-timing ideas that are simple, robust, and easy to replicate, please pass along. Also, refrain from sending us low-turnover, or low-frequency valuation timing evidence (e.g., show us that high P/E is associated with poor returns). These statistics, while interesting, provide little guidance when it pertains to managing a real-world portfolio.

Note: This site provides NO information on our value investing ETFs or our momentum investing ETFs. Please refer to this site.

Join thousands of other readers and subscribe to our blog.

Please remember that past performance is not an indicator of future results. Please read our full disclosures. The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. This material has been provided to you solely for information and educational purposes and does not constitute an offer or solicitation of an offer or any advice or recommendation to purchase any securities or other financial instruments and may not be construed as such. The factual information set forth herein has been obtained or derived from sources believed by the author and Alpha Architect to be reliable but it is not necessarily all-inclusive and is not guaranteed as to its accuracy and is not to be regarded as a representation or warranty, express or implied, as to the information’s accuracy or completeness, nor should the attached information serve as the basis of any investment decision. No part of this material may be reproduced in any form, or referred to in any other publication, without express written permission from Alpha Architect.

Definitions of common statistics used in our analysis are available here (towards the bottom)

About the Author

Wesley R. Gray, Ph.D.

After serving as a Captain in the United States Marine Corps, Dr. Gray earned a PhD, and worked as a finance professor at Drexel University. Dr. Gray’s interest in bridging the research gap between academia and industry led him to found Alpha Architect, an asset management that delivers affordable active exposures for tax-sensitive investors. Dr. Gray has published four books and a number of academic articles. Wes is a regular contributor to multiple industry outlets, to include the following: Wall Street Journal, Forbes, ETF.com, and the CFA Institute. Dr. Gray earned an MBA and a PhD in finance from the University of Chicago and graduated magna cum laude with a BS from The Wharton School of the University of Pennsylvania.

  • Which rule does MA(12,1) refer to above? Is this buying into the market when the 1-month average exceeds the 12-month average and being in cash otherwise?

  • Simple 12 month MA–compare current price to 12 month SMA. Monthly data, not daily.

  • Jan Vrot

    I have found that many quantitative strategies can be improved by varying the bet size rather than switching between 100% equity and 0% equity. I have not worked on MA, but what about something like this

    – P>MA20 + P>MA60 + P>MA200 then 100% equity
    – PMA60 + P>MA200 then 80% equity
    – P<MA20 + PMA200 then 60%% equity

    This all or nothing bet is a tough way to make money and opens one up to behavioral errors, “damn I sold all my shares and the market went up 10%”

  • Pingback: Quantocracy's Daily Wrap for 04/14/2015 | Quantocracy()

  • Mark

    Wes, the content in the web is pretty GOOD, but feel like too fanny!! TAKE A LOOK AT AQR WEBSITE!

  • Jan, great observation. We’ve done things like this in the past and they don’t seem to be that effective, however, they are much more realistic in practice because people can ACTUALLY FOLLOW these rules. 0/1 rules are too difficult from a psychology standpoint. Great point!
    We have another investigation into MA + Valuation if you’re interested

    We can work on the “bet varying stuff”

  • AQR has a great site and great content. I wish they’d improve their social sharing and create a listserv that is easy to subscribe too…hard to access their content without going to check in on their site all the time

  • pcavatore

    IMO valuation is better used in an asset allocation framework with a relative approach. Let me try to further elaborate as I guess I’m using the world “relative” in a different way compared to Gestaltu (relative to other asset classes rather than relative to its average). As a matter of fact I would be more than happy to switch from stocks to bonds as long as bond valuation is not richer than for stocks. This would allow to stay invested in expensive stocks as long as bonds are even more expensive. Having said that coming up with a comparable valuation metric for bonds may not be straightforward. I would use percentiles in order to do that…a basic example would be to calculate the percentile 1/CAPE and the percentile yield-to-maturity (or whatever other valuation metric you find relevant for bonds) and always be invested in the “relative” cheaper (higher percentile) asset class. Paolo

  • that definitely makes sense.

  • Ben

    Hi Wes,

    Very interesting post.

    It reminded me of a great article from JP Hussman (http://www.hussmanfunds.com/wmc/wmc130506.htm), where he uses 39-week moving average, shiller PE and Investor sentiment to categorize market environments. Based on those environments he calculated historical returns and Sharpe ratios.

    The conclusion is that (in order of progressive returns / Sharpes), one should use:
    – At a minimum a Moving average rule
    – Get better in conjunction with a PE 18 if the investor sentiment is not over bulish

    He also provides the %age of time each situation has occurred since 1940.

    He then graphs out the performance of the various approaches, including one where the market exposure is proportional to the Sharpe ratio… and this one performs best!

    Thought this could be interesting to test as an approach for you – or use some elements in the testing you are working on?

  • Stu

    Do you have a previous post or any links regarding your assertion that buying high quality dividend stocks does not work in practice? This seems like a surprising assertion to me and I don’t believe I’ve seen it discussed on your blog before.

  • Jack Vogel, PhD

    Here is a recent post on dividend stocks: http://blog.alphaarchitect.com/2015/03/24/investing-in-high-dividend-yield-stocks-is-a-sucker-bet/

    Also, research show that more complete measures of shareholder yield appears to work better than simple dividend yield: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2051101&download=yes

  • The problem with market timing is that you end up just pulling one lever — equities vs. bonds. That means you are effectively trading one market (equities minus bonds). I suspect that there is just not enough data to determine whether market timing based on value “works”, given the low frequency of trading; but even if there was, it is a daft idea to apply a trading strategy to just one market. This is especially true if you are a wealth manager investing the whole of someone’s financial wealth.

    “Value” works, applied across a large number of stocks and markets. “Trend” works too, applied across a large number of markets. So the reasonable thing to do is to have systematic value funds and systematic trend followers in your portfolio, and rebalance. People depart from this reasonable thing because they dream of reward without risk — “if only I could dodge every risk, just before it happens!” It’s nonsense.

  • John,
    I want a strategy that goes up when the market goes up, but doesn’t go down when the market goes down. And I also want a 10% yield, but I don’t want any risk…
    …my favorite client request 🙂

  • RT1C

    Earlier today I posted a comment on your 4/4 Quant Geek Weekend blog (catching up on my reading today!). Referencing the research paper linked to there, I suggested using secular cycles (based on P/E) to switch between trend-following MOM strategy and B&H strategy, since that research seemed to indicate that passive outperformed MOM in bull markets and reverse in bear. Maybe this is how you can exploit valuation. Please refer to my earlier post for details.

  • Antacular

    I know that that March 25 post you defined “high quality” dividend stocks as determined by their ROE, but what about using the payout ratio instead? As far as dividends (and buybacks) go, seems that whether there’s room for the dividend to grow/room to maneuver incase things go bad would be a better indicator of dividend “quality” than ROE.

  • Jan Vrot

    I only had a few minutes to spare using MA on sp500 to determine bet sizes. I used MA20, MA60, MA120 and MA200. Every day I count how many times price exceeds these moving averages. If count = 4 then 100% equity , 3 = 75% equity, 2=50%equity, 1=25% equity and 0 = 0%equity. Seemed to generate a promising result beating buyandhold and 200day moving average. I am sure it would be easy to improve this simple strategy, but, I may be accused of data mining. I may do another post depending on results I find for more sophisticated strategies.

  • Pingback: “Is the market overheated?” | Foundation Capital Advisors()

  • disqus_xKlysnzLJR

    Wes, nice article!
    How about valuation spreads compression/expansion? i.e. median EV/EBITDA for top and bottom decile. It will be hard or impossible to get the tops/bottoms inflection points, but I believe they tend to move in relatively long term trends, which makes sense, so applying a long term moving average to it would probably be a good way to determine when to be long value or not? It would be very interesting to see if you could publish the historical bands of say top and bottom decile for EV/EBITDA, EV/EBIT, P/E, etc.
    Obviously the extreme spreads such as the one in 2000 brought a boom to value, but just looking for extremes would probably keep you out of value for too long. I would be very curious to see historically how the trend in the spread has influenced future value returns vs the market.

  • Here is an old post related to timing on valuation spreads:

    We have some others but I can’t remember. We’ve done so much research at this point I can’t even remember what all we’ve tested. Just recently we retested a “novel idea” that we had tested 4 years ago. old age.

    I think our key takeaways on valuation timing are similar to those presented in Asness’ paper “sin a little”
    https://www.aqr.com/cliffs-perspective/sin-a-little. Trend seems to be pretty effective, valuation timing is marginal, at best…