Can Market Valuations Be Effective Market-Timing Signals?

/Can Market Valuations Be Effective Market-Timing Signals?

Can Market Valuations Be Effective Market-Timing Signals?

We know that valuation metrics such as the CAPE, or Shiller P/E, ratio are correlated with long-term returns (notice we didn’t say “predict” long-term returns–that is debatable).

Here is a brief background on the measure:

Here are some recent articles on the subject:

Some of our own ink on the subject:

There is a consistent drumbeat these days of “Valuations are too high; future returns will be poor.” I can’t say I disagree with the general sentiment, but Philosophical Economics has jumped into the debate and reminded all of us that correlation is not causation.

A few recent articles from “Jesse Livermore” are posted below:

All of these writings are educational and certainly thought-provoking. That said, we wanted to answer a different question:

How does one utilize valuation metrics information in the real-world?

If the P/E ratio on the market is in the top 5%, do we simply sit on our hands for 5, 10, or 15 years waiting for “the bubble to break?” Example: Internet Bubble. Meb makes a solid argument that maybe we should, but I doubt any human being could actually follow that model.

 

Or what if the P/E ratio is in the bottom 5%–we believe returns over the next 10 years will likely be favorable—but what about the digestibility of the 50% drawdown we endure when the market P/E goes from 10 to 5? Example: Great Depression. Cash might be king when your worried about paying for your next meal!

 

The goal of this research piece is to identify a simple and PRACTICAL way to integrate market valuations into a market-timing framework.

We test how valuation signals compare to, and interact with, another popular timing device, the moving average (MA) signal.

  • For the MA signals, we use monthly (1,12) MA rule on the total return for the SP500.
    • If the price for last month is above the past 12 months average, stay in the market, otherwise, invest in the risk-free rate.
  • For the valuation signal, we examine CAPE (Shiller’s Cyclically Adjusted PE ratio).
    • For robustness, we also test the Dividend Yield, Default Yield Spread, and GNP/Marketcap.

Summary Findings:

  1. Valuation-based trading rules add no value and underperform the simple (1,12) MA rule.
  2. Various combinations of MA and valuation rules cannot improve results relative to the (1,12) MA rule.
  3. MA rules starting in the late 1930’s are also questionable, albeit, at just about every other starting point they look more attractive. We start in 1938 because we need to burn 10 years of data establishing a “benchmark” valuation level (our data starts in 1927).
  4. The data suggest that market valuations have limited PRACTICAL use.

Strategy Background

  • The valuation rules are the 4 following measures: (Higher–>Cheaper; Lower–>More Expensive)
    • 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.
  • At the end of every month, we compute the (1,12) MA rule by comparing the previous month’s price series for the SP500 to the average over the past 12 months.
    • If last month’s price is above the past 12 month average, invest in the SP500.
    • If last month’s price is below the past 12 month average, buy U.S. Treasury Bills (RF).
  • For the valuation measures, we compute the percentile for the last month’s valuation measure relative to the past 10 years of each valuation measure.
    • Example: if the dividend yield was 3.2% last month, we compare this to the past 120 months and see what percentile this falls into.
      • If 3.2% is below the x-percentile (e.g., 5th percentile) for dividend yield, valuations are too high, so invest in RF.
      • If 3.2% is above the x-percentile (e.g., 5th percentile) for dividend yield, valuations are fine, so invest in stocks.
  • For the combination of MA rules and valuation rules we do the following:
    • If valuation rule says stocks are really cheap –> invest in the market
    • If valuation rule says stocks are really expensive –>  invest in risk-free
    • Otherwise, default to the MA rule.
      • Labeled “M&V_x%”
    • This strategy is effectively an MA rule with a “valuation override” component.
  • SP500 = S&P 500 Total Return Index
  • LTR = 10 year Treasury Total Return

Statistics Summary

One on One (Full Sample): MA vs. 1/CAPE (15 percentile expensive cutoff)

  • 1/1/1938 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 15% (expensive)
    • (1,12) MA outperforms 1/CAPE; tied with buy and hold.
1

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.

One on One (Full Sample): MA vs. 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1938 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 5%
    • (1,12) MA outperforms 1/CAPE; tied with buy and hold.
22

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.

One on One (First Half): MA vs. 1/CAPE (15 percentile expensive cutoff)

  • 1/1/1938 – 12/31/1959
  • 1/CAPE signals into RF if in bottom 15%
    • 1/CAPE signal works better than MA; worse than buy and hold.
33

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.

One on One (First Half): MA vs. 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1938 – 12/31/1959
  • 1/CAPE signals into RF if in bottom 5%
    • 1/CAPE signal works better than MA; worse than buy and hold.
Microsoft Excel - CALM_R&D_indicators_v04.xlsm_2014-06-12_20-38-21

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.

One on One (Second Half): MA vs. 1/CAPE (15 percentile expensive cutoff)

  • 1/1/1960 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 15%
    • (1,12) MA signal works better than CAPE and buy and hold.
55

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.

One on One (Second Half): MA vs. 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1960 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 5%
    • (1,12) MA signal works better than CAPE and buy and hold.
66

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.

Teaming Up (Full Sample): MA and 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1938 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 5%
  • 1/CAPE signals into stocks if in top 5%
  • Otherwise, use MA rule
    • (1,12) MA signal works; CAPE integration does not.
77

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.

Teaming Up (First Half): MA and 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1938 – 12/31/1959
  • 1/CAPE signals into RF if in bottom 5%
  • 1/CAPE signals into stocks if in top 5%
  • Otherwise, use MA rule
    • CAPE is ineffective; MA is ineffective.
88

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.

Teaming Up (Second Half): MA and 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1960 – 4/30/2014
  • 1/CAPE signals into RF if in bottom 5%
  • 1/CAPE signals into stocks if in top 5%
  • Otherwise, use MA rule
    • (1,12) MA signal works; CAPE integration does not.
99

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.

Valuation Metric Horse race: MA and 1/CAPE (5 percentile expensive cutoff)

  • 1/1/1960 – 4/30/2014 (sample where data is available for all measures)
  • Each valuation signals into RF if in bottom 5%
    • Dividend Yield, Default Yield, and GNP/Marketcap all work marginally better than CAPE.
    • All valuation timing strategies underperform buy and hold and the MA strategy.
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.

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.

 

Conclusion:

As a die hard value guy I want to believe there is a ***robust*** way to use valuations in a practical way for market timing. I haven’t seen anything out there that is convincing.

Please send ideas–I want to be convinced (Yes, I’m suffering from confirmation bias)!


  • 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. Our full disclosures are available here. Definitions of common statistics used in our analysis are available here (towards the bottom).
  • Join thousands of other readers and subscribe to our blog.
  • This site provides NO information on our value ETFs or our momentum ETFs. Please refer to this site.

Print Friendly, PDF & Email

About the Author:

Wes Gray
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.
  • janvrots

    I some times think determining the bet size is more important than the bet. Can you back test a strategy with a rule that bets $100, $80, $60….$20, $10 depending on how expensive the market is?

    I am guessing that this naïve scaling of the bet size will not improve the results so maybe I should scale the bet size differently eg I am 2stdev below average market pe bet $200, 1stdev bet $100.

    Just checking. Using these rules lets assume the average bet in the market is $50. Is it fair to compare that to a bet of $100 invested in the S&P? I assume the tables you show below is comparing a $100 bet in s&p to $100 ($50 in equity, $50 in RF)

  • Doug

    1. What about after-tax, after transaction cost returns? B/H would do well, but the valuation strategy may outperform the MA rule due to lower turnover.
    2. Timing just one market on valuation really hamstrings the study. I’m sure you’ve seen Meb’s country CAPE analysis, which looks pretty good. The problem with valuation timing is that the drag from cash returns gets pretty high over time. You need strategy where you can ALWAYS find something cheap to invest in. (I would also note that this applies equally to momentum timing).

    Full disclosure: I’m long Russia, Argentina and Greece. And am also implementing the FAA strategy you outlined earlier (Keller and Putten) for another chunk of my portfolio, so I’m agnostic toward value/momentum.

  • Eric Meier

    Enjoyed this analysis. Amusing that there’s a bloger with the pseudonym “Jesse Livermore”. I recently heard of this guy and read about his rollercoaster life on the wikipedia page on him (http://en.wikipedia.org/wiki/Jesse_Lauriston_Livermore).

    Wonder if it would be worth testing a strategy that gets in/out of the market when
    interest rates are high and about to drop vs. low and about to rise?

  • 1. B/H has a nice leg up on everything when you start thinking about reality.
    2. I agree. Cross-sectional, or relative valuation analysis, might be a better approach than time-series, or relative to one’s own historical valuation.

    In general, cheap is king…but it doesn’t seem to work on the S&P 500.

  • Done things like this in the past–not a real difference. Email us and I’d be happy to get you the underlying data–you can go buck wild on tweaks of the basic idea presented above.

  • Not a bad concept to explore

  • Remmelt

    It surprised me that by this data historically it was better to just go by the S&P500 instead of using CAPE as a “get out” indicator. Thank you for this interesting analysis.

    I wonder though if the same would apply using the Q Ratio, since it has an R^2 of 0.46 instead of 0.3 for CAPE against market returns. I’m not sure though how much of this is luck and whether you could rely on this if other value signals fail.

    Would that be something you could add to the test? Andrew Smithers has data on this.

  • Check out the chart in our paper on CAPE:
    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2329948

    All the valuation metrics are highly correlated. I’m not sure it would matter much. We can add to the R&D mix, however

  • Remmelt

    Thank you for taking the effort to answer.

    I think you’re likely right that Tobin’s Q would also not add value here, though I’d still be quite interested in seeing the results, given that the Q Ratio is noted in several papers as being a better predictor of returns than CAPE
    (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1990449, http://www.researchgate.net/publication/228313072_Tobins_Q_Versus_CAPE_Versus_CAPER_Predicting_Stock_Market_Returns_Using_Fundamentals_and_Momentum & http://public.econ.duke.edu/Papers//Other/Tower/Pessimism.pdf)

    Your cyclically adjusted book value measure did best on an individual firm level, and I wonder if Tobin’s Q (which has differences in underlying data, which I would argue could make it slightly more accurate – see http://csinvesting.org/2014/01/09/tobins-q-the-market-is-60-overvalued/), also does somewhat better at an index level.

  • A reader emailed and asked:

    “Regarding your article “Can Market Valuations Be Effective Market-Timing Signals?” how different are the results up to March 2009? Buy-n-hold looks better at peaks.”

    Below are the results from 1960 through February 28, 2009.

    CAPE is actually the worse timing metric among valuations metrics and all valuation metrics underperform buy and hold S&P 500.

  • Steven Ross

    I concur with your result that CAPE doesn’t work on its own with the S&P500 to actually outperform the market. One basic flaw with CAPE is that earnings tend to be high at the top of a bubble, and much worse, very low right after a crash, so P/Es can look more reasonable than they really are at the top of a bubble, and much worse than the really are (relative to actual earnings power) after a crash. The multi-year averaging helps, but not enough to counteract this effect, especially if the cyclical earnings impact extends for multiple years. Another issue is that bubbles tend to take multiple years to pop, and returns tend to be quite good up until shortly before the crash.

    I found an alternative metric that seems to correlate, though it’ll only get you out of the very worst bubbles, based on the assumption that the underlying earnings power of the S&P500 is always increasing:
    Take the annual earnings for the S&P 500 over the trailing 12 months. Call this your TTM earnings.
    If this TTM earnings is greater than the current max TTM earnings increased for inflation, it is the new max TTM earnings. Otherwise, increase max TTM earnings for an assumed baseline annual growth rate (inflation if nothing else; I used 4%).
    Create a ratio of price to max TTM earnings. If this ratio is greater than a threshold (I used 14.6), then the market is overpriced. Here’s where this approach differs though: hold until the market is overpriced for 53 consecutive months (or so). That’s when you know you are in a dangerously high bubble, and the expected return from holding longer is negative, so you should sell (and buy again once it is no longer overpriced by this measure). You can shorten the holding-during-bubble period some if you can provide a decent risk-free interest rate as an alternative, especially if this risk-free investment will benefit from a flight to quality after a crash.
    I can send you a spreadsheet with this computation based on the CAPE website data if you’re interested. The same approach (assuming bubbles take multiple years to build up) can be applied to other metrics, like dividends and GNP/Marketcap. My metric has the flaw that it doesn’t catch 1929, and has serious risk of overfitting (it only catches 2 recent bubbles), but if it has the advantage that it’s designed to eliminate false positives, which are very expensive to long-term returns.

    Have you considered more direct upcoming crash indicators, such as yield curve inversions (even for a day) or unemployment below 5%?

  • Steven,

    This certainly sounds like an interesting idea, but as you said–a large risk of overfitting. How does it work if you use 40 months or 60 months? Is it robust to changes in that parameter?

    We’ve looked at a ton of things. They all seem to work until they don’t. I’m becoming more and more a fan of keep it simple technical rules and/or buy and hold with smart diversification.

  • Steven Ross

    The baseline return for buying and holding from 1871 to early 2014 is 410X. With 53 months, it’s 680X. With 60 months it’s 623X. With 40 months it’s 465X. Note this assumes 0 return in the months where you’re out of the market; if you adjusted for that you could afford to be more conservative (like using 40 months). If you adjust the P/E threshold to 15, the return drops to 644X, with 16 it’s 544X, and with 14 it’s 431X (basically neutral, but less volatile).

  • seems pretty good then. Go ahead and fire the sheet/analysis our way and I’ll have one of my guys haze it for you.

  • Steven Ross

    Here’s the spreadsheet on google docs: https://docs.google.com/spreadsheet/ccc?key=0AvxNpd25VTAwdGw2REFIdDFSSHM3N2h0b0JkVnU3V3c&usp=sharing

    Here’s the data source:
    http://www.econ.yale.edu/~shiller/data/ie_data.xls

    I made some assumptions about when data is available to make decisions based on(earnings aren’t published immediately, do you make the sell decision at the beginning or end of the month), but those only modify the buy/sell time by a month or two. Feel free to download and modify it.

  • Steven Ross

    Another note: That has both the metric I mentioned earlier (Return tab), and one using the Shiller P/E (Shilller tab), but using 52 months and a threshold of 22.2 (as it’s a decade average, instead of a peak P/E). You may prefer the Shiller high for 52 months version; I’d forgotten about it but the returns are more impressive.

  • karl shewmake

    If the “price series” for last month is above the past 12 months average, stay in the market, otherwise, invest in the risk-free rate. (Quotes added), how is price series calculated?

  • Hi Karl,

    That is confusing language. I removed “series” and now it reads: “if the price for last month is above…”

    The reason is says “series” is an artifact of how we create the “price series” from underlying total return data (which is usually what we have). For example, we have total return data that ret1=5% and ret2=10% and ret3=-5%. We’d create a total return price series of 1, 1.05, 1.155, and 1.09725…