No. of Recommendations: 17
Thanks for confirming that neither the inflation-adjusted CAPE ratio nor the P/E ratio of the market can be used to say anything about future market returns during the next 3 years. There's no correlation with CAPE or P/E that you can take advantage of to time the market over time periods that are less than 3 years. It's not so much that there is no correlation, but that the error bars are so wide that it's pretty useless as a trading input.
Simply put, 3 years is a window short enough that it might just be a continuation of the current bull or bear market, or a horrible short-lived bear might be about to start. Stuff happens.
These are the forward returns by decile of starting trend real earnings yield. (like CAPE, just a bit smoother averaging method for the real earnings)
There is a nice progression from expensive=bad returns through to cheap=good returns, but note that the outcomes in a given bucket generally have a range of around 30%/year.
Good correlation, bad predictive model : )
Three year forward annualized rate of real total return from S&P 500 at various starting valuation levels based on trend real earnings yield:
Lowest Pctl 10 Average Pctl 90 Highest
Most expensive 5% of time -14.4% -13.2% -10.8% -7.3% 0.0%
Next 10% of time -10.9% -7.8% 0.1% 7.7% 12.4%
Next 10% of time -3.3% -0.5% 5.7% 12.9% 17.8%
Next 10% of time -9.7% 0.0% 6.1% 18.9% 22.0%
Next 10% of time -10.0% -8.8% 2.6% 10.7% 23.8%
Next 10% of time -8.9% -6.3% 6.2% 20.0% 24.7%
Next 10% of time -7.6% -3.1% 11.2% 25.1% 26.8%
Next 10% of time -2.9% 0.3% 12.5% 25.9% 27.3%
Next 10% of time 1.6% 10.8% 17.2% 26.7% 28.4%
Next 10% of time 6.3% 7.1% 11.2% 18.9% 28.9%
Cheapest 5% of time 7.8% 8.8% 12.9% 16.5% 21.4%
As before, the starts of each hold observation are a single date with a specific earnings yield slotted into a bucket, and the endpoints are smoothed.
This particular table is for starting dates January 1990 through January 2020, 30 years.
The best way to think of a valuation metric like CAPE is as a pretty good predictor of annualized returns from now until half way down the next bear market (or the one after that), or from now until half way up the next bull market (or the one after that): any ending date at which valuation levels aren't extremes but closer to typical.
But as it says nothing about how many years it might be until that next halfway point is reached, one has little idea how long the predicted rate of return will take.
Bottom line: CAPE cannot consistently forecast market returns over the next 10 years. Well, I think my table shows the reverse pretty conclusively. Even best observed return in the most expensive/worst starting date is worse than the worst outcome from the cheapest/best starting date.
But perhaps you have other data which supports the idea that it has no predictive power at those time frames.
"Attempts to time the stock market to make an above-average return is one of the most common practices among individual investors. Empirical evidence from behavioral finance suggests that investors who try to time the market often meet with disastrous outcomes.Yes, that's usually the case, I have to agree. Market tops in particular are very hard to see at the time.
Personally I have had some modest success spotting market bottoms, which has added a bit to my beer and pizza budget over time. My most recent "major market bottom" call 2022-09-30 was off by 8 trading days. The "buy" signal was only 0.2% above the lowest daily close within about ~1.5 years before the signal, or any time since. Better than random, anyway.
The signal was posted here www.mungofitch.com
I used the same model to call the absolute bottom 2009-03-04.
http://www.datahelper.com/mi/search.phtml?nofool=y...That time it was four trading days early, and only 2.9% above the lowest close between 1996 and today. It did give several dodgy buy signals in late 2008 on the way down--they were nicely profitable after a year, but it was (ahem) a bumpy year.
Jim
