This post was originally published on this site
CHAPEL HILL, N.C. — I have both good and bad news about the Fed Model’s current forecast for the stock market on the ninth anniversary of the bull market that began in March 2009.
The Fed Model, of course, is based on the simple comparison of the 10-year Treasury note yield TMUBMUSD10Y, +0.86% and the S&P 500’s earnings yield (the inverse of the P/E ratio). The Fed Model is bullish when the earnings yield exceeds the 10-year yield, and bearish otherwise.
Though fewer analysts today than a decade ago explicitly say they are relying on the Fed Model, virtually all of them do — and you probably do too. It’s the basis of the so-called TINA argument that’s been repeated like a mantra in recent years: There Is No Alternative, since interest rates are so low.
The good news is that the Fed Model remains very bullish on the stock market, despite the steady increase in recent months in the 10-year yield. In fact, that yield would have to jump to over 4% before the Fed Model would turn bearish — an increase of over a percentage point from Thursday’s yield around 2.86%. Virtually none of the analysts I follow think such a rise is likely anytime soon.
The bad news is that the Fed Model has a terrible track record. For example, it was bullish at the bull-market top in October 2007, thus failing to protect followers from the Great Recession. In fact, the last time the Fed Model was bearish was in 2003, which was actually a good time to be bullish.
(Astute readers will point out that the Fed Model was also bearish at the end of the Great Recession, just as the subsequent bull market was about to take off. But if you focus on operating earnings as opposed to as-reported earnings, the Fed Model was actually bullish in March 2009.)
Nor are these failures unusual. Upon measuring the Fed Model’s track record over the last 75 years at forecasting the S&P 500’s SPX, +0.45% subsequent 10-year return, Ned Davis of Ned Davis Research found that it was basically worthless. Its r-squared — the metric that reflects the degree to which the Fed Model predicts the S&P 500’s 10-year return — was a statistically significant 0.5% (less than 1%).
You would have done far better basing your forecasts on the plain and simple earnings yield (which is just the inverse of the P/E, after all). Its r-squared is 56%. In other words, adding interest rates to the equation makes thing worse — far worse.
Davis reached a similar conclusion upon comparing the Fed Model to any of a number of other valuation models. Each of the ones he analyzed did far better than the Fed Model, and each of those others is bearish. I list a few of those other models in this table:
|Valuation model||r-squared of correlation to S&P 500’s subsequent 10-year return||Current outlook|
|P/E ratio||56.3%||Very bearish|
|Price/sales ratio||67.2%||Very bearish|
|Households’ equity allocation as a % of total financial assets||88.4%||Very bearish|
The contrasts depicted in the table are glaring. The valuation model that is bullish has no statistical significance, while the competing models that are very bearish have impressive track records.
By the way, don’t think the Fed Model’s record would suddenly become better if we were to measure its ability to forecast one-year returns. Its r-squared at that forecasting horizon is even lower. In fact, the r-squareds for all these models are lower at the one-year horizon, which is just another way of saying that valuation models are more useful at longer-term horizons.
It’s possible, therefore, that the stock market will continue to perform well, for a while at least, in spite of these very bearish valuation models. But, unless this time really is different, the stock market will face very stiff valuation headwinds for the next decade.
Remember this the next time an analyst says that the stock market deserves to be high because of low interest rates. Ask for the historical evidence that supports that argument.
I wouldn’t hold your breath.