In July 2011, just before the market lost nearly 20% (but also the last time it corrected materially), I observed “Like Wile E. Coyote holding an anvil just past the edge of a cliff, here we are, looking down below as if there is much question about what happens next.” In my view, the stock market is hovering in what has a good chance of being seen in hindsight as the complacent lull before a period of steep losses. Meanwhile, we would require a certain amount of deterioration in stock prices, credit spreads, and employment growth to amplify our economic concerns, but even here we can say that there is little evidence of economic acceleration. Broad economic activity continues to hover at levels that have historically delineated the border of expansions and recessions.
With the S&P 500 price/revenue ratio more than twice its historical norm prior to the late 1990’s market bubble, the ratio of market capitalization to GDP also more than twice its historical norm, the most lopsided bullish sentiment in decades, an overbought market trading at a record highs – and two standard deviations above its 20-period moving averages at weekly and monthly resolutions, a “log-periodic” bubble at its most likely finite-time singularity (see Estimating the Risk of a Market Crash), bond yields well above their 6-month average, an economy where growth in real GDP, real final sales and employment are all near or below the growth rates at which historical recessions have started, and our own estimates of prospective market return/risk quite negative based on a broad ensemble of observable market conditions, we view prospective near-term and multi-year returns as strongly unfavorable, and prospective market risk as unusually elevated.
Emphatically, our investment stance here doesn’t presume or require a market crash (though we wouldn’t rule one out). As usual, we align our view with the prevailing return/risk profile that we estimate on the basis of observable evidence at each point in time. We have no downside target for the market, nor do we rely on any particular scenario. We estimate likely return/risk as conditions change over time. These estimates have been persistently negative in the face of the recent market advance because similarly extreme conditions throughout history have come to dismal ends. Our stance will change as the evidence does.
On valuation, we continue to see extremes in a broad range of measures that are very well correlated with actual subsequent market returns. Of course, there are some measures like the “Fed Model” (forward earnings yields less 10-year Treasury yields) that are fairly benign, and suggest no valuation concerns at all. The problem is that these alternative models don’t perform nearly as well in explaining actual subsequent market returns. Much of the reason is that they take profit margins at face value even when margins are elevated. At present, this is equivalent to assuming that the current extreme in profit margins will be permanent, and then fully reflecting that assumption in stock prices, even when all of history demonstrates that margins are cyclical.
We can calculate the historical errors of various valuation models in forecasting actual subsequent 7-10 year market returns. A good model should have random errors – that is, the errors should not themselves be highly predictable based on data that was readily available at the time. For the “equity risk premium” models that Janet Yellen and Alan Greenspan often reference as evidence that stocks are not overvalued, it turns out that theerrors of these models have a correlation of about 85% with profit margins that were observable at each point in time. In other words, these models make large and systematic errors because they fail to account for the cyclical variation of profit margins over time (see An Open Letter to the FOMC: Recognizing the Valuation Bubble in Equities, andThe Coming Retreat in Corporate Earnings).
A few quick charts will bring some of our present concerns up to date. The chart below shows the ratio of nonfinancial equity market capitalization to GDP. Again, this measure is about twice its pre-bubble norm, and is presently associated with an expectation of negative total returns for the S&P 500 over the coming decade. Measures based on properly normalized earnings are a little bit more favorable, with the overall outcome that we broadly expect nominal total returns for the S&P 500 of about 2.3% annually over the coming decade, with negative total returns on horizons of less than about 7 years.
On the economic front, we really don’t take a great deal of “signal” from the December payroll report, which fell short of expectations. The fact is that the seasonal adjustment for January alone was 876,000 jobs (and thechange in the seasonal adjustment was 320,000 jobs). It’s quite difficult to get a good read from monthly employment changes just before and after the holidays.
Normally, we can get useful economic signals from a broad range of leading measures such as various components of Fed surveys and purchasing managers indices, but in the face of repeated bouts of quantitative easing, the correlation between the most historically reliable leading measures and subsequent economic outcomes has been utterly destroyed in recent years (though probably only temporarily). Indeed, the correlation has actually turned negative (see When Economic Data is Worse than Useless).