Excerpt from the Hussman Funds' Weekly Market Comment (12/9/13):
Our estimate of prospective 10-year nominal S&P 500 total returns has eroded to just 2.3%, suggesting that equities are likely to underperform even the relatively low returns available on 10-year Treasury bonds in the coming decade. Those estimates have had a correlation of over 85% with subsequent 10-year S&P 500 total returns since 1940, and a higher correlation with subsequent returns more recently. We don’t rely on any single estimation method, and some are more complex discounting models, but a number of very good shorthand methods with similar historical reliability are reviewed in the Valuation section of Investment, Speculation, Valuation, and Tinker Bell. Note that these same valuation measures were quite favorable in 2009, and gave us sufficient room to shift to a constructive stance after the 2000-2002 bear market plunge, while warning of overvaluation at the 2000, 2007 and present instances. Despite other considerations that have been unique to this cycle, our valuation measures haven’t missed a beat.
As a side-note, it’s easy enough to evaluate the all of these methods versus actual subsequent equity returns in a century of historical data. One can demonstrate their reliability against alternatives such as the Fed Model, “equity risk premium” models, and forward-operating P/E ratios (see An Open Letter to the Fed: Recognizing the Valuation Bubble in Equities), as well as measures that use NIPA profits (where a rather egregious bit of analysis that I addressed months ago continues to stagger about like the living dead). Analysts who don’t show their data, and more importantly, don’t demonstrate that their valuation methods are tightly correlated with actual subsequent market returns over the next several years, are simply purveyors of noise. When you realize that the errors of the Fed Model in explaining actual subsequent market returns (in excess of bonds) are 86% correlated with the profit margins that existed at the time of the forecast, it should be clear that the model is not capturing regularities that ought to be captured.
Examining extreme market peaks across history, you’ll usually observe three things: a) the most historically reliable valuation measures will have modestly underestimated the advance over the preceding 10-years, b) those same valuation measures will be projecting dismal subsequent market returns, and c) those valuation measures will be proven right. The converse is generally true at extreme market troughs (see The Siren’s Song of the Unfinished Half-Cycle for a historical graph that illustrates these regularities). The reason for this is that any market extreme is the endpoint of an unfinished half-cycle. So not surprisingly, when we examine projections from a decade ago, or calculate a 13+year estimate from the 2000 peak, the S&P 500 has briefly outperformed those estimates by several percent annually. These “overshoots” are likely to be corrected in the completion of the full market cycle.
The median is not the messenger
The knee-jerk reaction to the fact that the median stock is now more overvalued than in 2000, and that prospective 10-year S&P 500 returns are dismal, is to dispense with the evidence because of our own challenges since 2009. I cannot emphasize strongly enough that those challenges trace primarily to the stress-testing decision that I made in 2009, even though our existing methods had enjoyed an outstanding record at that point. Moreover, the methods that resulted would have outperformed our pre-2009 methods both in historical data and in the recent cycle, had they been available at the time. I fully accept that missed returns in the half-cycle since 2009 will keep me open to criticism until we prove all of this out over the completion of this cycle and future ones.
Meanwhile, however, I fear that investors are going to do themselves a terrible disservice if they ignore the objective evidence on the basis of our subjective miss, or if they fail to distinguish the past year – a sustained period of overvalued, overbought, overbullish conditions that have always resulted in subsequent collapses that wipe out a significant fraction of the market’s value – from that stress testing decision. Understanding this narrative admittedly doesn’t make the half-cycle since 2009 less disappointing, but we can evaluate the resulting methods in a century of market cycles across history, including the most recent ones. Our confidence in their ability to navigate the completion of this cycle, and those in the future, isn't dimmed by the awkward transition that resulted from my insistence on stress-testing in 2009.
Everything looks easy in hindsight. At the end of every episode of “Deal or No Deal,” Howie Mandel asks the contestant to call out the numbers of the remaining unopened briefcases, to see whether the contestant had the million-dollar briefcase all along. It’s always fascinating that people judge whether a contestant “made a good deal” depending on how far they might have pushed their luck before losing everything. This is precisely what investors are doing here. They fear the regret of getting out with what is already a good deal because they might get more if they keep pushing their luck. One might get away with that strategy in a one-time game of “Deal or No Deal,” but in a repeated game, it’s a terrible approach. Investors should have discovered this during the 50% plunges following the 2000 and 2007 peaks, but they have again decided that “this time is different.”
Signs of a speculative peak
Last week, the percentage of bearish investment advisors plunged to just 14.3%, the lowest level in 25+ years. The Shiller P/E (the S&P 500 Index divided by the 10-year average of inflation-adjusted earnings) is now 25.4, and while we use the Shiller only as a readily-available shorthand for other measures, it is certainly better correlated with our actual valuation measures than alternatives like the Fed Model are. Looking over the historical record, the only other instance that bearish sentiment was this low while the Shiller P/E was over 19 was the exact peak of the S&P 500 in January 1973, before the market lost half its value. Looking at the overvalued, overbought, overbullish features that regularly define market peaks, there are only four instances in history when the Shiller P/E was over 19, the market was at a 5-year high, and bearish sentiment was anywhere below 18.5%. These instances were 1972, 1987, 2007, and today. The 2000 peak doesn’t appear because bearish sentiment never moved below 20% that year.
Among the litany of other classic features of a speculative bull market peak, margin debt on the NYSE has surged to the highest level in history, and at nearly 2.5% of GDP, exceeds all but two months in 2000 and 2007. The amount being borrowed to buy stocks on margin is now 26% the size of all commercial and industrial loans in the entire U.S. banking sector. As low-quality, high-risk borrowers rush to take advantage of the present speculative appetite, issuance of leveraged loans (particularly the junkier “covenant-lite” forms) has now hit a record high, already eclipsing the previous record in 2007 at the height of pre-crisis yield-seeking. New equity issuance is also running at the fastest pace since any point except the 2000 bubble peak. At the same time, Bloomberg reports that investors are plowing more into stock mutual funds than at any point since the 2000 bubble peak. Keep in mind how this works - every buyer is matched with a seller in equilibrium, so the same amount of stock is being sold by institutions and other non-retail investors. One doesn’t need to think long to answer who is likely to be the “smart money” in this trade, as the history of surges in retail participation provides a rather firm answer to that question.
Meanwhile, Value Line’s “Median 3-5 Year Appreciation Potential” for the stocks it covers is now the lowest since the 1960’s. Statistically, if one subtracts about 55% from the VLMAP, the result approximates the actual subsequent 4-year total return of the S&P 500. At a VLMAP of just 30%, the implication is a loss, in total return, of about -25% in the S&P 500 over the coming 4-year period. While my impression is that the market is likely to experience an even deeper interim loss, a 4-year overall market loss of -25%, followed by positive 9% average annual total returns for the S&P 500 over the subsequent 6-year period, would compound to produce a 10-year total return averaging 2.3%. That’s right in line with estimates from our own methods. Still, there is good news in that calculation: just because 10-year prospects are dismal at present doesn't necessarily mean that investors will need to be defensive for the next 10 years. I expect that we'll see much more attractive investment opportunities far sooner than that. In any event, when the books are closed on the next decade, we expect realized total returns on the S&P 500 to be even less than the depressed yield-to-maturity on 10-year Treasury bonds, with steep interim losses for stocks along the way.
All of this said, momentum and speculation continue to drive short-term outcomes, and valuation has very little near-term effect on that. It’s always problematic to gauge the endpoint of a parabolic advance, because very small changes in that endpoint can represent significant differences in price (see the chart above of the advance to the 1929 peak for a nice example of this). The gains tend to be surrendered very quickly after those peaks, though often with recovery attempts before severe losses follow. Any discussion about short-term market outcomes relies more on “impressions” and “tendencies” than reliable estimates. My impression is that the present advance remains consistent with a Sornette-type bubble that may have a bit of speculative reverie left in it.
The basic characteristic of a Sornette bubble is a “log periodic” pattern where market declines become progressively shallower and more frequent. I’ve described the investment behavior that drives this dynamic in terms of “increasingly immediate impulses to buy the dip.” What we’ve got now is a rather extreme version of this tendency, where nearly every pullback is seen as an opportunity.
We’ve been asked why the bubble appears to start in 2010 instead of at the 2009 low. The reason is that the advance from the 2009 low and into early 2010 was a legitimate non-bubble advance from a historically reasonable valuation to a historically elevated one. In the third-quarter of 2010, the Fed set a bubble in motion by encouraging the largely superstitious but – at least in the intermediate term – self-fulfilling belief in a central bank “backstop” against market risk. That’s when market dynamics shifted from investment to increasingly frantic speculation that has produced what are now historically obscene valuations.
We should not rely, but should allow for the tendency to buy every dip to move to an even more frantic pitch in the weeks ahead. In the chart below, I’ve enlarged the segment from recent months as a separate inset. Notice that log-periodic bubbles are “fractal” in that smaller segments are similar to larger segments. It will be important to watch for the emergence of fractal self-similarity at the smallest time-scales, which sometimes presents itself as increasing volatility even at 10-minute intervals (a characteristic I also noted at the 2007 peak, and again just before the market collapsed in 2008 – see Nervous Bunny). Remember, however, that this sort of fevered pitch is also terribly impermanent. We vividly recall the seeming “relentlessness” and “resilience” of the market approaching the 2000 and 2007 peaks. Investors must have felt nearly the same as prices approached their peak in 1929.