“We learn from history that we do not learn from history.”
Georg Wilhelm Friedrich Hegel
Last week, Investors Intelligence reported that bullish sentiment surged above 60%, coupled with a 5-year high in the S&P 500 and valuations beyond 18 times record trailing earnings. The same combination was last seen the week of the October 2007 market peak, last seen before that in January and May 1999 (which we should emphasize was good for only a 5% correction in the short run before a choppy run to the 2000 peak, but would still leave the S&P 500 more than 40% lower three years later), last seen before that the week of the August 1987 pre-crash peak, and last seen before that in January 1973, just before the S&P 500 lost half of its value.
Market conditions presently match those that have repeatedly preceded either market crashes or extended losses approaching 50% or more. Such losses have not always occurred immediately, but they have typically been significant enough to wipe out years of prior market gains. Aside from the 2000-2002 instance, they also have historically ended at valuations associated with prospective 10-year S&P 500 nominal total returns in excess of 10%. At present, reliable valuation measures are associated with estimated total returns for the S&P 500 of just 2.0% annually over the coming decade. On the basis of historically reliable measures, the S&P 500 would have to move slightly below the 1000 level to raise its prospective returns to a historically normal 10% annually. Given short-term interest rates near zero, economic disruptions would probably be required in order to produce that outcome over the completion of the current cycle, and we have no forecast or requirement for that to occur. Of course, there is no shortage of historically unreliable measures available to offer assurance that equity valuations are just fine.
Regardless of whether the market’s losses in this cycle turn out to be closer to 32% (which is the average run-of-the-mill bear market loss) or greater than 50% (which would be required to take historically reliable valuation measures to historical norms, though most bear markets have continued to undervalued levels), it’s going to be difficult to avoid steep losses without a plan of action. In our view, that action should be rather immediate even if the market’s losses are not. However uncomfortable it might be in the shorter-term, the historical evidence suggests that once overvalued, overbought, overbullish conditions become as extreme as they are today, it’s advisable to panic before everyone else does.
Meanwhile, our own approach remains to accept market risk in proportion to the return/risk profile we estimate based on observable conditions at each point in time. That has kept us out for quite a while, because history does not teach us to speculate until market conditions become as extreme as they are at present. It only teaches that steep losses typically follow once they do.
We are emphatic about two points here:
- The extreme conditions that we observe today do not necessarily resolve into near-term predictions about market direction. Historically, the most severe overvalued, overbought, overbullish syndromes do not precisely overlap market peaks, and occasionally precede them by months or quarters before they are resolved by steep losses. The eventuality of steep losses is predictable, but the timing is not. Speakingvery generally, similar extremes in history followed 5-year diagonal advances and were followed by 2-year collapses. Still, aside from the view that conditions are already extreme and monetary policy is an unreliable and receding support, yield-seeking speculation has played an enormous psychological role in the recent half-cycle, and marking turning points has not been our strong suit;
- Our own challenging experience since 2009 is not an adequate reason to ignore the objective historical evidence here. Our experience in recent years is not the reflection of a static investment method. It began with my insistence, at the height of our success in 2009, on ensuring that our methods were robust to Depression-era outcomes. While I saw that as a fiduciary responsibility, the decision immediately resulted in missed gains early in this half-cycle. And while the ensemble methods that we introduced in 2010 performed better in complete historical cycles than any approach we’ve tested, repeated bouts of quantitative easing then required us to reintroduce certain bubble-tolerant features of our pre-2009 methods as an overlay (mostly relating to what we called "trend uniformity" during the late-1990's bubble). The combined result was an unexpectedly difficult and awkward transition from our pre-2009 methods to our present methods. Ironically, both methods of classifying market return/risk conditions capably navigate market cycles across a century of history (though our present methods handle Depression-era data better), including the current cycle, had either method been in practice without that transition (SeeSetting the Record Straight).
We remain fully confident in the ability of our investment discipline to navigate the completion of the present market cycle and those that follow. My apologies to regular readers for repeating our discussion of our stress-testing narrative since 2009 – but the incorrect belief that our experience in the recent half-cycle traces to our strategy itself and not that stress-testing transition continues to be the source of endless second-guessing, not to mention well-meaning suggestions that we’ve already addressed in recent years to the greatest extent that the historical evidence supports.
