'Sitting Ducks' by John Hussman
If you flip a coin 100 times, you can expect to observe 50 heads on average, and two-thirds of the time you’ll flip somewhere between 45 and 55 heads (about 10% from your expected average one way or another). Flip a coin 1000 times, and two-thirds of the time the outcome will be within 16 heads of the expected 500 (about 3% from your expected average one way or another).
In finance, this feature is what drives the benefits from diversifying across multiple stocks and multiple security types (stocks, bonds, commodities, and the like). Even though the holdings may not actually be completely independent, each has a random component to its return, and the more those random components average out or offset, the greater the reduction in volatility. Essentially, diversification means accepting the randomness inherent in any segment of the portfolio, on the expectation that the randomness will tend to “average out” more often than not. A well-structured portfolio is one where each component is expected to achieve a desirable long-term return, but where the various holdings are also relatively uncorrelated.
This same feature of randomness shows up when we think about the market’s return in each period of time as having a random component. Even when the market is very undervalued or overvalued, the return in any given quarter or year is likely to be much different than the “expected return” we might estimate. What creates greater predictability from this sea of randomness is the constant discipline of aligning one’s position with the expected return at each point in time, and allowing “time diversification” to average out the random effects (see Aligning Market Exposure with the Expected Return/Risk Profile for an overview of how we generally approach this discipline).
Interestingly, when we examine the relationship between fundamental valuation measures and subsequent market returns (especially using reliable measures that don’t swing in lock-step with unstable profit margins), we also find greater predictability as the number of years in the investment horizon increases. This is particularly true when we examine market returns over horizons anywhere in the range of 7-14 years (see Investment, Speculation, Valuation, and Tinker Bell for a number of straightforward approaches that have a near-90% correlation with subsequent 10-year market returns). As one shortens the horizon, the range of error widens. Once you drop the horizon to less than about 3 years, the relationship between valuations and subsequent market returns becomes particularly noisy – still strong from a statistical point of view, but with enough variation that one has to start talking about “tendencies” instead of “expectations." On shorter horizons, factors such as market internals, sentiment, and trend-following measures become important. But because the long-term is made up of a series of short-terms, valuations remain a relevant feature of that broader context.
The present market context is this: from a valuation standpoint, virtually every reliablemeasure of market valuation we observe is now within the highest 1% of historical observations prior to the late-1990’s bubble. “Reliable” in this context refers to valuation measures that are well-correlated with actual subsequent market returns. These measures include price/revenue, price/book, various cyclically-adjusted price/earnings multiples, and market capitalization/GDP, among others. We’ll discuss valuations first, with the caveat that in practice, the most reliable effect of valuations is on long-term returns. The effect of valuations on shorter-horizon returns cannot be separated from other important factors; particularly the quality of market internals and the presence (or absence) of an overvalued, overbought, overbullish syndrome of conditions. Those factors can either mute or amplify the effect of valuations on subsequent market returns, but only for a while.
At its recent high of 24.6, the Shiller P/E (the S&P 500 divided by the 10-year average of inflation-adjusted earnings) matched the level that was observed in September 1929, exceeded the peak of 23 reached in March 1937 (the S&P 500 lost half of its value over the following year), matched the extreme of May 1965, which ushered in a 17-year secular bear market, and significantly exceeded the level of 19.8 seen at the August 1987 peak.
What draws complacency, however, is that valuations are significantly lower here than they were at the 2000 market peak. It doesn’t seem to matter that from the 2000 market peak to the present, the S&P 500 has achieved a nominal total return of just 2.7% annually, and even then only by re-establishing the present strenuous valuation extremes. It also doesn’t seem to matter that 2000 ushered in a decade that included two 50% market declines; one in 2000-2002 that wiped out the entire total return of the S&P 500, in excess of Treasury bills, all the way back to May 1996, and one in 2007-2009 that wiped out the entire total return of the S&P 500, in excess of Treasury bills, all the way back to June 1995.
A second feature that draws complacency is that the Shiller P/E broke above 27 at the late-2007 market peak, just before the market lost 55%, making present valuations seem not-so-high by comparison (“a Shiller P/E of just 24.6? We spit at your Shiller P/E of 24.6!”)
A third feature that draws complacency is that profit margins are about 70% above their historical norms, making raw P/E ratios (particularly those based on “forward operating earnings”) seem fairly reasonable. Understand that the use of raw forward operating P/E ratios implicitly assumes that these profit margins will remain at the most extreme levels in history forever. That’s the essential problem with using a single year of earnings as the basis for stock valuations. Stocks are a claim on a very long-term stream of future cash flows that will be distributed to shareholders over time, and P/E ratios are simply a shorthand. P/E ratios are useful only to the extent that the earnings measure being used is reasonably representative and informative about the long-term stream of cash flows – what might be called a “sufficient statistic.” I made the same observation (to no avail) at the 2007 peak (see Long-Term Evidence on the Fed Model and Forward Operating P/E Ratios). Investors who passively accept Wall Street’s never-ending pitch that “stocks are cheap on forward earnings” without attention to embedded profit margins are the sitting ducks of the investment world.
A final feature that draws complacency is that stocks were already overvalued three years ago, when the market was at lower levels, so to see stocks advance in the face of overvaluation (albeit only because of extraordinary monetary policy) seems to negate the entire basis for concern. I doubt that any of these gains will actually be retained by investors over the completion of the present market cycle. It isn’t difficult to see how badly this sort of complacency has injured investors in previous cycles, but there's no denying that stocks were overvalued some time ago, and stocks have advanced to even greater valuation extremes. The market could not have reached the level of valuation it set in 1929 if the same thing did not happen then. But, here we are. It's worth remembering how brutally those prior overvaluations punished speculators who held out for the last sip of punch.
As I’ve often noted, two features help to distinguish overvalued markets that subsequently move higher from those that subsequently drop like a rock. The main feature is the quality of market action across a broad range of internals (as measured by breadth, leadership, industry groups and security types). In my view, the importance of market internals is that they provide a robust signal about the willingness of investors to accept or avoid risk, and uniform strength across a very broad set of securities and asset types is a signal of risk-seeking preferences. But on average, even favorable market action has historically not been helpful once the other feature emerges – an overvalued, overbought, overbullish syndrome of conditions. The most unusual and frustrating aspect of the period between late-2011 and June 2013 (when our measures of market action finally shifted negative) was the tendency of the market to advance despite persistently overvalued, overbought, overbullish conditions that normally dominate even favorable market internals (see The Road to Easy Street to review how stocks have typically performed during these periods, compared with the past few years).
The best way to understand present market conditions is to clearly understand the tension and interplay between all of these factors: valuations, market internals, Federal Reserve policy, and what we call an “overvalued, overbought, overbullish” syndrome of conditions.
By mid-2013, the market had established extreme valuations that were among the richest 1% of historical observations prior to the late-1990’s market bubble. Following a period of strenuously overvalued, overbought, overbullish conditions, our core measures of market action shifted negative in June and have not yet recovered, creating a condition best described as a broken speculative peak. The initial selloff from that peak was enough to draw bullish sentiment down to less extreme levels. So our defensiveness lately has not been due to a continued overvalued, overbought, overbullish syndrome but instead due to the combination of unfavorable valuations and unfavorable market internals.
Less than two weeks ago, however, the decision by the Federal Reserve to defer any tapering of quantitative easing provoked a spike in the major indices. This spike was unconfirmed by broader measures of market action, but brought those measures closer to a positive turn. That created the potential that a favorable shift in market internals might occur in the absence of extremely bullish sentiment. As I’ve noted in prior weekly comments and should be clear from the foregoing discussion, the combination of rich valuation andfavorable market action, in the absence of an overvalued, overbought, overbullish syndrome, is typically associated with a moderately positive expected return/risk profile, at least over the short-run.
The bottom line is simple and not at all contradictory. First, we view market valuations as obscene, and likely to result in a market loss on the order of 40-50% over the next few years. Second, we view the recent inaction by the Federal Reserve as creating a risk – not an expectation by any stretch of the imagination – but a risk, of resumed speculation that might create a speculative blowoff over a period of weeks. In my view, given the present level of implied option volatility, very little seems needed to insure against that risk with “contingent assets” such as index call options, and given the low probability of such a blowoff, very little seems warranted to insure against that risk. Present conditions are what they are. The dominant risk is that of deep and extended market losses. A possible speculative blowoff would simply make the market’s subsequent losses that much worse. So the essential concern remains that of severe market losses. The secondary risk of a short-run blowoff is simply a source of potential frustration that is easy enough to mitigate.
In the end, we don’t really need to make any forecasts at all. Our discipline is straightforward – to take a greater exposure to risk, on average, in periods where we estimate a favorable market return/risk profile, and to take a smaller exposure to risk, on average, in periods where it is unfavorable. That discipline, repeated period-by-period, can help to elevate the average return/risk profile that is captured over the complete market cycle. We estimate that this would have been true in market cycles across history. Despite what has been a frustrating, extraordinary, and distorted period since the credit crisis, requiring an early period of stress-testing against Depression-era outcomes and a few lateradaptations to the relentless and misguided policy of quantitative easing, this is also – most importantly – an entirely unfinished cycle. I doubt that the S&P 500 will be materially higher a decade from now than it is today, and estimate that even a 40% market loss would only bring prospective 10-year total returns to historically average levels.
Notably, there is far less correlation between prospective stock market returns and 10-year Treasury yields than investors seem to imagine. Between the two, we view bonds as significantly more attractive, even though we estimate their prospective 10-year total returns to be nearly equal. In part, that’s because equities have always tended toward a significant risk-premium regardless of the level of bond yields, while bond yields themselves have been more closely tied to the level of short-term yields; the present yield spread between 10-year bonds and Treasury bills is about double the historical norm. A slowing in the pace of QE, or even a $500 billion reduction in the Fed’s balance sheet, would not provoke any meaningful increase in short-term interest rates by our estimates, particularly in a continuing tepid economic environment. While yield spreads already create a reasonable buffer for longer-term bonds, the same event would likely be devastating for stocks.
Meanwhile, remember how bond pricing works. The current 2.5% coupon, 10-year benchmark Treasury bond presently yields about 2.63%. A one-year holding period changes that security to a 9-year bond, and accrues one year of interest income. As a result, a 50 basis point increase in yield, over a period of a year, would result in a total return of about -1.3%, while a 50 basis point decrease in yield, over a period of a year, would result in a total return of about 6.7%. This is not a prediction or projection of one outcome or another, but it may be helpful in understanding how these securities work mathematically. We see much more fear about bonds at present than we do about stocks, and I believe that this is a profound mis-ordering of concerns. Our own strategy, as usual, is to align our outlook with the expected return/risk profile that we estimate at each point in time. While the secular decline in bond yields may be over, it doesn’t follow that bond yields will advance diagonally from here, and my impression is that the normal range of fluctuation in yields will be perfectly adequate to create good opportunities for approaches that can increase or decrease their exposure over time.
[Geek’s Note: In general, the price of a T-year bond with yield i, coupon C, and face value 100 is (C/i)*[1-1/(1+i)^T] + 100/(1+i)^T where i is expressed as a decimal (e.g. 0.0263). For semi-annual compounding, you simply double T, and halve C and i].
Finally, a few reminders regarding how market cycles are typically completed: First, the “catalysts” for a market decline typically become well-recognized only after a bear market decline is well underway, not as the market is peaking. Second, new unemployment claims typically reach their lowest level just as a bull market is peaking. Third, the presumed “support” that easy money provides for the stock market only exists provided that investors have no desire to hold defensive cash balances. That’s why aggressively easy monetary policy was of no use in preventing stocks from losing half their value in 2000-2002 and again in 2007-2009.
A significant number of voices within the Fed have openly described their recent inaction as a close call, a borderline decision, a missed opportunity because “costly steps had been taken to prepare markets,” and a choice that increased uncertainty and called the Fed’scredibility into question. But even if one believes that quantitative easing will continue indefinitely, it’s critical at least to understand the unreliable mechanism behind its effects. Quantitative easing exerts its only meaningful effect by creating so much revulsion and discomfort with short-term, zero-interest assets that investors feel forced to take risks that they would normally shun, accept dismal prospective returns that they would normally reject, and buy at prices that would normally provoke them to sell with a vengeance. Any event that increases the willingness of investors to hold defensive cash balances is identically an event that could provoke a market collapse.
All of these are lessons best learned from an examination of past data, and not from future experience.