The Good Without The Awful
By John Hussman
January 8, 2013
In the spirit of hope and optimism for the New Year, Iâ€™m going to depart a bit from my usual concerns (which are no less pressing at the moment), and instead discuss when to become bullish, why to become bullish, and how often to become bullish. Using the word â€śbullishâ€ť three times in a single sentence may be a record for me. Despite being aÂ lonely raging bullÂ for years coming out of the 1990 recession, and shifting positive in early 2003 after the 2000-2002 downturn, my defensiveness during the most recent cycle has lent far too much to my characterization as a â€śpermabear.â€ť Any previous bearishness I've had was validated by the 2000-2002 rout, and again by the 2007-2009 plunge, which wiped out the entire total return achieved by the S&P 500 - in excess of Treasury bill yields - all the way back to June 1995. While the S&P 500 - even with the recent advance - has underperformed Treasury bills for nearly 14 years, the stratospheric valuations of 2000 are well behind us. Valuations are still rich, but they are now in the range we've seen near moreÂ typicalÂ bull market highs, so I also expect a more typical frequency of bullish opportunities in the market cycles ahead. Looking over the full span of history, the return/risk estimates from our ensemble methods have been positive about 65% of the time, and would indeed have encouraged a leveraged position (unhedged, plus a few percent in call options) about 50% of the time.
Present conditions will change, and bullish opportunities will emerge, as they always have in other complete market cycles. Understandably, if one expects nothing but a defensive position at all times, even a moderate drawdown makes no sense to endure. But if one is pursuing a risk-managed strategy that seeks to take significant exposure over the course of the market cycle, and to significantly outperform the market over time, the drawdowns should be considered in the context of what the market itself typically experiences over the course of an ordinary cycle.
The average bear market loss is about 32%, and about 39% for cyclical bears that occur during secular bear markets. Given the extreme valuations that weâ€™ve experienced since the late-1990â€™s, the two most recent market plunges in 2000-2002 and 2007-2009 took the S&P 500 down to less than half of the preceding bull market peaks. A 50% drawdown requires a doubling to break even. An 85% drawdown, as the market experienced during the Depression, is even more intolerable because one needs to more than triple just to get back to a 50% drawdown.
A quick note on my reputation as a "permabear." The most recent cycle required us to seriously contemplate Depression-era outcomes, and that presented us with significant challenges. I still believe it would have been reckless to ignore Depression-era data as irrelevant, and I also believe that investors invite ruin if they pursue approaches that are not robust to that data (or worse, restrict their attention to data that primarily includes the bubble period since the mid-1990â€™s). By the spring of 2012, we had addressed what I view as the two extraordinary features of the most recent market cycle â€“ the need to contemplate Depression-era outcomes and incorporate the associated data into our methods, and the need to limit the use of actual put options in an environment where central banks have convinced investors that â€śvirtualâ€ť monetary put options are free. While ourÂ valuationÂ estimates were actually constructive in early 2009, my larger stress-testing concerns kept us defensive until I was convinced that our approach was robust to Depression-era outcomes (stocks lost about two-thirds of their value in the Depression evenÂ afterÂ historically normal valuations were established). The need to address that "two data sets problem" was the "extraordinary" aspect of the recent cycle.
In contrast, the period since spring 2012 is not one that I consider extraordinary. The S&P 500 is presently only a few percent above where it was last March, when our return/risk estimates dropped into the most negative 1% of historical data. Our defensiveness since then has not been rewarded, but that defensiveness has been of the ordinary, repeatable variety, as we also experienced during similar conditions approaching the 2000 and 2007 peaks. Given that I don't believe that market cycles have been permanently repealed, I also expect that the cycle ahead will include periods of well-rewarded defensiveness and ample constructive opportunities as well.
So with â€śextraordinaryâ€ť responses behind us, and a full market cycle ahead, it seems appropriate to put some focus on what is typical during a market cycle, which is actually not defensive positions, but constructive ones. None of this is to say that my concerns at the moment are any less pointed (indeed, weâ€™re at nearly the same set of conditions that Iâ€™ve often noted are among theÂ Whoâ€™s Who of Awful Times to Invest), but itâ€™s a new year, and itâ€™s reasonable to look ahead.
The Good without The Awful
Generally speaking, the very best times to be long are when a market decline to reasonable or depressed valuations is followed by an early improvement in market internals (breadth, leadership, positive divergences, price-volume behavior, and so forth). This is a version of a general principle: bullish investors should look for uniformly positive trends to be coupled with an absence of particularly hostile features such as overvalued, overbought, overbullish conditions. Put simply, we are looking for the good without the awful.
To illustrate this principle, letâ€™s put together a slightly crude model, but one that will serve well enough for instructional purposes. Iâ€™ll preface this by saying that this is not a model we use in practice, that past performance does not ensure future results, and that this model is for illustration only.
For â€śThe Goodâ€ť criteria, weâ€™ll use some simple indicators weâ€™ve used before in other market discussions. Give one point to each of the following:
1. S&P 500 dividend yield below its level of 6 months earlier (a trend-following measure)
2. Dow Utility average above its level of 6 months earlier (a multi-purpose indicator that has trend-following, divergence, and interest-sensitive functions) 3. 10-year Treasury yield below its level of 6 months earlier (an interest-rate trend measure)
Letâ€™s call 2-3 points â€śThe Goodâ€ť and only 0-1 points â€śNot Good.â€ť
As for â€śThe Awfulâ€ť criteria, weâ€™ll require a syndrome including all three of the following:
1. Shiller P/E above 18
2. Investment advisors (Investors Intelligence) bullish sentiment > 45%
3. S&P 500 more than 50% above its 4-year low
Clunky, simplistic, but enough to get some useful results. Most importantly for our purposes, if we look since 1960 at periods that couple the Good without the Awful, we find that 41% of history falls into that bucket. Assuming one was invested each week based on the previous weekâ€™s signal, this subset of periods produced a 21.4% annual total return, or 15.3% in excess of Treasury bill yields, on average.
In contrast, once the Good was joined by the Awful, the resulting 10% of market history featured market losses at a -1.3% annual rate (-4.8% short of Treasury bill yields). Indeed, of the 59% of periods not among the â€śThe Good without The Awfulâ€ť, the S&P 500 achieved an average annual total return of just 2.0%, and a shortfall of -3.1% versus Treasury bill yields, on average. If an investor does not seek to closely track market fluctuations â€“ and we donâ€™t â€“ the optimal strategy is to accept market risk only when the expected market return exceeds the risk-free rate.
The following chart presents the cumulative performance of The Good without The Awful, relative to the S&P 500 Index total return. In contrast to simple moving-average crossover methods, which typically donâ€™t survive transactions costs (seeÂ The Trend is Your Fickle Friend), these criteria would still have outperformed the S&P 500 after imposing a 0.25% cost on every round turn, even though the model moves in and out an average of every 10 weeks or so, and sometimes flip-flops from week to week. Itâ€™s certainly too clunky, binary, and simplistic to use in practice, but itâ€™s instructive.
The fact that the model is out of the market well over half the time is certainly beneficial from the standpoint of drawdowns. The deepest drawdown experienced by this model since 1960 is just -13% (compared with several separate drawdowns in the 45-55% range for the S&P 500). One has to proxy advisory sentiment to test the model prior to 1960, but thatâ€™s actually not terribly difficult given the strong correlation between sentiment and the size and volatility of recent price changes. Unfortunately, like many models that weâ€™ve tested (other than our present ensemble methods), the worst drawdown in Depression-era data exceeds 50%. Now, thatâ€™s certainly better than the 85% loss experienced by a passive buy-and-hold (where again, the market had to more thanÂ tripleÂ from its 1932 low just to get from an 85% lossÂ backÂ to a 50% loss). But itâ€™s still not acceptable, in our view.
One might ask why we donâ€™t simply use this model, despite its clunkiness, given a perma-bearish alternative. The answer is that my reputation as a permabear is an artifact of our 2009-early 2010 stress-testing period where we were forced to contemplate and incorporate Depression-era data. Also, the success of the simple model above to classify favorable/unfavorable market conditions is a fraction of what we observe from the broad ensemble methods that we presently use to make return/risk assessments (as much as I enjoy sharing research, nobody would publish a model that was competitive with what they rely upon). The ensembles produced positive return/risk estimates in a much larger portion of the 2009-early 2010 period, capture a larger set of positive market conditions more generally, and handle extreme periods such as Depression-era data more effectively.
Still, one might ask whether the Good/Awful model can be used as a â€śfilterâ€ť for our broader ensembles. For example, we can examine periods when our ensembles were negative enough to encourage strongly defensive hedges, and see how the Good/Awful criteria partition those outcomes. Among the most negative periods identified by the ensembles, even instances that wereÂ positiveÂ based on â€śThe Good without The Awfulâ€ť criteria were associated with average annual losses of -15.2% for the S&P 500 (-17.3% short of Treasury bill yields). In contrast, the instances that wereÂ negativeÂ on those criteria â€“ as is presently the case â€“ were associated with average annual losses of -48.1% for the S&P 500 (-49.5% short of Treasury bill yields). So while the Good/Awful criteria are helpful in distinguishing bad from abysmal, even positive readings on this model arenâ€™t enough to counter negative return/risk estimates from our ensembles.
In any event, regardless of the particular model we use, when we are looking for bullish investment opportunities, we are really seekingÂ some versionÂ of the good without the awful. We donâ€™t need deep undervaluation, but we are seeking the combination of more moderate valuations coupled with an early improvement in market action. There are many ways to define the basic convergence of reasonable valuations and market action, coupled with anÂ absenceÂ of awful conditions (overvalued, overbought, overbullish, yields rising). Presently, we observe the awful, and not enough of the good, but I have no doubt that we will observe a far more favorable convergence of market conditions over the course of the market cycle ahead.
Market returns vary with market conditions. A useful investment criterion is one that helps you to â€śpartitionâ€ť the data into two or more subsets, each having a different return/risk profile. Thatâ€™s essentially what the â€śGood without the Awfulâ€ť criteria are doing. They partition the S&P 500â€™s annual total return of 9.5% since 1960 into two subsets; 41% in a subset that is associated with a â€śconditional returnâ€ť averaging 21.4%, and 59% in a subset that is associated with a â€śconditional returnâ€ť averaging just 2% (and less than Treasury bills, on average). Geeks Note: (1.214)^0.41 * (1.020)^0.59 = 1.095.
The better the criteria, the greater the separation between returns across subsets. For example, even including 2012, the most negative 5% of historical periods from the standpoint of our ensemble methods is associated with an average annual loss of -40.5% (-42.0% short of Treasury bills), which is why we are willing to endure an uncomfortable defensiveness at present. We know we wonâ€™t remain in this partition for long. About 30% of periods since 1940 would be associated with a full hedge, and an average conditional return of about -0.1% for the S&P 500. About 65% of historical periods have produced positive return/risk estimates, and are associated with an average annual return on the S&P 500 of 22.6%. So the 11.2% total return of the S&P 500 since 1940 partitions out to (.595)^0.05 * (.999)^0.30 * (1.226)^0.65 = 1.112.
Needless to say, the period since 2000 has required far more defensiveness than most of market history, owing to extreme initial valuations. The Shiller P/E (S&P 500 divided by the 10-year average of inflation-adjusted earnings) touched 44 in 2000, and is presently at 22. That adjustment in valuations explains why the S&P 500 has underperformed Treasury bills for nearly 14 years. Present valuations may still be rich, but weâ€™ve approached similar extremes in other market cycles. Thereâ€™s every reason to expect numerous opportunities for bullish investment positions as we move forward from here, if not at the moment.
From the standpoint of investment discipline, itâ€™s instructive that even while â€śThe Good without The Awfulâ€ť criteria easily outperform the S&P 500, even after reasonable estimates of trading costs and slippage, and with far smaller drawdowns than the market even in Depression-era data, the model hasnâ€™t gained any net ground since October 2011. An unhedged position in the S&P 500 has performed much better since then, but that would hardly be a reason to abandon one for the other at what seems like a particularly extreme and inopportune point to accept market risk. This underscores the difference between long-term strategy and short-term performance. The importance of discipline, process, research, historical evidence and full-cycle characteristics canâ€™t be understated.
The past year was frustrating for risk-averse investors, particularly for investors who believe that large imbalances are eventually corrected (as they were in 2000-2002 and 2007-2009). The Federal Reserve and the European Central Bank were effective in kicking these imbalances down the road, at least for a quarter or two at a time, and the fiscal deal at year-end ensured that the short-run benefits of massive budget imbalances wonâ€™t be marred by a return to fiscal sustainability anytime soon. At least for investors who care both about risk and about the long-term, the magnitude of these distortions has been disconcerting. I donâ€™t believe that they will prove to be costless, but I do believe that we have adapted to a world where we can navigate both the distortions and their eventual consequences â€“ particularly over the complete market cycle.
Meanwhile, we are looking eagerly, but not impatiently, for bullish opportunities. This weekly comment could very well have been titled â€śA Renewed Whoâ€™s Who of Awful Times to Invest,â€ť because present conditions are largely back to â€śawfulâ€ť with an overvalued, overbought, overbullish, yields-rising syndrome. But having noted that, we also want to look ahead. The real message is that these conditions will change soon enough, and a sufficient amount of the good, without the awful, will be just fine.
The foregoing comments represent the general investment analysis and economic views of the Advisor, and are provided solely for the purpose of information, instruction and discourse. Only comments in the Fund Notes section relate specifically to the Hussman Funds and the investment positions of the Funds.
Past performance does not ensure future results, and there is no assurance that the Hussman Funds will achieve their investment objectives. An investor's shares, when redeemed, may be worth more or less than their original cost. Current performance may be higher or lower than the performance quoted above. More current performance data through the most recent month-end is available at www.hussmanfunds.com. Investors should consider the investment objectives, risks, and charges and expenses of the Funds carefully before investing. For this and other information, please obtain a Prospectus and read it carefully.
(c) Hussman Funds