HUSSMAN FUNDS

In mid-September, our estimates of prospective market return/risk dropped to the lowest figure we’ve observed in a century of market history (see Low Water Mark). That week turned out to be the high of the recent bull market, though it’s certainly too early to establish whether that was the ultimate peak. During the recent correction, I’ve noted a modest improvement in our return/risk estimates – which focus on a blended horizon looking out from 2-weeks to about 18-months. However, last week, the stock market experienced some significant damage to internals (breadth, leadership, price/volume measures, etc). As a result, our estimates of prospective return/risk have plunged lower again, to what is now the second most negative figure we’ve observed in a century of data – the September 14, 2012 weekly close of 1465.77 continues to mark the most negative estimate.

My intent in these weekly comments has always been to share what I am looking at, and what our analysis of the economy and financial market suggests – based on extensive historical data and every analytical tool we can bring to bear. There is no need to present the case as any better or worse than it is, but the simple fact is that our return/risk estimates for stocks dropped into the most negative 1% of historical data way back in March of this year, and the estimates we’ve seen since September have been even more extreme.

The S&P 500 has now underperformed Treasury bills for nearly 14 years, including dividends. The cycle since 2007 has been extraordinary in its economic, monetary and fiscal characteristics, not to mention the need to contemplate Depression-era data along the way. It’s undoubtedly easier to dismiss my present concerns as the rantings of a permabear than to understand the narrative of this particular market cycle. But for the benefit of those who do, I want to share my view that the statistical risk of severe market outcomes, given present observable data, has almost never been worse.

That’s not to say that we should necessarily expect market losses matching profound declines such as 2008-2009, 2000-2002, 1973-1974 and so on. There is a difference between the component of market returns that can be considered “predictable” ex-ante (before the fact), and actual market returns, which contain both that “predictable” component and a much larger random component. So to say that the “expected” return is among the most negative in history is a statement about the smaller predictable component, and it doesn’t follow that the full ex-post return will also be among the most negative in history – though we can’t really rule it out.

Also – and this is important – these concerns are based on the data that we observe at present. Could that data shift? Absolutely, and we would respond to more favorable return/risk estimates accordingly. It’s just that we don’t see that yet, and we generally don’t observe such abrupt shifts at rich valuations. Our approach is always to align ourselves with the return/risk estimate that we observe at each point in time.

It’s tempting to assume that last week’s market weakness was nothing more than a post-election letdown for Wall Street, or a transitory focus on the “fiscal cliff.” But that perspective would ignore the months of extreme indicator syndromes that were in place well in advance of the recent weakness. As for immediate catalysts, Germany reported a significant miss in industrial production the day after the election, and the European Union downgraded its expectations for 2013 growth. Given the clear indication of European recession, the weakness in the “backstop” country significantly complicates the prospects that Germany will continue to bail out its neighbors or embrace the monetary equivalents of those bailouts. Here in the U.S., forward revenue guidance from companies was very weak in the most recent quarter, and leadership from a number of growth darlings has deteriorated abruptly (see the notes on exponential revenue growth in Release the Kraken for my views on the inevitability of such disappointments). Given our continued view that the U.S. is already in an unrecognized recession that began in the third quarter of this year, the “recognition” risks remain significant, and extend well beyond concerns about the fiscal cliff.

As I emphasized in September, the negative expected return/risk estimates we observe at present can’t be traced to some single extreme factor. For example, though corporate profit margins remain at the highest level in history, which make valuations look misleadingly reasonable, valuations are certainly less extreme today than they were in 2000 or even 2007. Bullish advisory sentiment has backed off from recent highs, and is certainly nowhere near its historic peak. The intermediate-term overbought condition of the market has also eased in recent weeks, and is nowhere near historic extremes. So again, our concerns are not based on some obvious, extreme indicator. Rather, these concerns are driven by the entire ensemble of indicators we use, taken together.

The present list of concerns includes rich (but not singularly extreme) valuations, coupled with unfavorable market action and a breakdown in market internals and trend-following measures, coming immediately after a seemingly endless series of hostile indicator syndromes (e.g. overvalued, overbought, overbullish). We call these syndromes “Aunt Minnies” – combinations of market conditions that may not be terribly important when observed individually, but that have almost always been followed by a specific outcome in subsequent market data when all of the conditions are observed simultaneously (see for example the October 8 comment Number Five).

In March, when the market was at higher levels than today, I called the growing procession of warnings an Angry Army of Aunt Minnies. But the constant anticipation of further monetary policy announcements from the European Central Bank and the Federal Reserve made the intervening period miserable for us for a while (at least until the hope for further “announcement effects” was removed when the Fed and ECB went all-in with QEternity). In our view, the likelihood of anything but transitory benefit from further QE is limited (see What if the Fed Throws a QE3 and Nobody Comes?). Meanwhile, that sequence of overvalued, overbought, overbullish, technically exhausted setups – followed by a clear technical breakdown – is of greatest concern here, because we often observe that sequence at the beginning of deep and extended market losses.

In practice, we don’t think in terms of “bull market” and “bear market” distinctions, but instead focus on the prospective return/risk profile of the market at each point in time. That is because bull and bear markets can only be identified in hindsight, while prospective return/risk can be estimated at each point in time based on observable data. It’s possible that present conditions will resolve to more favorable return/risk estimates without a full-on bear market, and we’re open to a flexible outlook that moves with the prevailing evidence. To do so today, however, would require those improvements to swim against what we view as an unrecognized global recession in progress.

We can hardly wait for the point – invariably observed in prior market cycles – where depressed or at least moderate valuations are joined with an emerging firming of market internals. We seem to be 180 degrees from that point today.

Lopsided Risks

In the stock market, our ensemble models are comprised of dozens and dozens of individual “learners” – models that are each based on different subsets of indicators and random subsets of historical data. Every week, we examine the consensus of those learners (the proportion that are positive and negative), the dispersion of those learners (the variation across models), and the overall return/risk estimate that is produced by those learners. Consensus tells us something about how robust our conclusions may be – the greater the consensus one way or another, the more confidence we have that similar conditions in the past have resulted in a positive or negative outcome, as the case may be. Dispersion tells us something about uncertainty. Though it is related to consensus, it is more concerned with the overall “spread” between individual estimates. A large spread tells us that the outcome is uncertain – that prevailing market conditions have resulted in widely different outcomes depending on what historical period we examine, or what subset of indicators we consider. A narrow spread tells us that present conditions have produced fairly uniform outcomes regardless of what period and indicators we select.

It is really here where my greatest concern lies, because while our overall return/risk estimate for the stock market is the second most negative in history, that estimate also reflects relatively low dispersion, and the single most lopsided consensus (97% of individual learners negative) that we’ve observed in the historical data set. That means that based on our analytical metrics, the present syndrome of market conditions – taken in its entirety – has regularly and throughout history been associated with some of the worst investment outcomes on record. Again, I am most concerned about the combination of unfavorable valuation and unfavorable market action, including the breakdown in market internals and trend-following measures, immediately following an extended period of overvalued, overbought, overbullish conditions and other hostile syndromes.

Stocks certainly appear short-term oversold here, but the historical record doesn’t give much support to the idea of trading stocks for positive short-term mean-reversion in the face of such decidedly negative return/risk estimates. Then again, the “predictable” component of short-term market movements is small relative to the variation, so these comments shouldn’t be taken as a prediction of near-term outcomes, other than to emphasize what I see as unusual danger more broadly.