John Hussman, Ph.D.Pushing Luck 

There are certain points in history where the projections of S&P 500 total returns have differed somewhat depending on which fundamental measure one uses. At present, a wide range of valuation methods that are actually historically reliable show very little variation. Uniformly, and across fundamentals that have reliably correlated with actual subsequent market returns, we project likely S&P 500 total returns in the range of 1-3% annually over the coming decade. Given a 2% dividend yield, this implies that we fully expect the S&P 500 to be no higher a decade from today than it is at present.

These are, of course, the same methods that led us to correctly anticipate a decade of negative total returns in 2000, and significant market weakness in 2007. In contrast, note that these same methods were quite favorable toward equities in 2009 (our stress-testing concerns were not driven by valuations). A passive, equally balanced portfolio of stocks, bonds, and cash can be expected to return about 2% annually over the coming decade. If this seems to be an untenable long-term rate of return, understand that the security prices underlying those expected returns are equally untenable. Also, remember that historical evidence is sufficient to distinguish between competing valuation approaches.

John Hussman

None of this ensures that stocks will decline rather than advance over the near term, but any expectation of adequate longer-term returns in equities from present valuation levels taxes the historical evidence. What accelerates our concerns is the extreme level of speculation in the present market environment. For example, the chart below shows NYSE margin debt both in dollar terms and relative to nominal GDP. We use GDP here because margin debt to GDP has a much higher correlation with actual subsequent market returns than say, margin debt / market capitalization (which destroys information by muting the indicator exactly at points when prices are extremely elevated or depressed). That said, the main usefulness of this measure isn’t for any fixed correlation with subsequent returns – numerous valuation measures do much better – but for its extremes. This is particularly true when margin debt advances rapidly over a span of several quarters relative to prices, GDP and other measures.

John Hussman full article here.