Party Like It's 1999 (and 1929)

A quick update on near-term market action. Last week’s market retreat was a very minor example of the “unpleasant skew” I’ve discussed in recent months. Under present market conditions, the single most probable outcome in a given week remains a small advance, but with a smaller probability of a steep loss that can wipe out weeks or months of gains in one fell swoop. So the mode is positive, but the mean return is quite negative (see Impermanence and Full Cycle Thinking for a chart of what this distribution looks like). Again, last week was a very minor example. Prospective 10-12 year returns increased by only a few basis points.

In recent months, the compression of volatility has encouraged speculative put option writing by pension funds, coupled with increased market exposure by volatility-targeting strategies that buy as volatility falls and sell as volatility rises. Last week, JP Morgan’s quantitative derivatives analyst Marko Kolanovic observed “Given the low levels of volatility, leverage in systematic strategies such as Volatility Targeting and Risk Parity is now near all-time highs. The same is true for CTA funds who run near-record levels of equity exposure.” This setup is reminiscent of the “portfolio insurance” schemes that were popular before the 1987 crash, and rely on the same mechanism of risk-control - the necessity of executing sales as prices fall - that contributed to that collapse.

I continue to expect market risk to become decidedly more hostile in the event that various widely-followed moving-average thresholds are violated, as those breakdowns are likely to provoke concerted efforts by trend-following market participants to exit, at price levels now here near the levels where value-conscious investors would be interested in buying. For reference, the 100-day average of the S&P 500 is about 2120. The 200-day average is about 2057. For now, keep “unpleasant skew” in mind, to avoid becoming too complacent in the event of further marginal advances. I see the most preferable safety nets as those that don’t rely on the execution of stop-loss orders.

Looking out on a longer-term horizon, the following chart shows the ratio of nonfinancial market capitalization to nominal GDP, where we can reasonably proxy pre-war data back to the mid-1920’s. Our preferred measure is actually corporate gross-value added, including estimated foreign revenues (see The New Era is an Old Story), but the longer historical perspective we get from nominal GDP is also valuable. The chart shows this ratio on a log scale. To understand why, see the mathematical note at the end of this comment. The recent speculative episode has brought this ratio beyond every extreme in history with the exception of the 1929 and 2000 market peaks.

As I’ve often emphasized, one of the most important questions to ask about any indicator is: how strongly is this measure related to actual subsequent market returns? Investors could save themselves a great deal of confusion by asking that question. Hardly a week goes by that we don’t receive a note asking, for example, that “so-and-so says that earnings are going to strengthen in the second half - doesn’t that make stocks a buy?” Well, it might, if year-over-year earnings growth had any correlation at all with year-over-year market returns (it doesn’t), or if there was evidence that earnings tend to strengthen in an environment where unit labor costs are rising faster than the GDP deflator (they don’t). That’s not to say that we can be certain that earnings or stock prices won’t bounce, but we can already conclude that so-and-so hasn’t convincingly made their case. When investors ignore the correlation between indicators and outcomes, they make themselves the victims of anyone with an opinion.