By John McDonald
With some amusement, I read a recent article on MarketWatch by Mark Hulbert. The core concept behind the article was to use available data to make a relevant prediction about the severity of the downturn whenever the U.S. stock market next tumbles into a bear market phase.
Such prediction is, as Hulbert begins by admitting, something of a fool’s game. To spoil the surprise for you, I’ll go ahead and tell you that Hulbert concludes that the next bear market will sink the Dow Jones Industrial Average by a whopping 35.3%.
However, I strongly question the basis for Hulbert’s conclusion.
While this article inadvertently makes a great case for the wisdom of diversifying into precious metals, it fails miserably at giving a well-reasoned prediction about stocks. Let’s take a look at why this doomsday prediction has no statistical relevance whatsoever based on the data in Hulbert’s own article.
Hulbert describes his tests of a number of variables in order to identify “the one” which is most indicative of the severity of subsequent bear markets. To do this, Hulbert analyzed data from 36 different bear markets that have happened since 1900.
What he finds is a huge pile of nothing, but this does not stop him from making a prediction.
Hulbert begins by testing the question of whether the calendar length of a bull market is an indicator of how severe the subsequent bear market will be. If this hypothesis proved out, it could be important to us since the current bull market is among the longest (maybe the longest) bull market in history. That being the case, I’m happy to report to you that Hulbert found no substantive difference in the severity of bear markets on the basis of the duration of the immediately preceding bull market.
What about price/earnings (P/E) ratios? The data here is more interesting. Hulbert divided the 36 bear markets into two groups: He calculated the market’s P/E ratio at the start of the 36 most recent bear markets and found that the bear markets beginning when stocks are relatively cheap (“low” P/E ratios) tend to be less severe than bear markets beginning when stocks are relatively expensive (“high” P/E ratios). When stocks were cheap, the average decline was 27.8%. When stocks were expensive, the average decline was 34.5%.
This seems interesting and potentially valuable until Hulbert shares this tidbit: “Unfortunately, given the relatively small sample size and the variability in the data, this difference is not significant at the 95% confidence level that statisticians often use when determining whether a pattern is real.”
In plain English, this means that statisticians would not judge this analysis as reliable. Thus, no dice.
Unfortunately, the same result befalls an analysis of a specialized variant of the P/E ratio known as the CAPE (cyclically-adjusted price/earnings ratio): Cheaper stocks at the start of a bear market likely means a lesser decline whereas more expensive stocks likely mean a great decline. But again, there’s not enough statistical confidence to rely on this information.
The kicker? Despite Hulbert’s admission of the statistical irrelevance of this analysis, he proceeds to use the same data to posit a 35.3% stumble for the value of the Dow Jones Industrial Average on the basis of a nebulous model he describes as “a simple econometric model whose inputs are past bear markets and CAPE values”.
To me, this seems an inadequate basis for concern. I make no predictions about upcoming bear markets other than to say: At some point, there will be one. You should be ready when it happens.