- Be skeptical about the market’s alleged patterns -
How should you get ready for the worst month of the year for stocks?
The surprise answer: By doing nothing.
That’s the proper response because September’s awful record is, in effect, a crime in search of a motive. And without a motive, we have no business betting on the pattern’s persistence — no matter how impressive it might be from a purely statistical point of view.
To be sure, the statistics behind September’s awful record appear to be impressive indeed. Not only is the month’s average return far worse than every other month (an average loss of 1.05% since 1896, versus an average gain of 0.75% across the other 11 months), September’s average can’t be traced to just a few terrible instances. In fact, in every decade but one since the Dow Jones Industrial Average was created in 1896, September was one of the worst performers — as you can see in the accompanying chart.
That is remarkable consistency. Using traditional thresholds of statistical significance, it is unlikely this can be attributed to chance.
But it turns out that those standard thresholds are not appropriate for judging whether September’s record is statistically significant, according to Campbell Harvey, a finance professor at Duke University’s Fuqua School of Business. That’s because, in the process of “discovering” that September’s record is so awful, many other hypotheses over the years undoubtedly were explored by investors slicing and dicing stock-market history in search of seasonal patterns.
With each additional hypothesis, Harvey told me in an interview, it becomes more likely that a pattern that in fact is completely worthless will nevertheless appear to have a phenomenal record. In order to correct for that possibility, statisticians need to apply increasingly stringent standards of statistical significance.
How stringent? The answer depends on how many other hypotheses were also tested before “discovering” that September’s record is so awful — and that number of course is unknown. But it undoubtedly is a large number, making it doubtful that September’s otherwise impressive statistical record is significant.
But even if it were significant, Harvey added, there is an additional hurdle over which statistical patterns must jump in order to make them worth betting on: namely, can a plausible story be told for why that pattern should exist in the first place? If not, odds increase that the pattern is a fluke.
Can such a plausible story be told about September? I seriously doubt it. On each occasion over the last decade when I have written about the month’s awful record, I have challenged you to provide me a credible explanation for why it should be so bad for stocks. While some of you took up the challenge, none of your explanations has been able to withstand statistical scrutiny.
Let me review the most-suggested of these failed hypotheses so that you don’t email me to suggest them yet again. The four at the top of the list are:
By the way, September is not alone in its inability to jump over the two statistical hurdles that Harvey mentions. The same is true for most of the statistical patterns that capture Wall Street’s attention. In fact, Harvey argued in his address earlier this year to the American Finance Association, the premier association of finance academics, the same is also true for many of the patterns that up until now have enjoyed academic seals of approval.
In a roundabout way, therefore, September’s allegedly terrible record can help us to become better investors. That will happen if we use the month as the occasion to develop a healthy skepticism towards alleged patterns in the stock market, demanding that they scale high hurdles before risking our hard-earned assets by investing according to them.