It turns out that January may have special powers to predict the rest of the year, notwithstanding a Hulbert Financial Digest study suggesting that January’s predictive powers are not unique.

That HFD study found that most months of the year, not just January, do a credible job of foretelling the stock market’s direction over the subsequent 11 months. In fact, that study discovered April and November have a better statistical record in this regard than January. The study reached these conclusions after analyzing the Dow Jones Industrials Average back to 1896, when this benchmark was created.

These conclusions are placed in a different perspective by a just-released academic working paper by Michael J. Cooper and John J. McConnell of Purdue University and Alexei V. Ovtchinnikov of Virginia Tech. (Their study is not yet published, but Prof. Cooper tells me that you can obtain a copy by emailing him at mcooper@purdue.edu.)

According to Prof. Cooper and his colleagues, the HFD’s findings are largely a function of what happened during the 1930s. In that decade, January did a horrible job of predicting the market’s direction for the rest of the year: On average, positive Januaries were followed by terrible years, while negative Januaries were following by wonderful years.

Anyone who invested according to the so-called January Indicator during the 1930s, therefore, lost a lot of money. Any historical study whose focus includes that decade, therefore, will tend to be swayed away from the notion that January has special predictive power.

Another fascinating finding from the researchers: During the 1930s, some of the non-January months did an impressive job of foretelling the market’s direction over the subsequent 11 months. This was especially the case for April and November, for example.

If we ignore this decade, according to Prof. Cooper and his co-authors, two things simultaneously happen. First, the non-January months lose their apparent predictive power. And, January’s ability to forecast the rest of the year becomes stronger.

The net effect of ignoring the 1930s is to make January appear unique in its ability to predict what the market will do over the subsequent 11 months.

What importance should you as an investor put on the findings of this new study? It depends, in part, on whether you think the 1930s should be ignored. If you think they should be, then you are on relatively strong ground when betting on the January Indicator because Prof. Cooper and his colleagues report that January’s predictive abilities since 1940 have been statistically quite significant.

If, in contrast, you see no good statistical reason to ignore the 1930s (and I don’t), you will be less inclined to place importance on the January Indicator.

Even if you are willing to overlook the 1930s, however, there’s another hurdle you need to jump over before basing your investment strategy on the January Indicator: You need a plausible explanation for why the indicator should work. Without such an explanation, you run a big risk that the indicator is based on nothing more than a fluke of the data.

My favorite example of why this is so comes from David Leinweber, a visiting faculty member at CalTech’s economics department. Several years ago, wanting to illustrate the perils of mining the data for spurious correlations, he searched through all the data on a United Nations CD-ROM to find the indicator with the most statistically significant correlation with the S&P 500. His discovery: Butter production in Bangladesh.

So, statistical correlation is not, in itself, a sufficient enough reason to alter your strategy.

Is there a plausible explanation for why January should have special ability to predict the market’s direction over the subsequent 11 months? I am not aware of any. Nor are Prof. Cooper and his colleagues, who write that, after investigating any of a number of possible explanations, they essentially came “up empty.”

This does not guarantee that the January Indicator has no validity. But, in my opinion, it suggests that acting on it is very risky indeed.