Is the 'January Barometer' for Real?

By Heather Bell

The folks out there who believe in the January Barometer likely weren't that happy when February 1 rolled around. After all, the S&P 500 was down about 6%, one of the worst January performances in the history of the index.

The January Barometer theory is based on the idea that the direction of the market in January will dictate the direction of the market for the whole year. The Stock Trader's Almanac 2007, which is published by the Hirsch Organization, says the theory has a .750 batting average for accuracy. I don't know anything about baseball, but it sounds pretty good. If Wikipedia is explaining the definition of "batting average" clearly, that works out to 75% accuracy, right? Fifty percent would be the expected accuracy rate if the January Barometer theory were just random chance.

In any case, Ben R. Marshall and Nuttawat Visaltanachoti, both of New Zealand's Massey University, have recently had a study posted on the SSRN Web site that evaluates the theory's accuracy using "stochastic dominance and manipulation-proof performance measures to account for risk, and dummy variable persistence and data mining adjusted test procedures."

When I read that line out loud, it sounds like one of Charlie Brown's teachers is talking to me. But I'm taking it to mean that the folks from Massey University did something very mathematical and/or statistical to filter out any biases or misleading data.

Correct me if I'm wrong. Once you get past the introduction, the study first gives a rundown on previous examinations of the theory and its portrayal in the media, then describes the methodology used and the findings.

What's pretty cool about "How Accurate is the January Barometer?" is that the authors applied their methodology to 22 equity markets around the world in addition to the United States, looking at 23 markets in the MSCI World Market Index (as well as the world index itself) during the 1970-2006 time period.

The data reveals that—not including the results for the U.S.—initially only Switzerland, Spain and the World Index (which historically is dominated by the performance of the U.S.) show any sort of statistically significant predictive qualities for January performance with regard to the performance of the following 11 months.

After the stochastic dominance hocus pocus and dummy variable persistence mumbo jumbo are applied, any predictive qualities become statistically insignificant for all three indexes. So basically the January Barometer is a wash for the rest of the developed world, according to Marshall and Visaltanachoti. With the U.S., the verdict is not necessarily so clear-cut.

The study looks at the U.S. market as a whole (using more than 60 years' worth of CRSP (Center for Research in Security Prices) data, both value-weighted and equal-weighted) and also broken down into 10 size segments, 10 book-to-market equity ranges and 10 industries. In all cases, the January Barometer manifested statistically significant results (with the smallest size decile showing the strongest evidence of the theory's accuracy and the highest book-to-market equity decile showing the weakest evidence).

If the study did not go beyond this point, one might be tempted to throw in the towel on 2008, pull one's money from the stock market and stick it in the proverbial mason jar buried in the backyard. (Some people might be better off using one of those super-sized pickle jars from Wal-Mart or, like me, find the little jar the capers came in to be perfectly sufficient.) But Marshall and Visaltanachoti don't stop there: They decided to look at how the January Barometer theory works with individual stocks and with commodities.

When they looked at stocks that had positive performance in January, the data indicated that those stocks were not necessarily the ones that showed positive performance for the next 11 months. In some cases (though they were statistically insignificant), the data seemed to imply that an individual stock's positive (negative) performance in January actually might indicate that negative (positive) performance was likely for the following 11 months. So individual stock data did not back up the validity of the January Barometer theory, which is what one would expect it to do, even if the theory is more of a broad market concept.

The study also examines the performance of U.S.-traded commodities under the assumption that, if the January Barometer were the result of some behavioral quirk of the U.S. market, it would also apply to that particular asset class. And here's where I think the study may stray into some shaky territory—commodities are subject to so many global factors, no matter where they're traded, that it seems like they would instead show results similar to those of the MSCI World Index. So it really doesn't surprise me that the results for commodities don't support the January Barometer theory.

That said, I think the evidence that the January Barometer does not apply to individual stocks is enough to disprove the theory and support Marshall and Visaltanachoti's contention that the behavior is nothing more than the result of chance. Probably over time, as the years pass and the sample size grows, that 75% accuracy rate will trend toward 50%. I'd love to hear readers' reactions to the study and what they think the evidence supports. Based on January's rather horrific slide, do you think you can make a reasonable bet on the direction the market will take in the next 11 months?

You can access the article here.

Comments

[Gary Meade:]

What is that? Lies, lies, and stats or insert your favorite manipulation.

2008 Mar 15 09:20 PM