Move on, January: at least two more months have more predictive power than you do on stock market returns.
January has a reputation for predicting the direction of the US market for the next 11 months of a year. This supposed ability is known as the “January Predictor” and “January Barometer”.
You will see plenty of evidence of this indicator in the coming days, now that January is officially a “down” month in the record books – with the S&P 500
1.1% slip. I wrote earlier that the January predictor is based on a shaky statistical foundation. Financial headlines, however, will trumpet the supposedly negative impact of the January decline for the remainder of 2021.
Let me point out a few other ways that are not worth following the Predictor.
What is special about January?
A good place to start is to remember that January 2020 was also a bearish month (down 0.2%), and yet the following 11 months produced an above-average profit of 18.4% (assuming dividends reinvested were).
It’s just a data point. Another indication that January is nothing special is that other months have even greater predictive power in forecasting the direction of the stock market for the following 11 months. Since the S&P 500 was founded in 1954, June has shown the strongest forecasting capability, followed by February. January is in third place.
Then why not read about a June predictor or a February barometer? My guess is that followers are motivated less by statistical rigor than by stories and narratives that grab their attention. From a behavioral point of view, the calendar year is a more natural time span than the periods from February to February or June to June. However, the psychological significance is different from the statistical significance.
The importance of real-time testing
There’s another tell-tale sign that the January indicator isn’t doing what it says it is: it won’t pass real-time testing.
By that I mean tests that were done after they were originally “discovered”. If the January predictor had passed these tests, we would have a lot more confidence that this wasn’t just the result of a data mining exercise in which historical data is tortured long enough to create a pattern.
But it couldn’t. As far as I can tell, the real-time test of the January 1973 predictor begins. This is the earliest mention on Wall Street. according to an academic study on the subject. Unfortunately, his record has been far less impressive since then. Since 1973, not only is it insignificant at the 95% confidence level that statisticians often use to determine whether a pattern is real, it is also insignificant at the 85% level.
We shouldn’t be surprised; Indeed, the January predictor is in good company. Consider A study published in the Review of Financial Studies last May. It examined 452 putative statistical patterns (or “anomalies”) found by previous academic research. The authors of this recent study failed to repeat these results 82% of the time. The remaining 18% turned out to be much weaker than originally reported.
No correlation between the magnitude of the increase and the return in January over the next 11 months
Another indication that the January predictor is based on a shaky statistical foundation is that there is no correlation between the strength of the market in January and its gain over the next 11 months. If there was such a correlation, we could potentially create a plausible plot of investor confidence at the beginning of the year that carries over for the rest of the year.
However, there is no such correlation. To believe in the effectiveness of the January predictor, because of this absence, one would have to believe that an S&P 500 gain of just 0.01 has as much predictive power as a gain of 13.2%. That puts a strain on credibility.
Incidentally, I picked that 13.2% in my chart because that’s the biggest January gain for the S&P 500 since it was founded in the mid 1950s. That came in 1987. From January 31 of that year to the end of 1987, the S&P 500 lost 9.9%.
To benefit from a statistical pattern, you must religiously follow it for years
Even if the January predictor was based on a solid statistical foundation, you would have to respond to it for many years in a row to try rationally to take advantage of it. A good rule of thumb in statistics is that you need a sample of at least 30 before patterns become meaningful. In the case of the January predictor, it means you will have to follow it for three decades. Additionally, you wouldn’t be doing any other transactions during those 30 years other than switching to a 100% stock allocation every January 31, when the stock market rises in January, and a 0% allocation when the market falls in January.
Without this patience and discipline, there is little you can do to improve your chances of flipping a coin.
The final result? In any way, the January drop in the stock market doesn’t tell you anything about where it will be on December 31st.
Mark Hulbert is a regular contributor to MarketWatch. His Hulbert Ratings track investment newsletters that pay a flat fee for testing. He can be reached at [email protected]