Hoarfrost: Courtesy Wikimedia
Introduction
The “January effect,” in common with the “Halloween indicator” and “sell in May and go away”, is a catchy, get-rich-quick investment idea adored by financial commentators because it is so easy to explain to unsophisticated readers. It rests on the claim that the U.S. stock market performs better in January, compared to the other months in the year.
Unfortunately, financial reports promoting the “January effect” are often vague and confusing. One recent example is here, which, like others in this genre, lacks a specific actionable investment strategy. In fact, this particular report does not even clearly define the January effect.
Historical background
Wikipedia attributes the January effect to 20th century investment banker Sidney B. Wachtel, who noted in 1942 that since 1925, small stocks had outperformed the broader market in the month of January. He and others have further argued that historically the largest January effect occurs in year three of a U.S. president’s term.
However, Wachtel’s observation was based on only the 17 Januarys between 1925 and 1942, covering only four president cycles. With so few observable data and so many variable combinations, such observations are tantamount to backtest overfitting. As we have argued in several recent research studies (see, for example, Pseudo-mathematics and financial charlatanism: The effects of backtest overfitting on out-of-sample performance), overfitting is not merely futile and misleading; it is actually prone to lose money.
Related get-rich-quick promotions
Since the popularization of the January effect” numerous other similar get-rich-quick techniques have also been promoted in the financial press. We have already analyzed a few here in the Mathematical Investor blog (new blog and old blog):
- The myth of the Halloween indicator.
- Do large-cap stocks boost portfolio performance at year’s end?
- The scary chart fallacy.
- Is this stock market weaker during mid-term election years?
Analyzing the January effect
It is worth analyzing the January effect in greater detail. Let us follow Wachtel’s recipe, but use more recent data, say the record of the S&P 500 index for the past 34 years, divided into two parts, namely 1984-2000 and 2001-2017. If one examines this data, looking for some x-month effect in the years 1984-2000, one will observe a “December effect” — an average monthly return of 2.62% or an annualized return of 36.5%. In terms of monthly average return, January is a close second at 2.48% or 34% annualized. This seems to be a strong endorsement to the January effect. However, if one attempts to exploit the January effect, by purchasing an S&P 500 index fund on January 1 and selling it on January 31, keeping cash for the rest of the year, then over the next 17 years (2001-2017), one would lose 0.84% per year.
For such reasons, more careful journalists (see a Fortune report, a CNBC report and a USA Today report) have been quick to advise their readers to ignore the January effect.
Statistical reliability
The failure of the January effect shows the danger of jumping to conclusions based on limited data — indeed, this is the very essence of backtest overfitting. As pointed out above, overfitting difficulties are not limited to the “January effect.”
What can be done to avoid overfitting? Our paper The probability of backtest overfitting presents a method for estimating the probability of overfitting. While the technical details vary from phenomenon to phenomenon, as a general principle observations derived from small samples have relatively low statistical reliability. In individual cases, there are ways to estimate the statistical reliability, but all of this requires significantly more care and sophistication than the writers or readers of financial press reports are typically willing to offer.
The above analysis is still in the ideal world of paper trading. In reality, to utilize a real market anomaly involves much more. With regards to the January effect, even if it were a statistically reliable effect, the problem is that it provides an investment opportunity for only one month. What is one going to do with the investment capital for the other 11 months? Cash out? That is not likely to be a viable investment strategy in any environment.
Common sense
In such cases, a little common sense is useful. Consider that if the January effect were a real, statistically reliable phenomenon, and if there were a solid, no-nonsense means of capitalizing on it, then why doesn’t every self-respecting manager of a quant fund, mutual fund or endowment fund rush to take advantage of it? More significantly, why don’t the highly sophisticated computer programs employed by major quant investment operations automatically detect the phenomenon and then capitalize on it? And if they do, why wouldn’t all of these arbitrage activities quickly cancel each other out?
Why indeed? Any phenomenon that can be easily explained to unsophisticated readers of a financial news column surely cannot be real.