The “January Effect”: Science, Hocus-Pocus, or Bogus?
Some investors pay special attention to the first five trading days of the year. The reason is that stock-market returns over the month of January have gained a reputation as good predictors of returns over the rest of the year. January is also regarded as a period when returns are relatively higher than those in other months. As with many pieces of ‘market wisdom’, January-driven predictions sometimes come true (as in 2006), but oftentimes do not (as in 2011). For investors, the practical value of this view depends on the relative frequency with which those two outcomes happens to materialize.
The problem in trying to take advantage of this ‘knowledge' is twofold. First, it is not entirely clear if the regularity is statistically solid or what its main drivers may be. Second, even if it was an empirically solid phenomenon, the puzzle translates into explaining why the easy-profit opportunity has not been exhausted by arbitrage. These issues have interested academics for quite some time. There are many papers studying a variety of “anomalies” affecting the seasonality of stock-returns, like the day-of-the-week effect, the turn-of-the-month effect, the pre-holiday effect, the Friday-the-thirteenth effect, and the January effect, which is the focus of this daily.
The three basic postulates associated with the January effect are the following. First, that the sign of stock-returns over the course of the first month of the year are correlated with the sign of returns over the whole year. Second, that returns tend to be higher in January than in other months. And third, that the lion’s share of January’s predictive power lays in the first five trading days of the month. Sometimes, the effect is framed in the context of the small caps vs. large caps premium, or in discussions of the momentum patterns of equity returns.
The three postulates tend to take center stage in many discussions at the beginning of each year, which is why we decided to take a closer look at the January effect in this daily. With respect to each of those postulates, one can simultaneously find some degree of scientific evidence, some coincidences akin to hocus-pocus magic, and some forecasts that approximate prophetic bogus. Although we will not disentangle them all here, what we do is to analyze the long-history of returns in the US and other countries to obtain simple but well-founded conclusions of our own. We also discus some reasons that have been put forward to explain the effect and we highlight the risks involved in taking it too seriously.
The Effect is Somewhat Solid in the US, But Far from Certain
We analyzed the long-run history of S&P500 daily returns (since 1928 to the present time) to answer three basic questions: 1) Is there a positive correlation between returns during the first five trading days of the year (or January more broadly) and those on the rest of the year? 2) Do returns tend to be higher in January than in the other eleven months? And 3) Are returns higher than their historical average when January returns are positive? These are the answers we found:
• When stock-returns are positive during the first five trading days of the year, returns over the whole year are also positive about 60% of the time. The figure is similar when the condition is that returns be positive for the whole month of January. We went further in exploring whether the correlation changed depending on whether the given January was classified by the NBER as a recession month or not. As it turns out, the proportion remains more or less constant for the five-days type of conditioning (at 60% and 61%, respectively), but change somewhat for the whole-month conditioning (at 53% and 61%, respectively). More interestingly, conditioning on whether the analysis applies or not to an election year has a larger impact. The proportions for the five-days conditioning change to 70% and 57%, respectively, and for the whole-month conditioning change to 65% and 57%, in the same order.
• January does tend to show higher returns relative to the other eleven months of the year. Taking the long-run history, annualized returns over the month of January average approximately 38%, compared to the historical average of returns of about 8%. The annualized returns for the first five trading days are much higher. For example, during the post-WWII period, the annualized average returns over the first five trading days reach almost 130%. January surpasses returns in more than half of the other months of the year, and are especially high in comparison to the summer period.
• Returns over the course of the whole year tend to be higher than their historical average if returns in January were positive. The long-run average of stock returns is of approximately 8%. When returns during the five first trading days are positive (negative), the average return for the whole year is 11% (-1%). When returns during the month of January are positive (negative), the average return for the year is approximately 14% (-4%). These figures imply that when January returns are positive, returns over the course of the year are likely to lay above their historical average. In contrast, negative January returns tend to be associated with below-average yearly performances.
To verify the robustness of these findings, we looked at return series that are available for other composites, although they cover shorter histories. The correlation for the first five days of trading is weaker for Dow Jones composites weighted by capitalization, at 50% for the whole-month type of conditioning. In contrast, the correlation is higher for NASDAQ100, at about 64%. In most cases however, yearly returns tend to be higher than average when January returns are positive (at about 15% for the Dow, and 26% for NASDAQ100). Yet a broader test of the January effect is to explore the empirical evidence for other countries.
Patterns Are Similar in Other World Markets, But Data Is Imperfect
There are a handful of other countries for which sufficiently long and high-frequency series are available to carry out similar analyses. Even then, data start since the 1980s or later, which makes it difficult to draw very solid conclusions. In any event, it is surprising that similar patterns are visible along the lines of the questions we posed above for the US:
• The correlation of January returns and rest-of-the year returns is high in many other countries outside the US. Although the evidence is not uniform, we found various examples of countries where returns during the first five days of trading are indicative of subsequently positive returns for the year. In some cases, the correlation remains strong if the conditioning applies to the whole month of January, but not in all of them. Examples include (figures refer to the five-days and whole-month conditionings, respectively): Australia (38%, 38%), Brazil (50%, 38%), Germany (67%, 40%), Greece (63%, 44%), Japan (43%, 40%), Netherlands (63%, 50%), South Africa (67%, 44%), or and Spain (50%, 54%).
• Returns during January tend to be relatively higher with respect to other months in some developed countries, but the evidence is harder to confirm in emerging markets. Because stock markets outside the US and a few developed markets tend to be more volatile, it is harder to distinguish abnormally high returns in January with respect to other months of the year. Errors in the computation of nominal returns in emerging markets compound this problem. Still, we did find examples of countries where similar patterns hold, like Canada, Chile, Mexico, South Korea, South Africa, and Switzerland.
• When January returns are positive, yearly returns effectively tend to be higher than average in many other countries. This is a point where the relative nature of the comparison yields more reliable conclusions. The following are examples of cases where this holds (in parenthesis, we give the country’s historical return, followed by the average return when the January conditioning was positive): Canada (9%, 12%), Chile (11%, 19%), France (8%, 13%), and Switzerland (9%, 15%).
From the perspective of a global investor, it is helpful to know that the patterns more commonly associated to the January effect are visible outside the US as well. However, caveats remain on how strongly one can rely on these empirical regularities.
Exploiting Anomalies May Be Profitable, But Also Risky
The reason why these patterns are anomalies from the perspective of many academics and market participants is that, if they entail opportunities to profit-making in equity markets, arbitrage would eventually bring them down to zero. The survival of the patterns, especially those related to the seemingly abnormally high returns during the month of January, have been related to more ‘fundamental’ causes, especially tax-loss selling at the end of the previous year, transaction costs, year-end de-listings, and omitted risk factors in pricing models.
In fact, ‘risk’ turns out to be a key word when discussing the January effect or other seasonal anomalies of asset returns. The beginning of each year is a period of heightened uncertainty of what will come during the next 12 months. It is a stage of information gathering and strategy positioning by many investors, and early-year price action reflects the convergence of a mixture of expectations. In the end, although it is appealing to know the facts that can be extracted from the historical experience in terms of relative frequencies and average returns, a more careful analysis of fundamental drivers of equity returns is almost surely a better guide to predicting future returns—to the extent such thing can be achieved, of course.
-----------------------------------------------------------------
Discuss this post at the Hand of Scalpuman, our trading forums
Tweet
, in 2007 half of the revenue of the rating agencies came from the ratings of these CDO and other Asset-Backed Securities or ABS stuffs. Half of it, while once again the job of these chaps is to assess the fundamentals and inside information about corporates to give them a rating. Let me now tell you how the ratings of these CDO stuffs used to work then. Moody's and S&P provided the banks that arranged such CDO transactions with an Excel spreadsheet. The slaves of the CDO arranger had to fill it with the portfolio of names of the CDO to be rated with public details : their rating, their industry, etc. They had then to press on the "Rate me!" button to run Monte-Carlo simulations following a very basic gaussian-bell model (the same kind as the ones used by LTCM, works well...) in order to finally get a rating. They had then just to return the spreadsheet to the rating agency for some checks "Are you sure Coca Cola is in 'Aerospace and Defence' ??? " and finally a validation. "OK, it's AAA and for you mate, it's 100 grand mate, thanks!". 100 grand in a couple of hours and ten minutes of work, that was a good business. The trick is the CDO arrangers (Investment wankers) needed a rating as a selling argument to stuff their own clients with "well-rated" crap. And it worked really well as a lot of guys, teased by a sexy rating, got totally stuck with loads of toxic products. In that case, not only the rating agencies were useless, but they happened to be dangerous. 
Posts
