There is a popular mythology that January has particular significance as a leading indicator for the direction of the stock market for the entire year. One belief is that the first three (or the first five) trading days indicate the likely direction for the entire month.
The January effect is one myth that appears to have some foundation in fact. We'll take a close look at what more than 80 years of stock market data reveal.
The following table summarizes the various January effects. The bottom line in the table shows averages for all 82 Januarys and for all 81 years.
There is a January effect for every parameter examined. Only those that are about 40% or larger are considered significant and are shaded green. The most significant effects (shaded red) are more than double the base line.
Here are examples:
- 1. Sixty-seven percent of all 82 Januarys have had gains in the Dow Jones Industrial Average. When the first three days of January are up, 73% of the months finish higher. This is an improvement of 9%. That's not considered a "tradable" difference.
- 2. Thirty-three percent of all Januarys have had losses in the Dow. When the first three days of January are down, 48% of the months finish lower. This is a significant difference of 45%.
- Thirty-four percent of all 81 years are losing for the Dow. If January is down, 69% of those years have negative returns. The January effect here is 103%, a most significant effect.
The January effect has changed a little for different time periods. In the table below the effect is shown for three time periods: 82, 42, and 22 years.
The following graph shows that there is no relationship between the size of the market return for January and the size of the return for the rest of the year.
The linear trend line for all the data points has an R2 value very close to zero. A perfect relationship between January and the entire year return would have R2 = 1. Total lack of relationship is indicated by R2 = 0. The value of 0.08 is effectively zero.
An attempt to delete outliers wasn't successful in getting a major increase in R2. Deleting 1931 (down 53%) and 1933 (up 64%) increased R2 to only 0.11. Deleting those two years plus all other gains above 30% and losses below 20% produced R2 = 0.15. Neither of these shows any significant correlation of the data to the linear function. Attempting a fit to polynomial functions up to sixth power also produced R2 less than 0.15.
The size of the market change in January has no predictive ability regarding the amount of gain or loss for the year.
Don't ignore the January effect, especially when January is down. That is the case this year. The odds of 2010 being a losing year are double the odds for all years. The last six Januarys that were down saw three losing years for the Dow (2002, 2005 and 2008) and three winning years (2000, 2003 and 2009). While this is only 50%, it is still 48% above the 34% losing years for all 81 years, and more than 100% above the 24% losing years out of the most recent 21.
At the time of publication, Lounsbury had no position in