The January Effect, in short, says that January's performance is an indicator of the performance for the rest of the year. Does that still hold today? I'm specifically trying to replicate the results of this SeekingAlpha article. SeekingAlpha uses a much longer time frame than my algo, and while there are notable correlations with the larger frame, the correlations break down a bit later in time. The market is far more efficient today than it was in the 1950s, 1960s, 1970s, so I am not surprised by the result.
From SeekingAlpha: "The month of January has a very important significance in the financial world. It's the month when traders who had a good past year look forward to big bonuses, while traders who performed poorly get to wipe their slate clean and start afresh (if they are lucky enough to still have a job). Asset managers like mutual funds and pension funds analyze their portfolios from the previous year and make fresh allocations and redemptions from not just different asset classes but also into new strategies and new managers. The belief in the market is that these large annual allocations have an impact on not just January, but also set the tone for the rest of the year. This belief has resulted in the lore, 'As goes January, so goes the year.'"
I examined the correlations between the annual returns of the S&P 500 and the monthly returns of the S&P 500 from 2003-2013. SeekingAlpha performed the same analysis over a much longer time period (1952-2013). While my analysis and SeekingAlpha both find a correlation of 0.55 between January returns and annual returns, it is not clear in my analysis that January returns have more predictive power toward the annual returns than do the returns of other months. March, August, December all have stronger correlations with annual returns than does January. However, as mentioned in SeekingAlpha, August or December would be too late in the year to take advantage of an investment signal, since most of the year has already passed. I also looked deeper into the relationship and give a two factor analysis, the two factors being the direction of January returns and the size of those returns. We broke out the January returns in 1% increments and recorded the corresponding average annual returns and percentage of positive years for each period.
You should clone the algorithm and see for yourself. When you run the backtest, you can see the logs of the backtest for the correlations and factor analysis. I'm interested to hear what other people see in this work. Do you think there is a seasonal approach that makes sense in today's markets?
Ryan
