It seems that as inevitable as the sun rising, that the New Year forecasters attempt to sooth-say over what the year holds in store for equity markets. Additionally, there has been renewed interest recently surrounding the so called “January effect” which many people think drives the stock market in the year ahead.
Interest around the January performance has always been high, probably because investors look, hopingly, for clues which can give them a good start to the year. Many theories have evolved around the January effect but at its heart it basically states that should stock prices during primarily the first two weeks of January be buoyant that above average returns will be delivered over the remainder of the year.
A possible explanation for the January effect, certainly in the US is that individuals sell stocks before the end of the year in order to mitigate their capital gains tax issues and then re buy them again a few days after, in the New Year. Portfolio adjustments by fund managers and the receipt of year bonuses at the beginning of January are two additional explanations for these abnormal returns too. Our research however does not show that the January effect is a clear cut issue.
Digging into the data
In order to ascertain whether the “January effect” is myth or fact, we collected data for the S&P 500 from 1953 to 2012. Our database has 60 years of data and so corresponding to 60 periods. During this period, the S&P rose 43 times (72% of the cases), and recorded an average annual return of 8.30%. Each of the January’s recorded an overall rising proportion of 60:40 and recording an average monthly gain of 1.01%. So far so good, but one could say “with the S&P 500 returning 8.3% on average, it is only to be expected that January is a good month, as with any other during these rising years”. That’s right, the stock market is rising, and the probability of a monthly rise is always higher than that of a decline during a bull run. But, if we turn the 8.30% average annual return into a monthly average, we find that the average monthly gain is 0.67%, much less than our found 1.01% average gain for January, and so meaning that January really is a better month than average. Still in 40% of the cases where the stock market rose for the year, January ended with a decline and so we are in need of something more solid to validate the saying.
Testing the Data for a January Effect
What would be great as a trader is to look into the January data and then find a signal that we could use as a trigger to invest in equities. Following a similar methodology used by Tom Bowley in his article “Gauging the January Effect”, we ranked our dataset by January performance and then divided the data into four sub-groups, each with 15 observations. The first dataset includes the best January performances, the second dataset the next best, and so on, until we arrive at the fourth dataset which incorporates the worst January months.
Our goal is to correlate January performance and annual performance(February to December, a eleven month period). Of course, if a particular January is strong enough, the probability of a year being positive is higher but that doesn’t necessarily help us with our investment decisions. January may indeed be strong and say rise 6% but, if the year ends with a rise of only 4% we would end up losing money if we invested after January. We want to check if January is a bellwether for the rest of the year performance and not just for year performance.
Our dataset 1, shows an average January performance of 6.98%, with gains ranging from 4.10% to a tub thumping 13.18%. These are the strongest of all the January’s. Almost all recorded positive performances for the full year with 15 rising years and in 14 cases positive performances for the rest of the year. It means that after a strong January, in 14 out of 15 cases, the market continued to advance during the rest of the year, and on average rose an additional 16.55%. That is a really huge performance, especially if we consider that the annualised volatility for these returns is 38%.
Continuing to our dataset 2, here the average January performance is 2.93%, with a returns range from 1.73% and 4.05%. These aren’t as strong as in the first dataset but still above average. In 12 out of 15 cases, the market continued to rise, such that the average return for the rest of the year was 9.06%.
Our dataset 3 presents January returns ranging from -2.53% to 1.41% and showing an average negative return of -0.58%. In this case, there are only 6 positive Januarys and 9 negative ones. The average performance an investor would get from investing after a negative January return would be -0.64%.
In our final sub-group, dataset 4, we don’t have any positive January’s at all. The average return for the month is -5.31%, with returns ranging from -8.57% and 2.74% throughout this dataset. The average rest of year return is quite unexpectedly higher than for dataset 3 but still less than average, as the average for rest of year returns is 3.41%, with the overall average return for the year being dragged down by exceptionally poor January’s.
The above findings are intriguing. What is apparent is that a very strong January performance is almost a dead cert for a subsequently strong annual performance. Similarly, as with dataset 2, a merely strong January return still has very high predictive power for the remainder of the year. The confusing element is dataset 4 where it seems to mute the January effects supposition. However, thinking about this, it is apparent that particularly poor Januaries, conversely to expectations, should in fact also be bought as the rest of year return is positive overall implying a very strong rebound in subsequent months.
But this isn’t the end of the story. Sometimes, people also talk about the first five days of a year as also being a gauge for year performance. If the first five trading days of January are above average, then it may also be worth investing. Let’s now test this hypothesis too.
Testing for 5-Starting-Days of January Effect
From the same database, ranging from 1953 to 2012, we collected performance for the first 5 trading days of each year and similarly attempted to correlate it to the rest of year performance. This time however, “rest of the year” is from the sixth day to the end of the year instead of being from February to December. We also again divided the data into four sub-groups using the same methodology as above, ranking performance for those days that produced the highest returns to the lowest.
Dataset 5 shows 15 positive 5-starting-days with an average return of 3.11%. In 13 out of 15 cases, the rest of the year return was positive, averaging 17.03%. As in the above case for the January effect, this is pretty strong empirical evidence.
Dataset 6 also shows 15 positive 5-starting-days but now with an average return of just 0.99%. The rest of the year returns are still above average, of 9.47% but the number of positive rest of the year returns is now just 9 instead of 13.
In Datasets 7 and 8, the 5-starting-days returns further declines to -0.48% and -2.73% and the number of positive rest of the year returns and the average also decline. This clearly shows a very strong correlation between the five starting days of a given year and the rest of the year performance.
Conclusion
It is very clear to us that a combination of a positive five first days and then a modestly strong overall January is a good gauge for positive annual returns. The question is how best to use this data? We would personally look to purchase an index on any dip during the subsequent months below the flat-line for the year or indeed into negative territory. The odds of generating a profit under such a strategy seem exceptionally high.
So far this year, the S&P 500 recorded a 2.1% gain in the first five days of the month, a value that lies inside our dataset 5. We wait with interest to see how the balance of January plays out and look to position accordingly using our data.