« Announcing the Winner of the “No Excuses” Giveaway | Main | OT: Let’s Talk About the Weather… »
85 Years of the January Effect
By JLP | February 1, 2011
I have 85 year’s worth of monthly returns for the S&P 500 (and it’s precursor, the S&P 90). Lots of people like to use January as a sort of barometer for the rest of the year. After analyzing the numbers, I can understand why. Check out the two graphics below. The first one contains all 85 years of data, arranged by January’s returns along with the total return for that year and the total return for the eleven months (February – December) that followed that particular January.
The second graphic summarizes the first one.
That last graphic is pretty interesting. The median yearly return when the year starts off with a positive January is 19.68%. For those of you not familiar with “median,” it means that half the returns were above 19.68% and half were below. The geometric yearly average was 15.43%.
Guess what…
January 2011 was up 2.37%. What will 2011 bring? Nobody knows. But, consider this graphic:
In all 85 years of data, when the month of January was positive, the year’s return was negative only 8 times. I find that interesting.
Topics: Investing, S&P 500 Index | 12 Comments »
February 1st, 2011 at 12:21 pm
I’ve heard of the affect, but never seen it analyzed. Just a few more figures that summarizes you last graphic:
Positive January, year was positive: 85%.
Positive January, year was negative: 15%.
Negative January, year was positive: 50%.
Negative January, year was negative: 50%.
I wonder if this is specific to January, or the same type of observations would hold true if you picked some other month (March, for example) as your yearly starting point.
My guess is that there isn’t anything special about January (per-se), but more of a property of the mathematics behind this. If you start off negative, the next 11 months need to recoup those losses (plus some to end the year positive).
Exactly the same as if you get a bad grade on your first exam, you need to really outperform to bring your average up on the remaining exams.
February 1st, 2011 at 12:26 pm
Although the analysis is interesting, I’m not one to put much stock in any of this. I just thought the findings were worth sharing.
February 1st, 2011 at 4:41 pm
Your statistical analysis is lacking. Using this dataset, a linear fit between the January returns and the returns for the rest of the year generates an R^2 of 0.0014.
February 1st, 2011 at 5:05 pm
Thanks, Dan.
February 1st, 2011 at 5:10 pm
Dan,
When you say that, you are overstating what JLP is pointing out. He only points out that positive Januarys tend to be followed by positive overall years, not that the magnitude of returns in January is predictive of the magnitude of returns for the year. Linear regression analysis is not appropriate for this sort of claim, but a Chi-square analysis is.
A Chi Square test of the two factors (January pos/neg and Year pos/neg) yields a test statistic of nearly 12, which has a p-value of 0.005. This, as you know, means there’s a highly significant dependency between the two outcomes.
Had the claim been that you could really benefit from the January effect by timing the market, we could probably easily demonstrate that this is not the case, and your regression would be relevant in that discussion. Someone trying to use the effect would have to sit January out, and much of the time would miss out on significant gains by doing so– which would be worse than the losses acquired by holding the market all of the time.
February 1st, 2011 at 5:38 pm
RBK) That is kinda hitting on my point. The year’s return includes January’s return — so you aren’t looking at two independent variables.
I don’t think there is a real January-Effect, but is just a math puzzle and likely doesn’t matter which month you want to pick to see the same ‘effect’.
Similarly, the ChiSquare test for IBM and the DOW Index would also show extremely high dependence. Investigating further and you see that IBM stock price controls 10% of the movement of the DOW index (even though there are 30 companies in the index).
February 1st, 2011 at 5:59 pm
No doubt, BG. It’s an interesting observation on market returns but is far from actionable. From what I understand people became aware of it by doing some backtesting, then tried to turn it into a profitable timing strategy which fell completely on its face. That pattern of backtesting, action, and failure has been repeated throughout the history of markets. Eventually you’d think we’d learn!
It’s interesting nonetheless, and I do like the fact that history informs us that the the market is more likely to stay up than fall down between now and December. 1929, 30, and 31 are scary enough to keep me from seeing anything like a guarantee though.
February 1st, 2011 at 6:16 pm
Hey, I found a new one! January is more likely to be negative on decade boundaries, as in: 1940, 1960, 1970, 1990, 2000, 2010.
RBK) I hear ya on back-testing. I have a suspicious feeling that there is some hedge-fund somewhere that is doing exactly this sort of ‘investing’.
Now, I have heard of a more plausible technique, and it involves only investing on the 1st (and maybe the 15th) of each month — and staying out of the market on all the other days.
The reasoning goes: those are the days when the masses are likely buying shares do to automatic 401(k) contributions — so to take advantage of them: you purchase stocks the day before, the masses run up the prices on their payday, and you sell to them at an inflated price. Evil I know…
February 1st, 2011 at 8:08 pm
JLP points this out, but correlation does not mean causation. There is no reason to believe that a positive January has any influence on the rest of the year. The market is up more often when an NFC team wins the Super Bowl, but we know that a Packers win will not directly impact the markets. It is a statistical anomaly. The stock market’s history is to have 65% positive years. So is it really that much more statistically relevant because the market matches January’s result 80% of the time? I don’t think it is that much more significant considering the low number of results (85).
It is interesting though. I do hope the Packers win so the market goes up.
February 2nd, 2011 at 9:11 am
BG,
I did some analysis on that particular technique and found a few things about it make it impossible to make work.
1) It was hard to tell if this really had the desired effect at all. You’d think it would, but I believe there are enough 30th/31st/2nd and biweekly 401k contributions to dilute the effect of the 1st, and also enough institutional trading during the month to overwhelm it and make it minimal.
2) Even if there was a bump that day each month, trading costs in a taxable account (and taxes from ST capital gains) probably erode much of the bump away. If you do this in a retirement account, many of them have short term trading rules that would charge you a fee or prevent this entirely (such as Vanguard).
3) If you are invested only 2 days a month you’re probably going to miss out on the vast majority of the gains in the market. In all market timing strategies I think this is the biggest fatal flaw– time out of the market carries with it enormous risk.
The whole idea is very logical and well-conceived but practical considerations make impossible to actually profit from IMO.
February 2nd, 2011 at 1:54 pm
From a recent article on CNBC: “In 2010 alone, the first trading day (of the month) accounted for 123 of the 134 points the S&P gained.”
February 3rd, 2011 at 6:14 pm
Here is an academic paper on the effect
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=663563
In this paper the return is statistically significant.