The old adage Sell in May and Go Away suffers from one fatal flaw and several minor ones.
If you are anywhere near the investment world I am sure you have heard it.
“Sell in May and Go Away”
History shows that markets do not perform as well in the summer months as they do in winter. In fact, this Wall Street truism has gained so much traction that it has been picked up and analyzed by serious academic folks who say with confidence “the data backs it up”.
A recent article on CNBC.com titled “It sounds crazy, but ‘sell in may and go away’ is good advice” (link here) came across my screen recently. The article sites a study done in 2012 (link here) by Andrade, Chhaochharia, and Fuerst (ACF from here on out) which shows that the “sell in May and go away” effect has a long, enduring, and statistically significant implication; not just in the U.S. but across almost all global markets. It says that the half-year period from November to April has outperformed the May to October period by about 10% on average.
The authors were even more confident in their results given that a prior study was done by Boumen and Jacobsen (link here) in 2003 and the ACF study follow up confirms an out-of-sample continuation of the effect. Quoting from their conclusion the authors state “Sell in May and Go Away remains good investment advice”.
The whole line of reasoning of their argument is faulty as it falls victim to a classic error:
The Gambler’s Fallacy
Imagine you go to Las Vegas and you are excited to try your luck at the Roulette table. You will notice that next to the table is a board which shows you the past 20 or 30 results. The gambler believes that this board can help them make a decision at the table. As a smart gambler, you know that the addition of the two green spaces on the roulette wheel mean that your odds of hitting red or black correctly are 47.4%. On an even payout, that means the house has a 2.6% advantage. So in order to help your odds you examine the results posted on the board and wait to place your bet. After some time observing you see that red comes up 4 times in a row; so you do some quick calculations, and find that the odds of red coming up 5 times in a row is only 2.4%. You put your money on black, thinking the odds are now solidly in your favor but the wheel comes up red again and you lose.
You fell victim to the Gambler’s fallacy. You made a forecast about the future based on outcomes from the past. The next spin will always have a 47.4% chance of hitting red, a 47.4% chance of hitting black, and 5.2% chance of coming up green. It does not matter how many times black or red have come up in the past, your odds have not changed.
Sell in May and Go Away is the same thing, but in reverse. The gambler (er.. trader) looks at the board and says “May to October has statistically performed less well than November to April, so this must be a recurring thing.” It is more like a slot addict looking for the hot machine that has been paying out lately. The fact of the matter is that just because this was the case in the past, there is no credible way to believe that this will be the way it is in the future.
Gambler’s fallacy aside, the assumption that Sell in May “works” also suffers from some statistical shortcomings. For one, if you look at the S&P 500 going back 50 years, the May to October period, while lower than the November to April period, was still positive on average. In fact, in 34 years out of 50, if you sold in May, and then bought back November 1st, you would have bought in at a higher price. How can it be good advice to sell out of stocks and buy back in later if 68% of the past observations show you would have missed out on a gain?
First, one would say that the asset class you are replacing stocks with can outperform the modest average return that stocks saw of 0.644% (see chart). So you sell out of stocks, and buy government bonds which earn you a higher positive return and that is why it is still a winner. Unfortunately, that is not the case. The ACF study itself showed that a simple sell in May and buy back in October, does not beat a buy and hold S&P 500 index strategy. Over their time horizon of 1994 to 2012 the S&P 500 ETF earned 4.79% while the sell in May strategy earned only 4.64%. It turns out that in order to take advantage of this phenomenon you need to apply leverage into your portfolio.
The recommended trading strategy in the study is to underweight the market in May and then borrow to over invest for the November to April period. This has the effect of muting the returns from May to October and amplifying the returns in November through April. The introduction of borrowing has increased returns, but it has also increased the risk. What happens if next year a big correction hits the portfolio in April right when it is time to reduce your leveraged position? The losses would be magnified right at the time you are selling, and you would have no way to fully recover from it, because it was on borrowed money. This is what we call in the professional money management business as a permanent impairment of capital.
The ACF study also suggests that part of its robustness comes from the fact that the effect shows up in almost all the different countries studied. However, this should not be surprising given the increased interconnectedness of world markets. The stock market in the U.S. is not a separate isolated case study from the markets in Germany or Japan. In fact, the correlation between the S&P 500 and the All Country World Index (ACWI) excluding the U.S. has reached over 80% in recent years, . In other words, what happens in the U.S. stock market tends to impact what happens in other markets as well (we saw this in October 2008 where the financial crisis impacted markets around the world).
Finally, if you mine the monthly data going back 50 years, you can come up with better trading strategies than Sell in May. My numbers show that the only two months of the year that are negative on average are August and September. Why not sell in August and then buy back in October? It would have given you an even better return. But again, just because the last 50 years of data has come back this way, is there any real reason to believe that it will continue to behave this way in the future?
Gambler’s Fallacy Revisited
To help answer that last question I ran an experiment. Using my trusty excel spreadsheet, I labelled columns from January to December and did 50 virtual coin flips for each month. If heads win and tails lose than 25 heads and 25 tails would be a breakeven. Therefore, months where heads came up more than 25 times were winners and those with less than 25 heads were losers. As it turns out from the data, coin flipping was especially profitable in January, June, and August. It was not so good in April, July, and October. Would you place a bet based on these results? With the same Sell in May mentality, one might suggest to bet heads in June and tails in July.
To be sure, we need to recognize that financial markets have one very distinct difference from games of chance like roulette and coin flipping. In financial markets, investor expectations can affect the outcome. This is what legendary investor George Soros calls Reflexivity; meaning that the markets are impacted by the observations and expectations of it’s participants. For silly investing gimmicks like Sell in May, it is not unlike the magic of Tinkerbell. It might “work” simply because investors believe it exists.
The problem remains, however, that with any investment maxim based solely on investor belief and no other rational cause, it only “works” until it doesn’t anymore. If investors dutifully sold out in May as prescribed, and then we saw a summer like the one in 1982 (market up 19.5%) that would put a definite pause on any Sell in May talks for at least a couple years.
The Sell in May and Go Away study by ACF suffers from numerous shortfalls. The first and most important of these is the gambler’s fallacy. While that reason alone should be enough to dismiss it, other shortfalls include that it requires a use of leverage, and thereby an increase in risk, to utilize, it overstates the relevancy of numerous markets showing the effect, and finally does not state any cause as to the possibility for this phenomenon, other than that it has been there in the past, so it will continue to be there in the future.
The next time someone tells you to Sell in May and Go Away, ask them who won the Super Bowl this year. The Super Bowl indicator (where if the NFC team wins, it will be an up year in the market, and vice-versa for the AFC team) still boasts around an 80% accuracy rate. This blows away the Sell in May effect but you do not see serious academic studies trying to justify it as a workable investment strategy. That’s because it is widely acknowledged that the correlation between the winner of the Super Bowl and the performance of the S&P 500 is spurious… in other words, there is no rational cause and effect – it’s just coincidence. I do not see any difference with the Sell in May saying.
Intelligent investors invest according to a careful analysis of reasonable future expectations, not according to rhyming gimmicks based on past results.
Until Next Time….
“October: This is one of the peculiarly dangerous months to speculate in stocks. The others are July, January, September, April, November, May, March, June, December, August and February.” – Mark Twain
*for inquiries on any of the data used for this study, please comment or email at firstname.lastname@example.org. Monthly percentage returns for my numbers were based on the historical price table for the S&P 500 index using monthly adjusted close prices going back to 1966.