Seasonal Indicators – Sell in May – But When to Be Back? Does it Really Work?

There is an old saying “Sell in May and go away” which means that you could get better stock market returns if you stay away from markets over the summer. And then it goes something like “… but remember to be back in September” to describe when to enter the long position again. It’s also called “Halloween Indicator”.

The results of a little test are surprising: If you apply this strategy consequently with entering a long position on October 26 every year, and exiting the long position on May 2, then you would not have made so much more in bull markets, but the performance would have been very good in bear markets.

Take the Dow Jones (DJI) from 1982 to 2000: “Buy-and-Hold” gives +15.3% per year but only +13.5% for “Sell-in-May”. But in the bear market in the 1966 to 1982 you get -0.33% per year with “But-and-Hold”, and just +3.9% for “Sell in May”. The same is true for Germany’s DAX index. And also for Japan’s NIKKEI 225, where with “Buy-and-Hold” you would have lost -6.5% per year from 1990 till today (that’s down -74%) but would have received +2.5% for “Sell-in-May”.

The strategy works well in bear markets. Therefore, the volatility of returns is lower which means less stress for the investor.

Let’s apply the strategy to stocks. Take AMD in the last 26 years: “Sell-in-May” gives on average +28% per year while “Buy-and-Hold” performed poorly (-7.1% per year). 100 Dollar would now be 72,500 USD with “Sell-in-May” but just 14 USD for “Buy-and-Hold”. On the other hand, “Sell-in-May” performs poorly with Google (just +1.2% compared to +6.6% for “Buy-and-Hold”). But Google has been in a bull market, while AMD has been swinging sideways most of the time.

One can test forex data or oil, the pattern is always the same: “Sell-in-May” under performs in bull markets but does well in bear markets, resulting in reduced volatility of returns. This is a trend but in some markets like GOLD “Sell-in-May” does not work at all. You will find a link at the end of the article which allows to make your own experiments.

Doesn’t this violate the Efficient Market Hypothesis which postulates that there are no obvious patterns in price data? Yes, it does. Actually, there have been published more accurate and better market theories.

If you model markets based on complex system theory, the fractal patterned ups and downs are a result of local connections (networks) among traders and investors (e.g. direct conversation, watching the same TV shows, reading the same newspapers). If these local connections and habits of individual communication change over the year, this will directly impact the ups and downs of the market. And this causes seasonal price patterns.

Go to http://www.frog-numerics.com to experiment with out free interactive Java animation “Sell in May”.