Sell In May (3): A Strategy Worth More Than 25% A Year Since 2001

Includes: EWG, ICF, QQQ, SH, SPY, TLT
by: Fred Piard

This is the 3rd article of a trilogy about the "sell in may and go away" saying. In the precedent episodes (click here to read) you can find strategies using QQQ, EWG and TLT that would have returned 22% to 33% a year on average last 10 years. Taking SPY as a benchmark it's a nice performance for very, very simple strategies. Now I want to address some objections of high interest.

The main one is about using TLT: some readers wrote me that T-Bonds may be in a bubble (so it would be dangerous in the future) and that most of the strategy performance was due to that (so it would not be representative in the past).

The reality is that in the "EWG-TLT" strategy from 4/1/2001 to 5/19/2012, the total compound return due to EWG is 639% and the return due to TLT is 177%. So the seasonal effect on the German stock market was more than 3 times stronger than the interest rates effect in the performance.

If we reverse the roles of EWG and TLT for the same period, the EWG return is -81% (!) and the TLT return is -11%. Even in a "bubble", TLT had a negative average return during its bad seasons. EWG "full time" would have returned only 47%, and TLT "full time" 159%. That shows how powerful the seasonal effects have been on the last 11 years, not only on stocks but also on bonds.

However, I understand if you are afraid of a T-Bond bubble. One way to use seasonal bias is to take it as a framework. Consider that you have 2 bags, one for January, February and May to September, and another one for march, April and October to December. Feel free to fill them with tickers and strategies that are consistent with the seasonal effects.

Here is an example without bonds related ETF:

  • January, February and May to September: long ICF and SH
  • March, April and October to December: long QQQ and EWG

The return is compounded for each trade and the positions are initiated in equal parts for each season.

The following graph gives the return in %.

The average return is well over 25%, The max drawdown about 34% and the max duration of drawdown about 1 year.

You may wonder why I have selected ICF (a REIT ETF): if you do the math you will notice that its contribution to the performance is insignificant. First reason: to attract your attention to it because it is the only big sector ETF that globally had a positive return on the "bad stock seasons" the last 11 years. Second reason: you can find on this site information to make your own REIT fund outperforming ICF. Third reason: macroeconomics suggest that REITs might be a good bet for the future (but it's not my subject here).

I admit that the lack of logical reason for seasonal effects, and as a consequence the doubt about the future are very disturbing. If you want a historical point of view you may read this article by Mateo Blumer. I have no opinion about that. I'm just ready to include it in a set of reasonable bets based on reasonable bias.

As a conclusion of this "Sell In May" trilogy, I would like to answer a reader's objection that backtesting is dangerous. Yes it is, there are a lot of pitfalls. A big one is to focus on the performance instead of the probabilities (if you follow me, I'll write an article on why I look first at a game theory indicator).

Disclosure: I am long TLT.