According to the Halloween effect, investors should sell in May, getting out of the stock market on the 1st of the month, and step in again on the 1st of November. For readers who still doubt seasonal patterns in the stock market, recent academic research shows that the difference in return between summer and winter is as old as stock data: three centuries. The same researchers report that it has increased in the last decades and can be observed not only for the New York and London stock exchanges, but all over the world. More intriguing, the monthly patterns are similar in the northern and southern hemispheres, which indicates that the stock exchange seasons have nothing to do either with the climate or with a supposed legacy of agriculture cycles.
As this research suggests that the calendar might be among the best quantitative indicators on global indices, I wrote a full chapter of my book Quantitative Investing to help investors integrate the Halloween effect into their strategies and I want to share a few ideas on this here.
First, statistics show that the best seasonal pattern might have four seasons, not two. Indeed, October was among the best months of the last 15 years, despite infamous Octobers in 2000 and 2008. Moreover, January and February were the worst months in the same period. In fact, February has never been a good month on long time frames, and October is a good month for at least 50 years. The case of January is more questionable, but I don’t want to play a game when the odds are against me for 15 years.
So my first clue is to include October in the good stock season, and to exclude January and February.
The second clue is that countries have not been created equal for seasonal patterns. For example, the best market in Europe to play this game is also the biggest and one of the most volatile: Germany.
The third clue is that it is safer to play seasonal on general indices than on sectors. Indeed, the superposition of cycles of different and variable lengths makes the case of sectors more complicated.
The next chart shows the equity curve of investing in the S&P 500 index since 1/1/1999. Dividends are included. The blue line is just buying and holding the index. The red line is a four-season model, going out of the market from the 1st of May to the 30th of September, and from the 1st of January to the last day of February.
The seasonal model has a total return of 233.8% and a maximum drawdown of -33%, vs. a total return of 78% and a drawdown of -55% for the “buy-and-hold” investor. The seasonal model not only improves the return, it lowers the risk in terms of drawdown and volatility, and also reduces the exposition to five months a year. The results are even better for the most “seasonal” countries.
The fourth clue is that a seasonal strategy has idle capital during the bad seasons. It is possible to invest it in bonds, or even in shorting a stock index. Both have worked on the last 15 years. However, if you have doubts about bonds because of central bank policies, or are afraid of shorting, cash is sometimes the best place to be.
The fifth clue is about the fear that “this year may be different”. Of course, seasonals don’t work every single year. A pure seasonal strategy is clearly not safe enough to make a robust core portfolio. It is possible to add market timing rules, but my choice is combining a seasonal strategy with a momentum strategy. The result is a kind of “auto-hedging” portfolio. If the market is consistent with seasonals, both strategies are adding their gains. If it is not, they are hedging each other. That is what I call the “No Brain No Gain” model.
My book describes a very simple version of this model with super-liquid ETFs on the NYSE. A 10-year simulation shows an annualised return of 16% and a maximum drawdown of 13%. My own Core Portfolio*, which is a more elaborate and a little bit less liquid model based on a similar rationale, shows a simulated annualised return above 30%**.
Of course, past returns, real or simulated, are never a guarantee for the future. We live in a world of probabilities. But these numbers are enough to consider that seasonal patterns are worth more than a glance at academic publications. If you believe that human nature has a tendency to repeat the same behavioural patterns, studying the past is the best way to plan the future.
* Core portfolio positions are included in the weekly newsletter available on ypafi.com. This is a follow-up of an investor’s portfolio for informational purposes only. It cannot be considered investment advice under any jurisdiction. Full disclaimer on the website.
** CFTC RULE 4.41 – Hypothetical or simulated performance results have certain limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not been executed, the results may have under-or-over compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profit or losses similar to those shown.