Studies show that the "sell in May" effect may exist in markets, but that doesn't mean it is a tradable strategy

By Michael Nairne | June 2016

Calendar anomalies have been studied by financial academics and practitioners for decades. These irregular stock return patterns include the January Effect, the Holiday Effect, the Turn of the Month Effect and the weekend Effect. Every spring, one such anomaly - Sell in May and Go Away, also known as the Halloween Effect - is a recurring topic in the financial press.

The Sell in May effect entails selling stock positions at the start of May and re-entering the market at the beginning of November. The rationale is to avoid the negligible returns of May through October while profiting from the higher returns of November through April.

The Sell in May effect, originally ascribed as the conventional wisdom of English stockbrokers, was the subject of a pioneering study in 2002. That study found evidence of a Sell in May effect in 36 of 37 countries, including Canada, and that this effect is robust over time. In fact, in the U.K. the researchers found evidence of the Sell in May effect as far back as 1694.

The study, which was subject to fierce criticism, was vindicated in a subsequent 2013 study that verified the initial findings for stock returns from 1998 to 2012 in the same 37 countries, which were found to be, on average, about 10% higher in November through April than in May through October. A recent, more comprehensive study found that average returns were higher in November through April than in May through October in 82 of 109 countries.

Monthly returns studied

At Tacita Capital Inc., we analyzed the monthly returns of the S&P/TSX composite index for two periods: May 1956 to April 1986 and May 1986 to April 2016. The average monthly return in the May-October time span was less than the average monthly return in the November-April time span in both periods - and the differences were statistically significant. From 1956 to 1986, the average monthly return in May through October was 0.14%, vs 1.62% in November through April. From 1986 to 2016, the differential weakened, but manifested itself just the same: the respective numbers were 0.21% and 1.24%.

Critics of the Sell in May effect point out that a handful of tail market events in years such 1987 and 2008 were the primary causes of the phenomena. However, when we exclude from the analysis the years with the six worst September through October periods from May 1956 to April 2016, there still is a statistically significant difference. As for possible causes of this phenomenon, researchers have pointed to vacation-taking combined with low liquidity and trading volumes through the summer, then calendar yearend-related dividend hikes and stock splits.

Yet, even if the Sell in May effect exists, that does not mean that it is a tradable strategy. First, this effect has low predictive value in any given year. In 21 of the past 60 years, the May through October period had a higher return than the subsequent November through April period. In four of five years from May 1993 to April 1998, May through October returns outperformed those of November through April. Few investors can tolerate such protracted underperformance.

Also, once transactions costs and taxes are considered, Sell in May and Go Away rapidly pales in comparison to Buy, Hold and Rebalance. IE

Michael Nairne is president of Tacita Capital Inc. of Toronto, a private family office and investment counselling firm. The company, its principals, employees and clients may own securities mentioned in this article.

© 2016 Investment Executive. All rights reserved.