By David Larrabee, CFA
The Halloween indicator, popularized by the well-worn adage, "Sell in May and go away," holds that stocks offer their best returns during the November through April period. But is such a seasonal trading strategy really valid and thus an opportunity to be exploited, or is the Halloween indicator more trick than treat? Recent research seems to confirm what earlier studies found, which is that, remarkably, this stock market anomaly dating back to at least the 1930s has not been arbitraged away, and in fact, is stronger than ever.
The practice of abandoning stocks beginning in May of each year is widely thought to have its origins in the United Kingdom. The privileged class would leave London and head to their country estates for the summer months, where they would largely ignore their investment portfolios. To this day, many stock market watchers have postulated that the corresponding impact of summer vacations on market liquidity and investors' risk aversion is at least partly responsible for the difference in seasonal returns.
In what is considered to be a seminal piece of research on the subject, "The Halloween Indicator, 'Sell in May and Go Away': Another Puzzle," authors Sven Bouman and Ben Jacobsen were among the first to document a strong seasonal effect in global stock markets. In 36 of the 37 developed and emerging markets they studied between 1973 and 1998, the authors found returns in the November through April period to be, on average, significantly higher than those in the May through October period, even after taking transaction costs into account. What puzzled the authors was the fact that, while the anomaly was widely known and seemed to offer considerable economic rewards, it had not been arbitraged away.
More recently, Jacobsen partnered with Cherry Zhang on a follow up study, titled, "The Halloween Indicator: Everywhere and All the Time," and extended the research to 108 stock markets using all historical data available. The result was a sample of 55,425 monthly observations (including more than 300 years of UK data), which helped to rebut any criticisms of data mining and sample selection bias. The results were compelling, as the November through April "winter" period delivered returns that were, on average, 4.52% higher than the "summer" returns. The Halloween effect was evident in 81 out of 108 countries. The size of the Halloween effect varied across geographies. It was found to be stronger in developed and emerging markets than in frontier markets.
Importantly, Jacobsen and Zhang also found the effect to be increasing in strength: Over the past 50 years, the difference in returns widened to an average of 6.25%. The results also showed that a sell-in-May strategy was successful in beating the market more than 80% of the time when employed over a five-year horizon, and more than 90% successful in beating the market over a 10-year horizon.
When it comes to determining equity allocations, investors are best served by focusing on the market's underlying fundamentals. Even so, the Halloween indicator is a market anomaly that deserves investor attention because of the impressive returns it has provided, its persistence over time, and the fact that it cannot easily be explained away.