Sell in May and Go Away? What the Halloween Effect Really Shows

There is an old market saying: "Sell in May and go away." It sounds like a lazy piece of folklore. But when two researchers actually tested it across dozens of countries and decades of data, they found something that was hard to explain away: the market's gains really did pile up in the winter half of the year and nearly vanish in the summer half.

Who found it, and where (the who / what / when)

The landmark study is "The Halloween Indicator, 'Sell in May and Go Away': Another Puzzle" by Sven Bouman and Ben Jacobsen, published in the American Economic Review in 2002 — one of the most respected journals in all of economics. A follow-up by Andrade, Chhaochharia, and Fuerst in 2013, titled "Sell in May and Go Away: Still Good Advice," checked whether the pattern survived after the original paper was published. It largely did.

What they actually did (the how they studied it)

They split the calendar into two halves: the "winter" months of November through April and the "summer" months of May through October. Then they measured average stock market returns in each half, for 37 different countries, some with data going back decades. They asked a simple question: is one half of the year reliably better than the other?

What they found (the data points)

The answer was a clear yes. In the large majority of the 37 countries, returns from November to April were substantially higher than returns from May to October — and in many markets, almost all of the year's gains came from the winter half. The summer half, on average, delivered returns close to flat once you accounted for risk. The follow-up study found the effect was still there in the years after the original paper — unusual, because many market patterns fade once they are published.

The "equation" is just two windows

There is no math to fear here. The whole rule is a switch between two dates:

Be long from the close of October 31 through the end of April. Be flat (or in cash) from May through October.

That is it. "Halloween indicator" is just a nickname because you get in around Halloween. Some versions stay in a safer asset during summer instead of pure cash, but the core signal is the calendar date — nothing else.

Why it should work (the why)

Honest answer: nobody is completely sure, and that is part of what makes it a "puzzle." The leading ideas are behavioral and structural. Summer brings vacations and thinner trading — big institutions take time off, liquidity drops, and risk-taking cools. There may also be a mood and risk-appetite cycle tied to the seasons. And year-end effects — bonuses, new-year optimism, fresh fund allocations — cluster in the winter window. None of these is proven, so treat the "why" as a reasonable story, not a law.

Does it still hold — honestly?

Respect three caveats. First, it is a stock-index effect — it is strongest on equity index futures (like the E-mini S&P 500 or Nasdaq), and on forex it is weak and must be tested pair by pair, if at all. Second, it is a long-horizon, low-sample pattern: you only get one winter and one summer per year, so it takes many years of data to trust, and any single year can look nothing like the average. Third, well-known calendar edges can shrink as more people trade them. The upside: the rule has almost no parameters, so it is very hard to curve-fit — which makes it one of the cleanest seasonal ideas a beginner can test properly.

Build and test it in TapeScript, step by step

1. Create it in plain English. Type: "Build a seasonal strategy on daily ES: go long at the close of October 31 and exit at the close of April 30 each year; stay flat from May through October."
2. Classify and baseline. Type: "Classify this and run the baseline backtest over all available years." You will see an equity curve built only from the winter windows.
3. See the seasonal edge. Type: "Open the Time and Context view and show me average return by month — is winter really beating summer?" This is the visual heart of the study.
4. Prove it did not decay. Type: "Run a walk-forward test decade by decade — is the winter edge still there in the most recent years?" A seasonal edge must survive recent data, not just old data.
5. Compare markets. Type: "Clone this onto Nasdaq futures, gold, and EURUSD and compare — where is the November-to-April edge strongest?" This directly answers the futures-versus-forex question.
6. Final exam. Type: "Run Monte-Carlo and spend the untouched holdout." With few trades, be honest about how much confidence the sample really supports.

The bottom line

"Sell in May" is not just a rhyme — it is a puzzle documented in one of the world's top economics journals across 37 countries. But a once-a-year, few-sample pattern is exactly the kind of edge that fools people who do not test it carefully across decades and markets. TapeScript lets you measure the winter-versus-summer split with honest, no-lookahead numbers and a real holdout. Test the Halloween effect for yourself →

Citation: Bouman, S., & Jacobsen, B. (2002). "The Halloween Indicator, 'Sell in May and Go Away': Another Puzzle." American Economic Review, 92(5), 1618–1635. See also Andrade, Chhaochharia & Fuerst (2013), "Sell in May and Go Away: Still Good Advice." Free versions on Google Scholar.