Track 2 · Lesson 9 · 5 min
Calendar effects
Sell in May, the turn of the month and the Santa Claus rally: what the research found, and how to judge a pattern that might be luck.
Sell in May and go away
The old saying has real research behind it. In 2002, Sven Bouman and Ben Jacobsen tested it in 37 stock markets and found returns from November to April beat May to October in 36 of them. They called it a puzzle, because no obvious risk explains it. Jacobsen later extended the test to more than a hundred markets and centuries of U.K. data and found the gap persisted.
The Stock Trader's Almanac, founded by Yale Hirsch in 1967, calls the same window the "Best Six Months" and has tracked it since 1986. The Almanac and newsletter writer Sy Harding each refined it with MACD signals, so the switch dates move a few weeks depending on momentum.
The turn of the month
Robert Ariel found in 1987 that U.S. stock gains were concentrated in the first half of each month, starting on the last trading day of the previous month. A year later, Josef Lakonishok and Seymour Smidt looked at 90 years of the Dow and narrowed it to a four-day window: the last trading day of a month and the first three of the next. One common explanation is money flowing in at month-end from paycheques and pension contributions.
Be The Puck's turn-of-the-month play holds SPY for just those four days a month, about 19% of the time, and sits in T-bills otherwise.
The Santa Claus rally
Yale Hirsch defined it in 1972: the last five trading days of December and the first two of January. Seven days a year. It's a fun one to watch, but with so few days in the market it can't compound much.
The calendar plays on the S&P 500
On SPY since 2005, after trading costs. "Time in" is the share of days in the market.
| Play | Checks | Switches / yr | Time in | Annual return | Worst drawdown | Calls right |
|---|---|---|---|---|---|---|
| Halloween IndicatorSven Bouman & Ben Jacobsen & Yale Hirsch & Jeffrey Hirsch | Daily | 2.0 | 49% | +7.1% | −35.6% | 44% |
| Best Six + MACDYale Hirsch & Jeffrey Hirsch | Daily | 2.0 | 50% | +6.9% | −39.1% | 50% |
| Seasonal TimingSy Harding | Daily | 2.0 | 51% | +7.1% | −33.7% | 53% |
| Santa Claus RallyYale Hirsch & Jeffrey Hirsch | Daily | 2.0 | 3% | +2.2% | −6.7% | 41% |
| Turn of the MonthJosef Lakonishok & Seymour Smidt & Robert A. Ariel | Daily | 24.1 | 19% | +2.8% | −18.0% | 46% |
| Presidential Sweet SpotYale Hirsch & Jeffrey Hirsch | Daily | 0.5 | 17% | +4.3% | −19.5% | 55% |
| Pre-HolidayRobert A. Ariel & Josef Lakonishok & Seymour Smidt | Daily | 18.2 | 4% | +2.1% | −7.5% | 47% |
| January BarometerYale Hirsch & Jeffrey Hirsch | Month-end | 0.9 | 58% | +7.1% | −30.0% | 25% |
| Weekend EffectKenneth French & Josef Lakonishok & Seymour Smidt | Daily | 104.5 | 79% | +4.5% | −46.8% | 52% |
| Pre-FOMC DriftDavid Lucca & Emanuel Moench | Daily | 15.9 | 3% | +2.8% | −4.3% | 43% |
| Lunar CycleKathy Yuan, Lu Zheng & Qiaoqiao Zhu | Daily | 24.7 | 49% | +5.3% | −38.0% | 50% |
| Buy & holdbenchmark | Never | 0 | 100% | +11.0% | −55.2% | – |
Luck or law?
Calendar effects are the easiest patterns to find by accident. There are thousands of possible windows (the third Tuesday, the week before a holiday, odd years after an election), and if you test enough of them some will look brilliant by chance. That's data mining.
Three questions help. Was the pattern found before the test period, or after? Does it show up across many markets, not just one? And has it held since it was published, when people could trade on it? The Halloween effect does reasonably well on all three; many newer calendar "rules" don't.
Check yourself