Skip to main content

Sell in May / Halloween Indicator

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,305 words

"Sell in May and go away" is the best-known calendar anomaly in equities: the observation that stock returns tend to be markedly higher during the "winter" half of the year (roughly November through April) than the "summer" half (May through October). The full adage — "Sell in May and go away, come back on St. Leger's Day" — is old London market folklore, but it was formalized by academics Sven Bouman and Ben Jacobsen, who relabelled the buy-side mirror of it the Halloween indicator (buy at the end of October). The core tension is genuine and unresolved: the seasonal return gap is one of the most statistically persistent and geographically widespread anomalies ever documented, yet no risk-based or behavioral explanation convincingly accounts for it, and skeptics argue it is partly an artifact of a few outlier crashes.

How it's formed (the seasonal split)

The indicator is a simple binary calendar rule, not a price-derived signal:

  • Winter / "Best Six Months": hold equities from the close of October 31 through April 30 (6 months).
  • Summer / "Worst Six Months": exit equities (move to cash or bonds) from May 1 through October 31.

The "Halloween indicator" framing emphasizes the buy date (end of October); "Sell in May" emphasizes the sell date. They describe the same six-month split. There is no formula, threshold, or parameter to optimize — the only inputs are the two calendar dates, which is part of why proponents argue it is hard to dismiss as curve-fitting.

A widely used enhancement comes from the Stock Trader's Almanac (Yale and Jeffrey Hirsch), who independently identified the "Best Six Months" in 1986. Sy Harding later overlaid Gerald Appel's MACD in 1999 so that the switch is not made on a fixed date but on a MACD signal occurring near the seasonal boundary (tracking for a confirming MACD sell in April or later, and a confirming buy in October or later). This MACD-timed version is the form most often traded in practice, since it avoids whipsaw on the exact calendar day.

How it's used in practice

The classic implementation is a seasonal switching strategy: be long a broad equity index (DJIA, S&P 500) in the Best Six Months and rotate to short-term bonds, T-bills, or cash in the Worst Six Months. Per the Stock Trader's Almanac, this captures most of the market's long-run gains while spending half the year in lower-risk assets — the stated appeal is comparable returns to buy-and-hold with materially lower drawdown and volatility, not necessarily higher raw returns.

In real money management, very few professionals trade the rule mechanically as a pure all-in / all-out switch. It is more commonly used as a risk-posture tilt: leaning more aggressive in the favorable window and trimming exposure, tightening risk, or rotating toward defensive sectors in the unfavorable window. Practitioners also fold it into a confluence view — seasonality as one weak input alongside trend, breadth, and macro, never a standalone trigger.

Adoption, debate & evidence

This is one of the few "folklore" effects with a serious academic footprint. Bouman and Jacobsen (2002, American Economic Review) found that winter (Nov–Apr) returns exceeded summer (May–Oct) returns in 36 of 37 countries studied, with the gap statistically significant in many and especially strong in Europe; they could not explain it away with risk, the January effect, data errors, or shifting interest rates, and called it "another puzzle." A follow-up by Jacobsen and co-author Cherry Zhang ("Everywhere and all the time," published 2021 in the Journal of International Money and Finance) extended this to a very large set of markets spanning up to ~300+ years of data — the UK series reaching back to 1694 — and reported that average six-month returns were roughly 4% higher in the winter half, with summer excess returns often near zero or negative. They argue the breadth and longevity make pure data-mining implausible.

The skeptical case is equally credible and must be stated. Maberly and Pierce (2004, Econ Journal Watch) re-examined the U.S. S&P 500 and found Bouman–Jacobsen's significance was driven heavily by two outliers — the October 1987 crash and the August 1998 LTCM collapse (both summer-half events). With those dummied out, the U.S. effect was no longer statistically robust. This is the central, unresolved tension: globally the effect looks strong, but for the headline U.S. index much of the historical edge clusters in a handful of summer crashes, so the exploitable, after-cost, U.S.-only edge is far weaker than the raw averages suggest. Commonly cited figures (e.g. S&P 500 averaging roughly ~7% Nov–Apr vs. ~2% May–Oct since mid-century, or the often-quoted Dow figures of more than ~7% Nov–Apr vs. less than ~1% May–Oct over a ~50-year window, per StockCharts/Almanac framing) are real but sensitive to sample window, index, and whether dividends and dividend-bearing cash legs are included. Treat any single precise number as window-dependent.

Honest summary: the statistical phenomenon is well-documented and persistent; whether it is a tradable edge net of costs, taxes, and the opportunity cost of being out of a rising market is genuinely contested. Note also that being in T-bills for six months historically earned positive carry, so part of the switching strategy's risk-adjusted appeal comes from interest income, not stock-picking skill — an effect that shrinks in zero-rate regimes.

Strengths & limitations

When it helps: as a risk-management lens — the summer half has historically hosted a disproportionate share of crashes and corrections, so a defensive tilt has real downside-control logic. It is simple, transparent, and not over-fitted.

When it fails: plenty of individual summers (e.g. strong May–October rallies in many bull years) crush the rule, and being out of equities in a trending bull market is a large opportunity cost. The effect is weaker and more outlier-dependent in the U.S. than the global averages imply.

The #1 misuse: treating it as a market-timing certainty — selling everything on May 1 regardless of trend, valuation, or macro regime. It is a seasonal probability tilt averaged over decades, not a forecast for any given year. Acting on the slogan literally, in a single year, is closer to superstition than to the evidence.

Sources

Confidence: medium. Dispute flagged: global persistence is strong, but the exploitable U.S.-only, after-cost edge is contested and outlier-sensitive; widely quoted percentage figures vary with sample window and index.