Market seasonality and calendar effects are the family of empirical regularities in which average returns, volatility, volume, or liquidity vary systematically with when on the calendar a trade occurs — by time of year (Sell in May), by point in a multi-year political cycle (the presidential cycle), by position within the month (turn-of-the-month), or even by day of the week and time of day. They are the oldest and most-studied class of market "anomalies": cheap to compute, intuitive, and folklore-rich. The defining tension of the entire domain is that these effects are simultaneously well-documented in historical averages and deeply suspect as exploitable edges — many are partly or wholly products of data-snooping (testing thousands of calendar rules until some look special) and most that were genuinely real have decayed once published and arbitraged. A master of this domain holds both facts at once: the patterns are real enough to be worth knowing as context, and fragile enough that almost none should ever be a standalone trade trigger.

What this section covers

This section organizes the principal recurring calendar patterns in equity markets. The unifying questions across every node are the same four: (1) what is the exact definition (calendar windows differ between studies and casual usage, and the difference matters); (2) what does the measured historical record show, separated from the folklore; (3) is there a plausible mechanism (liquidity flows, tax code, monetary-policy timing) or is it likely a statistical artifact; and (4) how robust is it out-of-sample and after costs. The honest answer to (4) is usually "weaker than the headline average suggests."

Map of the sub-topics

• Sell in May / Halloween Indicator — the seasonal split between a strong "winter" half (≈Nov–Apr) and a weak "summer" half (≈May–Oct). The single most-studied calendar effect, with a serious academic footprint (Bouman & Jacobsen) showing global persistence — yet the headline U.S. edge is heavily driven by a few summer crashes (Maberly & Pierce). Best used as a risk-posture tilt, not an all-in/all-out switch.

• Santa Claus Rally — a precisely-defined seven-trading-day window (last five sessions of December plus first two of January) with a positive historical skew, popularized by Yale Hirsch's Stock Trader's Almanac. Real on average but tiny, easily swamped by macro news, and most over-interpreted in its "forecasting" / bearish-omen use.

• Turn-of-the-Month Effect — the finding that historically all of the equity risk premium was earned in a ~four-day window straddling month-end ([-1,+3]), with the rest of the month flat. Among the better-documented and internationally-replicated anomalies (McConnell & Xu: 31 of 35 countries), with a plausible liquidity/payday/rebalancing mechanism — but materially decayed in recent U.S. large-cap data.

• Presidential / Election Cycle — the four-year term rhythm with the third (pre-election) year historically strongest, attributed largely to accommodative Fed policy ahead of elections. A real descriptive regularity on a tiny sample (~24 cycles since 1928), with a weak recent forward record. Must be kept distinct from the separate, more strongly contested party-in-power return premium.

• Tax-Loss Selling Season — the late-year (Nov–Dec) clustering of loss-harvesting sales driven by the U.S. tax code (loss offsets, the $3,000 ordinary-income cap, the 30-day wash-sale rule), and the early-January rebound in the most-sold names. This node owns the tax mechanism behind the broader January Effect; the measured edge is small, decayed, and concentrated in illiquid small-cap losers.

• Day-of-Week & Time-of-Day Effects — two phenomena that must be kept separate: the fragile, largely-faded day-of-week return anomalies (e.g. the weekend/Monday effect, now weak and even reversed in some samples) versus the extremely robust intraday volume/volatility U-shape and the migration of liquidity into the closing auction. The first is contested lore; the second is durable microstructure used operationally every day.

The core tension: folklore vs. measured edge

The single most important lesson of this domain is the data-snooping critique. Sullivan, Timmermann & White (2001, Journal of Econometrics) showed that while individual calendar rules had highly significant nominal p-values, once you account for the full universe of calendar rules that could have been tested, the apparent significance of the best-performing ones largely evaporates. With enough candidate windows, some stretch of the calendar will always look special by chance. Compounding this, McLean & Pontiff (2016) found that documented anomalies in general decay by roughly a third to a half after publication, as sophisticated capital trades against them — a pattern visible across several of these effects (turn-of-the-month, the January/turn-of-year effect, the weekend effect) where the modern edge is a shadow of the original-sample edge.

The corollary distinction a domain expert insists on: behavioral/return calendar effects are fragile and arbitrageable; structural microstructure patterns are not. The intraday U-shape and closing-auction concentration persist precisely because they are generated by the architecture of the trading session (overnight information, scheduled open/close, benchmark-driven passive flow), not by mispricing — so they are features to model, not anomalies to exploit directionally.

When it matters vs. when it doesn't

Seasonality matters as a low-weight contextual backdrop over the relevant horizon — leaning marginally more constructive or defensive when several independent calendar tilts and the macro regime agree, and using the robust time-of-day patterns to reason about execution timing. It does not matter as a forecast for any single year, month, or day: every one of these effects is a long-run average that masks enormous dispersion, and any dominant macro shock (Fed surprise, recession, crisis) overrides the calendar completely. The recurring #1 misuse across the whole section is treating a multi-decade average as a probability for this instance — acting on the slogan rather than the (heavily qualified) evidence.

Sources

• Sullivan, R., Timmermann, A. & White, H. (2001), "Dangers of Data Mining: The Case of Calendar Effects in Stock Returns," Journal of Econometrics 105(1): 249–286. https://escholarship.org/content/qt2z02z6d9/qt2z02z6d9.pdf
• McLean, R. D. & Pontiff, J. (2016), "Does Academic Research Destroy Stock Return Predictability?" Journal of Finance — post-publication anomaly decay. https://www.researchgate.net/publication/254926004_Does_Academic_Research_Destroy_Stock_Return_Predictability
• Jacobs, H. & Müller, S. (2020), "Anomalies Across the Globe: Once Public, No Longer Existent?" Journal of Financial Economics — decay concentrated in U.S. markets. https://www.sciencedirect.com/science/article/abs/pii/S0304405X19301618
• Hirsch, Y. & J., Stock Trader's Almanac — origin/curation of Best Six Months, Santa Claus Rally, January Trifecta, presidential cycle framings.
• Child nodes (this section) carry the primary per-effect sources: Bouman & Jacobsen (2002); Maberly & Pierce (2004); McConnell & Xu (2008); Lakonishok & Smidt (1988); Beyer, Jensen & Johnson; Rozeff & Kinney (1976); Cross (1973) / French (1980).

Confidence: medium. The section-level framing (data-snooping per Sullivan-Timmermann-White; post-publication decay per McLean-Pontiff and Jacobs-Müller) is well-sourced; per-effect figures and disputes are carried in the child nodes and flagged there. The overarching honest verdict — real historical averages, fragile/decayed exploitable edges — is the consensus position, not a contested one.