Calendar Anomalies

Some of the most persistent patterns in equity returns are tied to calendar dates rather than fundamentals or momentum. Earnings drift, FOMC drift, the January effect, and weekend effects have all been documented in academic research, sometimes for decades. This article covers what's real, what's decayed, and how to use calendar effects honestly in QuanterLab.

Post-Earnings Announcement Drift (PEAD)

Stocks that beat earnings expectations continue to outperform for several weeks; stocks that miss continue to underperform. Bernard & Thomas (1989) documented this on US data; it has been replicated on every developed equity market and across decades. The drift typically lasts 30–60 trading days and earns 1–3% per quarter for the long-short version.

The mechanism: market participants underreact to earnings information, with the slow incorporation of news producing the drift. Even after decades of academic attention, PEAD has only partially decayed — it's smaller now than in the 1990s but still detectable.

FOMC Drift / Pre-FOMC Announcement Drift

Lucca & Moench (2015) documented that US equity markets earn most of their cumulative excess returns in the 24 hours before scheduled FOMC announcements. The "pre-FOMC drift" averaged about 33 basis points per FOMC day in their sample (1994–2011), nearly the equity risk premium for the year arriving in 8 trading days.

This is one of the most striking calendar effects in modern data. The mechanism is debated — possibly information leakage, possibly hedging flows, possibly compensation for some unobserved risk — but the pattern is robust enough that many systematic strategies treat FOMC days as scheduled positive-drift events.

Day-of-Week Effects

Historically: Monday returns were lower on average; Friday returns were higher. The effect has weakened substantially since the 1990s and may be largely arbitraged away in liquid US large-caps. Still detectable in some international markets and small-caps.

The "weekend effect" (lower Monday returns) was attributed to weekend information flow concentrating bad news in Monday opens. The strength of the effect declined sharply after 1990.

The January Effect

Small-cap stocks have historically outperformed in January, particularly the first 5–10 trading days. The effect was strongest from the 1940s through the 1980s and has weakened substantially since. Tax-loss selling at year-end (creating December weakness) and rebalancing flows in January are the standard explanations.

Lakonishok & Smidt (1988) provides 90-year evidence of the effect; Thaler (1987) is the canonical economic-perspective discussion. Today the January effect is small in US large-caps; still detectable in micro-caps and some international markets.

Turn-of-Month Effect

Equity returns have historically been concentrated in the last few trading days of one month and the first few of the next. The "turn of the month" (roughly day -1 to +3) accounts for a disproportionate fraction of total monthly returns. Mechanisms involve pension and 401(k) inflows, mutual fund window dressing, and end-of-month rebalancing.

This effect has been more persistent than the January effect, partly because the underlying flows continue to occur regularly.

How to Use Calendar Effects Honestly

1. Verify in your sample. Don't rely on academic papers — re-test the effect in the period you actually care about. Effects that worked in 1980–2010 may have decayed.
2. Use as filters, not signals. "Don't enter a mean-reversion trade in the 3 days before FOMC" (filter) is more robust than "buy SPX at 3 PM the day before FOMC" (signal).
3. Combine with strategy-level edge. Pure calendar strategies are heavily competed and have low capacity. Calendar effects are most useful as conditioning on top of strategies with their own alpha.
4. Adjust for transaction costs. Many calendar effects are 5–30 bps in magnitude — comparable to or smaller than realistic round-trip costs. After-cost performance often differs sharply from gross.
5. Walk-forward validate. If a calendar-conditioned strategy has WF Sharpe much higher than unconditioned, the conditioning has real edge. If WF is only marginally better, the calendar adjustment is mostly fitting noise.

What's Decayed and What's Not

The Bottom Line

Calendar effects are real but smaller than they were, and the temptation to data-mine across hundreds of date-based filters is enormous. Stick to a small number of well-documented, mechanism-grounded effects (PEAD, FOMC drift, turn-of-month), use them as conditioning rather than as standalone signals, and walk-forward to confirm the value-add hasn't decayed in your sample.