The hidden calendar. Why seasonality is the most underrated edge in equity investing.

1. What the academic literature actually says

Seasonality in equity markets refers to the systematic tendency of asset returns to differ across calendar periods: months, quarters, day-of-week, intra-month, even pre-holiday. It is not a single "effect" but a family of related anomalies. The most rigorously studied are the Halloween indicator (also known as "Sell in May and go away"), the January effect, the turn-of-the-month effect, and the presidential cycle.

The benchmark paper is Bouman & Jacobsen (2002), published in the American Economic Review, the most prestigious peer-reviewed economics journal. They examined 37 markets between 1970 and 1998 and found that returns from November through April were statistically and economically significantly higher than returns from May through October in 36 of those 37 markets. The result was not a small statistical wobble : the average difference exceeded 10 percentage points per year in many cases, robust to risk adjustment, sub-period splits, and crisis years.

The follow-up by Andrade, Chhaochharia & Fuerst (2013) replicated the effect out of sample using data Bouman & Jacobsen had not seen. The pattern persisted, with monthly returns in the "winter" period averaging roughly 10 % annualized, against barely 2 % during "summer". In a financial-economics context where most anomalies decay or disappear after publication, this stability is unusual.

2. The January effect: small in headlines, real in small caps

The January effect, first formalised by Rozeff & Kinney (1976), observed that small-capitalisation US stocks generated abnormally high returns in early January, concentrated in the first five trading days. The mechanism proposed by Keim (1983) was tax-loss harvesting : US investors sell losing positions in December to crystallise capital losses, depressing prices ; the rebound in early January generates a structural return.

The effect is real but narrow : it is concentrated in the smallest decile of stocks (illiquid, high-friction names) and has materially decayed since the late 1990s in the US: likely because professional arbitrage, including ETFs and tax-aware strategies, now front-runs it. International evidence (Agrawal & Tandon, 1994) shows analogous patterns in 18 countries, with persistence depending on local tax regimes. In the absence of capital-gains tax (think Hong Kong, Singapore), the effect is weaker: direct evidence that the mechanism, not market sentiment, drives the pattern.

3. The turn-of-the-month effect : where the bulk of equity returns hide

Lakonishok & Smidt (1988) and Ariel (1987) documented that the US equity market generated virtually all of its positive returns during a four-day window straddling the turn of the month, the last trading day of month n through the third trading day of month n+1. Outside this window, the average return was statistically indistinguishable from zero.

Subsequent replications across 31 countries (McConnell & Xu, 2008) confirmed the effect is global. The proposed mechanism is institutional cash flows : pension contributions, mutual-fund inflows, dividend reinvestment plans and 401(k) auto-deposits concentrate at month-end. The price impact precedes consumption by retail. For a long-only investor, exposure during these days is not optional.

4. So why doesn't arbitrage erase it?

This is the question that should make any serious investor uncomfortable. If the effect is real and well known, why does it persist?

Three structural answers, each with academic support:

• Behavioural and seasonal mood biases: Kamstra et al. (2003) link the Halloween effect to seasonal affective disorder (SAD). Reduced daylight in autumn raises risk aversion. The pattern reverses in winter latitudes when daylight bottoms out and turns up.
• Institutional liquidity calendars: pension contributions, year-end window dressing, IPO syndication windows, and buyback authorisations cluster in specific months by regulatory and corporate habit, not by random distribution.
• Limits to arbitrage, most retail investors and many institutions cannot or will not switch in and out twice a year. Tax friction, trading costs, mandate constraints, career risk on "tracking-error" deviations all keep capital from fully eliminating the anomaly.

Zhang & Jacobsen (2018) document the Halloween effect across 68 markets with the longest sample available, going back to 1693 in the UK. The effect is present in 65 of 68 markets, has not weakened post-publication, and shows no sign of being arbitraged away. That kind of stability is exceptional in empirical finance.

5. How a serious investor uses seasonality (without fooling themselves)

Three principles, learned from cycles of practitioner overreach :

1. 01Use seasonality as a tilt, not a switch. Going to cash from May to October implies leaving substantial summer rallies on the table when they happen, and they do happen, just less reliably. A measured reduction of beta during weak months, not a binary in-out, captures most of the edge while preserving optionality.
2. 02Stratify by sector. Jacobsen & Visaltanachoti (2009) show that the Halloween effect is concentrated in production, manufacturing and consumer-discretionary sectors: and absent or reversed in defensives. Indexing the effect at the sector level multiplies its information value compared to applying it to the broad market.
3. 03Combine with regime detection. Seasonal patterns weaken or invert during crises (2008, 2020). A useful filter is to disable seasonal tilts when realised volatility breaks above its 95th percentile or when macro indicators flag recession. The seasonal "winter" was deeply negative in 2008–2009 ; mechanical Halloween rules lost money, while a regime-aware version sat in cash.

6. The bottom line

Seasonality is one of the few empirical regularities in equity markets that has survived decades of peer-reviewed scrutiny, out-of-sample replication, and publication. It is not a mystical pattern in the calendar, it reflects identifiable liquidity calendars, tax structures, and mood-driven risk aversion. Used as a tilt, refined by sector, and gated by regime, it is one of the most cost-effective edges available to a long-only investor.

It will not make you rich on its own. It will modestly improve your risk-adjusted returns over a decade: which is what compounded edges look like in real markets. The investors who ignore it pay the cost. The ones who treat it like a switch lose money. The ones who model it carefully, alongside fundamentals and macro state, quietly outperform.