Cross-Sectional Seasonality in Factor Returns

Everyone knows aggregate calendar anomalies like the January effect, where the whole market tends to behave differently in one month. Cross-sectional seasonality is a related but distinct and stronger effect at the individual-stock level: a given stock's return in, say, March of this year is unusually well predicted by that same stock's own return in March of prior years, more so than by its returns in any other month. Rank stocks by their historical same-calendar-month return, and a long-short portfolio (long the top decile, short the bottom decile) built purely on that ranking earns a statistically significant premium, even after controlling for standard risk factors.

What makes this odd is that it isn't really about the calendar per se — it survives controls for well-known seasonal effects like tax-loss selling around year-end, and it isn't explained by exposure to standard risk factors like size, value, or momentum. The leading explanations lean on investor and institutional behaviour that recurs on an annual cycle for reasons specific to individual firms — earnings announcement dates, index rebalancing dates, and other firm-specific events that themselves recur in the same month year after year, rather than any single unifying risk-based story.

Because the effect is measured across thousands of individual stock histories rather than one aggregate time series, it is also a useful test case for how weak, noisy patterns can nonetheless be statistically robust in a cross-section — a reminder that "seasonality" in factor research is not just about January.