Topic survey
Size
Small stocks once beat large — the premium that launched factor investing, then largely vanished, and now survives mainly as a conditioning variable.
TL;DR
Small-capitalization stocks historically out-earned large ones — the size effect (SMB) helped launch factor investing. But the standalone premium largely disappeared after its 1981 publication, is concentrated in illiquid microcaps and in January, and today survives mostly as a conditioning variable: it reappears once you control for quality/junk, and it amplifies other anomalies among small names.
A 40-year arc
Banz (1981) — the founding result: small firms earn higher risk-adjusted returns than the CAPM predicts.
Fama & French (1992, 1993) — size becomes a canonical factor (SMB) in the 3-factor model.
Post-publication decay — the standalone size premium weakened sharply after 1981; many argued the "size effect is dead," driven by microcaps, January, and survivorship/data issues.
Asness, Frazzini, Israel, Moskowitz & Pedersen (2018) — Size Matters, If You Control Your Junk: size revives strongly once you neutralize quality, because small firms are disproportionately junk.
Sub-threads
Standalone small-minus-big · the January/turn-of-year concentration · microcap vs all-cap construction · size as a conditioning variable (anomalies are larger among small stocks) · size-within-quality.
Why it works (or did)
Risk-based — small firms carry distress, illiquidity, and information risk.
Microstructure / data — much of the raw premium is a microcap, illiquidity, and January artifact.
Quality confound — small firms skew toward low-quality "junk"; controlling for quality restores a clean size premium (AFIMP).
The dark side
Mostly gone standalone — realized SMB has been near zero for decades in large/all-cap universes.
Illiquidity & capacity — the premium lives in tiny, costly-to-trade names.
Fragility — sensitive to universe (microcap inclusion), weighting, and the January effect.
Does it survive out of sample?
The standalone size premium is the most contested of the classic factors — largely vanished out of sample on its own, but it returns conditionally (within-quality) and as a risk amplifier. Our replications recompute SMB from Ken-French building blocks and score it on the holdout.
Run it yourself
Replications — Fama-French SMB.
Curriculum — the synthetic market ships a size_cap characteristic; the cost-of-trading mission shows why microcap-concentrated premia are hard to harvest net of costs.
Playground / Competitions — build a size-conditioned signal and score it OOS.
