Size
Not really, at least not the way it was originally sold. Banz (1981) found small stocks beating large ones by about 5%/year over 1936–1975. A from-scratch, value-weighted replication of that same size factor over the full 1963–2026 record comes back with a t-statistic of 1.60 — below even the textbook 1-in-20 bar — and a post-publication return of roughly 0.4%/year with a t-stat of 0.29, indistinguishable from zero. It’s the cleanest decay story in this site’s catalog: the premium essentially vanished the year Banz published.
Banz equal-weighted his portfolios, which put outsized weight on the smallest, thinnest, often-illiquid names, and his sample also carries known January and delisting-return biases that later research uncovered. This site instead uses the value-weighted small-minus-big factor — the version an investor who isn’t chasing micro-caps could actually trade — and it is much weaker: 1.7%/year annualized over the full sample, against Banz’s reported 5%.
Splitting the record at publication makes the collapse explicit. Before Banz published — 1963 to 1981, the earliest this site’s data runs — the value-weighted premium ran 6.1%/year with a t-stat of 2.23. After 1981, once the effect was public, it fell to 0.4%/year with a t-stat of 0.29 — effectively noise.
Decade by decade, the annualized return goes 9.5% (1960s), 4.7% (1970s), then negative in the 1980s (−0.4%) and 1990s (−2.1%), a rebound in the 2000s (+7.1%), and back to negative or flat in the 2010s (−0.4%) and 2020s so far (−1.6%). Three of the last four decades have lost money outright. The 2000s rebound is the main thing keeping the debate alive.
This site’s catalog notes size as its clearest decay story: the premium essentially vanished right after Banz published, and it survives today mainly as a control variable and as an interaction — small-and-value, small-and-quality — rather than as a standalone bet. Neither the risk-based story (small firms carry more funding and liquidity risk) nor the behavioral one (small caps are under-researched and neglected) has produced a premium that shows up reliably once you weight by market value and look past the paper’s own sample.
See the full record
The Anomaly Replication Lab has the complete size-factor history — in-sample vs post-publication, the decade table, the factor-model alpha, and whether it survives a multiple-testing correction.
Open the size factor →Not necessarily — small caps can still play a role for diversification or as one leg of combined tilts like small-and-value or small-and-quality. What the data doesn't support is the standalone story that small beats large by several points a year just for being small.
Banz equal-weighted his portfolios and leaned heavily on the very smallest, often illiquid stocks, plus a January seasonal and delisting-return bias that later research identified. A value-weighted version — which is what an investor without unlimited capacity for micro-caps can actually trade — is far weaker.
There were pockets, most visibly the 2000s (+7.1%/year annualized in this replication), which is enough to keep the debate alive. But four of the seven decades since the 1960s show negative returns, and the average since 1981 rounds to roughly nothing.
Educational replication, not advice. Figures reflect this site’s value-weighted size factor construction and move as the data updates.