Stats › Long run returns › Size premium
The size premium
The size premium is the claim that small companies earn higher returns than large ones, beyond what their extra risk explains. It traces back to Banz's 1981 paper, it launched an entire industry of small cap funds, and it's one of the four classic factors people build portfolios around. It's also the weakest of the four, by a distance.
This page sets out the evidence on both sides. The long run factor numbers, the recent investable record, the seven separate objections that have been raised against the claim, and the one defence that has held up. Small minus big returned 2.99% a year from 1926 to 2011 with the lowest risk adjusted score of the four classic factors. More recently the Russell 2000 lost to the S&P 500 in every calendar year from 2021 through 2025. Then over the twelve months to June 2026 it beat it by eighteen points.
That last sentence is really why the page exists. This isn't a settled question in either direction, and I've tried to write it that way rather than pick a side. There's also a note at the bottom listing the numbers I could not verify and therefore did not use, which is unusual for a page like this and I think worth your time. Everything I did use is linked back to its paper or its fund page.
The size premium is the weakest of the classic factors. Over 1926 to 2011 the small minus big factor returned 2.99% a year with a Sharpe ratio of 0.26, the lowest of market, value, size and momentum. In practice small caps have been losing: the Russell 2000 trailed the S&P 500 in every calendar year from 2021 to 2025 and by 3.88 percentage points a year over the decade to June 2026. AQR's defence is that the premium is real but obscured by junk, and that controlling for quality restores a stable premium. That is a live dispute, not a settled question, and several of the numbers most often quoted on both sides have no accessible source.
Members only
The rest of this page is for members
Below this point there are 8 sections, 1 chart, 1 table and 4 named sources, roughly 1700 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.