What are size, value and momentum, and who found them?
Three sorts of shares earned more than beta said they should, in the long US record, and each has a paper attached to it that every practitioner cites and few have read. Here are the three facts, the three papers, and the caveat that travels with each.
Size: small beat big
Banz reported it in 1981; Fama and French put it in the 1992 cross-section and made it a factor, small minus big, in their 1993 three-factor model. Companies in the smallest capitalisation groups earned more per year than the largest, over 1963 onward, by more than their higher betas explained.
The caveat is that the premium is concentrated. Much of it sits in the very smallest companies — the ones an index fund cannot hold in size — and much of it arrived in January. Measured on the funds a portfolio can actually own: from 2 September 2016 to 3 September 2026 the Russell 2000 fund IWM.US returned about 169% against about 316% for SPY.US. Ten years in which small did not beat big at all. The premium is a long-run average, and long-run averages contain decades that run the other way.
Value: cheap beat expensive
Sort shares by book value per share divided by price. The cheap group — high book-to-market — earned more than the expensive group, and Fama and French's high-minus-low factor became the second leg of the 1993 model. Cheap relative to what you get is the oldest idea in investing; the paper's contribution was showing it worked as a rule, on the average, across thousands of names, and that beta did not explain it.
The caveat is the same shape as size, larger. From 3 January 2017 to 31 August 2020 the large-cap value fund IWD.US returned about 18% while its growth twin IWF.US returned about 124% — three and a half years in which cheap was punished on the same exchange, in the same index family. The lesson on drawdowns is that window in full.
Momentum: what went up kept going
Jegadeesh and Titman, 1993: buy the shares that rose most over the past three to twelve months, sell the ones that fell most, hold for three to twelve months, and the winners kept beating the losers by roughly a percent a month in their 1965–1989 sample. Carhart added it as a fourth factor in 1997. It is the anomaly hardest to explain as a risk and easiest to explain as behaviour, and it has the highest turnover of the three, which the implementing unit will charge for.
The caveat is the crash. Momentum's losers are the shares that fell hardest, and when a bear market turns they rebound hardest; a long-winners short-losers portfolio is short the rebound. Spring 2009 was the textbook case, and it has its own lesson too.
What the three have in common
Each is a rule, not a view. Each earned its premium on the average over decades and lost it for years at a time. And each is measured as a long-short spread — which is why a fund that only buys the top group, which is what you can actually own, captures part of the story and not all of it. That gap is the next unit.
In the data
The sorts are yours to reproduce from ordinary published figures: market capitalisation for size, price against book value for value, the price history for momentum. The value ratio for two companies at opposite ends of the sort, Verizon and NVIDIA:
Read the price-to-book lines. Book-to-market, the ratio in the papers, is the same number turned upside down, so the lower price-to-book is the cheaper share. On 2026-09-04 the two were about eleven times apart.
Try it now
- The size sort you can hold, over five years, both funds on the same footing:
Measure both from the same start date and write the gap down. Then say what the size premium being "about 3% a year on the long-run average" would predict for this window, and what happened instead. 2. Using the price-to-book of Verizon and NVIDIA above, write which one a value sort buys and which it sells — before reading the next lesson's argument about whether that is a good idea.