An investor who bought cheap US stocks and sold expensive ones at the end of 2006 lost 57.8% of that position by September 2020. The same long-short portfolio had returned an average of 5.9% a year from 1927 to 2006. Both numbers come from the value factor that Eugene Fama and Kenneth French built in the early 1990s.
Factor investing is the practice of choosing securities by measurable characteristics, such as cheapness, recent price trend, profitability or volatility, that research has linked to differences in average returns. A factor portfolio buys the stocks that score high on a characteristic and, in the academic version, sells short the stocks that score low. The return of that long-short portfolio is the factor's premium.
From CAPM to the Fama-French three-factor model
The capital asset pricing model, published by William Sharpe in 1964, has one factor. A stock's expected return depends only on its beta, its sensitivity to the whole market. Risk specific to one company can be diversified away, so the model pays no premium for bearing it. Sharpe shared the 1990 Nobel prize in economics for pioneering work in financial economics.
Stephen Ross proposed in 1976, with arbitrage pricing theory, that several sources of risk could be priced at once. Rolf Banz found in 1981 that smaller firms had earned higher risk-adjusted returns than larger ones for at least forty years. Then Fama and French showed in 1992 that size and book-to-market equity captured the cross-section of average US stock returns, while the relation between beta and average return was flat once size was accounted for.
The three-factor model of 1993 explains a portfolio's return with three factors:
- Market (Mkt-RF): the return of all US stocks minus the one-month Treasury bill rate, the only factor in CAPM.
- Size (SMB, small minus big): small stocks minus large stocks.
- Value (HML, high minus low): stocks with a high book-to-market ratio minus stocks with a low one, which in practice means cheap stocks minus expensive growth stocks.
In a 2004 review, Fama and French judged CAPM's empirical record poor enough to invalidate the way it is used in applications. Fama received a share of the 2013 Nobel prize for his empirical work on asset prices.
Momentum and the Carhart four-factor model
In 1993 Narasimhan Jegadeesh and Sheridan Titman showed that buying recent winners and selling recent losers produced significant positive returns over 3 to 12 month holding periods, which systematic risk did not explain. Part of the gain reversed in the following two years.
Mark Carhart added momentum as a fourth factor in a 1997 study of mutual funds. His four-factor model, market, size, value and momentum, together with fund expenses, almost completely explained why some funds kept outperforming. The "hot hands" of winning funds came mostly from happening to hold last year's winning stocks.
The momentum factor in the Kenneth French Data Library ranks stocks on their return from 12 months ago to 2 months ago, skipping the most recent month. From 1927 to 2025 it averaged 8.6% a year and lost money in 21 of 99 years. One of those years was 2009, when it lost 83.5%.
From March to May 2009, as the market rebounded 26%, the decile of past losers, stocks that had fallen furthest in the crisis, rose 163% while past winners gained 8%. Kent Daniel and Tobias Moskowitz found the same pattern in the summer of 1932 and called these episodes momentum crashes: they arrive after market declines, when volatility is high, as the market turns up.
Profitability, investment and the quality factor
In 2013 Robert Novy-Marx showed that gross profits divided by assets predicted returns about as well as book-to-market, even though profitable firms tend to trade at higher valuations. Profitable firms earned more than unprofitable ones, and controlling for profitability improved value strategies, especially among the largest stocks.
Fama and French responded in 2015 with a five-factor model that adds two factors to the original three:
- RMW, robust minus weak
- Firms with high operating profitability minus firms with low operating profitability.
- CMA, conservative minus aggressive
- Firms that grow their assets slowly minus firms that grow them fast.
With those two factors included, the value factor became redundant for describing average returns in the 1963 to 2013 sample Fama and French examined. Momentum and low volatility were left out of the model.
Fund providers call the same territory "quality", with broader definitions. Clifford Asness, Andrea Frazzini and Lasse Pedersen defined quality as profitability, growth and safety, and found that a quality-minus-junk factor earned significant risk-adjusted returns in the US and across 24 countries. High-quality stocks traded at higher prices on average, by a modest margin.
The low-volatility anomaly and betting against beta
CAPM says riskier stocks should pay more. In the data, the least volatile US stocks did better. A dollar placed in the lowest-volatility quintile in January 1968 grew to 59.55 dollars by the end of 2008. A dollar in the highest-volatility quintile was worth 58 cents, before transaction costs and before 41 years of inflation.
Malcolm Baker, Brendan Bradley and Jeffrey Wurgler traced the anomaly to benchmarks: a mandate to beat a fixed index discourages arbitrage in low-beta, high-alpha stocks and in high-beta, low-alpha stocks. Andrea Frazzini and Lasse Pedersen pointed to leverage constraints, which push investors who cannot borrow toward high-beta assets. Their betting-against-beta factor goes long leveraged low-beta assets and short high-beta ones. High beta went with low alpha in US stocks, 20 international stock markets, Treasury bonds, corporate bonds and futures.
Factor premiums in the Fama-French data, 1927 to 2025
The Kenneth French Data Library publishes the monthly and annual returns of every Fama-French factor, updated from the CRSP database. The table below averages the annual returns through 2025. Each factor is a long-short portfolio of US stocks, before trading costs, fees and taxes.
| Factor | Long minus short | Years | Average annual return | Losing years |
|---|---|---|---|---|
| Market (Mkt-RF) | All stocks minus Treasury bills | 1927 to 2025 | 8.9% | 29 of 99 |
| Size (SMB) | Small minus big | 1927 to 2025 | 2.5% | 50 of 99 |
| Value (HML) | Cheap minus expensive | 1927 to 2025 | 4.2% | 40 of 99 |
| Momentum (Mom) | Past winners minus past losers | 1927 to 2025 | 8.6% | 21 of 99 |
| Profitability (RMW) | Robust minus weak | 1964 to 2025 | 3.4% | 20 of 62 |
| Investment (CMA) | Conservative minus aggressive | 1964 to 2025 | 2.9% | 26 of 62 |
The size premium belongs almost entirely to the first half of the sample. SMB averaged 4.6% a year from 1927 to 1980 and minus 0.1% from 1981 to 2025. Banz published the size effect in 1981, the first year of the flat stretch.
Value and momentum move against each other. Their monthly returns have a correlation of minus 0.41 from 1927 to July 2026, and Asness, Moskowitz and Pedersen found the same negative correlation in eight markets and asset classes, from country stock indices to currencies and commodities. An equal mix of the two had an annualized volatility of 7.9% over that period, against 12.3% for value alone and 16.3% for momentum alone.
The value factor's 57.8% drawdown, 2007 to 2020
The value factor peaked in December 2006. Over the next 13 years and 9 months it lost 57.8%, and the first nine months of 2020 alone took 33.7%. Compounding its monthly returns, HML then gained 22.2% in 2021 and 31.7% in 2022, lost 11.1% in 2023 and 7.0% in 2024, and in July 2026 stood 29.6% below its 2006 peak.
Fama and French found that value premiums were much lower in the second half of the July 1963 to June 2019 period than in the first. Monthly premiums are so volatile, though, that the data could not reject the hypothesis that the expected premium was the same in both halves.
Rob Arnott, Campbell Harvey, Vitali Kalesnik and Juhani Linnainmaa traced the loss to two causes. Book value misses intangible assets such as software, brands and research, and a value measure that capitalizes intangibles did far better than the book-to-market ratio. The second cause is valuation: the price gap between growth and value stocks widened so much that the relative valuation of value fell from the top quartile of its history in early 2007 to the bottom percentile in June 2020. Their decomposition attributes the whole drawdown to that widening spread. The paper discloses that Research Affiliates, where Arnott is chairman and Kalesnik a partner, has a commercial interest in the subject.
The factor zoo and the replication debate
In his 2011 presidential address to the American Finance Association, John Cochrane described a zoo of new factors replacing CAPM. In 2016 Campbell Harvey, Yan Liu and Heqing Zhu counted it. They catalogued 316 factors from 313 papers, working papers included, proposed between the first tests in 1967 and 2012, and argued that with so many tested, the usual threshold for significance, a t-statistic around 2, lets through too many false discoveries. They set the bar for a new factor at a t-statistic of 3.0, and by their reckoning most claimed findings in financial economics are likely false.
Kewei Hou, Chen Xue and Lu Zhang rebuilt 452 anomalies with microcaps controlled through NYSE size breakpoints and value-weighted returns. 65% failed a t-statistic of 1.96, and 82% failed the multiple-testing hurdle of 2.78. The survivors earned less than originally reported.
Theis Jensen, Bryan Kelly and Lasse Pedersen reached the opposite conclusion with a Bayesian model. In their data most factors replicate, cluster into 13 themes, and work out of sample across 93 countries, and the large number of factors strengthens the evidence for the main themes.
What publication and trading costs take from the premiums
David McLean and Jeffrey Pontiff followed 97 published return predictors and found portfolio returns 26% lower out of sample and 58% lower after publication. They read the first figure as an upper bound on data mining and the gap between the two, about 32%, as investors trading on the published result.
Andrew Patton and Brian Weller compared the returns of factors on paper with the factor returns mutual funds actually earned. After implementation costs, typical mutual funds earned low returns to value and none to momentum.
Norway's finance ministry commissioned Andrew Ang, William Goetzmann and Stephen Schaefer to evaluate the active management of its sovereign wealth fund, whose active returns had been sharply negative in 2008 and early 2009. They found that about 70% of the fund's active returns could be explained by exposure to systematic factors, and recommended managing those exposures deliberately.
Smart beta ETFs and what they cost
The retail form of factor investing is the smart beta ETF: a long-only index fund that selects or weights stocks by a factor score. It holds only the long side of the academic factor, so its return combines the whole market's return with a partial exposure to the factor. BlackRock builds its iShares factor range around five factors: value, quality, momentum, size and minimum volatility.
Three iShares factor ETFs, with the firm's S&P 500 fund for comparison, as of September 23, 2026:
| Fund | Factor | Expense ratio | Net assets |
|---|---|---|---|
| iShares MSCI USA Quality Factor ETF (QUAL) | Quality | 0.15% | $47.8bn |
| iShares MSCI USA Min Vol Factor ETF (USMV) | Low volatility | 0.15% | $23.5bn |
| iShares MSCI USA Momentum Factor ETF (MTUM) | Momentum | 0.15% | $21.3bn |
| iShares Core S&P 500 ETF (IVV) | Market only | 0.03% | For comparison |
The expense ratio is the visible cost: 0.12 percentage points a year more than a market index fund, or 120 dollars a year on 100,000 dollars. Trading costs at each rebalance come out of the fund's return and do not appear in the expense ratio, and they are the type of implementation cost Patton and Weller estimated for mutual funds. A factor strategy can also lose for long stretches: through July 2026, the value factor had spent more than 19 years below its December 2006 peak.
Risk premium or mispricing: factors and market efficiency
A factor premium poses a question for the efficient market hypothesis. If cheap stocks or recent winners earn more, either they carry a risk that the premium pays for, or investors are making a repeated mistake that others could exploit.
Fama and French argued in 1996 that the three-factor model absorbed most known anomalies in a way consistent with rational asset pricing, with one exception: the continuation of short-term returns, which is momentum. Josef Lakonishok, Andrei Shleifer and Robert Vishny found in 1994 that value strategies earn more by exploiting the suboptimal behavior of the typical investor, and not because value stocks are fundamentally riskier. McLean and Pontiff read the post-publication decline in their 97 predictors as a sign that investors learn about mispricing from academic papers.
Our comparison of active and passive investing covers the evidence on managers who try to beat the market by other means, and our introduction to econophysics for investors covers the fat-tailed return distributions behind crashes like momentum's in 2009.
This article is general information, not investment advice.
Frequently asked questions
What are the 5 factors in factor investing?
The best-known list is the Fama-French five-factor model of 2015: market, size (small minus big), value (high minus low book-to-market), profitability (robust minus weak) and investment (conservative minus aggressive). BlackRock builds its iShares factor funds around a different five: value, quality, momentum, size and minimum volatility. Momentum and low volatility are absent from the Fama-French model.
Why has factor investing underperformed?
Mostly because of value and size. The Fama-French value factor fell 57.8% from December 2006 to September 2020, as cheap stocks lagged expensive growth stocks, and the size factor has averaged close to zero since 1981. Research published since 2016 has also found that many documented factors earn less once they are published and once trading costs are paid.
Is factor investing active or passive?
A factor ETF follows published rules and an index, like a passive fund, but it chooses stocks by characteristics and departs from market weights, which is an active decision. Factor funds usually charge more than market index funds: the large iShares factor ETFs charge 0.15% a year, against 0.03% for the iShares Core S&P 500 ETF.
Does factor investing work?
The long-run data show positive average premiums for value, momentum, profitability and investment in US stocks since at least 1963, with long losing stretches for each. How much of that reaches an investor depends on trading costs, fund fees and whether the premium survives after publication. One study of 97 predictors found returns 58% lower after publication, and another found that mutual funds earned low returns to value and none to momentum after implementation costs.
What is the difference between factor investing and smart beta?
Smart beta is the fund-product version of factor investing: long-only index funds, usually ETFs, that weight or select stocks by a factor score. Factor investing is the broader idea and also covers long-short portfolios, like the academic factors, and factors in bonds, currencies and commodities.