Fama-French Three-Factor Model

The Fama-French three-factor model explains equity excess returns using market, size, and value factor returns plus alpha and residual return.

The Fama-French three-factor model is an equity return model that adds size and value factor returns to the market excess return used in the Capital Asset Pricing Model. It is commonly used to estimate portfolio exposures, attribute historical performance, and test whether average returns are explained by market, small-minus-big, and high-minus-low factor loadings. It does not guarantee that those factors will earn positive future returns.

Key Takeaways

  • The model uses three return series: the market excess return, SMB for size, and HML for book-to-market value.
  • SMB and HML are returns on constructed long-short portfolios, not company characteristics or guaranteed risk premiums.
  • A factor loading measures historical co-movement under a specified regression; it is not the same as a portfolio weight.
  • Alpha is conditional on the factors, benchmark, sample, frequency, and costs in the analysis.
  • Factor definitions and source data can be revised, so reproducible work should record the exact dataset and access date.

Model Equation

For asset or portfolio (i) in period (t), the time-series regression is:

$$ R_{i,t}-R_{f,t}=\alpha_i+\beta_{i,M}(R_{M,t}-R_{f,t}) +\beta_{i,S}SMB_t+\beta_{i,H}HML_t+\varepsilon_{i,t} $$

where:

  • (R_{i,t}-R_{f,t}) is the asset’s excess return
  • (R_{M,t}-R_{f,t}) is the market excess return
  • (SMB_t) is the return on small-stock portfolios minus big-stock portfolios
  • (HML_t) is the return on high-book-to-market portfolios minus low-book-to-market portfolios
  • (\beta_{i,M}), (\beta_{i,S}), and (\beta_{i,H}) are estimated factor loadings
  • (\alpha_i) is the regression intercept
  • (\varepsilon_{i,t}) is the residual return

The equation explains realized excess returns over the estimation sample. A separate asset-pricing interpretation asks whether factor exposures help explain differences in average expected returns across assets.

What the Three Factors Represent

FactorConstruction conceptPositive portfolio loading generally indicatesImportant boundary
Market excess returnBroad value-weighted equity return minus the risk-free returnGreater co-movement with the equity marketMarket beta does not measure every source of risk
SMBSmall-stock portfolio returns minus big-stock portfolio returnsSmall-cap-like return behaviorNot ownership of a literal security called SMB
HMLHigh-book-to-market portfolio returns minus low-book-to-market portfolio returnsValue-like return behaviorBook-to-market is an accounting-price characteristic, not intrinsic value

A negative HML loading indicates growth-like co-movement relative to this factor definition. It does not establish that every holding is a growth stock.

How SMB and HML Are Constructed

The U.S. factors in the Fama-French Data Library use six value-weighted portfolios formed from independent size and book-to-market sorts:

PortfolioSize groupBook-to-market group
SVSmallHigh, or value
SNSmallNeutral
SGSmallLow, or growth
BVBigHigh, or value
BNBigNeutral
BGBigLow, or growth

The size factor is the average return on the three small portfolios minus the average return on the three big portfolios:

$$ SMB=\frac{SV+SN+SG}{3}-\frac{BV+BN+BG}{3} $$

The value factor is the average return on the two high-book-to-market portfolios minus the average return on the two low-book-to-market portfolios:

$$ HML=\frac{SV+BV}{2}-\frac{SG+BG}{2} $$

The official U.S. construction notes specify the eligible exchanges and data requirements, use NYSE breakpoints for the size and book-to-market groups, and form portfolios at the end of June. Book equity and market equity are aligned with lags intended to avoid using financial-statement information before it is available. These details matter when attempting to reproduce the factors.

Worked Attribution Example

Assume a hypothetical monthly regression has estimated these loadings:

  • market beta: (1.10)
  • SMB loading: (0.40)
  • HML loading: (-0.20)
  • monthly alpha: (0.10%)

During one month, assume the market excess return is (1.00%), SMB is (-0.50%), and HML is (0.80%). The model-fitted excess return is:

$$ 0.10\%+(1.10)(1.00\%)+(0.40)(-0.50\%)+(-0.20)(0.80\%)=0.84\% $$

If the portfolio’s actual excess return was (1.40%), the monthly residual is:

$$ \varepsilon_t=1.40\%-0.84\%=0.56\% $$

The negative HML contribution does not mean the value factor was harmful in general. It means this portfolio had a negative HML loading during a month when HML was positive. Likewise, the 0.56% residual is not automatically skill or persistent alpha.

Interpreting the Coefficients

Market Loading

A market loading above one indicates that the portfolio historically moved more than one-for-one with the market excess-return series, holding the other included factors constant. It does not cap gains or losses and can change through time.

SMB Loading

A positive SMB loading indicates small-cap-like co-movement. The loading can arise from explicit small-company holdings or from correlated portfolio characteristics. It should be compared with actual market-cap exposure.

HML Loading

A positive HML loading indicates value-like co-movement under the model’s book-to-market construction. The loading depends on the sample and is not interchangeable with a price-to-earnings screen or a fundamental estimate of undervaluation.

Alpha

Alpha is the intercept after controlling for the included factors. Its interpretation requires a standard error, economic significance, model diagnostics, and an assessment of fees and trading costs. Omitting a relevant factor can shift return into alpha.

Fama-French vs. CAPM

FeatureCAPMFama-French three-factor model
Common return driversMarket excess returnMarket, SMB, and HML
Central exposureMarket betaMarket, size, and value loadings
Typical useExpected-return benchmark and market-risk modelEquity attribution and multi-factor asset-pricing tests
Main simplificationOne priced market factorThree empirically motivated stock factors
Important limitationCan leave size and value patterns unexplainedCan leave momentum and other return patterns unexplained

The three-factor model extends the Capital Asset Pricing Model, but better in-sample explanatory power does not prove that the added factors are causal, correctly priced, or persistent.

Practical Uses

Performance Attribution

An analyst can estimate whether a manager’s returns resemble broad market, small-cap, or value exposure. This helps distinguish factor-driven performance from residual performance, but the result remains conditional on the model.

Portfolio Comparison

Two funds with similar broad-market benchmarks can have materially different SMB and HML loadings. Comparing loadings with holdings can reveal whether style labels match actual behavior.

Model Testing

Researchers use portfolio returns and factor regressions to test whether intercepts are statistically distinguishable from zero. Test results depend on the sample, test assets, standard errors, data revisions, and model specification.

Risk Review

Loadings can support scenario analysis, but a return-regression model is not a complete risk system. Liquidity, leverage, concentration, options, nonlinear payoffs, credit exposure, and tail risk may require separate measures.

How to Estimate the Model

  1. Select the asset or portfolio total-return series, currency, and frequency.
  2. Download matching market, SMB, HML, and risk-free series from a documented source.
  3. Confirm whether returns are percentages or decimals and whether they are simple or continuously compounded.
  4. Align dates without silently filling unavailable observations.
  5. Subtract the risk-free return from the asset return once, not twice.
  6. Run the time-series regression analysis.
  7. Review coefficient uncertainty, residual autocorrelation, heteroskedasticity, influential observations, and stability across subperiods.
  8. Record the source file, download date, factor definitions, code version, and any data transformations.

Risks and Limitations

  • Historical estimation: loadings and average factor returns can change across samples and regimes.
  • Factor interpretation: return co-movement does not prove one unique economic cause.
  • Model omission: momentum, profitability, investment, term, credit, and other drivers are outside the basic stock model.
  • Linear specification: nonlinear exposures and options may not be captured well by constant betas.
  • Data revision: the underlying research series can change when source data or construction methods change.
  • Implementation gap: long-short research factors are not identical to long-only investable products.
  • Multiple testing: selecting a model after comparing many alternatives can overstate significance.

Common Mistakes

  • Defining SMB as small-company return rather than small minus big portfolio return.
  • Defining HML as a high book value portfolio rather than high minus low book-to-market return.
  • Treating a loading as a portfolio weight.
  • Mixing daily asset returns with monthly factor returns.
  • Using total asset return on the left side while also including a market excess-return factor without handling the risk-free term consistently.
  • Annualizing monthly alpha by multiplying without considering compounding or uncertainty.
  • Calling a statistically insignificant intercept managerial skill.
  • Assuming the model’s historical fit guarantees future factor premiums.

Authoritative References

  • Factor Models: The broader family of return and risk decomposition models.
  • Factor Investing: Portfolio implementation designed to target selected characteristics or exposures.
  • Beta: Estimated sensitivity to a market or other factor.
  • Book-to-Market Ratio: The accounting-to-market characteristic used in the HML construction.
  • Arbitrage Pricing Theory: A broader multi-factor asset-pricing framework that does not prescribe these exact factors.

FAQs

Does a positive HML loading mean a portfolio will outperform?

No. It indicates value-like historical co-movement under this model. HML can be negative, the loading can change, and other factor, residual, fee, and cost effects can dominate performance.

Is the Fama-French model a trading strategy?

No. It is a return and asset-pricing model. A strategy that targets SMB or HML exposure requires separate investable rules, constraints, trading, and risk controls.

Does the three-factor model include momentum?

No. The standard three-factor specification includes the market excess return, SMB, and HML. Momentum is a separate factor in other models and datasets.

This article provides general financial education. It does not recommend a security, factor allocation, fund, manager, or trading strategy, and historical model estimates do not guarantee future returns.

Browse Investing