Fama-French Three-Factor Model: Beyond CAPM

The Capital Asset Pricing Model (CAPM) revolutionized finance by linking expected returns to a single risk factor: the market. But decades of research have shown that market beta alone leaves significant return patterns unexplained. The Fama-French three-factor model addresses this gap by adding two additional factors — company size and value characteristics — that historically have explained roughly 90% of diversified portfolio return variation in sorted portfolios, compared to approximately 70% for CAPM alone.

What is the Fama-French Three-Factor Model?

The Fama-French three-factor model is a multifactor asset pricing model developed by economists Eugene Fama and Kenneth French in their landmark 1993 paper. It extends the CAPM by recognizing that two additional characteristics — firm size and book-to-market ratio — systematically predict stock returns beyond what market risk alone can explain.

Fama and French observed that small-cap stocks and stocks with high book-to-market ratios (value stocks) consistently earned returns higher than the CAPM predicted. Rather than treating these patterns as anomalies, they proposed that size and value proxy for additional sources of systematic risk that investors are compensated for bearing.

The Fama-French Model Formula

The model can be expressed in two forms, each serving a different purpose:

Where:

• β_M — sensitivity to market risk (same as CAPM beta)
• β_S — sensitivity to the size factor (positive = tilts toward small caps)
• β_V — sensitivity to the value factor (positive = tilts toward value stocks; negative = tilts toward growth)
• R_M − R_f — market risk premium
• SMB — size factor premium (historically ~1.5–3.5% annually in U.S. data since 1926)
• HML — value factor premium (historically ~3–5% annually, though it varies substantially by sample period)
• α_i — intercept (alpha); if the model fully explains returns, this should be zero

The regression form is used to estimate factor loadings (β coefficients) from historical data. The expected return form then uses those loadings, combined with expected factor premiums, to forecast required returns — much like CAPM, but with two additional dimensions of risk.

SMB and HML Factors Explained

The two additional factors in the Fama-French model are constructed as long/short portfolios designed to isolate the size and value premiums:

SMB (Small Minus Big)

SMB is the return difference between a diversified portfolio of small-cap stocks and a diversified portfolio of large-cap stocks. It captures the historical tendency of small companies to outperform large companies over long periods. The size factor relates directly to market capitalization classifications — small-cap stocks have historically carried a return premium that CAPM cannot explain through market beta alone.

HML (High Minus Low)

HML is the return difference between a portfolio of high book-to-market stocks (value stocks) and a portfolio of low book-to-market stocks (growth stocks). It captures the value premium — the historical tendency of inexpensive stocks (relative to book value) to outperform expensive ones.

Factor Construction Methodology

Fama and French construct these factors using a double-sort of all U.S. stocks. They first sort stocks into two size groups using the NYSE median market cap as the breakpoint. They then independently sort stocks into three book-to-market groups: the bottom 30% (growth), middle 40% (neutral), and top 30% (value). The intersection creates six value-weighted portfolios (Small/Growth, Small/Neutral, Small/Value, Big/Growth, Big/Neutral, Big/Value).

SMB and HML are then computed from these six portfolios. Both factors are zero-net-investment portfolios — they represent return differences between two sides of a long/short position, not returns on a single investable portfolio. Factor return data is freely available from the Kenneth French Data Library, updated monthly.

Fama-French Three-Factor Example

Contrasting Example: Large-Cap Growth Fund

Fama-French Model vs CAPM

Understanding when to use each model is essential for practitioners:

CAPM remains the standard model for estimating cost of equity in corporate valuation and capital budgeting. The Fama-French model is preferred when the goal is evaluating fund manager skill, understanding portfolio factor tilts, or explaining cross-sectional return patterns in academic and quantitative research.

How to Analyze Factor Exposures

Using the Fama-French model in practice involves estimating a portfolio’s factor loadings and interpreting what they reveal about the portfolio’s risk profile:

1. Obtain factor return data from the Kenneth French Data Library (monthly returns for the market, SMB, and HML factors)
2. Gather portfolio returns for the fund or portfolio you want to analyze (same frequency as factor data)
3. Run a multiple regression of portfolio excess returns on the three factors to estimate β_M, β_S, and β_V
4. Interpret the loadings to understand the portfolio’s risk exposures

Typical factor loadings for common portfolio types:

A fund that claims to generate alpha but has high β_S and β_V loadings may simply be harvesting known factor premiums. The Fama-French model helps investors distinguish between genuine skill and systematic factor exposure.

Risk vs Behavioral Debate

One of the most important unresolved questions in finance is why the size and value premiums exist. Two competing schools of thought offer different explanations:

The risk-based explanation (Fama and French’s view) argues that small-cap and value stocks are fundamentally riskier. Small firms are more vulnerable to business cycle deterioration and have less access to capital markets during downturns. High book-to-market firms may be in financial distress or face uncertain future prospects. The higher average returns on these stocks compensate investors for bearing these additional risks.

The behavioral explanation argues that these premiums arise from systematic investor mistakes. Investors may overreact to bad news, pushing value stocks below their fundamental worth. Small stocks may be neglected by analysts and institutional investors, creating pricing inefficiencies. Limits to arbitrage — such as short-selling constraints and transaction costs — prevent sophisticated investors from fully correcting these mispricings. For more on how cognitive biases affect markets, see our guide to behavioral finance.

Extensions: Four-Factor and Five-Factor Models

Since the original three-factor model, researchers have proposed extensions to capture additional return patterns:

Carhart Four-Factor Model (1997) adds a momentum factor called UMD (Up Minus Down) — the return difference between recent winners and recent losers. This fourth factor captures the well-documented tendency of stocks with strong recent performance to continue outperforming in the short term.

Fama-French Five-Factor Model (2015) adds two more factors: RMW (Robust Minus Weak), which captures the profitability premium — firms with higher operating profitability tend to earn higher returns — and CMA (Conservative Minus Aggressive), which captures the investment premium — firms that invest conservatively tend to outperform aggressive investors. Notably, including RMW and CMA in some samples reduces the standalone significance of HML, suggesting that part of the value premium may be explained by profitability and investment patterns.

Each extension improves the model’s explanatory power but adds complexity and data requirements. The proliferation of proposed factors has led to what researchers call the “factor zoo” — hundreds of candidate factors, many of which may be redundant or the result of data mining rather than genuine risk premiums.

Common Mistakes

1. Treating factor premiums as guaranteed. SMB and HML have earned positive premiums on average over long historical periods, but they can underperform for extended stretches. The value premium, for example, was negative for much of 2010–2020 as growth stocks dramatically outperformed. Factor investing requires patience measured in decades, not years.

2. Applying the model to individual stocks. The Fama-French model is designed for diversified portfolios. Individual stocks carry substantial idiosyncratic (company-specific) risk that the model does not capture. Factor loadings estimated for a single stock tend to be noisy and unreliable.

3. Confusing book-to-market with P/E ratio. The HML factor uses book-to-market ratio (book value of equity divided by market value), not the price-to-earnings ratio. While both metrics relate to valuation, they measure different things and can give conflicting signals. A stock can have a high P/E but a low book-to-market ratio.

4. Ignoring transaction costs and investability. Fama-French factor returns are computed from theoretical long/short portfolios that assume frictionless trading. In practice, implementing factor strategies involves significant transaction costs, short-selling constraints, and capacity limits that erode realized premiums relative to the theoretical returns.

Limitations of the Fama-French Model

1. The risk vs behavioral debate remains unresolved. Without consensus on why size and value premiums exist, it is difficult to predict whether they will continue. If they are compensation for risk, they should persist. If they are behavioral anomalies, they may decay as more investors exploit them.

2. Factors can have extended negative periods. The size premium has been weak or negative in U.S. large-cap stocks for several decades. The value premium was sharply negative during the growth-dominated 2010s. Investors must be prepared for long stretches of underperformance.

3. Factor definitions are somewhat arbitrary. Why book-to-market and not other value metrics? Why the specific sort breakpoints? Different construction choices can produce meaningfully different factor returns, though the broad patterns tend to hold across reasonable specifications.

4. Data-snooping concerns. Critics argue that size and value factors were discovered by mining historical data for patterns. However, Fama and French have shown that these factors predict returns across different time periods and in international markets worldwide, substantially mitigating data-snooping concerns.

5. Momentum is not captured. The three-factor model does not explain the momentum effect — the tendency of recent winners to keep winning. This requires the Carhart four-factor extension, suggesting the original model is incomplete.