What the Fama-French Three-Factor Model Actually Explains
Two stocks can carry an identical market beta and still deliver very different long-run returns, a gap the capital asset pricing model has no way to account for. The three-factor model closes most of that gap by adding size and value as separate, priced sources of risk.
The core idea: two more risk dimensions
The single-factor capital asset pricing model says a stock's expected return depends on exactly one thing beyond the risk-free rate: its sensitivity to the overall market, measured by beta. For decades this was the working assumption of academic finance, and it is a clean, intuitive story. The trouble is that when researchers sorted decades of stock returns by size and by valuation, they found systematic, persistent patterns that market beta alone could not explain. Small-capitalization stocks, as a group, earned higher average returns than large-capitalization stocks with similar betas. Stocks trading at low prices relative to their book value, generally called value stocks, earned higher average returns than stocks trading at high multiples of book value, generally called growth stocks, again after controlling for market beta.
The three-factor model responds to this by adding two more variables to the return equation, each capturing a distinct, economically motivated source of risk. The first addition is a size factor, often labeled SMB for "small minus big," which measures the historical tendency of small-cap stocks to outperform large-cap stocks. The second is a value factor, labeled HML for "high minus low," referring to high versus low book-to-market ratios, which measures the tendency of statistically cheap stocks to outperform statistically expensive ones. Each factor has its own risk premium, estimated from long-run historical spreads between the relevant portfolios, and each stock has its own sensitivity, or loading, to each factor, estimated separately from its sensitivity to the market as a whole.
The full equation reads: expected return = risk-free rate + (beta_market x market premium) + (beta_size x SMB premium) + (beta_value x HML premium). Each of the three betas is estimated by regressing a stock's or fund's historical returns against the three factor portfolios simultaneously, which separates out how much of its return pattern is genuine market exposure, how much is size exposure, and how much is value exposure. A fund manager who appears to "beat the market" using a simple market-only comparison may, once this decomposition is run, turn out to have simply held a portfolio tilted toward small and value stocks, a pattern that resembles skill in a one-factor world but resembles factor exposure once you control for it properly.
It helps to know, at least in outline, how the SMB and HML portfolios are actually constructed, because the mechanics explain why the factors behave the way they do. Researchers sort the full universe of traded stocks into groups by market capitalization and, separately, by book-to-market ratio, then build long-short portfolios: SMB is the return of a basket of small stocks minus the return of a basket of large stocks, and HML is the return of a basket of high book-to-market (cheap) stocks minus a basket of low book-to-market (expensive) stocks. Because these are long-short constructions, a factor's monthly return can be positive or negative regardless of what the overall market did that month, which is exactly what allows a regression to isolate size and value effects independently of the broad market's direction. A stock does not need to be small or cheap in an absolute sense to carry a meaningful loading on either factor; what matters is how its returns move relative to these two constructed spread portfolios over the estimation period.
The math, worked in full
Worked example 1: decomposing a small-value fund's return. Suppose a mutual fund specializing in small, statistically cheap companies has an estimated market beta of 1.05, a size beta of 0.60 (meaningfully tilted toward small caps), and a value beta of 0.45 (meaningfully tilted toward cheap stocks). Assume a risk-free rate of 4%, a market risk premium of 5.5%, a historical SMB premium of 2.5%, and a historical HML premium of 3.0%. Plugging in: expected return = 4% + (1.05 x 5.5%) + (0.60 x 2.5%) + (0.45 x 3.0%).
Working through each term: the market contribution is 1.05 x 5.5% = 5.775%, the size contribution is 0.60 x 2.5% = 1.5%, and the value contribution is 0.45 x 3.0% = 1.35%. Summing: 4% + 5.775% + 1.5% + 1.35% = 12.625%. A single-factor CAPM estimate using only the 1.05 market beta would have produced 4% + (1.05 x 5.5%) = 9.775%, understating the fund's expected return by nearly three full percentage points, precisely the portion attributable to its deliberate size and value tilts rather than to raw market exposure or manager skill.
Worked example 2: separating alpha from factor exposure. Now suppose that small-value fund actually delivered a 14% return over the year in question. Under a naive single-factor comparison against a 9.775% CAPM benchmark, the fund appears to have generated 14% - 9.775% = 4.225% of outperformance, which a casual observer might read as manager skill worth paying an active fee for. Under the three-factor benchmark of 12.625% computed above, the true unexplained return, or alpha, is only 14% - 12.625% = 1.375%. Most of what looked like skill was in fact compensation for size and value risk the fund was already taking on, exposure an investor could replicate directly and far more cheaply through a low-cost small-value index fund rather than paying an active manager to deliver it.
Worked example 3: a large-growth fund on the other side of the tilt. The same logic runs in reverse for a fund tilted the opposite way. Consider a large-growth fund with a market beta of 1.10, a size beta of negative 0.35 (tilted toward large caps, since SMB is small minus big), and a value beta of negative 0.50 (tilted toward growth, since HML is high minus low book-to-market). Using the same risk-free rate and premiums: market contribution is 1.10 x 5.5% = 6.05%, size contribution is -0.35 x 2.5% = -0.875%, and value contribution is -0.50 x 3.0% = -1.5%. The three-factor expected return is 4% + 6.05% - 0.875% - 1.5% = 7.675%, noticeably below the single-factor CAPM estimate of 4% + 6.05% = 10.05%. If this fund actually returned 11% over the period, its three-factor alpha is 11% - 7.675% = 3.325%, a genuinely larger and more interesting result than the naive single-factor comparison of 11% - 10.05% = 0.95% would suggest, because the fund's growth tilt was working against it, not for it, and it still delivered a strong result net of that headwind.
What the historical record shows
Across long stretches of U.S. and international stock market history, the size and value premiums have shown up with enough consistency to be treated as genuine, priced risk factors rather than statistical artifacts of a single data sample, though both premiums have also gone through extended stretches, sometimes a full decade or longer, where they delivered flat or even negative returns relative to the broad market. The 2010s were a notably difficult period for the value factor specifically, with growth stocks, particularly a small number of very large technology companies, outperforming value stocks by a wide and persistent margin, a stretch long enough that some researchers publicly questioned whether the value premium had been arbitraged away by widespread awareness of the finding itself.
The size premium has shown a similar pattern of inconsistency across different multi-decade windows, appearing strong in some periods and largely absent in others, which is one reason serious factor researchers now describe both premiums as real but noisy, meaningfully positive over very long horizons while entirely capable of producing a decade of disappointment for any investor who adopted the tilt expecting smooth, dependable outperformance. This variability is itself consistent with a genuine risk-based explanation: a risk premium that always paid off on a predictable schedule would not really be compensation for risk at all, since risk by definition involves the possibility of extended underperformance precisely when it would hurt most.
International evidence adds an important cross-check. Researchers who extended the same size and value sorts to stock markets outside the United States, across developed markets in Europe and Asia and, with more caution given thinner and less liquid data, in emerging markets, found broadly similar patterns: small and value stocks earning a premium over long horizons, with the same tendency toward multi-year stretches of underperformance. Finding a similar pattern across many independent markets and time periods is one of the stronger pieces of evidence that the size and value premiums reflect something structural about how markets price risk, rather than a quirk specific to one country's data or one historical era. It is also worth being direct about the model's limits: the original three factors leave a meaningful share of return variation unexplained, which is exactly why later academic work added a profitability factor and an investment factor, and why a momentum factor is frequently bolted on as a fourth or fifth addition in practice. The three-factor model was a major advance over CAPM, not a final, complete description of what drives stock returns.
Using the model in a real portfolio
For most individual investors, the three-factor model is more useful as a diagnostic lens than as a literal blueprint for portfolio construction. It explains why two funds with the same headline market benchmark can behave so differently, why a "value" fund can lag the broad market for years without anything having gone wrong with its process, and why comparing a manager's return only to a broad market index can flatter or unfairly penalize a fund depending on which factors it happens to be tilted toward in a given period. Reading a fund's factsheet with this framework in mind, checking its size and value tilt rather than only its category label, is a more precise way to understand what you actually own.
Investors who do want deliberate factor exposure, most commonly a modest tilt toward small and value stocks alongside a core broad-market holding, can generally get it today through low-cost index and exchange-traded funds built explicitly around these factors, without needing to pay active management fees to access the same underlying exposure. The honest caveat is that a factor tilt is a long-horizon bet on a risk premium persisting going forward the way it has, on average, in the past, and any investor adopting one should be prepared to hold it through a stretch of underperformance lasting years, not months, since that is exactly the kind of discomfort the premium is thought to compensate for.
The model is also a useful sanity check when evaluating a financial advisor's or fund company's marketing claims. If a proposed portfolio's back-tested outperformance can be substantially reproduced by tilting a plain index fund toward small and value stocks, that outperformance is not evidence of a proprietary edge worth an extra layer of fees; it is evidence of a factor tilt available far more cheaply elsewhere. Running, or asking an advisor to run, a three-factor regression on a fund's historical returns before committing capital is a more rigorous due-diligence step than reading a glossy brochure of trailing returns, and it takes only a modest amount of publicly available data to do.
Actionable breakdown
- Read fund factsheets for size and value factor loadings, not just category.
- Compare active manager returns to a factor benchmark, not just the market.
- Adopt a size or value tilt only if you can hold it for a full decade.
- Prefer low-cost index funds over active funds for factor exposure.
- Expect factor premiums to disappoint for years at a time.
Common pitfalls
- Crediting manager skill for what is really unpaid factor exposure.
- Abandoning a factor tilt after a single disappointing decade.
- Assuming size and value premiums are guaranteed rather than risk-based.
- Paying active fees for factor exposure available cheaply in index form.
The bottom line
The three-factor model shows that a meaningful share of what looks like investment skill is really compensation for taking on identifiable size and value risk, which is best captured cheaply rather than paid for through an active manager.
All articles · Factor investing · Multifactor models overview · The capital asset pricing model · Arbitrage pricing theory