Guide

Fama-French five-factor model explained

Harbor Capital's U.S. equity sleeve screened for high gross profitability and low asset growth. It beat the Russell 1000 by 190 basis points annualized over eight years. A three-factor regression on SMB and HML left a stubborn positive residual — the PM insisted it was stock selection. Running the Fama-French five-factor model (FF5) changed the story: loadings of 0.64 on RMW (robust minus weak profitability) and 0.58 on CMA (conservative minus aggressive investment) absorbed most of the gap. Jensen's alpha shrank to 12 bps with a p-value above 0.40. The sleeve was a profitability-and-investment tilt dressed as pure skill — valuable exposure, but not mysterious alpha.

Eugene Fama and Kenneth French extended their 1993 three-factor framework in 2015 with two additional systematic factors motivated by decades of anomalies: profitable firms earn higher returns than unprofitable ones, and firms that invest conservatively (low asset growth) outperform aggressive expanders. FF5 is the academic workhorse for equity attribution, factor ETF design, and asking whether active managers add value after controlling for size, value, profitability, and investment style. This guide defines all five factors, walks through the regression equation, explains how RMW and CMA portfolios are constructed, covers the Harbor Capital attribution refactor, compares FF5 to CAPM, FF3, and practitioner quality factors, provides a model decision table, common pitfalls, and a production checklist alongside our factor investing guide and quality factor guide.

Why FF3 was not enough

The Fama-French three-factor model added size (SMB) and value (HML) to market beta. It explained much more cross-sectional return variation than CAPM, but two patterns persisted:

Profitability: Firms with high operating profitability relative to book equity tended to outperform, even among stocks with similar size and book-to-market ratios (Novy-Marx, 2013).
Investment: Firms with low total asset growth (conservative investment) outperformed high-growth (aggressive) firms — the corporate investment anomaly (Titman, Wei, and Xie; Cooper, Gulen, and Schill).

FF5 absorbs these anomalies into priced factors rather than leaving them as unexplained alpha or hidden style bets. For allocators, the practical question is not whether FF5 is “true” in a philosophical sense, but whether attributing returns with five factors gives a fairer read on manager skill than three.

What FF5 does not include

Momentum (UMD) is the most cited missing sixth factor; French data libraries publish it separately. Low volatility, liquidity, and short-term reversal are also outside the core FF5 specification. Practitioners often run FF5 plus momentum in a six-factor regression when evaluating tactical or quantitative funds.

The five factors: market, SMB, HML, RMW, and CMA

The time-series regression for portfolio or fund excess returns is:

R_i − R_f = α_i + β_i(R_m − R_f) + s_i SMB + h_i HML + r_i RMW + c_i CMA + ε_i

Market (MKT-RF): Excess return of the market portfolio over the risk-free rate. Same role as in CAPM and FF3.
SMB (Small Minus Big): Return spread between diversified small-cap and large-cap stock portfolios, controlling for other characteristics in the 2×3×2 sort.
HML (High Minus Low): Return spread between high and low book-to-market (value minus growth) portfolios.
RMW (Robust Minus Weak): Return spread between firms with robust (high) and weak (low) operating profitability. Profitability is typically operating profit minus interest expense, divided by book equity.
CMA (Conservative Minus Aggressive): Return spread between low-investment (conservative) and high-investment (aggressive) firms. Investment is measured by growth in total assets or similar capex-plus-R&D proxies.

Factor returns are long-short portfolio spreads, not macro indicators. A fund with r = 0.5 behaves like it is 50% invested in the RMW long-short factor on top of its other exposures.

Reading factor loadings together

Common profiles:

s > 0, h > 0 — small-cap value (classic FF3 tilt).
r > 0, c > 0 — profitable, conservative-investment (“quality” adjacent; overlaps quality factor screens).
r > 0, c < 0 — profitable but aggressive growers (some tech growth profiles).
h > 0, r < 0 — deep value with weak profitability (distressed value trap risk).

Loadings are estimated jointly. Adding RMW and CMA often reduces apparent HML exposure because value and profitability are correlated but not identical.

How RMW and CMA portfolios are built

French's methodology uses independent sorts on size, book-to-market, profitability, and investment, then forms factor-mimicking portfolios. At a high level:

Universe: U.S. common stocks on NYSE, AMEX, and NASDAQ with required accounting data (lagged to avoid look-ahead bias).
Size breakpoint: Median NYSE market cap splits small vs big.
BM, profitability, investment breakpoints: 30th and 70th percentiles of book-to-market, operating profitability, and asset growth (for investment).
Portfolio formation: Intersecting sorts create many small portfolios; SMB, HML, RMW, and CMA are weighted combinations of long and short legs with dollar-neutral construction.
Value-weighting: Stocks within portfolios are typically value-weighted; factors are rebalanced annually or monthly per the data library specification.

RMW goes long robust-profitability firms and short weak-profitability firms, holding size and value characteristics roughly neutral in the combined factor. CMA goes long conservative (low asset growth) and short aggressive (high growth). The economic story: profitable firms earn persistent excess returns; firms that retain and reinvest earnings into low-return projects destroy shareholder value on average.

Practitioner proxies vs academic series

Live funds rarely replicate French portfolios exactly. Common proxies:

RMW proxy: gross profitability (revenue minus COGS, scaled by assets) or ROE screens.
CMA proxy: low capex-to-assets, low net issuance, or negative asset growth quintiles.
ETF blends: value + profitability ETFs as a rough RMW/HML substitute.

Regression results depend on which factor return series you use. For external reporting, stick to Kenneth French's downloadable monthly factors or a single vendor (Bloomberg, FactSet) end to end.

Interpreting alpha, t-stats, and R-squared

The intercept α is Jensen's alpha after controlling for five factors. A manager beating the S&P 500 may show zero FF5 alpha if returns come from RMW and CMA tilts.

Statistical significance: Report t-statistics on α and each loading. With 60 monthly observations, |t| > 2 is a rough 95% heuristic; use Newey-West standard errors for autocorrelated fund returns.
R²: FF5 often pushes explained variance from ~0.85 (FF3) toward 0.90+ for diversified equity funds. Low R² may mean alternative exposures (bonds, options, international) or timing skill not captured by static loadings.
Rolling regressions: Loadings drift as style rotates. A five-year full-sample regression can hide two-year windows where the fund looked like pure momentum.

Pair FF5 attribution with performance attribution on holdings when you have position-level data — regression tells you what factor exposures explain returns; holdings bridge tells you why.

Harbor Capital FF5 attribution refactor

Harbor's U.S. all-cap equity sleeve used a quality screen: top-quartile gross profitability, bottom-quintile asset growth, plus a mild value overlay. Marketing labeled it “active stock selection with quality bias.” Investor due diligence requested factor attribution against public benchmarks.

FF3 baseline — Positive α of 78 bps (t = 2.1) over 96 months; HML loading 0.41, SMB 0.18. Looked like modest skill.
FF5 extension — RMW loading 0.64, CMA 0.58; α fell to 12 bps (t = 0.4). HML loading dropped to 0.19 as profitability explained overlap with value.
Holdings bridge — Top holdings clustered in high-OP consumer staples and software with buyback-heavy capital allocation (low asset growth). Confirmed regression story.
Fee disclosure update — Prospectus language shifted from “alpha generation” to “systematic exposure to profitability and conservative investment factors with security selection overlay.”
Benchmark change — Custom blended benchmark: 50% Russell 1000, 25% profitability factor proxy ETF, 25% min-volatility ETF; reduced reported tracking error without changing portfolio.

Outcome: No strategy change required, but LP satisfaction improved because expectations aligned with measurable factor bets. Lesson: FF5 is as much a communication tool as a risk model.

Model decision table

Model	Factors	Best when	Watch out for
CAPM	Market beta only	Quick beta estimate, CAPM-era legacy reports	Misses size, value, profitability tilts; inflates alpha
FF3	MKT, SMB, HML	Traditional value/small-cap funds, simpler LP reporting	Leaves profitability/investment exposures as fake alpha
FF5	MKT, SMB, HML, RMW, CMA	Quality/profitability funds, broad equity attribution (2015+ standard)	Omits momentum; correlated factors muddy individual loadings
FF5 + UMD	Above plus momentum	Quant, tactical, or growth-momentum sleeves	Momentum crashes; short history in some regions
Practitioner multi-factor	Custom (vol, liquidity, etc.)	Proprietary risk systems, ETF mapping	Not comparable across managers without shared factor defs

Common pitfalls

Claiming alpha after FF3 when the fund is quality-tilted — Run FF5 before marketing skill; RMW and CMA often explain “mysterious” outperformance.
Confusing RMW with a full quality definition — RMW is profitability only; leverage stability and earnings quality are separate (see quality factor guide).
Ignoring CMA on growth funds — High-growth tech can show negative CMA loading; that is a style bet, not necessarily bad performance.
Using ETF regressions on monthly data with 24 observations — Too few points for stable loadings; prefer 60+ months or Bayesian shrinkage.
Survivorship-biased fund databases — Dead funds with positive FF3 alpha disappear; FF5 post-mortems on live funds only overstate manager skill (see survivorship bias guide).
Mixing factor return vendors — French SMB with a proprietary RMW series breaks orthogonality assumptions.
Static loadings on dynamic funds — Managers who rotate style quarterly need rolling or regime-switching attribution.
International equities with U.S. FF5 factors — Use regional factor libraries (Europe, developed ex-U.S.) when available.

Production checklist

Download consistent monthly factor returns (French data library or single vendor).
Align fund returns and factors on same calendar (month-end, USD, net of fees).
Run FF3 and FF5 side by side; document how α and loadings change.
Use Newey-West (or equivalent) standard errors on time-series regressions.
Report R², adjusted R², and number of observations with every table.
Plot rolling 36-month loadings for RMW and CMA if the mandate allows style drift.
Bridge regression to holdings: top quintile profitability, asset growth distribution.
Add UMD when evaluating momentum-heavy or quant funds.
Disclose whether returns are gross or net of fees in alpha interpretation.
Reconcile FF5 with Brinson-style attribution when both are available.

Key takeaways

FF5 adds profitability (RMW) and conservative investment (CMA) to the classic size and value factors, absorbing anomalies that FF3 left as unexplained alpha.
Factor loadings are interpreted as long-short portfolio exposures; high RMW and CMA often describe quality-tilted funds previously labeled as pure stock pickers.
Academic factor portfolios use independent sorts on size, book-to-market, profitability, and asset growth; practitioner proxies should not be mixed across vendors.
Harbor Capital's refactor showed FF5 attribution aligns investor expectations with measurable factor bets more honestly than FF3 alone.
FF5 is the baseline for modern equity factor attribution but omits momentum; extend to six factors when the mandate includes tactical or momentum exposure.