The Mathematics of Multi-Factor Asset Pricing: From CAPM to Fama-French 5-Factor & Momentum

01. The Breakdown of Single-Factor CAPM

For three decades following the seminal work of William Sharpe (1964) and John Lintner (1965), modern portfolio theory rested upon the Capital Asset Pricing Model (CAPM). The model asserted that the expected excess return of any equity asset $i$ is solely a linear function of its non-diversifiable systematic sensitivity (beta) to the broader market portfolio:

Under pure CAPM, any security or managed portfolio exhibiting a non-zero Jensen's Alpha (alpha != 0) represented an arbitrage opportunity or an exceptional display of managerial skill. However, extensive empirical testing throughout the 1980s and 1990s demonstrated that the CAPM beta explained less than 70% of the cross-sectional variation in common stock returns. Two glaring empirical anomalies repeatedly falsified the model:

• The Size Effect (Banz, 1981): Small market capitalization equities consistently generated higher risk-adjusted returns than predicted by their market beta.
• The Value Anomaly (Stattman, 1980; Rosenberg et al., 1985): Equities with high book-to-market (B/M) ratios systematically outperformed low B/M growth equities on a risk-adjusted basis.

02. Factor Portfolio Sorting Mechanics (2x3 Sorts)

In their foundational 1993 paper, Eugene Fama and Kenneth French resolved this divergence by proposing that Size and Value represented systematic state-variable risk factors, rather than market inefficiencies. To construct orthogonal, investable factor long/short portfolios without confounding size and value effects, Fama and French established the authoritative 2x3 independent sorting methodology:

03. The Fama-French 5-Factor Mathematical Model

In 2015, Fama and French extended their model to five factors, incorporating Operating Profitability (RMW: Robust Minus Weak) and Investment Intensity (CMA: Conservative Minus Aggressive). The complete regression model specification is expressed as:

Remarkably, empirical testing across the CRSP/Compustat universe revealed that upon the introduction of RMW and CMA, the traditional Value factor (HML) becomes redundant in explaining cross-sectional returns for the modern era (t-statistic on h drops below significance when r and c are included).

04. Carhart Momentum & The Cross-Section of Returns

Mark Carhart (1997) documented that mutual fund managers who appeared to generate persistent alpha were actually capturing Jegadeesh and Titman's (1993) 12-month relative-strength momentum effect. The Carhart 4-Factor (and extended 6-factor) model incorporates the Up Minus Down (UMD) or Winners Minus Losers (WML) factor:

05. Fama-MacBeth Two-Pass Regression Methodology

To test whether candidate factors carry genuine positive risk premiums across the broader economy, institutional quantitative researchers utilize Eugene Fama and James MacBeth's (1973) two-pass cross-sectional regression:

1. Pass 1 (Time-Series OLS): For each asset or test portfolio $i$, regress time-series excess returns against the factors to obtain factor loadings: $$R_{it} - R_{ft} = \alpha_i + \sum_{k=1}^K \beta_{ik} F_{kt} + \varepsilon_{it}$$
2. Pass 2 (Cross-Sectional OLS per Period): For each time period $t$, regress cross-sectional asset returns across the estimated betas to determine the period factor risk price $\lambda_{kt}$: $$R_{it} = \lambda_{0t} + \sum_{k=1}^K \hat{\beta}_{ik} \lambda_{kt} + \eta_{it}$$
3. Time-Series Averaging: Compute the time-series mean and Newey-West standard errors for each factor price: $$\bar{\lambda}_k = \frac{1}{T}\sum_{t=1}^T \lambda_{kt}$$

06. True Alpha vs. Disguised Factor Beta in Portfolio Underwriting

For institutional asset allocators, family offices, and endowment chief investment officers, multi-factor regression is the ultimate truth engine. Historical hedge fund outperformance often collapses into leveraged factor tilts: