1.3 Fundamental Multi-Factor Models

Basic Prerequisites: Linear Algebra & Matrices (OLS, projections, eigen-decomposition) and Econometrics & Time Series (regression, covariance, stationarity). (these are the folder-level prerequisites for pages 02–06; page 01 states its own, smaller, entry requirements)

1. Intuition & Practical Objective

Rather than predicting each of 3,000 stocks independently, quantitative asset pricing projects stocks onto a low-dimensional basis of systematic risk factors. The claim of the whole folder, from Fama–French 1992: two easily measured variables - size and book-to-market - capture the cross-sectional variation in average stock returns that the CAPM’s single beta alone could not. Once you know a stock’s loadings (exposures, $\beta$s) onto a handful of priced factors - market, size, value, profitability, investment - you can explain up to ~90% of its return variance and, crucially, isolate whether an active manager’s performance is true idiosyncratic skill (alpha) or merely repackaged systematic factor risk (beta).

This folder is the quantitative engine layer beneath the accounting-based factor catalog of Quantitative Fundamental Investing. That folder asks which fundamental characteristics pay (value, profitability, investment, quality); this folder asks how you turn those characteristics into an econometrically valid, tradeable, risk-budgeted factor model. It is a hub: it (a) gives the model & formula lookup below, and (b) routes you to six sub-pages that walk from raw intuition through the Fama–French construction, the cross-sectional (Barra/Axioma) machinery, the failure modes, and the extensions.

The one-sentence essence. “Regress returns onto a few priced factors to decompose total return into systematic (factor-compensated beta) and idiosyncratic (alpha + noise) pieces - where the factors are built by sorting the cross-section on real fundamental characteristics and then long-shorting the extreme deciles.”

Audience arc: beginner learns why a single beta fails and what a factor is; intermediate learns how SMB/HML/RMW/CMA are constructed and estimated by time-series regression; expert reads cross-sectional GLS/WLS models, covariance factorization, and the factor-zoo/crowding failure modes. The sub-pages below are ordered for exactly that climb.

2. Mathematical Ground Truth & Model Lookup

Quick lookup. The folder’s key formulas in one table; each is derived on the sub-pages.

Notation: $R_{it}$ asset $i$ return at $t$; $R_{ft}$ risk-free rate; $R_{Mt}$ market return; $MK{T}_t=R_{Mt}-R_{ft}$ market excess; $\text{SMB},\text{HML},\text{RMW},\text{CMA}$ the Fama–French factor returns; $X_t$ the $N\times K$ cross-sectional matrix of standardized factor exposures; $f_t\in {\mathbb{R}}^K$ factor returns; $\Omega$ factor covariance; $\Delta$ diagonal idiosyncratic covariance; ${\alpha }_i$ the pricing-error/intercept.

Fama–French 3-factor time-series regression (Fama & French 1993; Tsay §9.3.2):

Fama–French 5-factor model (FF 2015, eq. 5; add profitability and investment):

with $RMW$ = robust-minus-weak operating profitability and $CMA$ = conservative-minus-aggressive investment.

Factor construction by 2×3 independent sorts (FF 2015; NYSE-only breakpoints): split on median size (S/B), and on 30th/70th percentile B/M → six value-weighted portfolios; then

and SMB averages the size factors from the B/M, profitability, and investment sorts. (Verified in §3.)

Cross-sectional factor model (Barra/Axioma; Tsay §9.3.1 eq. 9.1–9.8): each period

a WLS/GLS (Tsay eq. 9.7) with the factor-mimicking portfolio weights $\omega =(X^{\top }{V}^{-1}X{)}^{-1}{X}^{\top }{V}^{-1}$ (Tsay §9.3.1).

General factor-model covariance (Tsay eq. 9.1–9.4): if $r_t=\alpha +B{f}_t+{\epsilon }_t$ with $\text{Cov}(f_t)=\Omega$, $\text{Cov}({\epsilon }_t)=D=\text{diag}\{{\sigma }_i^2\}$,

which replaces the $\frac{N(N-1)}{2}$ pairwise covariances with $K\ll N$ factor covariances (the practical point of a risk model).

Statistical factor model / PCA (Tsay §9.5; ESL Ch 14): orthogonal-factor model $r_t-\mu =\beta f_t+{\epsilon }_t$ with $\text{Cov}(f)=I$, $\text{Cov}(\epsilon )=D$, so ${\Sigma }_r=\beta {\beta }^{\prime }+D$ (Tsay eq. 9.17); communality $c_i^2={\sum }_j{\beta }_{ij}^2$ plus unique variance ${\sigma }_i^2$ splits each asset’s variance, and PCA loadings are $\sqrt{{\lambda }_j}{e}_j$ (Tsay eq. 9.19).

3. Computational Implementation - the factor engine

This runs on the standard library only (no numpy/scipy). It builds SMB and HML by a 2×3 independent sort on a synthetic cross-section and reports the realized long-short spreads - the machine that every factor model in this folder reduces to.

(Every number on every sub-page was re-executed and reproduced exactly from this engine and its siblings; the full construction is derived on 03 · Factor Construction.)

4. Failure Modes & First-Principles Breakdowns

Hub signposts - the folder’s failure-mode analysis lives in 05 · Failure Modes & Practice. In one line each:

1. Factor crowding & liquidity black holes - when many funds crowd the same factor (the Quant Quake of August 2007), a forced deleveraging by one fund cascades into same-side liquidations across the whole factor, producing unprecedented drawdowns in supposedly market-neutral books.
2. Multicollinearity in style factors - overlapping factors (five variants of value/momentum) make $X^{\top }X$ nearly singular: betas swing wildly and their standard errors explode (the factor zoo).
3. Data-mining / multiple testing - with hundreds of candidate characteristics, some premium will look significant by chance; backtested factor premiums must be discounted (Green–Hand–Zhang; see Backtesting Hygiene & Deflated Sharpe).

5. References

• Fama, Eugene & French, Kenneth: “The Cross-Section of Expected Stock Returns” (Journal of Finance, 1992)
• Fama & French: “Common Risk Factors in the Returns on Stocks and Bonds” (JFE, 1993)
• Fama & French: “A Five-Factor Asset Pricing Model” (JFE, 2015)
• Tsay, Ruey S.: Analysis of Financial Time Series (3rd ed., 2010)
• Hastie, Tibshirani & Friedman: The Elements of Statistical Learning (2nd ed., 2009)

6. Connected Graph Bridges

• Foundational base: Linear Algebra & Matrices · Econometrics & Time Series (the regression and covariance toolkit this model runs on)
• Fundamentals source: Quantitative Fundamental Investing (the characteristics - value, profitability, investment - that this folder turns into priced factors) · Core Financial Ratios
• Sibling topic: Cross-Sectional & Time-Series Momentum (the momentum factor family) · Statistical Arbitrage & Pairs Trading
• Downstream risk: Portfolio Optimization (factor tilts → factor-risk-budgeted portfolios) · Quantitative Risk (factor covariance $\Sigma =B\Omega B^{\prime }+D$ is the risk model)
• Sub-pages (in-folder): 01 From Zero · 02 Fama–French Factor Model · 03 Factor Construction · 04 Cross-Sectional Models · 05 Failure Modes & Practice · 06 Advanced Extensions

Beginner: start at 01 · Practitioner: start at 05