Factor Model
A factor model is a statistical framework that explains the returns of a security or portfolio as a function of a set of common risk factors, separating systematic return sources from idiosyncratic, stock-specific variation.
Factor models are the foundational tool of modern quantitative portfolio management and risk analysis. The underlying idea is that the returns of any individual security can be decomposed into two components: returns attributable to broad systematic forces (factors) that affect many securities simultaneously, and returns driven by company-specific events that are independent across stocks.
The simplest factor model is the single-factor Capital Asset Pricing Model (CAPM), which attributes all systematic variation in returns to a single market factor — the excess return of the broad stock market over the risk-free rate. Each stock has a beta coefficient that measures its sensitivity to the market factor. The residual return after removing the market component is idiosyncratic risk, which in theory can be diversified away.
Multi-factor models extend this framework by identifying additional systematic sources of return variation. The Fama-French Three-Factor Model added size and value factors to the market factor. Subsequent research added momentum, quality, low volatility, and dozens of other factors. Commercial risk models from MSCI Barra, Axioma, and Northfield use hundreds of style, industry, and macroeconomic factors to decompose portfolio risk.
In portfolio construction, factor models serve two distinct roles. First, they are attribution tools: a portfolio manager can use a factor model to decompose historical returns into factor contributions and residual stock selection, understanding how much return came from being tilted toward value versus having picked the right individual stocks. Second, they are forecasting tools: if the factor structure of returns is stable over time, factor exposures today predict the factor-driven component of future returns.
Factor model construction requires statistical care. Factors must be defined precisely, estimated from sufficiently long data series, and tested for stability across market regimes.