Barra risk models (developed by Barra, now part of MSCI) decompose the cross-sectional variance of equity returns into two components: systematic variance driven by common risk factors, and idiosyncratic variance specific to each individual stock.
The model structure is: r_i = Σ_k (X_ik × f_k) + ε_i, where r_i is the return of stock i, X_ik are its exposures to factors k (style, industry, country), f_k are the factor returns estimated from cross-sectional regressions, and ε_i is the idiosyncratic return.
Factor categories in a typical Barra equity model
- Style factors — value, momentum, size, quality, volatility, liquidity
- Industry factors — sector and industry exposures using GICS or similar classification
- Country/region factors — relevant for global or regional models
Uses in quantitative research
- Portfolio construction — control factor exposures while maximizing signal exposure
- Risk attribution — decompose portfolio variance into factor vs idiosyncratic sources
- Return attribution — identify which factor tilts drove performance vs pure selection