Windjam

Conceptually, when you introduce structure (such as a factor model) the number of parameters that you need to estimate goes down. That reduces impact of noise (spurious correlations) and simplifies covariance estimation. for example, if you have n assets, without factor models you would have to estimate n*(n+1)/2 parameters. With one factor model, you would need to estimate n + n + 1 = 2n+1 (n betas, n residuals and market variance). Does that help?