Using, Taming or Avoiding the Factor Zoo?
A Double-Shrinkage Estimator for Covariance Matrices
The applied research study shows that a systematic incorporation of certain factors, such as those from the CAPM or Fama-French model, allows a more accurate estimation of the covariance matrix and thus of the minimum-variance portfolio. The proposed "double-shrinkage" procedure smooths not only the estimate of the volatilities and correlations of stock returns, but also that of the factors. This new robust estimation can be used in portfolio optimization to reduce risks, as well as to compute a more stable and better-balanced portfolio with less extreme allocations.