Impact of the Reference Covariance Matrix in Portfolio Optimization Techniques with Shrinkage

This work focuses on efficient portfolio computation methods and more specifically on the estimation of the covariance matrix. First, theoretical reminders about portfolio optimization are presented. A special focus is made on shrinkage-based method (i.e. Ledoit \& Wolf) and principal component analysis (PCA). Next, different strategies are compared and new shrinkage- and PCA-based portfolio optimization methods are presented. Then, the performance achieved with these methods and those from the literature are compared and discussed.