A unified model for regularized and robust portfolio optimization

Mean-variance optimization severely suffers from model- and estimation errors. Two commonly isolated but complementary concepts to overcome the corresponding limitations are problem regularization and robustification. I introduce a joint method for covariance regularization and robust optimization which exploits this complementarity and I show that both the regularization- as well as the robust optimization part can be achieved through systematic manipulations of the correlation matrix’ eigenvalues. An application of the method to equity markets reveals similarly attractive behavior as pure covariance regularization during normal times and improved performance if a jump in systematic risk occurs. Furthermore, the model constitutes a framework for the logically consistent incorporation of systematic risk expectations into the portfolio selection problem and thereby complements similar models for individual return expectations, such as the Black and Litterman model.