Portfolio risk hedging: Currency exposure and Shrinkage techniques

Abstract

We apply different shrinkage techniques to the covariance and concentration matrices used for the minimum variance currency risk hedging of a globally diversified portfolio. The techniques are applied with the aim to induce sparsity in the estimator and by that to reduce the multicollinearity and the noise, resulting in a more robust estimator and decreased out-of-sample portfolio risk. We show that the application of such techniques leads to a worsening of the risk characteristics of the portfolio. We argue that this is likely due to the structure of the minimum variance hedge as well as to the shrinkage methods which seem to disturb the balance and optimality of the minimum variance hedge.