Robust Portfolio OptimizationRobust Portfolio OptimizationQiu, Huitong and Han, Fang and Liu, Han and Caffo, Brian2015
Paper summarynipsreviewsThe authors derive an estimator of a "proxy" of the covariance matrix of a stationary stochastic process (in their case asset returns) which is robust to data outliers and does not make assumptions on the tails of the distribution. They show that for elliptical distributions, which includes Gaussians, this proxy is consistent with true covariance matrix up to a scaling factor; and that their proposed estimator of the proxy has bounded error.
The authors derive an estimator of a "proxy" of the covariance matrix of a stationary stochastic process (in their case asset returns) which is robust to data outliers and does not make assumptions on the tails of the distribution. They show that for elliptical distributions, which includes Gaussians, this proxy is consistent with true covariance matrix up to a scaling factor; and that their proposed estimator of the proxy has bounded error.