The benefits of improved covariance estimation

Abstract

Recent advances in covariance estimation can improve portfolio formation strategies aimed at avoiding high risk market environments. We consider a covariance specification with information variables that include both historical firm specific variables and an ex ante measure of macro volatility (CBOE VIX). We compare the in-sample and predictive out-of-sample performance of the information instrument model relative to three alternative approaches. Out-of-sample, a risk-on, risk-off strategy that optimally weights the global minimum variance (GMV) portfolio and a riskless asset shows the information instrument model provides effective exit signals during the financial crisis and other high risk environments.