Estimating Covariance Matrices for Investments Whose Histories Differ in Length

Abstract

For this paper, the team considered the daily returns on the basket of stocks that compose the S&P 100 Index. We performed Principal Components Analysis on the data using Ordinary PCA and Asymptotic PCA. We also used PCA and APCA with a monotone missing data adjustment. Our results produced a minimum variance portfolio that could be used to diversify risk. This portfolio was calibrated using each method and then tested via a back-testing methodology.