High Dimensional Minimum Variance Portfolio Estimation


10:30:00 - 12:30:00

103 , Mathematics Research Center Building (ori. New Math. Bldg.)

We study the estimation of high dimensional minimum variance portfolio (MVP). Two settings are considered: the low frequency setting where returns are modeled as \hbox{i.i.d.}, and the high frequency setting where returns can exhibit heteroskedasticity \emph{and} possibly be contaminated by microstructure noise. We first propose an estimator of the minimum variance, which provides a reference for comparison. More importantly, under some sparsity assumptions on the precision matrix, we propose an estimator of MVP, which asymptotically achieves the minimum variance. Simulation and empirical studies demonstrate that our proposed portfolio performs favorably. Based on joint work with Tony Cai, Jianchang Hu and Xinghua Zheng.