Abstract

Modeling and forecasting high-dimensional time series is a challenging task in economics and finance, with the conditional covariance matrix being of great interest to both academics and practitioners. In this work, we utilize the General Dynamic Factor Model, one of the most effective methodologies for forecasting high-dimensional time series, in combination with the Principal Volatility Component framework to introduce a new approach for forecasting high-dimensional conditional covariance matrices. This novel procedure is applied to portfolio allocation problems, yielding competitive results