Estimation of a dynamic multi-level factor model with possible long-range dependence

A dynamic multi-level factor model with possible stochastic time trends is proposed. In the model, long-range dependence and short memory dynamics are allowed in global local common factors as well innovations. Estimation of performed on prewhitened series, for which prewhitening parameter estimated semiparametrically from cross-sectional average observable series. Employing canonical correlation analysis a sequential least-squares algorithm resulting estimates have centered asymptotic normal distributions under certain rate conditions depending bandwidth cross-section size. Asymptotic results components also established. The selection number discussed. methodology shown to lead good small-sample performance via Monte Carlo simulations. method then applied Nord Pool electricity market price comovements among different regions within power grid. identified be system price, fractional cointegration relationships found between prices motivating long-run equilibrium relationship. Two forecasting exercises