Multivariate Hyper-Rotated GARCH-BEKK

Abstract For large multivariate models of generalized autoregressive conditional heteroskedasticity (GARCH), it is important to reduce the number parameters cope with ‘curse dimensionality’. Recently, Laurent, Rombouts and Violante (2014 “Multivariate Rotated ARCH Models” Journal Econometrics 179 : 16–30) developed rotated GARCH model, which focuses on for standardized variables. This paper extends model by considering a hyper-rotation, uses more flexible structure rotation matrix. The shows an alternative representation based random coefficient vector moving-average (VARMA) process, provides regularity conditions consistency asymptotic normality quasi-maximum likelihood (QML) estimator VARMA hyper-rotated GARCH. investigates finite sample properties QML new model. Empirical results four exchange rate returns show specifications works satisfactory reducing parameters.