Realized BEKK-CAW Models

Abstract Estimating time-varying conditional covariance matrices of financial returns play important role in portfolio analysis, risk management, and econometrics research. The availability high-frequency data can provide an additional source for dynamic modeling. In this paper, we propose to use the information asset return vector realized measures simultaneously develop a new matrix model. We derive stationary condition normal Wishart distributions construct quasi-log-likelihood function. also consider variance targeting (VT) method, which plugs weighted average sample mean measure unconditional matrix, order maximize show consistency asymptotic normality quasi-maximum likelihood (QML) VT estimators. investigate finite property these estimators via Monte Carlo experiments. empirical example bivariate Nikkei 225 index its futures indicates that first-step estimation could have non-negligible effects on standard errors second-step estimates.