Hansen, Peter Reinhard Janus, Pawel Koopman, Siem Jan
Year of Publication:
Tinbergen Institute Discussion Paper 16-061/III
We propose a novel multivariate GARCH model that incorporates realized measures for the variance matrix of returns. The key novelty is the joint formulation of a multivariate dynamic model for outer-products of returns, realized variances and realized covariances. The updating of the variance matrix relies on the score function of the joint likelihood function based on Gaussian and Wishart densities. The dynamic model is parsimonious while each innovation still impacts all elements of the variance matrix. Monte Carlo evidence for parameter estimation based on different small sample sizes is provided. We illustrate the model with an empirical application to a portfolio of 15 U.S. financial assets.
high-frequency data multivariate GARCH multivariate volatility realised covariance score Wishart density