Munich Personal RePEc Archive

Estimating multivariate GARCH and stochastic correlation models equation by equation

Francq, Christian and Zakoian, Jean-Michel (2014): Estimating multivariate GARCH and stochastic correlation models equation by equation.

[img]
Preview
PDF
MPRA_paper_54250.pdf

Download (576kB) | Preview

Abstract

A new approach is proposed to estimate a large class of multivariate volatility models. The method is based on estimating equation-by-equation the volatility parameters of the individual returns by quasi-maximum likelihood in a first step, and estimating the correlations based on volatility-standardized returns in a second step. Instead of estimating a $d$-multivariate volatility model we thus estimate $d$ univariate GARCH-type equations plus a correlation matrix, which is generally much simpler and numerically efficient. The strong consistency and asymptotic normality of the first-step estimator is established in a very general framework. For generalized constant conditional correlation models, and also for some time-varying conditional correlation models, we obtain the asymptotic properties of the two-step estimator. Our estimator can also be used to test the restrictions imposed by a particular MGARCH specification. An application to financial series illustrates the interest of the approach.

UB_LMU-Logo
MPRA is a RePEc service hosted by
the Munich University Library in Germany.