Consistent Estimation for Aggregated GARCH Processes

We study the properties of a quasi-maximum likelihood (QML) for the parameters of a quot;weakquot; GARCH process obtained by contemporaneous aggregation of two independent quot;strongquot; GARCH processes. The inconsistency of the Gaussian quasi-likelihood estimator (QMLE) has already been reported by Nijman amp; Sentana (1996) but has not yet been solved. In this paper we identify the causes of inconsistency of QMLE in the quot;weakquot; GARCH case and compare the performance of QMLE when the innovations are assumed to have Gaussian, Laplace (double exponential) or alpha-stable distribution