Likelihood inference in BL-GARCH models

The paper presents a procedure based on the EM algorithm for the indirect estimation of the parameters of BiLinear GARCH (BL-GARCH) models. BL-GARCH generalize the class of GARCH models by considering interactions of past shocks and volatilities in the conditional variance equation. In this way the response of the conditional variance to past information becomes asymmetric allowing to account for the so called leverage effect, typically characterizing the behaviour of financial time series. The results of an application to a time series of stock market returns are presented.