Likelihood inference in BL-GARCH models (Q1424647): Difference between revisions

The authors consider a bilinear (BL) GARCH model for the observed time series \(u\): \[ u=h_t z_t;\quad h_t^2=a_0+\sum_{i=1}^r a_i u_{t-i}^2+\sum_{j=1}^s b_j h_{t-j}^2+ \sum_{k=1}^{r^*} c_k u_{t-k}h_{t-k}. \] Conditions of stationarity and representations for the fourth unconditional moment and kurtosis of the BL-GARCH(1,1) process are presented. An EM-algorithm is proposed for the calculation of maximum likelihood estimators for unknown coefficients of BL-GARCH models. It is based on a state-space representation of BL-GARCH processes. An application to the S\,\&\,P500 stock market index from 1.03.1999 to 31.01.2001 is discussed.