Процедура уточнения ICSS алгоритма обнаружения структурных сдвигов в GARCH-моделях
We suggest a hybrid algorithm for structural breaks detection when using a class of piecewise-specified GARCH(1,1) models. The algorithm comprises two steps. In the first step the moments of structural breaks are detected using KL-ICSS method based on (Kokoszka, Leipus, 1999) and (Inclán, Tiao, 1994). In the second step previously detected moments of structural breaks are refined with the help of a modified MML method. Therefore, the whole procedure is called ML-KL-ICSS algorithm. We also provide five numeric experiments to show the overall performance of the proposed procedure. Four of five experiments show that ML-KL-ICSS method is significantly more accurate in detecting structural breaks as opposed to one-step procedures. In one experiment the accuracy of both methods was comparable but ML-KL-ICSS method performed slightly better. Finally, we test our method using real data. In order to do that we detect structural breaks in common stocks returns volatility for the Russian “Gazprom” company. Detected moments of structural breaks correspond to significant events in the Russian economy.