SWARCH modeli

Zbog velikog utjecaja šokova na procjenu ARCH familije modela bila potrebna nova specifikacija koja dozvoljava promjene u režimu generirajućeg procesa. Cai [3] i Hamilton i Susmel [17], nezavisno predlažu SWARCH model, koji je dizajniran da obuhvati promjene režima u volatilnosti pomoću neopažene varijable stanja koja prati Markovljev lanac prvog reda sa diskretnim skupom stanja. U ovom radu, prezentirani su SWARCH modeli, s fokusom na SWARCH model razvijen od strane Hamiltona i Susmela [17], za koji je dan i prikaz procjene parametara i značajki modela te na temelju kojeg je i napravljen primjer upotrebe modela, naveden u ovom radu. Ukratko su opisani E-SWARCH model te multivarijantni SWARCH model, kao glavna proširenja osnovnog modela. Glavni zaključak je da model primjereno modelira pojave tijekom kojih dolazi do promjena u volatilnosti, bitno reducirajući ustrajnost utjecaja šokova i time dodatno naglašava potrebu za modelom uvjetne varijance koji dozvoljava stohastička prebacivanja. Također, model dopušta datiranje i time identifikaciju glavnih epizoda i događaja u pozadini visoko volatilnih perioda na financijskim tržištima.Because of the large influence of shocks on the estimation of ARCH family of models, a new specification was needed, which will allow changes in the regime of the generating process. Cai [3] and Hamilton and Susmel [17], independently suggested SWARCH model, which was designed to capture regime switches in volatility with unobserved state variable that follows first order, discrete state Markov chain. In this paper, SWARCH models are presented, focusing on SWARCH model developed by Hamilton and Susmel [17], for which is given the parameter estimation procedure and features of the model, and on which basis was made an example of usage of the model, referred to in this paper. Also, there are brief descriptions of the E-SWARCH model and multivariate SWARCH model, as major extensions of the basic model. The main conclusion is that the model adequately models phenomenona in which occurred the change in volatility, significantly reducing the persistence of shocks and thus further emphasizes the need for a model of conditional variance that allows the stochastic switching. Also, the model allows for time dating and identification of key episodes and events in the background of highly volatile periods in financial markets