A1896
Title: Modelling switching regimes with score-driven time series models
Authors: Frederik Krabbe - Aarhus University (Denmark) [presenting]
Abstract: A new autoregressive mixture model is proposed with time-varying mixture probabilities driven by the score to model switching regimes in time series. Although the model belongs to the class of score-driven models, it nests the Markov-switching autoregressive model proposed in a prior study. The statistical properties of the model as well as the asymptotic properties of the maximum likelihood estimator are studied. Moreover, the two models are compared in an empirical application which shows that the proposed model is able to capture dynamics that the Markov-switching autoregressive model is not able to.