Dynamic mixture vector autoregressions with score-driven weights

We propose a novel dynamic mixture vector autoregressive (VAR) model in which time-varying mixture weights are driven by the predictive likelihood score. Intuitively, the state weight of the k-th component VAR model in the subsequent period is increased if the current observation is more likely to be drawn from this particular state. The model is not limited to a specific distributional assumption and allows for straight-forward likelihood-based estimation and inference. We conduct a Monte Carlo study and find that the score-driven mixture VAR model is able to adequately filter and predict the mixture dynamics from a variety of different data generating processes, which other observation-driven dynamic mixture VAR models cannot appropriately handle. Finally, we illustrate our approach by an application where we model the conditional joint distribution of economic and financial conditions and derive generalized impulse responses.