2/2018 Wenjuan Chen and Aleksei Netšunajev. Structural vector autoregression with time varying transition probabilities: identifying uncertainty shocks via changes in volatility
Working Papers of Eesti Pank 2/2018
Structural vector autoregressive models with regime-switching variances have been used to test structural identification strategies. In these models the transition probabilities are assumed to be constant over time. In reality these probabilities may depend on certain economic fundamentals that help predicting turning points. This paper is the first to introduce time-varying probabilities into structural VAR model that is identified via volatility. A generalized Expectation-Maximization algorithm is developed for estimation of the model. For empirical illustration the model is applied to test two sets of assumptions used for identification of uncertainty shocks. A formal test rejects the hypothesis that uncertainty shocks do not influence macroeconomics variables on impact but supports the alternative of non-negligible contemporaneous effects.
JEL Codes: C32, D80, E24
Keywords: structural vector autoregression; Markov switching; time varying transition probabilities; identification via heteroscedasticity; uncertainty shocks; unemployment dynamics