Identifying Structural Vector Autoregressions Via Non-Gaussianity of Potentially Dependent Structural Shocks
The Institute of Economics will hold a seminar meeting as part of its Seminar Series on Tuesday, April 16, 2024: Markku Lanne will present the paper “Identifying Structural Vector Autoregressions Via Non-Gaussianity of Potentially Dependent Structural Shocks".
Abstract:
We show that all shocks in an n-dimensional structural vector autoregression (SVAR) are globally identified up to their order and signs if they are orthogonal and either (i) have zero co-skewness and at most one of them is not skewed or (ii) exhibit no excess co-kurtosis and at least n − 1 of them are leptokurtic. The former case covers SVAR models with errors following dependent volatility processes. Moreover, if the numbers of both skewed and leptokurtic shocks are smaller than n − 1, the skewed and leptokurtic shocks are globally identified, while the remaining shocks are set identified. To capture the non-Gaussian features of the data, versatile error distributions are needed. We discuss the Bayesian implementation of an SVAR-GARCH model with skewed t-distributed errors, including the assessment of the strength of identification and checking the validity of exogenous instruments potentially used for identification. The methods are illustrated in an empirical application to the oil market.
The Seminar will be held in Aula 6.
For online partecipation use this link.