A0812
Title: Bayesian inference on fully and partially identified structural vector autoregressions
Authors: Jetro Anttonen - University of Helsinki (Finland)
Markku Lanne - University of Helsinki (Finland) [presenting]
Jani Luoto - University of Helsinki (Finland)
Abstract: It is shown that because the elements of the impact matrix of the structural vector autoregression (SVAR) are always at least set identified under standard assumptions, valid Bayesian inference is possible without additional restrictions even if only some (or none) of the columns of the impact matrix are point identified due to non-Gaussianity or heteroskedasticity. This facilitates the assessment of the properties of the shocks to find out which of them (if any) are point identified. Identification results in the previous literature are also expanded to models where all or part of the structural shocks are orthogonal but mutually dependent. To exploit deviations from Gaussianity, employing versatile shock distributions is recommended and their implementation in Bayesian analysis is discussed. Simulation results and an empirical application to U.S. fiscal policy highlight the usefulness of the proposed methods and lend support to efficiently accounting for non-Gaussianity.
