CMStatistics 2018: Start Registration
View Submission - CFE
Title: Set identification in non-Gaussian SVARs: A refinement of the sign restriction approach Authors:  Simone Maxand - University of Helsinki (Finland) [presenting]
Helmut Herwartz - Georg-August-University Goettingen (Germany)
Abstract: In identifying structural vector autoregressive (SVAR) models, theory based sign restrictions have become a prominent means to reduce a universe of potential models to a set of structural relations which fulfil supposedly consensual economic beliefs. Typically, sign identified SVARs are obtained from tracing the symmetric effects of unit impulses hitting a system of interest in isolation. Noticing an important recent literature that targets at the identification of unique non-Gaussian structural shocks, set identification by means of isolated impulses is at the risk to neglect higher order dependence and possible asymmetries among orthogonalized model residuals. Accounting for higher order dependence in set identification of non-Gaussian SVARs, we (i) take advantage of conditional moment profiles to design `stylized unit shocks', and (ii) rely on selected `empirical shocks'.