Title: The dark side of the SVAR: A trip into the local identification world
Authors: Emanuele Bacchiocchi - University of Milan (Italy) [presenting]
Toru Kitagawa - University College London (United Kingdom)
Abstract: The focus is on structural vector autoregressions where the identification issue can be addressed only locally. In this particular case, there are different isolated points in the parametric space satisfying the imposed restrictions. Unfortunately, all these points are observationally equivalent and it is impossible to chose among them simply by considering the likelihood function. The estimation of the parameters, thus, is subject to the algorithm and the related starting values used for maximizing the likelihood function. We address this problem by considering all the locally identified parameters and use Bayesian techniques to make inference on the estimated impulse responses.