Advances in Structural Vector Autoregressions with Imperfect Identifying Information
This paper examines methods for structural interpretation of vector autoregressions when the identifying information is regarded as imperfect or incomplete. We suggest that a Bayesian approach offers a unifying theme for guiding inference in such settings. Among other advantages, the unified approach solves a problem with calculating elasticities that appears not to have been recognized by earlier researchers. We also call attention to some computational concerns of which researchers who approach this problem using other methods should be aware.
We thank Diego Kaenzig and Lam Nguyen for helpful comments on an earlier draft of this paper. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.