A new algorithm for sign restrictions in vector autoregressions

This paper proposes a new, comprehensive Bayesian sampling scheme for inference in vector autoregressions (VARs) using sign restrictions. I build on a factor model decomposition of the reduced-form VAR disturbances, which are specified to be driven by a few common factors/shocks. The outcome is a computationally efficient algorithm that allows to jointly sample VAR parameters as well as decompositions of the covariance matrix satisfying desired sign or even zero restrictions. Unlike existing algorithms, the proposed framework allows to statistically test the plausibility of sign and other restrictions in VARs, for example, in situations where there is little guidance from economic theory about the effect of a certain shock. Using artificial and real data I show that the new algorithm works well and is multiple times more efficient than existing accept/reject algorithms for sign restrictions