Statistical identification in SVARs - Monte Carlo experiments and a comparative assessment of the role of economic uncertainties for the US business cycle

Abstract

Structural vector autoregressive analysis aims to trace the contemporaneous linkages among (macroeconomic) variables back to underlying orthogonal structural shocks. In homoskedastic Gaussian models the identification of these linkages deserves external and typically notdata-based information. Statistical data characteristics (e.g, heteroskedasticity or non-Gaussian independent components) allow for unique identification. Studying distinct covariance changes and distributional frameworks, we compare alternative data-driven identification procedures and identification by means of sign restrictions. The application of sign restrictions results in estimation biases as a reflection of censored sampling from a space of covariance decompositions. Statistical identification schemes are robust under distinct data structures to some extent. The detection of independent components appears most flexible unless the underlying shocks are (close to) Gaussianity. For analyzing linkages among the US business cycle and distinct sources of uncertainty we benefit from simulation-based evidence to point at two most suitable identification schemes. We detect a unidirectional effect of financial uncertainty on real economic activity and mutual causality between macroeconomic uncertainty and business cycles.