This paper examines the usefulness of applying structural vector autor
egressions (SVARs) to the study of business cycles. The SVAR approach
aims to provide robust inferences, by imposing only weak theoretical r
estrictions. We illustrate that the robustness of conclusions drawn fr
om SVAR exercises are questionable. We also examine the problem of ide
ntification failure in structural VAR models. (C) 1998 Elsevier Scienc
e S.A.