A Discussion of the Reliability of Results Obtained with Long-Run Identifying Restrictions
In a recent article, Faust and Leeper (1997) discuss reasons why inference from structural VARs identified with long-run restrictions may not be reliable. In this paper, the authors argue that there are reasons to believe that Faust and Leeper's arguments are not devastating in practice. First, simulation exercises suggest that this approach does well when used with data generated with standard macroeconomic models. Second, empirical applications suggest that it gives results that are much more robust than would be implied by Faust and Leeper's main proposition. A reasonable approach would appear to be, therefore, to follow Sims' (1971; 1972) and Dufour's (1997) recommendation and to present robustness checks, allowing readers to judge for themselves what the effects of possible approximation errors might be.