Do Bivariate SVAR Models with Long-Run Identifying Restrictions Yield Reliable Results? An Investigation into the Case of Germany

Bivariate SVAR models employing long-run identifying restrictions are popular tools to investigate the source of business cycle fluctuations. Their advantage is the simplicity in use and interpretation. However, their low dimension may also lead to a failure of the identification procedure, with the result that the identified shocks are a mixture of the 'true' shocks. To investigate this issue, the consistency of results from different bivariate SVAR models estimated for German data is evaluated using the FAUST and LEEPER (1997) test procedure. The principal result is that these models do not allow reliable inference on the sources of output fluctuations.