Testing for Spurious Dynamics in Structural Models with Applications to Monetary Policy

We propose a universal and straightforward test for validating assumptions in the structural models. Structural models impose a causal structure, take data as an input, and then produce exact structural parameters. We simulate the new data while breaking the original causal structure. We then feed the model the simulated data and then see whether it produces different results. If its conclusions are the same, then the models' implications are not sensitive to the underlying data, and the model fails the test. We then apply our test to the models analyzing monetary policy. We find out that simple SVARs successfully pass the test and can be used to identify monetary policy effects. On the other hand, DSGE models estimated via full-information methods such as Smets and Wouters (2007) fail the test and potentially force their conclusions on the data.