Dynare forums

Hi,

I am trying to simulate a model and see how the results compare to a recursive VAR for monetary policy shocks that i have already estimated. There is an issue with the timing, since recursive VARs (like Christiano Eichenbaum and Evans) impose a timing restriction (for example output not reacting contemporaneously to a monetary shock) which is typically not satisfied in standard models. Ravn, Schmitt-Grohe and Uribe (2010) (page 347) propose a method to get around this issue, which essentially involves taking simulated data and running a VAR on it and then computing impulse responses thereoff. I am wondering if there is a simpler way to implement this within dynare, i.e imposing restrictions on contemporaneous behavior of IRFs. Any help would be appreciated.

thanks
nikhil

In the end, you want to consider your model as the data generating process. Thus, the restrictions on IRFs need to come from the model and not from some trick. How problematic the violation of your identification assumption in the model is can be seen using the Monte Carlo experiment suggested. Simulate data from the model and see whether your identification scheme recovers the shocks.