by kyri82 » Fri Jul 13, 2012 3:26 pm
Theoretical IRFs are derived from the solution of the model, from the policy and transition functions. That is, when the model is solved each variable x is expressed as:
x = g(Y, e)
where Y is the vector of the model's state variables (pre-determined) and shock e. In a linearised model the function g(.) is linear. Thus, given the shock e you get the IRF, which is exact.
What IRFs are you talking about exactly?