Dear Dynare community,
I have a question about the timing of a monetary policy shock in standard Dynare models versus the model in Christiano, Eichenbaum and Trabandt (Econometrica 2016). In that study, the authors estimate their model using an IRF-matching methodology. They first evaluate a VAR model with an identification scheme that assumes that the monetary policy shock only affects the Federal funds rate (R) in the period of the shock. Thus, in period zero, none of the other variables respond except R. Their model therefore assumes that agents take their decisions before observing the MP shock in period t. In their MOD file, therefore, all the (+1) variables are redefined using model-local variables using the expectation operator as follows:
X_tp1 = EXPECTATION(-1)(X(+1))
The nominal interest rate is likewise redefined such that the period t interest rate R is based on t-1 expectations in the agents’ decision rule.
R_tp1 = EXPECTATION(-1)(R)
But in the Taylor rule, R is current. Thus, in principle, the Taylor rule responds on impact to the shock, but in the rest of the model, variable response lags by one period.
My understanding of standard Dynare timing is that shocks are observed at the start of time zero, and all variables respond in time zero. My questions therefore are as follows:
- Are the IRFs resulting from the CET approach equivalent to the IRFs of a model with standard timing, but with an additional time zero period added to every variable except R?
- I have a model that works perfectly under standard Dynare timing. However, when I make the CET style modifications (i.e., shifting expectations one period behind except in the Taylor rule), I get a BK violation “Rank Failure” indicating a timing issue. Does this imply that the CET model does not work under standard timing?
Hope my questions make sense! Thank you in advance!