Page 1 of 1

Blanchard-Kahn Condition

PostPosted: Wed Jul 03, 2013 7:28 pm
by justus
Dear forum,

I am trying to replicate "Monetary Policy and Asset Price Fluctuations" by Bernanke/Gertler. I do get similar results compared for one (out of four) policy rules, but for the other ones dynare comes up with a warning that the Blanchard-Kahn condition does not hold (although only the response to expected inflation in the policy rule is increases by a factor of 2): There are only 5 eigenvalues above 1 although 7 forward-looking variables are used. The variance of the output gap and inflation are pretty close (output: 2.9954 vs. 2.221 and inflation: 8.0590 vs. 9.676).

BG uses steady state levels in the log-linerized equations. But I am uncertain of how to implement that in the equations. So far I have made guesses / assumptions of which I am not certain that they hold.

Also, BG write in their paper that the bubble is exogenously implemented in period 1 by setting the stock price some percent above the fundamental value. the model determines the following behavior. in period five the bubble is exogenously killed off. Is it possible in Dynare to set the value of variables at a certain time period?

Can anybody help me with these problems? Thank you!

Re: Blanchard-Kahn Condition

PostPosted: Thu Jul 04, 2013 6:51 am
by jpfeifer
Please concentrate on one model and try to replicate this exactly. Looking similar does not suffice.

Your mod-file looks as if you are violating Dynare's timing condition for predetermined stock variables like capitel. This is responsible for the Blanchard-Kahn condition problem.

Regarding the steady state values used in the equations: they are parameters and have to be consistent with the underlying nonlinear model. The only way to compute them is to use pencil and paper.

Specifying a path for state variables is only possible for deterministic simulations. If you need to do deterministic simulations, you have to program it yourself using Dynare's simult_ function.