Dynare forums

Dear Michel,

I have 2 questions:

1)I am working on a incomplete market IRBC model. I am trying to reproduce the results of Heathcote and Perri (Journal of Monetary Economics 2002, Financial autarchy and international business cycle) for the bond economy.
My code runs but the impulse responses I get, even if they have the same shape of the ones they get, have a different order of magnitude. I use their same calibration. When I run the same code with Matlab I get their same results. Moreover the steady states I get with Dynare are different from the ones I get with matlab.
I think, even if I am not sure, that the problem could be with the timing convention of the variables.

2)How can I replicate the model 100 times and get the average simulated moments?

I would really appreciate your help,
thanks,
Corto Maltese 79

1) By default stoch_simul computes second order approximation. Is it what you want?
2) timing shouldn't have an influence on the steady state. Can you also post your Matlab program that gives the same results as the authors?
3) why do you want to average simulation path when you can have theoretical moments (without the periods option)?

Best

Michel

Dear Michel,
first of all thanks a lot for your fast reply.

1) By default stoch_simul computes second order approximation. Is it what you want?

Yes, that is what i want.

2) timing shouldn't have an influence on the steady state. Can you also post your Matlab program that gives the same results as the authors?

You are right. Sorry I made a mistake. Steady states are not influenced by timing. When I checked again my matlab code I realized that the specification of the Armington aggregator that I used there was slightly different.
Thus now I am sending you the Matlab code which reproduces the results of the paper and a new Dynare code adjusted to the new specification. However, again, the irfs produced by Dynare have the same shape and are very similar to the ones produced by Matlab but mainly the responses of investment, imports and exports have a different magnitude. I think it could really be some mistake in the timing specification.
In the attachment there is my Matlab code and the adjusted Dynare code.
All of them are inside the zipped file.

3) why do you want to average simulation path when you can have theoretical moments (without the periods option)?

I am not sure I understand what you mean when you say theoretical moments.
However I am trying to get the averages of the moments of 100 simulations to compare them with the ones in the Heathcote and Perri's paper and also because then I will modify the original model and i will need to compare the simulated averages of the standard deviations with the actual statistics.

Thank you very much again,
Corto Maltese 79

I guess it's too late, but you can replicate your IRF by using exp('variable') instead of 'variable' in all equations of your model part (except for the exogenous process equation). And the ss equations get a log.
Check out these notes:
http://www3.nd.edu/~esims1/using_dynare.pdf