Page 1 of 1

Indeterminacy of estimated results

PostPosted: Sun Jul 29, 2012 2:05 pm
by m_53cek
Hello,

I estimated a small-open economy DSGE model and estimate it on data for Slovenia. When I try to simulate the model using estimated parameter values as calibrated values in stock_simul command, I get the message that I have indeterminacy problem. Is it possible that the estimation results in such parameter values that the Blanchard-Kahn conditions are not varified?

Thank you and kind regards,
Matic

Re: Indeterminacy of estimated results

PostPosted: Mon Jul 30, 2012 2:54 pm
by jpfeifer
That should not be possible. Could you please post the mod-file and the parameter values from estimation you calibrate.

Re: Indeterminacy of estimated results

PostPosted: Mon Jul 30, 2012 4:39 pm
by m_53cek
Hey,

I posted the mod file, I used for estimation. The parameter values I get are:

beta = 0.99;
eta = 1.5401;
sigma = 2.2648;
gamma = 0.2980;
a = 0.3996;
thetap = 0.6686;
alphae = 0.7682;
alphay = 0.8211;
kappah = 0.2375;
kappaw = 0;
kappay = 0;


If I set this parameter values in the parameters section and use the: stoch_simul y pih omegah pi gap delta; command, I get this message:


??? Error using ==> print_info at 42
Blanchard Kahn conditions are not satisfied: indeterminacy

Thank you and kind regards,
Matic

Re: Indeterminacy of estimated results

PostPosted: Mon Jul 30, 2012 7:21 pm
by jpfeifer
The reason is that your model features a unit root. One eigenvalue is exactly 1. If you want this eigenvalue to count as an unstable one as in estimation (i.e. you think your model is correct with the unit root), where lik_init=1 by default sets qz_criterium to 1-1e6, put
Code: Select all
options_.qz_criterium=1-1e-6;

before stoch_simul