Dynare forums

Hi
I am working on with a two sector (consumption and durable goods, Housing) NK model with financial frictions. I have dynare 4.2.0 and 4.2.2 on my desktop at home while I am working with 4.2.4 at my university office desktop. I have observed a strange problem while working at my relatively big model baseline model (56 equations) on these versions. Yesterday I ran a model with 4.2.2 and got the following message "There are 14 eigenvalue(s) larger than 1 in modulus for 14 forward-looking variable(s) The rank condition is verified" but with no impulse responses and moments output. While with 4.2.0 version, I am getting the impulse responses with BK and rank conditions met. However, the problem is that I am getting "NAN" against moments and also getting the following warning.

"There are 14 eigenvalue(s) larger than 1 in modulus
for 14 forward-looking variable(s)
The rank condition is verified.
Warning: Matrix is close to singular or badly scaled.
Results may be inaccurate. RCOND = 7.060769e-033.
> In dr1 at 433
In resol at 145
In stoch_simul at 66
In DSGE_FAplus6312 at 606
In dynare at 132

However, with 4.2.4 the same model has returned me the following error message "There are 14 eigenvalue(s) larger than 1 in modulus for 14 forward-looking variable(s). The rank condition is verified. ??? Error using ==> print_info at 45 Blanchard Kahn conditions are not satisfied: indeterminacy due to rank failure". model_diagnostics(M_,options_,oo_) is not giving me any help either. I have consulted the various relevant posts on rank condition and BK condition and in light of these post I have tried to alter my calibration values but the aforementioned problem still persists. IRFs from 4.2.0 are making some sense but I am worried about the above warning message. My main concern is that can I carry out the main extension in the baseline model in the presence of the aforementioned warning. Can someone suggest me a solution how to get rid of this warning message that would help me in my main extended model (with additional 10 equations). Suggestions/tips in this regards are welcome.

Cheers
Zulfiqar

It seems that your model is rank-deficient. There usually is one redundant equation, often due to Walras Law, and a corresponding missing equation. For larger models, detecting such a singularity is numerically tricky. Sometimes the rank condition seems to be satisfied, but the inversion fails. Hence, the singularity warning. If you invert a singular matrix, this sometimes yields NaNs such that Dynare will not show moments or IRFs. If the problem persists with different parametrizations, this often suggests a serious problem with the underlying model.

Thanks, jpfeifer for your response. Yep, I have re-worked my model in the light of your suggestions/observations. Then I un-commented one equation of the model, now my model is generating IRFs and some of the theoretical moments with NAN for others. If I use Drop and period option, then I am getting simulated moments for all endogenous variables. I have compared the IRFs with "stoch_simul(order=1,irf=40)" and "stoch_simul(order=1,irf=40, periods=2000, drop=200)" and these are identical. Can you please let me know why this is happening? Is this still a serious problem? What can I do to get theoretical moments? Another question is about the nominal shock of my model. My model works well with 3 out of 4 main shocks (shock to networth, productivity shock and capital quality shock) but with nominal shock, I am getting the following message "All endogenous are constant or non-stationary, not displaying correlations and auto-correlations". I read the various post in the forum and consult the user manual for this which talked about stationarity issue. I have log-linearized my model around steady state by hand and input the model in mod file as "model (linear)". Can I still expect stationarity problem with log-linearization model around the steady state? What should I do to get rid of this problem?