rank condition is verified but Blanchard Kahn conditions not

[shuixing]

dear professor peifer
I run my mod file and the first time the rank condition is not verified but the program can run and get results, then I modify the mod file, this time the order condition is verified but the program can not run and get a error message: Blanchard Kahn conditions are not satisfied: no stable equilibrium.

that seems weird because I thought once rank condition is verified then BK condition must be hold. what seems to be the problem.

looking forward to hearing your reply

best

[jpfeifer]

Almost any change in the model can affect the determinacy conditions. Therefore, what you describe is not that surprising.

[shuixing]

Almost any change in the model can affect the determinacy conditions. Therefore, what you describe is not that surprising.

thanks professor Pfeifer, so rank condition and Blanchard Kahn condition are quite different? when Blanchard Kahn condition holds while rank condition does not, the estimation results is not realiable?

[jpfeifer]

Dynare does not strictly distinguish the two. Please describe exactly what you are doing. Now you are talking about estimation. Depending on the model, there you might encounter issues with unit roots that affect determinacy. In this case, you need the

Code: Select all

diffuse_filter

option

[shuixing]

Dynare does not strictly distinguish the two. Please describe exactly what you are doing. Now you are talking about estimation. Depending on the model, there you might encounter issues with unit roots that affect determinacy. In this case, you need the

Code: Select all

diffuse_filter

option

thanks professor, I am trying to estimate the model, and the first time rank condition is not verified but Blanchard Kahn conditions is hold, and I can still get a result choose mode_compute=6, but I think the results may be misleading so I modify the timing of one equation, then it shows rank condition is verified but Blanchard Kahn conditions not.

[jpfeifer]

If you can get results during estimation, this suggests that the problem is caused by the parameterization of the model. For some parameter sets the BK conditions are satisfied, for others not.

[shuixing]

If you can get results during estimation, this suggests that the problem is caused by the parameterization of the model. For some parameter sets the BK conditions are satisfied, for others not.

OK， thanks professor Peifer, now I see what's happening~