Explosive roots=forward-looking variables but still indeterminacy (rank failure)

ZBCPA · October 11, 2020, 2:07pm

Dear Johannes,
Could I ask when I run simulation with Dynare, why when the explosive roots=forward-looking variables in number already, it is still indeterminacy (as shown below)? And how to deal with this issue generally?
Many thanks in advance！
Cheers
Huan

"There are 5 eigenvalue(s) larger than 1 in modulus
for 5 forward-looking variable(s)

The rank condition ISN’T verified!

MODEL_DIAGNOSTICS: No obvious problems with this mod-file were detected.
Error using print_info (line 32)
Blanchard & Kahn conditions are not satisfied: indeterminacy due to rank failure."

wupeifan · October 12, 2020, 3:21am

Probably because your model is specified incorrectly, as not all of the equations are linearly independent.

jpfeifer · October 12, 2020, 11:46am

Exactly. Try running model_diagnostics. Maybe it shows something.

ZBCPA · October 13, 2020, 9:18am

Thanks, I have run it but it shows “MODEL_DIAGNOSTICS: No obvious problems with this mod-file were detected.”…

jpfeifer · October 13, 2020, 12:51pm

Without the mod-file it is impossible to say anything more.

ZBCPA · October 14, 2020, 4:18pm

ZBCPA:

Sorry， I have uploaded the mod file. Could you please take a look at it on your convenience?
I use Bianchi and Nicolo’s method to make one degree of indeterminacy model to be determinant… so I have equation “z=1.5*z(-1)+sunspot-(y-expectation(-1)(y))” and set interest rate responds to inflation psipi <1. This method sucessfully works for indeterminacy model (already tried) like the one in Lubik and Shorfield (2004) when psipi<1. The attached model can also have two degree of indeterminacy when I keep psipi<1 and change some other calibration, under two degree of indetermincy with Bianchi and Nicolo’s method I can make the model determinant and sucessfully make it run.
But now , at one degree of indeterminacy, even I have make the explosive roots = forward-looking variables, dynare shows rank failure…

model_test.mod (1.4 KB)

jpfeifer · October 16, 2020, 7:58am

Are you sure that putting the expectational error on y (which is always determinate) is correct? Shouldn’t it be

z=1.5*z(-1)+sunspot-(pi-expectation(-1)(pi));

ZBCPA · October 20, 2020, 4:14pm

Thanks. The problem may still be some equations linearly independent. I am wondering if it can be called ‘source’ in the case that ‘Blanchard & Kahn conditions are not satisfied: indeterminacy due to rank failure’.