mode compute=6

[ZBCPA]

Dear Johannes,

For a very simple NKM with the Taylor rule Rt=psi* Piet + shock1, we know psi >1 then the NKM is determinate;while psi<1 the model is indeterminate.
Then I can add an additional equation to the system-----wt=alpha*wt-1+ shock2; when alpha >1 , we know this equation is explosive.(Francesco Bianchi and Giovanni Nicolo 2016)
Therefore, the augmented model is determinate and can be solved or estimated Either when psi>1 and alpha<1 or psi<1 and alpha>1

Then I use data which has been proved supporting psi>1(true global mode) as observed data to estimate the model, but I set initial value : alpha=1.0001 and psi<1 . Using

Code: Select all

mode_compute=6

, I find that when initial value of psi is set very close to 1, like 0.9, the mode can jump across the boundary----I get mode of psi >1 and mode of alpha<1 , and this is just what I hope to see since mode of psi is already close to its true value;
But if initial value of psi is set a little far from boundary , like 0.7, then the mode finding process can be stuck and mode of psi just stays <1(and of course alpha stays>1). Could I ask if Dynare has any option for mode_compute=6 to solve this issue?

Many thanks in advance,
Huan

[jpfeifer]

No. This is a general problem with mode-finders. They always depend on the initial starting values. The problem is somewhat less pronounced for "global" mode-finders like 6 or 9.