I am trying to program the investment delay described in Bernanke Gertler Gilchrist what '99 into their financial accelerator model (or one like it anyway). In the paper, they claim that it is sufficient to change q=psi*(i-k(-1)) to

E_t[q(+1)-psi*(i(+1)-k)]=0. Of course when i try that in Dynare, the rank condition is no longer satisfied. I tried stepping them forward and then using a past expectation too:

dummy=q(+1)-psi*(i(+1)-k)

dummy(-1) = 0

In this case, Dynare tells me that I have 8 forward looking variables and 8 eigenvalues grater than one in absolute value but that the rank condition is NOT satisfied.

Anybody have any idea what is going on and how to get past it? Thanks!

Cheers

Rob

the simplified code for bgg is attached.

bggwork.mod (1.7 KB)