Imaginary eigenvalues and Blanchard-Kahn condition

[szms]

Hello!

I have two questions, please answer them

- 1: what does it mean when the imaginary part of the eigenvalue is not equal to zero? Is that a problem?
- 2: I have for example 5 forward-looking variables and just 4 eigenvalues larger than 1, then I change the timing of an equation by decreasing it by 1, and then there are 4 forward-looking variables and 4 eigenvalues larger than 1, so the rank condition is now verified. Is that an approppriate procedure?
For example:
c0=1/(beta*(1+rk(+1)-delta)/c1(+1));
c0(-1)=1/(beta*(1+rk-delta)/c1);

Thanks for Your help!!!

[jpfeifer]

1) As we are talking about generalized eigenvalues, this is not a problem - unless you get oscillating IRFs.
2) No, this is not appropriate. There is one unique correct timing and you cannot arbitarily shift timings.