Weird IRFs for eps_pco shock

Hi everyone!
I have the following issue here:

I’m interested in the irf (for an eps_pco shock) for the variable wage_prem (wage premium) which is defined as wH/wL, that is, the ratio between high (H) skilled wage rate and low (L) skilled wage rate. The reasonable thing that I was expecting to obtain is that

IRF.wage_prem = IRF.wH/IRF.wL

but this is not the case. In fact, Dynare spits that while the IRF for an eps_pco shock for wH and wL increase, wH increasing more than wL, the IRF for wage_prem goes down.

I can’t see where this is coming.

I appreciate any help!
pcu3_sim2.mod (3.9 KB)

IRF.wage_prem = IRF.wH/IRF.wL

would hold in nonlinear model. But you will have a linearized version hold, i.e. the difference between the two.

Thanks, Johannes!
Though, i’m sure that my model was non-linear, since I didn’t linearize the equations, or am I getting something in the wrong way?

stoch_simul uses a polynomial approximation. a/b is not a polynomial.