TANK Model: Residuals of static equations

I am running a calibrated TANK model based on the lecture notes of Valerio Nispi Landi/Gali et al (2007). Despite huge residuals in the static equations I get a (reasonable) solution to the problem, I wonder why. Moreover, I am trying to get the IRFs as percentage deviations. I am aware of the options on how to get them, however, e.g. the loglinear option (see TANK5_revised.mod) and the use of logged auxiliary variables (see TANK7_revised.mod) yield different results, again, I wonder why and where I am going wrong. Any help is very much appreciated.

TANK7_revised.mod (7.2 KB)
TANK5_revised.mod (6.8 KB)

  1. If you just provide sensible starting values given labor equal to 1/3, then residuals are expected if the true steady state is 0.52.
  2. How are the results different?

Thanks for the hint with the steady states. I am not sure as to why these deviations come along. I will have to check that.

I induce a negative productivity shock. Wages (all three wage variables) respond in an oscillating manner when using logged auxiliary variables (in the first period wages exhibit a negative response, then jump to a positive value in period two before, from the third period on, the convergence behavior is as expected and values are all negative again). That is not the case when using the loglinear option (i.e. the IRFs are all negative and converge as expected). The results are similar otherwise.

model_diagnostics tells you there is a singularity issue. Maybe that explains the slight differences.