Estimation error: in computing likelihood for initial parameter

Estimation ML/Bayesian estimation
1

Hi. I’m trying to estimate the Gali-Monacelli (2005) open small economy based on Pfeifer’s code, but I’m receiving this message:

Error in computing likelihood for initial parameter values

ESTIMATION_CHECKS: There was an error in computing the likelihood for initial parameter values.
ESTIMATION_CHECKS: If this is not a problem with the setting of options (check the error message below),
ESTIMATION_CHECKS: you should try using the calibrated version of the model as starting values. To do
ESTIMATION_CHECKS: this, add an empty estimated_params_init-block with use_calibration option immediately before the estimation
ESTIMATION_CHECKS: command (and after the estimated_params-block so that it does not get overwritten):

Error using print_info (line 32) Blanchard & Kahn conditions are not satisfied: no stable equilibrium.

I’ve already checked the equations and try to used the use_calibration in estimated_params_init, but it didn’t solve it.

Would you help me, please?Gali_orig_Est.zip (14.2 KB)

2

Please see

3

Thank you, professor Pfeifer.
I used the diffuse_filter and now it is working. I also read some other topics in the forum about it but I have some more questions:

1 - If I estimate this a model with an unit root, there are infinite steady states. Which one Dynare will give me as a solution depends on the initial values in estimated_params_init?
2 - Does the inexistence of a unit root assure an unique steady state?
3 - With a finite number of steady state (say 2), will Dynare give me one of them or both?
4 - In that case above, will the impulse response function be relative to the same steady state? Or is it possible to see same transition path between two differents steady states (maybe using another feature of Dynare)?

Finally, if you have some literature about these questions I will really appreciate it.
Thank you again.

4
  1. If there are infinitely many steady states, you should select the correct one using a steady_state_model-block. Note that the estimated parameters have nothing to do with the issue. For any given parameter sets, there is a continuum of steady states.
  2. No, it does not. Take the standard Solow model. It has two steady states. The normal one we always consider and the trivial one where everything is zero.
  3. If you let Dynare solve for the steady state numerically, Dynare’s solver will (hopefully) converge to one steady state. As a local solver is used, it’s typically the nearest one. But solving can also fail because the unit root may imply a singular derivative matrix singular and solvers can then fail.
  4. The decision rules will be based on an approximation around the steady state that was selected/found. If you want control over this, again see point 1.
5

That’s perfect. Everything’s clear.
Thank you so much, professor.