IV-GMM, replication for Gali & Gertler (1999)

Hi there,

I am new to Dynare, now I want to replicate the difference equation specification of the hybrid NKPC by using GMM. z_t in the equation is a vector of variables dated t and earlier, it contains 6 variables with 4 lags. I don’t know how to set the instrument variables in this model and how to do this replication. I also upload what I did currently, the results show decomposition failed, I don’t know how to correct this.

Any suggestion will be useful. Thanks in advance.

Warm regards,
Gali_GMM.mod (1.2 KB)

I am pretty sure that won’t work in Dynare. You are trying to estimate a single equation using GMM instead of a full model.
The program was not designed to do this. In Eviews or Gretl (or Stata for that matter), this should be straighforward.

@wmutschl Do you agree?

Thanks for your reply.
Can you please be more specific? If I want to run it successfully in Dynare, what should I do? Does that means I need more models in the “model section”? My goal is to estimate the parameters: phi, omega, theta, beta, lambda, gamma_f and gamma_b.
I am really appreciate for your help.

The problem is that you are using moment conditions involving instruments that are not endogenous variables of the model. That will not work. Dynare is not intended to be a general purpose estimation package.

Hi professor, I changed my code, but now it shows that “The generalized Schur (QZ) decomposition failed”, could you please help me with that?Gali_GMM1.mod (960 Bytes)

Thanks for your time.

Again, Dynare is not an GMM-IV econometrics package.

Yes I agree, you are trying to do what is known in the literature as “single-equation GMM estimation” in contrast to the GMM estimation in Dynare which is based on the reduced-form perturbation solution of all model equations.

So for your goal, you should just use a standard GMM toolbox and not Dynare, GitHub - tholden/gmmtbx: An updated version of Kostas N. Kyriakoulis's GMM Toolbox for MATLAB or https://cran.r-project.org/web/packages/gmm/vignettes/gmm_with_R.pdf come to my mind. There are many replication codes you can use when you google “GMM NKPC” or “GMM New-Keynesian Phillips curve”.