Hi, owing to your support, I have made some progress in reproducing Lubik and Schorfheide (2007).
The reason is that TFP has a unit root, but the model is entered in detrended form. To get the permanent IRFs to TFP shock, you need to reconstruct the underlying non-stationary series. An example for this can be found in the AguiarGopinath2007.mod on at DSGE_mod/Aguiar_Gopinath_2007.mod at master · JohannesPfeifer/DSGE_mod · GitHub
Thank you very much.
I will try with reference to this mod file !
There are three things in my modified code compared to original code (detrended form).
model(linear);
...
...
z = a - a(-1) ; //z_t is growth rate of technology (A)
yy = y + a ; // y_t = Y_t/A_t
However, the response from dynare is that “unstable”.
I searched a lot to solve this problem and concluded that I should use trend_var command or .
Unfortunately, these do not work as well.
And finally, dynare works when I add diffuse_filter command in estimation()
and I got the IRF that I wanted.
But I am not sure whether I solved well.
Please give me more comments.
-Jerry
You are introducing a unit root in a. To deal with this non-stationarity in estimation, you need the diffuse filter or lik_init=2.