CET (2016) vs. SW (2007): Irf sizes rescaling?

General DSGE Modeling Replication Attempts
1

Hello,

I am simulating the model of CET (2016) and I observed very small responses of the irfs to the shocks (10^-3) compared to the ones of for example SW (2007) which is loglinearized as well.
In the CET model, all variables are in logs and the authors performed an exp() substitution to loglinearize it. The shock sizes are set to 1.

Now I am wondering if I might have to rescale the irfs by 100 to make them comparable?

Thanks!

cet.mod (61.6 KB)

2

The shock sizes in CET are not set to 1. Rather, they estimated the standard deviation. If you look at the code, it shows

sig_mupsi*mupsi_eps/100

Thus, they have a factor 100 less than SW (2007). That is also reflected in the posterior results tables in both papers. To make the IRFS comparable you need to multiply the ones in CET by 100.