Dear Johannes, could you explain in which sense 100*0.0018916 captures the mean growth rate in the data?
Thank you very much
The data grows by 0.0018916, i.e .0.2 percent per quarter, which is multiplied by 100 as everything is scaled by 100.
1 Like
Hi!
I am trying to replicate GNSS too and actually I have a doubt regarding the observation equations. I have treated the data in the same way of GNSS therefore by detrending the real variables with the HP filter and by expressing interest rates as absolute deviations from the sample mean. For what concern the observation equations I am expressing them in growth rate terms too. In order to use the observation equations in this form e.g. 100*(0.0007868 + I - I(-1)) scaled up by 100 should I scale also my data by 100 before running the estimation right?
Thank you
I don’t really understand what you are trying to do.
- You are not supposed to use two-sided HP-filtered data for estimation. It’s something GNSS should not have done.
- But if you just want to replicate their work, then go for it. But if your data is HP-filtered, it will be mean 0 and you should not pretend to have growth rates. That does not make sense.
- If you scale your data by 100, you need to do so as well for the shock standard deviations. If you want to use the prior for the shock standard deviations given in their paper, then you cannot multiply the data by 100 as this will be inconsistent.
Dear Pfeifer thank you for your reply,
I am bit confused by the replication mod file of GNSS that you can find on mmb.
First the observation equations are expressed in a way that I find strange for a non linear model and considering how they treated the data (i.e. log deviations from HP-filter trend and in absolute deviations from sample mean).
in the original mod file observation equations are written as :
interestPol = 400*exp(r_ib);
interestH = 400*exp(r_bh);
interestF = 400*exp(r_be);
inflation = pie*100;
loansH = BH*100;
loansF = BE*100;
output = Y1*100;
consumption = C*100;
investment = I*100;
deposits = D*100;
interestDep = 400*(exp(r_d));
bankcapital = 100*K_b;
Shouldn’t be e.g. consumption = 100*(C-steady_state(C))?
Second sorry again for the trivial question again, if in the observation equation you multiply everything by 100 to have them in percentage terms, should also the detrended/demeaned data multiplied by 100 before running the estimation to match the observation equations?
Third, if this “100*(C-steady_state(C))” is the correct way of specifying the observation equations in case the steady state has been estimated by the authors it is possible to use the estimated steady state instead of the dynare operator in the equation?
e.g. interestPol = 400*(r_ib-r_ib_ss).
Thank you again, attached you can find the transformed data, the graphs from them and those of GNSS, the original mod file and my mod file for the estimation.
myestimation.mod (25.9 KB)
EA_GNSS10_rep.mod (24.9 KB)
correct_data.mat (5.1 KB)
I have no clue what is going on in that file. The observation equations make no sense and parameter dependence is not correctly handled. You may have to ask the authors that the MMB forum.
Thank you for your answer. I have written directly to the authors asking them if they could share their original code.
Best