Observable variables in a DSGE model

Estimation ML/Bayesian estimation
21

Yes, of course, unless you demean your empirical growth rates.

22

Thanks so much professor for your helpful comments. The other question is that when we multiply the variables by 100 and transform them into the growth rates you indicated that we should multiply standard deviation of shocks by 100.

My question is that we enter standard deviation of shocks in Dynare in DSGE estimation we have two different blocks namely:

shocks
stderr u_a , 0.01
end

and the other block is :

estimated_params
stderr u_a , inv_gamma_pdf , 0.01 , 0.05
end

u_a is innovation of TFP shock

Is it necessary to multiply the standard deviation of model shocks by 100 in theses two aforementioned blocks when we estimate a DSGE model ??

23

The first one is usually for calibration and will be superseded by the second one for estimation. Thus, strictly speaking only the second one needs to be adjusted. But I would always recommend to be consistent within a file. Otherwise, you yourself might get confused.

24

Thanks professor.

Therefore only the standard deviation in the second block should be adjusted.

25

Hi Prof,
In estimated_params block of Smets2007.mod file, the average growth rate γ is assigned with the value of 0.3982 for INITVAL (ctrend,0.3982,0.1,0.8,NORMAL_PDF,0.4,0.10;). Could you please explain where this number comes from? I have the similar concerns with steady state inflation rate (constepinf) and steady state interest rate (conster).
Thanks.

26

It’s just a starting value the authors picked. I don’t know the rationale behind assigning that number. In principle, this should not matter.