I’m trying to estimate a model with IRS production and a deterministic growth trend using non-stationary data. As such, I’ve separated the equations linking model variables to data variables into [dynamic] and [static] equations. When the steady state is calculated at the initial parameter values, I get the error message “index exceeds dimensions” in dsge_likelihood at 299. I’m unsure of what I need to manipulate in my code to avoid this error. Thanks! irsdata11.m (6.6 KB) fgIRS.mod (1.61 KB)

Replace the Dynare file for 4.4.2 with the attached one. The problem is that your steady_state_model block does not solve the dynamic model. In the dynamic model the steady for the c_obs is g while it is 1 on the static model. dsge_likelihood.m (31.4 KB)

Thank you! I’m not sure I understand quite how to deal with this though, since the observed variables are non stationary. From reading around, the sense I got was that I needed to define the observables as dummy variables in the steady state, but apparently I’m missing something about how to do this correctly. Is it possible to do this or do I need to make the data stationary?

I am not following. Does you model feature a unit root? Data usually is non-stationary and you have to link the data to the model. From what you are doing, it seems to me as if you are essentially trying to use growth rates (log first differences) for your data. That can easily be done, see Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models” sites.google.com/site/pfeiferecon/Pfeifer_2013_Observation_Equations.pdf