Hello Dynare Community,

I’m trying to replicate Ferroni’s 2011 paper “Trend agnostic one-step estimation of DSGE models”, and I would like to know how would you implement the idea of the paper in Dynare.

The idea, to sum up, is the following: Instead of filtering the observables first and later doing the estimation, Ferroni proposes the observables being filtered inside the model allowing it to model freely the trend specification and estimating at the same time the structural parameters and the trend parameters.

However, in Section 2.2.1 (which I upload as an attachment) he uses a deterministic trend -t- that I don’t know how to use in Dynare, or how this variable t would have a steady state.

I would really appreciate your help. Sincerely,

Jose Niño

Ferroni_paper.pdf (104 KB)

Dear Jose,

I haven’t yet found the time to replicate the paper (still on my admittedly long list). If you use the unstable version, you would be using the

observation_trends

command. See e.g. github.com/DynareTeam/dynare/blob/master/tests/observation_trends_and_prefiltering/Trend_model_no_prefilter_common.inc

Thanks Dr. Pfeifer for your answer.

To make clear the link between the observation_trends command that you added ( github.com/DynareTeam/dynare/blob/master/tests/observation_trends_and_prefiltering/Trend_model_no_prefilter_common.inc ) and Ferroni’s paper, I have another question for you.

The g_y and g_p parameters you have in observation_trends are the elements of the B matrix in the Linear-Trend-DSGE setup in Ferroni? The observation_trends command automatically adds the linear trend t and multiplies it for the B matrix?

The const_y and const_p are the elements of the A matrix in Ferroni’s setup?

Thanks again for your help!

Yes, that is correct. The constant and trend coefficients represent the elements of A and B in his equation (3)