Nonlinear Model and Observation Equation

[aaronsun]

Hi all,

I am estimating a nonlinear DSGE model. When I take the variables to the data, I match the detrended model variables to detrended data (not demeaned). However, the steady state values of the observables implied by the model are not consistent with the corresponding sample means.

I read the paper "A Guide to Specifying Observation Equations for the Estimation of DSGE Models" by Johannes Pfeifer, especially on Section 1.4, where the author provides two ways of taking models to data. The first way is to write down a stationary model and enter data made stationary. This is exactly what I am doing. Following this way, am I supposed to obtain a steady state value for any observable that is consistent with the sample mean?

Thanks a lot in advance.

Aaron

[jpfeifer]

No, in that case you are not supposed to work with demeaned data. The steady state does not matter for the data related work except for variable definitions. Essentially, you use undemeaned data and match this to undemeaned model variables (for this you might have to add the steady state to the model variables). See "Listing 10: Observation Equation in Nonlinear Model for Log-Linearization using first differences for non-stationary variables". That is the way to go.