Parameter dependence and steady state model block

Estimation ML/Bayesian estimation
1

Hi! I’m trying to learn how to do Bayesian estimation in dynare.
I have this NK model which is very similar to Sims and Wu (2020, JME) and Karadi and Nakov (2021, JME).
I have two related questions.

  1. As the authors of those papers do, I simulate the model. To do that, I have a separate file where I compute the steady state recursively out of some calibrated parameters. I understand that parameter dependence can be an issue in estimating the model, so I’m wondering if the way I call the steady state file in the mod file allows parameters to be updated. Alternatively, how should I define a steady_state_model block?
  2. I’m having a little issue defining parameter dependence. I understand that the value of some parameters will affect the steady state, but how should I treat this in the model?

These file runs, but I wanna make sure the output is reasonable, which, at the moment, is not

Thanks!
model_estimation.mod (8.2 KB)
data_use_ACGM.mat (12.6 KB)
estimation_params.m (7.9 KB)

2

I would not recommend proceeding in the current way. Neither the implied parameters will be correctly updated nor will the initial values. The latter typically results in falsely discarded parameter draws due poor starting values. AT the same time, it seems straightforward to just move the content of the separate m file to a steady_state_model block. That would get rid of the two problems.

3

Thanks for the tip! As I’m new in Dynare, would you mind showing me/pointing me towards creating a steady_state_model block?
Thanks!

4

An example is

You can sequentially define computations within the block to define the value of the steady states of the endogenous variables and the parameter values conditional on the current parameter draw.