One question about Bayesian estimation with nonstationary observable variables

Dear all,

I have set up one model which cannot be run by Dynare. It tells me that my endogenous variables are non-stationary. However, in my model all the autoregressive coefficients are less than one. Also, dfuller test points out that all the observable variables in my data are stationary. After adding the option “diffuse_fitler” in the estimate command, my model can be estimated successfully but I cannot see the unconditional variance decomposition.

I have transformed my model into the log-linearization form to avoid setting the initial value. This is the first time for me to do research in dynamic macroeconomics, so your guidance is really helpful to me. Thank you very much!

Best wishes.
HC_loglinear.mod (4.2 KB) obs_log_Y.xls (10 KB)

If you add check; after steady, you will see that your model indeed has a unit root. Using model_diagnostics shows the variables involved. There must be a mistake in the linearization.