Dear dynare community,

I have tried to replicate the model in Adolfson et al. (2007) for a developing economy. I managed to estimate it, but it seems I have a couple of issues in my results:

- The model contains a stochastic trend as in the original paper, but the model block is made up of log-linearized equations (variables are in log deviation from steady state and thus stationary. observable variables are in growth rates and thus stationary as well). Do I have to add an option with diffuse_filter in the estimation command?
- Smoothed values of model variables do not seem to make much sense, since some of them have a little bit unrealistic values like -5 or 0.6 (which cannot really be possible since the variables are in log deviations from steady state. In addition, forecasts for observed variables are unrealistic as well). What could the problem be? Perhaps observation equations? But I think I have specified them correctly.
- If we turn on dynare_sensitivity command, then it says that only 0.1% of draws from the priors give stable saddle path solutions whereas the rest give either explosive or indeterministic solutions. Does it really matter as soon as I get some results from the Bayesian estimation (which in fact give me unrealistic values for smoothed variables, and hence for historical decomposition)?

Best,

Nurdaulet

P.S. I have observed variables such as ‘‘infl’’ and ‘‘inflru’’ appearing as gross inflation rates in the data whereas other price inflation rates like ‘‘gdpdef’’ and ‘‘invdef’’ appear as net inflation rates.

EA_SR07_rep.mod (29.8 KB) EA_SR07_rep_mode.mat (17.0 KB) SR_07.xlsx (23.1 KB)