Explosive shock behavior

[klg123]

Hello all,

I am in the midst of estimating my first DSGE model and am a new user of DYNARE. I am in the early stages of writing my code and was wondering if someone might take a look at my code to help identify the problematic areas. The IRFs seem to be exhibiting explosive growth for several variables. My model is written in log-linear form already. I suspect there is some issue with one of my equations (perhaps e_star / e) but am unsure and would greatly appreciate some advice on how to move forward.

Many thanks!

[jpfeifer]

Your model has three unit roots. The variables you are referring to are not explosive, but your shocks have permanent effects. You are correct that the equation

% Foreign nominal exchange rate
e_star - e_star(-1) = e(-1) - e;

is the source of the problem. In many models the level of the nominal exchange rates (or price levels in general) is indeterminate. Only inflation or the growth rate of those variables can be determined.

[klg123]

Thank you for the response! What, then, is this a matter of modifying? I know it is not simply removing the equation. Must I deal with other instances of e and e_star in some way or remove them from the model? I have characterized my monetary policy rule to respond to changes in the nominal exchange rate, so I am unsure how to get around these errors.

[klg123]

I have tried to re-write the model without e or e_star and it seems like epsilon_pif and epsilon_pih are the permanent shocks. Any suggestions?

[jpfeifer]

What are those shocks? It might be that they simply have permanent effects. Without knowledge of the model, it is impossible to tell whether permanent effects is a bug or a feature.

[klg123]

Those are shocks to domestic inflation and foreign inflation. The model is based of Liu (2006) and Justiniano and Preston (2010)...I believe it is a bug.

[klg123]

My mistake -- one of my equations was mistyped !

[donihue]

For future reference, if you replace (e - e(-1)) in your model by "dele", and use this name for the old variable "e", your first version will work.
Regards
Donihue