unit root variable in the model

[jill1009]

Hi,
I'm estimating a DSGE model using Dynare. When I use "steady; check;" command, I find that there is an eigenvalue equal to 1. Does that mean one of the endogenous variables is unit root? How to find that variable? What should I deal with it? Is there any consequence if there is a unit root variable in the model?
Thank you,
Ji

[SébastienVillemot]

If you have a unit eigenvalue, then there is indeed a unit-root in your model. You can see which variables are explosive by looking at the theoretical simulated moments: the unit-root variables have infinite variance.

Unit roots have a lot of consequences: there is no unique steady state, stochastic simulations may diverge, estimation needs to be done using a specialized (diffuse) Kalman filter.

All these problems can be circumvented, but you need to have a good understanding of these issues. Search the forum and read books/papers.

If unsure about the consequences, it is better to stationarize your model by removing the unit root.

[jill1009]

Thank you! I checked the theoretical moments, none of the variances is infinite. what's the matter?

[SébastienVillemot]

Hard to tell. Please post your MOD file.