DSGE model highly sensitive to parameter changes

Stochastic simulations
1

Hello,

I am working on a DSGE model in Dynare. The model currently runs correctly with my baseline calibration and steady-state values. However, the model is extremely sensitive: even very small changes in some parameters or steady-state values cause the model to fail (steady state does not converge, residuals explode, or Dynare returns errors).

For example, changing a parameter only slightly can immediately prevent convergence, although the baseline model works. This makes calibration and robustness analysis very difficult.

I would like to understand what this usually indicates in DSGE models.

Also, what are the best ways to diagnose and reduce this sensitivity?

Thank you very much for your help.

Dalal

2

Can you provide more details or an example showing the issue? How do you define “slightly”? From what you describe, it sounds the issue is with numerical steady state finding. That can indeed be very sensitive to starting values. Have you tried computing the steady state analytically?