Dear Dynare Team,
I would like to construct a time series of (ideally identified) shocks. I was wondering if a Bayesian estimation of a DSGE model would be appropriate? Therefore, I would like to run a bayesian estimation and take the shock sizes from the estimation. Latter on I would like to use the shock series and construct empirical irfs in a VAR or local projection. Is this a common approach to check model and empirical irfs? Or is irf matching a more common approach?
Thank you for your help!
What exactly are you aiming at? The Kalman smoother can be used to extract identified structural shocks. But why would use want those for an empirical analysis? After all, the shocks are estimated based on a model structure that already provides IRFs.
I want to check whether my DSGE model irfs coincide with empirical irfs. The hard part is the construction of an empirical time series for this shock, so I thought similar to the uncertainty shocks in your paper “Uncertainty-driven Business Cycles: Assessing the Markup Channel”, I could take the series from the estimation.
But the shocks in that paper were estimated separately using a particle filter without estimating the rest of the model.