IRF's: Andreasen et al (2013) toolkit vs Dynare

[yoyo]

Hi everyone,

I have a question regarding the calculation of IRF's when there is time-varying volatility (an uncertainty shock to e.g. an AR(1) process of technology).

When I calculate the impulse response functions using the toolkit of Andreasen, Martin M., Jesús Fernández-Villaverde and Juan F. Rubio-Ramírez (2013): "The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications", the IRF's are both qualitatively and quantitatively different from calculating them based on the deviation of the ergodic mean, as suggested in this (very helpful) thread viewtopic.php?f=1&t=3697

Attached is the code I used for the comparison. It is originally the code of Fernandez-Villaverde, obtained from http://faculty.wcas.northwestern.edu/~l ... labus.html, modified just with respect to the different ways of computing the IRF.

[jpfeifer]

This is expected as they are two very different concepts. See the appendix for Born/Pfeifer (2014): "Risk Matters: A comment" on my homepage.