I know there are no recipes to make calibration, but I would like to learn from your experience.

I am trying to replicate a simple model like the one in the book of Cooley 1995 (Ch. 7). I mean a CIA monetary economy and then compare the detrended series performance (correlations and variances) with the real data of one country.

My doubt is: How should I treat the series in order to use them to calibrate?

For example. for the simplest case: the calibration of “rho” in an AR(1) process for the technological shock.
It is not difficult estimate rho from : z= rho*z(-1)+ e

But I have some alternatives:

  1. use the trend of z?
  2. use cycle of z?
  3. use the variable without transformations?

Which one is more plausible or more used empirically?

Thank you in advance.

Mauricio V.

I’m struggling with the same problem. Did you find a good method for calibration?

I guess that you have to write some code to utilize Dynare instead of using Dynare Keywords directly. The idea is to minimize a measure of “distance” between the moments in your model and the moments in the data. Ideally, the distance should be zero, but it is possible you can’t reach it.

My understanding that the minimum distance estimation used in Christiano et al (2005 JPE) is just an extension to the calibration technology.