Detrending variables and steady state file

Hello, I would like to ask if I can still supply steady state file but when I allow dynare to do the detrending for me, using, e.g. var(deflator=A) K? If so, does that mean that estimates in the supplied steady state file has to be detrended by hand? Thanks

Sorry for the delay. Dynare will remove the trend from the model by dividing it out. Thus, no trend variables will appear in the detrended model. If you employ a steady state file, your job is to provide the steady state of the detrended model (for the non-detrended one, no steady state exists). See e.g. examples/fs2000_nonstationary.mod · master · Dynare / dynare · GitLab

Dear Prof Pfeifer, thank you a lot for the response.

Just in addition to the previous question, is there an easy way (some type of test) to detect whether the detrending has been done appropriately? I happen to find that Dynare does second derivatives to check if equations are in line with a stable path. However, I happen to notice that if I switch some variables (e.g. marginal costs) from the variable block, that it is subject to detrending, to the block where variables are not deflated, dynare does not give any error.

The check is a numerical one. For small trend growth rates it may fail. Have you tried what happens if you change the growth rate?

Thanks for the valuable comment, it helped me to address some issues when computing steady-state. Initially I’ve put no growth rate (s.t. TFP is a random walk), the code ran. Now I’ve put 1% quarterly growth, the steady-state check found some issues. I fixed them.

However, the dynare does not highlight any problem if I switch, e.g. government spending of being subject to TFP trend to a variable that it is not detrended.

I might have some my own small mistakes, but I was just wondering if the user can be agnostic and rely on dynare to highlight if issues exist or if there are any ways to use dynare (via simulations, for example) to investigate if model’s dynamics is somewhat off

Thanks!

The detrending engine is an important sanity check, but it does not replace human input. All numerical checks come with a margin of error.

1 Like