Variance Decomposition Help

[Nice]

Hello jpfeifer,

I have performed Bayesian estimation already and I would like to additionally obtain the (1) forecast error variance decomposition and (2) historical decomposition as in Smets and Wouters 2007. I have read the forum on this topic and understand that I can achieve this via the following commands:

1. For variance decomposition at specific periods: conditional_variance_decomposition = [ integer];

I also understand that there is a difference between specifying this command within the estimation command and specifying it in stoch_simul AFTER the estimation command. My first question is, if I want to follow Smets and Wouters (2007) where do I specify the conditional_variance_decomposition?

My second question is, if I do not want to run my estimation again ( as this will take much time) and instead I simulate my model by setting my parameters at the mean values obtained from my estimation, will specifying conditional_variance_decomposition in stoch_simul serve the purpose of following the method in Smets and Wouters (2007) to obtain variance decomposition?

For historical decomposition: shock_decomposition command is only for estimation?

Many Thanks!

[jpfeifer]

The Smets/Wouters variance decomposition is the one at the mode. Thus, it is obtained by using stoch_simul with the parameters displayed after mode-finding.

shock_decomposition also works on calibrated models. See https://github.com/DynareTeam/dynare/blob/master/tests/shock_decomposition/example1_calib_shock_decomp.mod for an example.