Variance decomposition and measurement errors in estimation

[britney0109]

Dear all,

I set up a model with some structural and a few measurement errors in the observable variables. The Bayesian estimation works fine, but I get strange results when I compute the conditional variance decomposition of the observables following my estimation. Each structural shock explains 100% of the variation of one variable, the measurement errors explain nothing.

Searching the dynare files, I discovered that in the file conditional_variance_decomposition.m it says in line 17

% SPECIAL REQUIREMENTS
%
% [1] In this version, absence of measurement errors is assumed...

(This is Dynare 4.3.1.)

Obviously, this requirement is not met by my model. Is there an easy and fast way around this problem? Is it possible to apply variance decompositions in the presence of measurement errors?

I really appreciate any help!

[jpfeifer]

This should only be an issue, if you use Dynare'S internal measurement errors. If you treat the measurement errors like structural shocks, i.e. for example you have a measurement equation like

Code: Select all

y_obs=y+eps_meas_y;


the conditional variance decomposition should be correct.

[britney0109]

That helps. Thanks a lot!