Variance decomposition and shock decomposition

Dynare help (legacy posts)
1

Hi All,
I have done both variance decomposition and shock decomposition in dynare, however, i have the following questions.

  1. in my variance decomposition and shock decomposition, some observed variables’ variances are attributed to a variety of structural shocks, while several observed variables’ variances are attributed to only one single structural shock, for example, education is the observed variable, its variance is only influenced by human capital shock (100% in variance decomposition is attributed to human capital shock), while other variables such as gdp, consumption, their variances are influenced by a number of structural shocks, is there a potential problem about the variance decomposition for education? is this an indication of model misspecification of the dsge model?
  2. i find some structural shocks contribute to shock decomposition in large proportion but did not contribute variance decomposition or just a little,is this normal or does this also indicate model misspecification of the dsge model?
    Thank you and looking forward to hearing from you.
    Kind regards,
    Jesse
2
  1. That question is impossible to answer. Sometimes one shock being the only one to impact a particular variable is by construction. Moreover, we don’t know the true data generating process. Thus, it is very hard to say which shock plausibly should be driving variables.
  2. Not contributing at all to the variance decomposition but contributing to the shock_decomposition should never happen. Having them diverge in terms of quantity can happen, particularly when you filter the data and the variance decomposition is only in a particular frequency range. Moreover, the shock decomposition is an in-sample one and within short samples the actual realizations can differ from their theoretical properties (albeit this is rather unusual)