I have a question. Suppose that after running “shock_decomposition” I compute for every period the contribution of each shock to the deviation of an endogenous variable. Then, for every shock I compute the sum of the contributions over all periods in percentage.
Shouldn’t this number be close to the result of unconditional variance decomposition I get after running “stoch_simul”?
the unconditional variance decomposition is based on the average size of the shock standard deviations, and the fact that the dynamic system has been allowed to run for very many periods. the shock decomposition is constructed from the sequence of estimated shock innovations for each period and their role in building the time series considered. so we are talking of very different things here, even though they are computed using the same model solution.
Right, the shock decomposition explains all deviations from steady state at every point in time, while the variance decomposition is interested in the unforecastable part at every point in time.