Empirical shock decomposition


Why do I get these strange shock decompositions? I used the command shock_decomposition after the estimation command, because I wanted to do an empirical shock decomposition.

Please let me know whether this graph is correct, because the theoretical one (using simulated data before the estimation) looks quite different.

Have a look at the data, i.e. the black line. There are at least two issues here. The first one is the unhandled seasonal pattern. The second one is that your data seems to be non-mean zero while your model variable implies a different steady state. This forces (close to) unit root shocks to account for the different mean. This suggests your observation equation is wrong.

1 Like