How do variables drive each other?


I would like to determine by how much some variables drives the response of a specific variable when I shock a variable. I thought about looking at the variance decompositon in the output, but it seems like it only reports by how much the shock variable itself can explain the variance of each variable in the model. How do I get a measure for how much the variables drives each other?



So it is not possible at all?

In general no. Sometimes it is possible to do counterfactuals where you shut off a particular channel.