A Cumulative Variance Decomposition?

[macro_man]

Dear Estima:

Let's say I estimate a structural VAR with the variables in log-differences. I then accumulate the impulse response functions so that they are back in levels. Now if I estimate the variance decomposition for the same model the forecast error decomposition will based on the log-differenced data used in the VAR. Thus, I am creating VDC for the differenced data, correct? If so, is there a way to estimate the VDC for the level data using the VAR model with the diffferenced data? In other words, can I estimate the forecast error decomposition for the levels even if I estimate the model in differences?

Thanks.

[TomDoan]

The simplest way to do that is to specify the VAR in log levels, then add an ECT statement, but with a blank list; for instance,

Code: Select all

system(model=ectmodel)
variables ftbs3 ftb12 fcm7
lags 1 to 6
ect 
end(system)

That will estimate in first differences, but compute impulse responses and error decompositions for the original levels.