VAR-Model - Absurd Impulse Responses

[kristin]

Hey Guys,

I´ve got a problem with the impulse responses of my VAR-model.
I estimated a VAR-model including 4 endogenous variables with their 5 lags und 2 exogenous variables. I took first or second differences in case the time series was not stationary. I want to know the responses if nothing happens to the first variable, the second decreases, the third increases and the fourth decreases. So I constructed a shock typing
matrix(4,1) shock
shock.fill(by=c) 0,-1,1,-1
and chose "user specified" as decomposition method.
The graphs of the impulse response functions show an absurd picture as they do not converge to zero but rather become extremer after 10 or 20 periods.

Unfortunatelly I have no idea of the reason for this outcome. I´m doing my first course in time series analysis and so I´m neither well schooled in EViews nor in econometrics.

I hope for helpful answers.

Many thanks in advance.