IMPULSE RESPONSE VECM

[msound]

Hello there,

Please, I have a question and would like to see if anyone can help me. I'm running a VAR model with I(1) and cointegrated variables (VEC) and I my impulse response graphs seem to be really unstable. Moreover, the graphs do not show the confidence bands. My impulse response functions are estimated directly from the VEC estimation using variables in levels (they are used in levels since the VEC automatically difference them for the VAR estimation). The lag length seems to be good using AIC and SC, so I don't believe is a lag problem.

Please, I would like to know how to solve the instability. Should I estimate the Impulse response using a VAR estimation with differences instead of the VEC?

Thank you in advance!