Robustness of VECM

[gobbble]

I would like you to ask about the implications of a bad VECM model.

I constructed a pairvise VECM model. Checking on robustness I found the VECM model very weak with violations against normality, heteroscedasdecity and autocorrelation..

Now I did PROC => MAKE SYSTEM => ORDER BY VARIABLE

I copied the variables to run an OLS to check short run and long run causalities.

This OLS model is robust in all respects.

My question:

Can I ignore the bad VECM model if the final OLS model is robust, or what is the effect of ignoring?

AS I have to finish my research soon, I would appreciate your quick help here. Thanks!