I have a couple of question regarding the use ECM/VECM and Johansen test for cointegration. 

In my model (6 variables), I have only one variable that is Stationary in its level, the remaining five are only stationary when differenced. This gives me a mixture of I(1) and I(0) variables, which is my first issue.

I conducted the selection-order criteria for the variables and found out that the majority of the criteria's (LR, FPE, AIC) suggest we include 4 lags in the model. Based on this I performed a Johansen test for Cointegration with 4 lags, which resulted in: the "trace statistics" suggesting that there are four or fewer cointegration equations.

My question relates to the issue regarding the mixture of I(1) and I(0) variables. Will the ECM/VECM be sufficient went estimating our model? Or should I start thinking of perhaps using an ARDL model proposed by Pesaran et al. (2001). The authors specifically point out that when we have a set of regressors which have the characteristic of our mix of variables, this model is optimal.

I am a STATA (13) user and greatly appreciate any guidance.

All the best!!

  • $\begingroup$ your question looks very similar to questions already asked on this forum. Can you search/link to these, and tell us how the answers elsewhere do not answer your question? $\endgroup$ – Matifou Jun 26 '17 at 6:23
  • $\begingroup$ Would you be so kind and attach a link to the answer for this question? $\endgroup$ – Arash Jun 28 '17 at 22:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.