differencing and coint rank in VECMX

[Sujin Choi]

Hi,

I'm estimating VECMX(5,5) and all variables are nonstationary.

I understand that when I estimate VECM, the (all the endogenous) variables are not differenced.

But for VECMX, should I difference x variables before I estimate? or, like vecm, just use the original version of all the variables?

For cointegration rank, should I check the rank only for engodenous variables or for all the variables?

Thanks in advance! Happy new year.