I noticed that the way variables are entered into a VECM affects the speed of adjustment, the output from Granger causality tests and even variance decompositions. How can we them select the best way to order the variables?
Obi Ebere Before you estimate a VECM, maybe try to do a simple OLS test to check on your R and ajdusted R -squared, try to play with variables and choose the variables that give you the highest value for both coefficients. In my opinion, to get a better model, it takes time to play with the software in order to keep track of the result you get and choose the one that fits your aim and the existing literature.