Testing for cointegration
Posted: Tue Feb 02, 2010 8:02 pm
I have an equation of the following form, where all variables are I(1):
log(y) = log(x1) + log(x2) + C @trend
This means that the equations should be estimated using first differences, however I have been told that if it is proven that the data is cointegrated, I can use the normal estimations instead.
How would I go about detecting cointegration in the above model? Can I use the simple Error correcion model? or must I use the more complicated Johansen approach, because I have more then 1 explanatory variable?
thanks
log(y) = log(x1) + log(x2) + C @trend
This means that the equations should be estimated using first differences, however I have been told that if it is proven that the data is cointegrated, I can use the normal estimations instead.
How would I go about detecting cointegration in the above model? Can I use the simple Error correcion model? or must I use the more complicated Johansen approach, because I have more then 1 explanatory variable?
thanks