DOLS/FMOLS/CCR estimation of the LR relationship

[bobreednz]

BACKGROUND: Given two variables y and x, for which one suspects there is a LR relationship such that y = beta*x, the specification for the ARDL bounds test looks like this (ignoring deterministic regressors and lagged differenced regressors):

D.yt = b1*yt-1 + b2*xt-1 + b3*D.xt + et .

---------------------------

QUESTION: If I have a combination of I(0) and I(1) variables, and I conclude that a LR relationship exists using the Pesaran, Shin and Smith (2001) ARDL bounds test, can I then estimate, and do hypothesis testing, on the LR relationship using DOLS, FMOLS, and/or CCR? Or can I only use the DOLS, FMOLS, and CCR procedures if all the variables are I(1)?

[TomDoan]

I believe that is correct. Certainly in the case of DOLS, it relies heavily on the fact that the differences are asympotically uncorrelated with the levels which won't happen if the levels are stationary.