Error-Correction Model ECM

[saccc011]

Hi all, I am trying to estimate an Error-Correction model, with the current account as my dependent variable and the labour share, the real effective exchange rate, government saving and some other variables as independent variables. All variables are I(1) and when I pick particular variables, the residuals are I(0), the regression is not spurious, thus I think it's ok to use the ECM.

I am aware that in the LR, the ECM works in logs and in the SR, variables have to be in d(logs). However, some of the variables might not make sense when transformed into their logarithmic form (e.g. the dependency ratio), or some values of a particular time series (especially the current account and government saving) are negative. How do I go about estimating such model in the LR and SR? Is there any transformation which I can possibly do? Or is it not mandatory to have the variables in log form?

Many thanks.