Stationarity, Cointegration test and ECM

[gr031]

Hi! I am constructing a model of GDP and 6 independent variables.

I test for stationarity and if variables are non-stationary I take differences to bring them stationary and run my regression. Then I would like to test for cointegration using the Johansen technique by entering the variables in levels. However how do I know the lag length to enter for this test? Furthermore, if I find there is a cointegrating relationship of 1, how does that factor in into building my ECM?

Thanks :)

[mr.ngoc]

You need to test AIC or SC and choice lag length is smallest. Example,