ECM and Granger causality test for combination of I(1) and I(2) variables

[Frances Manzano]

I am currently testing for the cointegration between export competitiveness (EC) and labor productivity (LP). After conducting stationarity tests using Augmented Dickey-Fuller test, I found out that EC is I(1) while LP is I(2). Would it be okay to proceed with the ECM and and Granger causality test or should I try another cointegration tests?

[Sebastian Kripfganz]

You cannot have cointegration between I(1) and I(2) variables. The integration order needs to be the same. If you believe that labor productivity is indeed I(2), then you can only check for cointegration between export competitiveness and the labor productivity change (i.e. the first difference of labor productivity). Alternatively, taking the log of labor productivity might solve your problem.