Im trying to find a cointegration relationship between a spot and futures series as I'd like to compute the optimal hedge ratios. As I find the spot and futures at the price level to be nonstationary, but at the return level (first difference) to be stationary indicating that both series are integrated of the same order I was wondering if I have to run the Engle Granger regression for the price changes or at the level of the returns. The residuals of this regression I would then check for stationarity to find a cointegration relationship in case of stationarity. I was wondering since when calculating the returns of the portfolio i.e. to hedge a long position(S-h*F), where S is the return of the spot, h the hedge ratio and F the return on the futures, would I be using the beta from the Engle-Granger regression?