Conintegration question

General
1

Course Name: Mean Reversion Strategies In Python, Section No: 11, Unit No: 14, Unit type: Notebook

The notebook run Johansen Test on 2006 to 2010 dataset and showed cointegration. Since the calculation of spread only used first 90 days data, I tried to use only 90 days data to run Johansen Test and failed to show cointegration.

If that period tested is not cointegrated, can we still use the calculated spread to trade mean reversion? Would the statistical ground be weakened? During the backtest, we already looked ahead that cointegration existed in the future.

2

Hi,

You are right that the statistical ground could be weakened, but we are illustrating an example of the reason for including USO and forming a triplet. Before adding USO, we checked whether it is co-integrated in the period or not. And then we move further towards backtesting.
I hope this helps.