What is the ideal time-frame for cointegration testing?
The answer is…there isn’t one.
A pair cointegrated over 10 years may not be so under any shorter resolutions. Another cointegrated over 3 months at 15 minute intervals might not be so easy to find similar results for over longer periods.
Thus scanning, experimentation with different parameters and testing are the only methods of assembling a promising portfolio of relative value pairs to trade.
Does cointegration lend itself to such wide-ranging methods – particularly in terms of intraday results? Yes – and this paper is an excellent read on that very subject.
We’ve added it to our Useful Links section.
