Yes, it is true, go look at the defintion of 'cointegrated'. Has nothing to do with correlation, as correlation is a very poor measure for modelling co-movements. Things can be very correlated but not be cointegrated.
You have created an artificial standard for defining a tradeable pair... That it must satisfy the term "cointegrated"... Which, frankly, I couldn't care less exactly specifically what it means in statistics... Because I trade and follow about 400 highly corelated securities... And have more action and profits than I can possibly trade at this time. There is a miguided tendency here... To equate power with complexity. Complexity rarely results in powerful strategies... But it sure keeps the "rocket scientists" among us occupied. rm+
I've created no standard. I never said that it has to be cointegrated to be tradeable, I was simply pointing out that if series are truly cointegrated then they are the 'best' series to trade as the spread is stationary (constant mean and constant variance). Trading based on correlation is definately profitble, but you say you are following 400 correlated securities.. I'd say that is "complex" even if all that math is going on in your head, I prefer not to be bothered staring at screens all day, let the computer do what it is good at doing, and it lets me think about the strategy and guide it along.
You get an account at IB. IB pays YOU about 3% on your short position. My typical portfolio consists of about 40 positions and 60/40 long... Leveraged fairly close to the 2:1 max that IB's 50% overnight margin allows. This results in POSITIVE interest every month... Specifically on a ** net basis **... I do not pay any interest... But actually receive interest every month. IB doesn't have to do this... It's highly unusual for retail... but also highly ethical. Just an example of how IB does a lot of little things right. rm+
From a Google search it seems to me... That corelation and cointegration are 2 similar ways of establishing the same thing for pair trading purposes. Give me an example of a pair that might have 0.900 corelation ** for a rational reason **... And not be "cointegrated". Or give me a specific example where calculating "cointegration" is superior to corelation. And while building computer systems that track and rank 400 stocks in real-time is complex... The underlying or core idea for profitable hedging strategies is usually quite simple. So when I see people talk of tick-by-tick "data mining"... I just roll my eyes. rm+
So, let me get this straight please. All you do is open up an excel sheet, put the past historic prices of two stocks in there (How far back do you go?). Check the correlation (How many days/months do you check it for?), and then I guess you have a benchmark for the correlation (ie. >.9). Then you just take the ratio of the two prices, plot it with its mean and standard deviation. If the ratio is more than 2 standard deviations or something, you take a position and you just keep this data sheet running until the ratio returns back to it's current mean then close the position.
These guys made a study and found a 12% return. You can download the paper at the end of the page. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=141615
Wow, the more complex strategy produces no better actual returns than S&P500 index. Something to think about.