wilmott.com

I am looking to prove that FX rates (at least the major currency pairs) are mean reverting. Has anybody got an idea how to best go about it? Did anyone do a similar analysis or knows about an article or research paper about the subject? Or do people disagree with this assumption?Thanks!!

Sure, check whether a euro(mark) is more likely to go down when it is above the 100-day moving average than when it is below.And then you have the Yen. And how long has this sample of three been operating under the free-floating/2%-inflation central bank target?I disagree with the assumption. I think that by the time it would mean revert, Alan Greenspan will be dead, and most cars will be made in China.

There is a good article from lehman brothers' research, "Mean Reversion in FX rates" published in LMR Quaterly vol.2003-Q2I am trying to upload the PDF file but it doesn't allow me, does anybody knows why?

mencey,Did you zip the file before attempting upload? Only winzipped files allowed.

Mencey,thanks, I would really like to see this article. If you can't upload it, where do I find it?

I assume there must be some type of mean reversion, but this is probably only over very long time periods (many years) and during such a time period also the central bank policy could change a lot, or even the currency pair you are looking at could simply disappear . So limited how much one can trade on this...but still somewhat useful....

That is exactly what I would be interested to find some research on. I am not worried about short term changes. What I would like to prove is that if you simulated an FX rate using a lognormal process you would not describe it as well as using some kind of a mean reverting process. Anybody out there got any ideas?

Well, for what it's worth, I disagree with this assumption. I cannot see any fundamental reason why FX rates should be mean reverting. I'd imagine that foreign exchange would be the most trending asset, along with commodities.Nevertheless (and I'm not sure here - any other ideas are welcome) I would suppose that a Dickey-Fuller test for the presence of a unit root would do your job. Try to run a regression and see how negative the t-test is.

Also, I couldn't find the Lehman document that Mencey mentioned in his message, but while searching for it, I did find a July 2002 Lehman Brothers article on "Mean Reversion in Implied Volatility" which relates specifically to FX times series. The study clearly supports that "while spot time series are NOT mean-reverting, it is a well documented fact that implied volatility series are".I'm attaching the file; hope it helps.