let's say i have data for volatilities to construct vol surfaces for USD/KRW and for EUR/USD, but i do not have EUR/KRW data. I guess i can then try to derive EUR/KRW vol if i imply certain correlation levels. Could someone please explain to me if/how this is done in practice, and more specifically how i should estimate correlations (e.g on fx spot, or imply them from some other type of data) to do so?Thanks a lot!Dook

Thats a really interesting problem. I have worked on the same issue but still got no satisfying solution. I worked with a formula published amongst others in Wystup (2006), fx options and structured products:,where is the implied volatility of the cross. The problem is that I could not figure out how to estimate the implied correlation (for USD/KRW and EUR/USD). Using historical data leaded to strange volatilities (for EUR/KRW) and thus option prices. Maybe there is somebody knowing more about implied correlations or different approaches to that problem?

implied FX correlation calculations described in my 1996 paperhttp://www.espenhaug.com/Haug1996a.pdfunfortunately only in Norwegian (Beta Norwegian quant journal), and I would not agree on everything in that paper anymore due to my increased knowledge about options and women (both are much more complex to understand than I thought in the first place, and still so simple)the formula you mention was known among FX option traders I worked with at least back in 1993, and they had been in market for quite some time already so probably much older, first known written source is to my knowledge Tompkins 1994 "Options Explained" implied correlations are very unstable, and they have a much wider meaning than I once stated in my 1996 paper, as you grow older you grow wiser, until you loose it all...

Could the use of implied betas help? Not really sure if this technique is applicable here.http://papers.ssrn.com/sol3/papers.cfm? ... id=1416753

Thank you for your responses!First replying to Collector:According to my findings the use of historical correlations does not lead to accurate results. My Norwegian is not that good ... but I guess you find that by statistics historical correlations are more like implied correlations for longer time periods.Replying to taneururer:I dont see how your technique can help to solve that problem. Maybe you can explain how you thought it can?Well, I thought maybe there is an approach common used in practice to price fx-options on a illiquid currency pair with unkown implied volatility . All that is known is . What I really want to know is how to get . One way is to get a good estimator for and apply the equation above. But maybe there is a complete different approach to get (Without using that equation)?

"would not agree on everything in that paper anymore due to my increased knowledge about options and women (both are much more complex to understand than I thought in the first place, and still so simple)"=====> Collector: So when is your next paper about women ?

A multi-cultural monograph would be very good."About Chinese Women", by Julia Kristeva would be an interesting model, with plenty of room for history, philosophy, and semiotics, including Lacanian equations!***"It is probably necessary to be a woman (ultimate guarantee of sociality beyond the wreckage of the paternal symbolic function, as well as the inexhaustible generator of its renewal, of its expansion) not to renounce theoretical reason but to compel it to increase its power by giving it an object beyond its limits." ~Julia Kristeva - "From One Identity to Another"

By the way a paper about women needs a lot of stochastic calculus. It might help me with my problem ...I think about the role which implied volatilities and correlation would assume in that paper.

Convexity and duration also.

I'm also interested in this fx topic (the women topic too, but that's a different thing...).So, Collector, I understand your paper in the same manner as Quantosaurus did. Therefore my question to Collector: Can you make it a bit more precise in what way you do not agree with the paper anymore?