Hi,I am wondering if people in the street are looking at the convexity adjustment for Euro$ futures and the potential impact of discounting at OIS.When I talk around I hear the same thing: Kirikos [amended: and Novak] formula or Hull White adj formula with... mean rev at 0.03... I feel that there is something inconsistent but I cannot exactly figure out how it should be done nor if it actually makes sense from a price impact standpoint. Maybe it is too small to bother.Any insight or info on this would be much appreciated!Thanks

People in the street, no. People on the Street, maybe. What is somewhat imprecisely called the convexity adjustment is in the case of ED futures actually compensate for two completely different phenomena. The simplest is the fact that on the maturity date, the futures price is set equal to 3M Libor, rather than to 3M Libor discounted for three months. This is the same effect as you get in a swap where the Libor leg pays in advance, and is a true "convexity effect". It is primarily based on the term volatility of the forward Libor, which should not be materially affected by OIS discounting. The other effect arises from the covariance between the daily mark to market payments and the rate at which margin can be funded/reinvested. This covariance will probably be slightly lower by assuming OIS discounting rather than discounting at the short end of a curve that is built from futures/swaps referencing 3M Libor, but my gut feeling is that the difference is pretty minor. I have not seen anybody try to estimate the size of the difference, but that may just mean that I am out of the loop.

Thanks. Maybe people on the Street actually don't look at it... or maybe they just don't say it and we don't know who does it right and who gets picked off.

It is correct to refer to the two effects contributing the futures rate adjustment as convexity adjustments. This is fairly well explained in Hull's book. The first effect you describe is a plain convexity adjustment, whereas the second effect is what hull calls a timing adjustment, but it is positively an effect due to the convexity of the time value of money.QuoteOriginally posted by: bearishPeople in the street, no. People on the Street, maybe. What is somewhat imprecisely called the convexity adjustment is in the case of ED futures actually compensate for two completely different phenomena. The simplest is the fact that on the maturity date, the futures price is set equal to 3M Libor, rather than to 3M Libor discounted for three months. This is the same effect as you get in a swap where the Libor leg pays in advance, and is a true "convexity effect". It is primarily based on the term volatility of the forward Libor, which should not be materially affected by OIS discounting. The other effect arises from the covariance between the daily mark to market payments and the rate at which margin can be funded/reinvested. This covariance will probably be slightly lower by assuming OIS discounting rather than discounting at the short end of a curve that is built from futures/swaps referencing 3M Libor, but my gut feeling is that the difference is pretty minor. I have not seen anybody try to estimate the size of the difference, but that may just mean that I am out of the loop.

lequocle, Both bearish and mtsm have provided clear descriptions of the two effects that I wrote about in the 1997 paper that you ungenerously ascribed to only my co-author. Your lack of due consideration aside, you will please note that in that paper I suggested that the convexity correction is a chicken and egg problem. The true zero curve requires a convexity correction but the convexity correction assumes you know the zero curve. Hence, as I described in that paper, both the zero curve and the convexity correction need to be jointly estimated. This is not a difficult thing to effect give the closed form solutions available in a normal interest rate model. One can do it trivially in Excel using the built-in Solver functionality, for example. bearish's comments about the impact of OIS discounting strike me as cogent; OIS discounting does not change the basic nature of the need for a futures convexity correction. Futures contracts still pay a fixed amount per tick size regardless whether yields are zero or one hundred percent.You have noted that some people have implemented the convexity correction in a rather sloppy fashion. I imagine you are referring to Bloomberg, for example. Is it worth taking a more considered approach to the problem? Go ahead and trade without it and find out.

I apologize and please do not take it as a lack of consideration. I shall refer to the formula as Kirikos-Novak from now on.Thank you for your explanation.

QuoteOriginally posted by: DavidJNlequocle, Both bearish and mtsm have provided clear descriptions of the two effects that I wrote about in the 1997 paper that you ungenerously ascribed to only my co-author. Your lack of due consideration aside...Possibly Relevant

I doubt it. Is that Mark E. Smith you've chosen as your public face on this website?

The man is something of a hero to me. Not that he's much of a role model.

Does anybody know any (good) paper about futures convexity adjustment in a multicurve setting?

QuoteOriginally posted by: BerndSchmitzDoes anybody know any (good) paper about futures convexity adjustment in a multicurve setting?Do you mean that you found out why Mercurio removed the relevant appendix? That'd be interesting for all of us.See also Ametrano&Bianchetti 2013.

Not really sure what you mean with Mercurio.The problem is that all papers cited in Ametrano&Bianchetti 2013 are still singleCurve. One can probably use the same arguments in a multiCurve setUp if one assumes a deterministic Libor-OIS-spread but I haven't thought that through. So I was hoping that somebody has already written this down.Cheers,Bernd

QuoteOriginally posted by: BerndSchmitzNot really sure what you mean with Mercurio.The problem is that all papers cited in Ametrano&Bianchetti 2013 are still singleCurve. One can probably use the same arguments in a multiCurve setUp if one assumes a deterministic Libor-OIS-spread but I haven't thought that through. So I was hoping that somebody has already written this down.I have done some analysis in my recent book: Interest Rate Modelling in the Multi-curve Framework: Foundations, Evolution, and Implementation, in particular Section 2.6 (deterministic spread) and 7.8 (stochastic spread). (You can also find the arguments in the corresponding working papers on SSRN: The Irony in the Derivatives Discounting Part II: The Crisis and Multi-Curves Framework with Stochastic Spread: A Coherent Approach to STIR Futures and Their Options, but don't tell it to anybody, I prefer to sell the book).There is also some extension when combining multi-curve and collateral frameworks and to new exchange trade products. But this is still in preparation. I will post it when I have something clean.Hope this help.