My understanding is that whenever we use Eurodollar futures to build a curve, we must subtract the convexity adjustment that is implied in the futures prices. This is so as to ensure the yield curve is free from convexity noise.But in the world of collateralised swaps, where collateral is paid on a daily basis, shouldn't there also be convexity embedded in the swap rate compared with non-collateralised swaps?Then should we or should we not compute this convexity if we use the swaps to build the curve?Or, should we also stop removing the convexity adjustment in futures, to be in line with the swaps--whose convexity is not removed either?Thanks for any enlightenment.

Perhaps I should clarify what I mean by convexity adjustment of Futures. (By futures, I mean Eurodollars in the context of interest rate world)I understand it to be in relation to FRA, where a FRA payoff valued today has convexity due to discounting effects, whereas a Future has no discounting effects.But if we value a FRA payoff at maturity T, then FRA payoff is linear. There is no convexity. It is instead the Future which has convexity, due to early repayment and then reinvestment of the cash. It is very similar in concept to the libor-in-arrear adjustment.So convexity lies with FRA or with Future?

Futures are not convex instruments, FRAs and swaps are. Once you invert the argument in your initial post, everything should fall into place.

Hi, thanks for reply.What I understand now is that the margins for futures are computed on the non-discounted value, thereby giving rise to *convexity while the margins for its corresponding FRA are computed on the discounted value.I am still confused if this convexity adjustment has anything to do with the convexity arising out of early repayment seen in, say, payment-upfront FRA?This question is relevant, as I wld like to understand heuristically if collateralised swaps have any convexity relative to uncollateralised swaps.*I am speaking of convexity as a relative term here. Futures have linear payoff, hence no convexity relative to FRA, but only at t0. If we consider final payment date, it is the FRA which has linear payoff, and correspondingly, the convexity now lies in the futures payoff, due to possibility of re-investment.

I don't really understand what you mean...What is a "payment-upfront" FRA? How is the payoff of a FRA on the final payment date different to that of the futures? Pls offer a specific example. As to collateralised vs uncollateralized swaps, I need to think about that, but, frankly, what's the point?

revisiting this topic:http://www.stirfutures.co.uk/?p=214so collateralised swaps/FRAs behave in the same way as futures, so there are no convexity bias between them.so this follows that:Removing convexity bias when using Eurodollar futures prices when building LIBOR curves are an outmoded concept?Is this what is currently observed in the market?Thanks.

No, nobody suggests that swaps/FRAs have stopped being convex...However, it is true that the convexity issue has become very complicated. That's not just a function of margin netting, but also of balance sheet available to "earn" convexity, treatment of negative rates, etc.To my knowledge, this has not stopped people from applying the convexity bias when building curves.