Hi,I am looking at a CMS spread option that is priced using a joint pdf model. Historically spot correlations are up towards 100% but the forward correlations fall off quite a bit. I am hesitant to use the lower forward correlation in the pricing given that generally the rates mean revert. Can anyone suggest an interesting way to look at this problem?Thankspabo

Do you have arbitrage-free condition?How correlation could be greater than 100%???

I think historically it's the other way around: spot correlations are lower than forward correlations.

Sorry - I was being unclear.Suppose that we have the swap curves going back for a period of time (say 5 years) we use these curves to run historical analysis.Also that we are pricing an instrument that has a payoff dependent of the quoted 2 year rate and 5 year rate on fixings going say 5-10yrs in the future.I'm looking at the structure of the correlation of the forward implied CMS rates of the 2 vs 5.I have taken the spot rates of the 2 and 5 year and found the correlation up towards 100% (say 95%).Also the other extreme is taking the 10 year forward implied 2 and 5 year rates on my set of historical yield curves and running correlation analysis on this.The correlations on these are much lower.

Hi Pabo,which joint pdf model do you use? Do you incorporate the smile?

If your forward correlations are lower than your spot correlations you are almost certainly using bad data. Try looking at the correlation of weekly moves rather than daily, this should remove some noise (the problem here is that your forward rates are very sensitive to the par rates you are inputting ).