Consider a spread option or range accural where the spread is the 30/2 yr swap ratesHow do I estimate volatility using historical data?The spread has not been negative in the last 4 + yrs. I cannot take the lognormal returns since the spread can be negative. If I assume that the 2 and 30 yr rates are lognormal and work out the compositve vol for the process as Vol(c) = sqrt(vol(a)^2+vol(b)^2-2*vol(a)*vol(b)*corr(a,b)) then I get a pretty high volatility number. Attached are the workingsGrateful for any comments - honestly I am not after a wonderfully martingale model just some helpful comments on volatility estimates when the process is not log-normal.

Sorry - my attempt at attaching the spreadsht failed due the hard work of the wonderful IT people here at the bank. Anyhow, if you just get the last 4 yrs of data from Bloomberg the vols, correlations and composite vols are all straight forward.