The method I'm using only relies on this formula: IV^2 = 1/DTE * EV^2 + (1 - 1/DTE)*AV^2 We're assuming that the implied variance is the weighted sum of the ambient and event variances (got this from formula from Colin Bennett's book). Given several implied vols, we're basically "inverting" the formula to compute the ambient vol and event vol. In Bennett's book, he uses the forward vol to approximate "AV", but I've found, empirically, that the forward vol estimates are quite unreliable.
All estimators are from the day before earnings Garch 39% Close-Close 41% Garman Klass 38% An average of say ~ 40% vol. Broken down into 12 days is a daily standard deviation of 1.83%. If we increase the post earnings date sd by 50 % we end up with an estimated vol of 41.7%. Meaning there is a 8.3% risk premium on MBUU (opened at 50% imp vol), still not good enough for my estimation (but much better than forward vol est). Can you reference me the page? Lets walk through a real life example
It's on page 126 of the book. As an example, consider ADBE, which has earnings on Sep 18. The implied vols of all expirations from Sep 21 to Nov 16 (including weeklies) are 0.413,0.37,0.343,0.328,0.332,0.295,0.31. I'm using aggregate IVs here, for convenience, but you can use ATM IVs if you prefer. Using my method, I get an earnings vol of 1.26 and an ambient vol of 0.26. The forward vols, on the other hand, for the same expirations, are 0.2544644 0.2390520 0.2566830 0.3518539 0.3430847.
I used to determine a forward vol. but I don’t think it was statistically better. The MBUU situation could be something idiosyncratic and just the exception to your formula