Here are my assumptions for Berkshire's future cost of float which I peg at -0.8%. I use this estimate for a float based valuation but (perhaps inconsistently) I assume underwriting break-even for the two-column method due to a desire to be conservative.

http://www.rationalwalk.com/wp-content/uploads/2012/10/BRKFl...

I arrive at the -0.8% cost of float by estimating the cost of float for each of the four reporting groups separately and then coming up with a weighted average. I still penalize General Re for its (now long past) sins and only assume break-even for BH reinsurance. The slight negative cost comes from GEICO and the primary insurers.

It is possible that I'm being way too conservative here but I've been burned with overly bullish float based valuation assumptions in the past.