Insurance Float Models
high confidence — compiled with source citation from the investment brain.
Insurance Float Models
Insurance Float Models
Float — money an insurer holds that belongs to policyholders but which it may invest for its own benefit — is the shared engine behind berkshire-hathaway, Fairfax, and Markel. The three converge on one goal: run underwriting at or below break-even so float becomes free, even negative-cost, leverage. They diverge sharply in how they grow that float and how much underwriting volatility they tolerate to do it.
Berkshire: float as a cost-free financing engine
Buffett calls insurance the core business of Berkshire. In the 2008 letter he reports that $58.5 billion of float cost less than zero — Berkshire was in effect paid to hold and invest the money — and states his expectation that Berkshire will average an underwriting profit over the years, whereas most insurers incur substantial underwriting losses over time.
The engine began with a single small purchase: National Indemnity bought for $8.6 million in 1967, the seed of what Buffett later calls the world leader in insurance float.
Berkshire's two growth levers are GEICO and General Re. GEICO is the low-cost auto insurer that grew its market share from 2.0% in 1993 to 7.7% in 2008 through disciplined underwriting and cost efficiency; Buffett attributes the edge to saving motorists money on a product no one enjoys buying. General Re, by contrast, is Berkshire's cautionary tale: purchased in 1998, its underwriting, reserving and expense discipline had deteriorated by 2001, and it was restored only after new management re-imposed discipline.
The lesson Berkshire draws is that float is only valuable if underwriting discipline keeps its cost at or below zero. See geico and insurance-float.
Fairfax: discipline expressed through the combined ratio
Fairfax frames its model around the combined ratio — the sum of the loss and expense ratios, where below 100% means an underwriting profit. Its 2019 letter reports combined ratios "well below 100%" across its major operations alongside reserve redundancies (reserves released as prior-year claims came in better than expected).
Fairfax enumerates results subsidiary by subsidiary: Odyssey at a 93% combined ratio over five and ten years, Zenith at 85%, Allied World, Northbridge and Crum & Forster at 96–98%, most with reserve redundancies in nearly every year. The discipline is presented as durable, not a single-year artifact.
The difference from Berkshire is emphasis rather than substance: Fairfax runs a decentralized stable of insurance and reinsurance platforms and reports the combined ratio as its primary proof of underwriting quality, while Berkshire reports the aggregate cost of float. Both measure the same thing — whether the insurer is paid to hold float.
Markel: specialty niches as the source of underwriting profit
Markel's stated financial model is "to earn consistent underwriting profits and superior investment returns", and it describes having "many different specialty products, over 90 at last count" as a source of stability.
Markel's 2005 letter is candid that the model can break in a single year: hurricane losses of $246 million pushed the year to a $12 million underwriting loss while the long-run record stayed good. The specialty strategy accepts that one catastrophic event can overwhelm results, and Markel's answer is a spread of risk across many niches plus its own underwriting models and "an even greater margin for safety."
What the three models reveal
All three treat float as the prize and underwriting discipline as the gate that decides whether float is an asset or a liability. Berkshire scales float through GEICO's cost advantage and General Re's reinsurance reach; Fairfax scales through a portfolio of platforms measured by combined ratio; Markel scales through hundreds of specialty niches. The structural difference is tolerance for underwriting volatility — Markel and Fairfax openly accept occasional catastrophe-driven losses, while Berkshire's stated goal is an underwriting profit on average over the years. See capital-allocation for how float-derived capital is then deployed.
See also
Sources
berkshire-hathaway/Letters-Markdown/2008 - Letter to Shareholders.mdberkshire-hathaway/Letters-Markdown/2021 - Letter to Shareholders.mdinvestor-letters/FairfaxFinancial/Letters-Markdown/2019 - Shareholders Letter.mdinvestor-letters/Markel/Letters-Markdown/2005 - Shareholder Letter.md