Insurance is one of the few businesses that collects revenue before it incurs its cost of goods. Premiums arrive when a policy is written; claims are paid months or years later. The pool of policyholder money an insurer holds in the interim — money it does not own but gets to invest — is called float. The investment returns earned on that float (plus on the insurer's own capital) are investment income. The central tension of insurance analysis is this: an insurer can run a losing underwriting business and still be highly profitable if the income from float more than covers underwriting losses — or it can run a profitable underwriting business and earn float essentially for free, multiplying the value of every dollar of capital. Understanding which lever is driving earnings, and how durable it is, is the core of valuing an insurer.

How float & investment income are formed

Float is read off the balance sheet, not the income statement. Berkshire and most analysts compute it roughly as:

> Float ≈ unpaid losses + loss-adjustment-expense reserves + unearned premium reserves + other policyholder liabilities − premiums receivable − reinsurance recoverables − deferred acquisition costs.

In essence, it is the net of insurance liabilities the insurer is holding against assets owed to it. Berkshire Hathaway's float was roughly $176 billion at year-end 2025, up from ~$114 billion in 2017 (Berkshire reports; Insurance Business, Artemis).

The cost of float is determined by underwriting result. The bridge is the combined ratio = (incurred losses + expenses) ÷ earned premiums:

• Combined ratio < 100% → underwriting profit → negative cost of float (the insurer is paid to hold the money).
• Combined ratio > 100% → underwriting loss → that loss, expressed as a percent of average float, is the cost of float.

Buffett's framing in the 2002 Berkshire letter: float "has value if its cost over time is less than the cost the company would otherwise incur to obtain funds" (berkshirehathaway.com/letters/2002pdf.pdf).

Investment income has two components that analysts keep separate:

• Net investment income (NII) — recurring coupons, dividends, and rents net of investment expense. This is the stable, predictable line.
• Realized/unrealized capital gains — lumpy and market-driven; generally excluded from "core" or operating earnings.

How it's used in practice

The analytical workflow:

1. Decompose earnings. Split pre-tax income into underwriting result (driven by combined ratio) and investment result (driven by NII + gains). A P&C insurer reporting a 99% combined ratio is barely profitable on underwriting — its earnings quality depends heavily on NII.

2. Measure investment leverage = invested assets ÷ equity. This tells you how much the investment book amplifies returns on capital. P&C insurers typically run ~1.5–3.0x (NEAM, S&P Global), while life insurers run ~10x because long-dated policy liabilities generate enormous, slow-turning float. A given investment yield therefore moves a life insurer's ROE far more than a P&C insurer's.

3. Match duration to liabilities (ALM). Asset-liability management dictates the investment book. Life insurers hold long-duration bonds (industry asset durations commonly cited around 5–6 years and longer) to match decades-long policy obligations. P&C insurers hold shorter portfolios (commonly ~3–4 years) because claims are lumpier and shorter-tailed. Mismatch is interest-rate risk; "reaching for yield" by extending duration or buying credit is a recurring source of insurer blow-ups.

4. Track the new-money vs. book-yield gap. Because portfolios turn over slowly, NII lags interest rates. After the 2022–2024 rate rise, U.S. P&C industry book yield climbed to a decade-high ~4.20% in 2024 (NEAM), and net investment income grew ~18% to ~$85.4 billion (AM Best), as maturing low-coupon bonds were reinvested at higher rates — a multi-year tailwind. The same dynamic runs in reverse when rates fall.

Adoption, debate & evidence

Float is universally understood as a mechanic but Buffett-coined as a value framework — the term and its "negative cost of float" emphasis trace to Berkshire's letters. The accounting reality (premiums-before-claims) is uncontested; the valuation spin is where judgment enters.

Honest caveats the folklore glosses over:

• Negative-cost float is rare and cyclical. Berkshire is the standout case — Buffett has stated its insurance operations have generated an underwriting profit (combined ratio below 100%) in most years, and its P&C combined ratio was ~82.9% in 2024 (Carrier Management). But the U.S. P&C industry as a whole frequently runs combined ratios above 100% in soft markets, so most insurers' float carries a positive cost most of the time.
• Reserves can be wrong. Float is built on estimated future claims. Under-reserving inflates current earnings and float quality, then reverses. Buffett's 2002 results absorbed a $1.31 billion reserve-strengthening charge to fix prior-year errors — from the very company held up as the gold standard.
• Catastrophe and tail risk spike the cost of float discontinuously. Buffett: "We're certain to get one of these disasters periodically, and when we do our float-cost will spike."
• Investment income is not free alpha. It is largely a bet on rates and credit. Higher market yields lifted NII in 2022–2024 but simultaneously pushed bond fair values below book value, forcing realized losses in those same years (NAIC, NEAM). The gain and the loss are two sides of the same coin.

Strengths & limitations

When the framework illuminates: It correctly explains why a low-combined-ratio insurer with high investment leverage compounds capital exceptionally well, and why two insurers with identical premiums can have wildly different earnings power. It is the right lens for separating durable underwriting skill from borrowed tailwinds (a rate environment, a benign cat year).

When it misleads: The single biggest misuse is treating float as costless leverage and reported reserves as fact. Float is leverage whose cost is unknown until claims fully develop, sometimes a decade out. An insurer can grow float aggressively by underpricing risk — which looks like success (rising float, rising NII) until the loss-development tail arrives. A second misuse is crediting management for investment income that is really just the prevailing yield curve.

Sources

• Warren Buffett, Berkshire Hathaway 2002 Chairman's Letter — berkshirehathaway.com/letters/2002pdf.pdf (cost-of-float definition; reserve charge; cat-risk caveat)
• FinMasters, "The Insurance Float: The Secret Behind Warren Buffett's Wealth" — finmasters.com (definition, combined-ratio link)
• Insurance Business — "Berkshire's 2025 operating earnings slip as float climbs to $176 billion" (insurancebusinessmag.com), confirming ~$176B float at YE2025
• brk-b.com, "Calculating Berkshire Hathaway's Float"; einvestingforbeginners.com — float formula components (matches Berkshire's own definition: net loss & LAE reserves + unearned premiums + funds held under reinsurance − premiums receivable − DAC − deferred charges)
• NEAM Group, "2024 P&C Industry Investment Highlights" — 2024 book yield ~4.20% (decade high), NII yield ~3.63%
• AM Best (via riskandinsurance.com) — 2024 P&C net investment income +18% to ~$85.4B
• S&P Global Market Intelligence — P&C and Life insurance investment KPIs; investment leverage ranges (P&C ~1.5–3.0x, life ~10x); new-money vs book-yield
• NAIC, "The Impact of Rising Rates on U.S. Insurer Investments" — fair-value vs book-value, realized losses 2022–2024
• Financial Edge / AnalystPrep — P&C vs Life duration and ALM differences

Dispute flags: precise cost-of-float figures and the "negative cost" emphasis are a Berkshire-specific framing, not an industry norm; insurer duration figures vary by source and year and are cited as commonly-quoted ranges, not exact constants.