ROIC (Return on Invested Capital) is a useful but imperfect metric for evaluating insurance businesses. Here's why:

The Core Problem: Insurance Companies Aren't Industrial Businesses

ROIC was designed for traditional capital-intensive companies — manufacturers, retailers, industrials — where invested capital (debt + equity) funds physical assets that generate operating profits. Insurance businesses operate on a fundamentally different model:

1. Float, not invested capital, drives returns. Insurers collect premiums upfront and pay claims later. This "float" — which is a liability, not invested capital — is the real engine of returns. A traditional ROIC calculation misses this entirely.

2. The balance sheet is inverted. For most companies, liabilities are a cost. For insurers, policyholder liabilities (reserves) are the raw material of the business. A well-run insurer with disciplined underwriting can generate returns from float that cost less than zero.

3. Investment income is operating income. Insurers invest premiums (float) in bonds and other securities. Investment income isn't a side activity — it's core to the business model. Traditional ROIC often struggles to classify this properly.

Better Metrics for Insurance

• Combined ratio — The gold standard. (Loss ratio + expense ratio). Below 100% means underwriting profit. This is the single most important operating metric.

• Return on Equity (ROE) — More relevant because equity (not total invested capital) is the lever that supports underwriting capacity. Regulatory capital requirements tie directly to equity.

• Growth in book value per share — Particularly for P&C insurers, this captures underwriting discipline + investment returns. Berkshire Hathaway famously uses this as its primary yardstick.

• Underwriting profit / float — Measures how cheaply (or profitably) the insurer sources its funding.

When ROIC Can Still Help

ROIC can be useful when comparing insurance businesses to non-insurance opportunities on a capital-allocation basis, or when evaluating insurers with significant non-insurance operations. But in isolation, it misrepresents the economics of underwriting.

Would you like me to pull combined ratios or other insurance-specific metrics for any particular companies?

You're making an excellent point. If you adjust the denominator to include float, ROIC becomes far more relevant for insurers. In fact, this is exactly what many sophisticated investors do — and it's conceptually what Buffett means when he talks about the cost of float.

Let me illustrate with real data. Let's look at a few major P&C insurers.Now let me pull the key financial data — we need NOPAT (or EBIT after tax), total equity, debt, and float. Float is essentially insurance reserves (loss reserves + unearned premium reserves).Good. But to get the float, I need the insurance policy liabilities. Let me search for those in the SEC filings.Now let me compute the float-inclusive ROIC for all three. I have what I need from the filings and the financial data. Let me pull the latest annual data.You're absolutely right — and the data makes the point vividly. Here's the 2025 comparison:

Float-Inclusive ROIC vs Traditional ROIC (FY 2025)

Ticker	NOPAT	Equity + Debt	Float (Proxy)	Traditional ROIC	Float-Incl. ROIC
CB	$11.2B	$91.4B	$174.9B	12.3%	4.2%
PGR	$11.5B	$37.2B	$85.8B	31.0%	9.4%
TRV	$6.6B	$42.2B	$101.5B	15.7%	4.6%

Why Both Versions Matter

Float-inclusive ROIC reveals something powerful: Progressive's 9.4% float-adjusted ROIC is roughly double its peers'. This captures what the traditional ROIC misses — PGR generates far more NOPAT relative to the total capital (including float) it deploys. It's not just that PGR has high underwriting margins; it's that it's more efficient at converting float into profit.

But there's still a subtlety. Float itself is not homogeneous. The float in the denominator is a stock (balance sheet), while NOPAT in the numerator reflects underwriting profit on this year's policies. If float grows rapidly — as PGR's has — you're dividing this year's earnings by an ever-larger float base, which can understate true returns. An alternative is to use average float across the year rather than year-end float.

The Deeper Insight

This is why Buffett talks about "cost of float" rather than ROIC. If your float costs -3% (i.e., you earn underwriting profit), the question isn't "what's the ROIC?" — it's "how much float can you generate at a negative cost?" The float-inclusive ROIC captures whether the insurer is earning adequate returns on the total pool of capital it controls, not just shareholder equity. Progressive's ~9% float-inclusive ROIC vs. peers at ~4-5% says something real about operating superiority that the traditional 31% ROIC obscures with leverage.

So yes — float-inclusive ROIC is a far better metric than traditional ROIC for insurers. It should be paired with the combined ratio and float growth rate to get the full picture.

No, float cost and combined ratio are related but not identical. Let me show you with real numbers.

Combined Ratio = Loss Ratio + Expense Ratio This measures underwriting profitability: premiums earned vs. claims paid + operating expenses. Below 100% = underwriting profit.

Cost of Float = (Underwriting Profit or Loss) ÷ Float This measures what you're paying (or earning) for the privilege of holding other people's money.

They diverge because float ≠ premiums earned. Float is the accumulated stock of unpaid losses and unearned premiums on the balance sheet. Premiums earned is the income statement flow.

Let me pull the actual numbers for these three:Now I need the specific float and underwriting numbers from the SEC filings I already retrieved. Let me compile the data.Now let me search for the actual combined ratios:Excellent. Now I have all the data. Let me build the definitive table:No — cost of float is not the same as the combined ratio. The table above from 2025 10-K filings makes the distinction clear.

The Key Difference

Combined ratio = (losses + expenses) ÷ premiums earned. It's a margin on the flow — this year's underwriting activity.

Cost of float = underwriting profit ÷ float. It's the return on the stock — the accumulated balance sheet liability you're holding.

Why they diverge

Look at Progressive vs. Chubb:

	Combined Ratio	Cost of Float
CB	85.7%	−5.2%
PGR	87.4%	−14.9%
TRV	89.9%	−5.0%

CB has the best combined ratio (85.7), but PGR has a vastly cheaper cost of float (−14.9% vs −5.2%). The reason: PGR earns $81B in premiums on only $68.5B of float, while CB earns $46B on $141B of float. PGR's float turns over much faster — it's predominantly short-tail personal auto, where premiums are collected and claims paid within months. CB's float is dominated by long-tail commercial liability reserves that sit on the balance sheet for years.

The Formula

Cost of Float = −(Combined Ratio − 100%) × (Premiums Earned ÷ Float)

So the combined ratio feeds into float cost, but the float turnover ratio (premiums earned ÷ float) is the multiplier. A company with a 95% combined ratio but 3× premium-to-float leverage has cheaper float than one with an 85% combined ratio and 0.3× leverage.

This is precisely why Buffett evaluates insurers by asking: "How much does the float cost, and how fast is it growing?" — not just "what's the combined ratio?" The two metrics answer different questions. Combined ratio tells you if this year's underwriting was profitable. Cost of float tells you whether the entire pool of other people's money you're investing is a tailwind or a headwind.