FINANCE, POLICY & MARKETSPublished by Paul Ivinskas
fc.The Financial CurrentDAILY INTELLIGENCEWhat matters across finance
Deep-dive library

Insurance underwriting: the combined ratio, reserve development and investment income

6 min read · estimatedAI-generated analysis · Methodology
Current version · 1 version · Publication details

First published . This version published .

Initial full research explaining the mechanism, worked examples, competing interpretations and material limitations.

Related research, policy & entities ↓

At a glance

Excerpts from this version
What it covers
A property-and-casualty insurer’s combined ratio measures underwriting, while reserve revisions and investment results explain why reported profit can tell a different story.
0% through article

Tap a dotted-underlined term for a definition; terms are highlighted once per section. Use Aa in the navigation for reading preferences.

In this article

Two earnings engines, one balance sheet

Property-and-casualty insurance combines a promise to pay covered claims with investment of money held before those claims are settled. The combined ratio describes the first activity: losses, claims-handling costs and underwriting expenses relative to premium revenue. It does not measure the return on the investment portfolio, the shareholder’s return on capital or the insurer’s ability to survive an extreme loss. A ratio below 100% indicates an underwriting profit on the stated basis; above 100% indicates an underwriting loss. [1]

The distinction is visible in Travelers’ full-year 2025 release, published January 21, 2026. It reported an 89.9% combined ratio, versus 92.5% in 2024, alongside $3.959 billion of pretax net investment income and $6.288 billion of net income. These are separate measures with different denominators and tax treatment. They cannot be added together as percentages. The release is a dated illustration, not a claim about the company’s subsequent quarters. [2]

Written premium is a sale; earned premium is elapsed coverage

Written premium records business booked; earned premium recognizes the portion of coverage provided during the reporting period. W. R. Berkley’s 2025 filing says its premiums are primarily earned pro rata over the policy term. Gross premium includes business before ceded reinsurance; net premium reflects the part remaining after reinsurance. Comparing a gross loss numerator with a net premium denominator would mix two different exposures. [1]

For a deliberately simplified annual policy beginning October 1, a $1,200 premium creates $300 of earned premium by December 31 if exposure is earned evenly. The remaining $900 relates to future coverage. A portfolio can therefore have strong written-premium growth while much of the associated revenue has yet to reach the income statement. Collection timing is another dimension: the premium’s accounting recognition does not establish that every dollar has already arrived in cash.

This timing also delays the visible effect of repricing. In a hypothetical book that renews evenly through the year, a rate increase affects successive renewals rather than the entire book at once. The earned-premium benefit then develops during each renewed policy’s coverage period. A comparison of today’s claim costs with today’s announced renewal increases can overstate how quickly those increases repair the margin.

A combined-ratio reconciliation in dollars

Consider a hypothetical insurer with $100 million of net earned premium. Estimated current-accident-year losses and loss-adjustment expenses are $65 million, and underwriting expenses are $28 million. The illustrative loss ratio is 65%, expense ratio 28% and combined ratio 93%. Underwriting profit is $7 million. All figures use the same earned-premium denominator and omit policyholder dividends and other potential complications.

Now suppose new information reduces the estimated remaining cost of claims from earlier accident years by $4 million. Calendar-year incurred losses become $61 million; the combined ratio becomes 89%, and underwriting profit rises to $11 million. If the old claims instead require $6 million of strengthening, losses become $71 million, the combined ratio becomes 99% and underwriting profit falls to $1 million. Neither change means the current year’s policies suddenly became four or six percentage points better or worse.

The arithmetic depends on definitions. A GAAP presentation may use earned premium for both ratios, while a statutory presentation can use a different expense denominator. Adjusted or underlying ratios can exclude specified items. The reconciliation and accounting basis determine comparability; a familiar metric name alone does not. Berkley expressly defines its GAAP expense ratio using net earned premium. [1]

An accident year develops across several calendar years

An accident year groups losses by when the insured event occurred. A calendar-year result includes the current estimate for that year’s events and revisions recognized during the year for older events. An insurer therefore carries several generations of claims estimates at once. Case reserves address reported claims; incurred-but-not-reported estimates also address claims not yet reported and development beyond existing case estimates. These are estimates of liabilities, not a segregated pile of cash. [1]

Travelers attributed its 2025 combined-ratio improvement to a 2.3-point improvement in the underlying ratio and a 0.7-point greater benefit from favorable prior-year development, partly offset by a 0.4-point increase in catastrophe losses. Its underlying ratio was 83.9%. The bridge separates changes in recurring underwriting from changes in estimates and large events; it does not make catastrophe costs economically unreal. [2]

In the hypothetical example, a $4 million reserve release reduces a liability and increases current earnings. It does not itself collect $4 million from a customer. If final payments ultimately match the revised estimate, the release was recognition of better experience. If later information reverses that conclusion, some of the apparent benefit disappears. A favorable revision is neither automatically manipulation nor conclusive proof of superior reserving.

Investment income can offset an underwriting loss

Suppose the hypothetical insurer instead has a 103% combined ratio on $100 million of premium, producing a $3 million underwriting loss. Add $8 million of net investment income and subtract $2 million of other costs: the simplified pretax result is positive $3 million before investment gains, losses and other omitted items. Calling that result a 97% combined ratio would erase the business distinction. A profitable company can have unprofitable underwriting.

The reverse is possible as well: a 95% ratio yields $5 million of underwriting profit, but $8 million of investment losses can more than offset it. Cash needs matter independently. An asset with a positive expected long-run return might be sold at an unfavorable price if claims require cash now. NAIC identifies insufficient reserves, uncollectible assets, reinsurance problems and risky investments among potential causes of insurer insolvency. None is ruled out by a single strong ratio. [3]

Why the same ratio can support different interpretations

A 95% ratio on a stable, quickly settling property portfolio and a 95% ratio on newly written liability risks are not equivalent evidence. The second contains more unresolved uncertainty if claims emerge slowly. Similarly, changing business mix can lower an aggregate ratio without improving any individual line. These are analytical possibilities, not findings about Travelers or Berkley.

The optimistic interpretation of falling ratios is that pricing, selection and claims management have improved. The more cautious interpretation is that favorable weather, old-year releases or changes in mix account for much of the improvement. Both can be partly true. More mature accident-year experience, consistent segment definitions and reconciliation of actual claims payments with successive estimates help separate them.

What subsequent evidence can clarify

The useful evidence is the evolution of claims, not merely the next headline percentage. Paid and incurred development by accident year can reveal whether earlier estimates have held up. Changes in coverage terms, exposure counts and renewal pricing can explain premium growth. Investment cash flows and can explain whether earnings are available when obligations come due. These observations address different questions and should retain their original periods.

The combined ratio remains an effective compact description of underwriting. Its limitation is precisely what makes it useful: it isolates one economic activity. A complete interpretation connects that activity to the uncertainty of claims estimates, the cost of reinsurance, asset performance and capital, rather than asking one ratio to represent the entire insurer.

Sources

  1. W. R. Berkley; 2025 Annual Report and Form 10-K; year ended December 31, 2025, premium recognition, reserves and GAAP ratio definitionsFiling / report · PDFBack to text: ↑1↑2↑3↑4
  2. Travelers; full-year 2025 results; January 21, 2026SourceBack to text: ↑1↑2
  3. NAIC; Receivership; updated March 3, 2026SourceBack to text: ↑

Flag an error or suggest a correction →Public corrections log →