Reading insurer profits through reserve revisions

On October 16, 2026, Travelers reported a third-quarter combined ratio of 86.8%, improving by 0.5 percentage points, while its underlying ratio rose by 0.2 points. The company release provides the figures. Their different directions invite a question about which group of risks an insurance result describes. This article draws no conclusion about the share price and does not treat an estimate revision as evidence of an accounting problem.

The Insurance Information Institute’s explanation describes the combined ratio as claims and expenses relative to premiums, excluding investment income.

An improving aggregate can coexist with pressure elsewhere in current business. Underwriting accounts describe the cost of risks in the present period while also incorporating new knowledge about earlier accidents. Readers need to distinguish improvements in new business from information becoming more complete about old business. Both affect shareholders, but they support different expectations.

Put unsettled claims back into time

The NAIC glossary describes loss reserves as estimated liabilities for incurred, unpaid losses, including unreported claims. The starting point is a payment obligation, rather than a pot of spare money.

Consider an entirely fictional insurer with claims from earlier accidents. At the start of the year it expects to pay another 100 currency units; after further investigation it revises that estimate to 90. To isolate the mechanism, ignore intervening payments, reinsurance, taxes and other changes. In this simplified example, reducing the liability estimate by ten improves current earnings. It does not bring in ten units of new premium. Subsequent evidence of larger losses could cause another upward revision.

Now assume the same insurer's current-period business generates exactly the same revenue and expenses as before. That ten-unit revision alone can improve the aggregate result. Conversely, better cost control in new business can coexist with a worse overall result if earlier claims develop adversely. This example does not represent Travelers' actual accounts. It shows why a period's total change is insufficient evidence about the quality of new business.

Interpreting favorable revisions

One possible explanation is that earlier estimates allowed room for outcomes that proved less costly than feared. The improvement has economic meaning and should not be erased simply because it relates to an earlier period. Another possibility is that information remains incomplete and the revised estimate has not yet been tested by sufficiently mature payment experience. The favorable direction alone cannot distinguish these interpretations.

Nor should readers assume that a falling reserve demonstrates management is polishing earnings. That would turn a question requiring evidence into an allegation. A better approach follows the same claims over time: how the earlier estimate was formed, why it changed, and whether later outcomes support the revision. The reasonableness of an adjustment and the repeatability of its earnings benefit are separate questions.

Consistency matters. Removing adverse prior-period developments while treating all favorable ones as present operating strength improves the apparent trend by construction. Yet dismissing every revision also overlooks the gradual emergence of outcomes from past risk selection. A transparent reading preserves the overall result and explains its components instead of selecting the definition that produces the most attractive number.

Underlying measures also need interpretation

Separating earlier-period changes and catastrophe effects can sharpen a question. It cannot automatically reveal a profit figure free of estimates. Even measures closer to current business require an understanding of their definitions, risk mix and maturity of information. Similar names do not guarantee an identical scope of comparison.

In another hypothetical, two insurers receive the same premiums. One covers many small risks whose losses become clear quickly; the other accepts a few liabilities whose eventual payments take much longer to establish. An identical ratio at the reporting date cannot fully express that difference in uncertainty. What one insurer already knows may still be an estimate for the other. Before comparing business quality, ask how far the evidence has developed.

Research hosted by the Casualty Actuarial Society cautions that measures of reserve uncertainty can depend on strong assumptions. Our analytical implication is to examine what information could move an estimate, not merely its central value. A small numerical movement under unchanged assumptions and a judgment revised after new claims evidence need not carry the same implications.

Make the quarterly interpretation testable

The useful output is an explanation that can be checked later, rather than an immediate good-or-bad label for earnings quality. If an improvement is attributed mainly to experience on older accidents, subsequent analysis should return to those accidents, not just the next quarter's profit. If better pricing of new business is the explanation, it needs evidence from the corresponding losses and expenses. Gains on older business cannot supply that evidence by themselves.

This leaves room to acknowledge good news. Reported earnings remain earnings, and a downward estimate revision may reflect genuinely lower obligations than previously expected. Its power to explain future new business is nevertheless limited. Readers can recognize the present result while keeping its persistence an open question.

The practical lesson from this disclosure is modest: identify the period of risk behind each change before discussing sustainability. Insurers make commitments to pay under uncertain future circumstances, while outsiders receive evidence in installments. Updating an interpretation as evidence emerges about the same claims is more faithful to the business than treating every quarterly movement as a turning point in the operating trend.