The Hartford booked a Business Insurance combined ratio of 91.4 in the second quarter of 2026, down four points from 94.8 in the first quarter (Q2 2026 earnings release).

Net favorable prior-year development fell to $111 million pretax from $187 million a year earlier, a $76 million swing driven by strengthening in general liability and commercial auto that offset releases in workers' compensation, catastrophes, personal insurance and bond.

Key Takeaways

  • The underlying combined ratio was 89.2 in Q1 and 89.3 in Q2. Nearly all the four-point sequential improvement is the absence of a $70 million one-time charge rather than margin gain.
  • Year over year the ratio deteriorated from 87.0 to 91.4, which the release attributes to 2.9 points of less-favorable prior-year development and 0.2 points of higher current-accident-year catastrophe losses.
  • General liability rate rose 9.7% in Q1, 50 basis points ahead of the fourth quarter's 9.2%, which is not the pricing posture of a carrier that considers its GL loss trend settled.
  • CFO Beth Costello said in April there was "no change in our net GL reserves, excluding the legacy item." GL sits on the unfavorable side again in Q2 with no legacy exposure named.
  • Swiss Re cites $15.8 billion of adverse US casualty development in 2024 while cutting casualty reinsurance volume, on the same two lines.

Two Comparisons, Two Different Quarters

The sequential framing is the flattering one: 91.4 against 94.8 reads as a clean turn. But quarter-over-quarter combined ratio comparisons in a book with weather seasonality and lumpy reserve reviews are a poor lens on reserve quality. Q1 absorbed a $70 million general liability charge for legacy sexual molestation and abuse exposures on policies written in the 1970s and 1980s, including a provision for a settlement in principle in a bankruptcy proceeding involving a religious institution.

ComponentQ1 2026Q2 2026Read
Business Insurance combined ratio94.891.4Improved 3.4 points sequentially
Underlying combined ratio89.289.3Essentially flat; underlying margin did not improve
One-time legacy GL charge$70MNone disclosedAbsence alone explains most of the sequential swing
Companywide net favorable PYD$75M (ex-legacy charge)$111MHigher in dollars, but GL and commercial auto now sit on the unfavorable side

Comparing against Q2 2025 isolates the reserve trend from that one-time item. On that basis the Business Insurance combined ratio moved from 87.0 to 91.4, a four-point deterioration the release attributes to 2.9 points of less-favorable prior-year development and 0.2 points of higher current-accident-year catastrophe losses. The underlying ratio, stripping both out, rose to 89.3 from 88.0.

The composition matters more than the dollars. Excluding the legacy charge, Q1 companywide net favorable development was $75 million, concentrated in workers' compensation, homeowners and personal auto. Q2's $111 million is not a larger release from the same sources; it is a different mix, with catastrophes, bond and Personal Insurance joining workers' compensation on the favorable side while general liability and commercial auto crossed to the unfavorable column.

The net figure recovered sequentially. The composition shifted onto the two lines with the longest tail and the least forgiving correction cost if the picks are wrong.

The Line Management Said Needed No Action

The Q1 charge was explicitly legacy business, and management drew a sharp line around it. Costello told analysts that excluding it, "we feel very good about our loss picks in the GL book, both from the standpoint of prior year reserves and the loss trend that we've embedded in our 2026 picks," adding there was "no change in our net GL reserves, excluding the legacy item" (Q1 call transcript). That is a falsifiable claim: the current book, as of April, needed no reserve action.

Three months later general liability appears on the unfavorable side again, without a bankruptcy settlement or named legacy exposure attached in the public disclosure. Either the Q2 movement continues the same legacy tail, in which case the April framing needs updating, or it reflects the current book Costello described as settled.

The pricing points the same way. GL rate rose 9.7% in Q1, 50 basis points ahead of the fourth quarter's 9.2%. Carriers do not push rate at high single digits on a line whose reserve picks they consider adequate; they push rate where loss trend keeps outrunning the prior assumption.

Commercial auto strengthening alongside it is close to an industry constant since roughly 2017, and the mechanism explains why the line strengthens even when frequency looks benign. Severity is driven by a small share of large claims, nuclear verdicts and above-limit settlements concentrated in a handful of venues, rather than by claim volume. Frequency can run flat or improve on telematics adoption and hours-of-service compliance while tail severity re-accelerates, because litigation financing and jury awards operate independently of how many accidents occur.

That decoupling is why development factors selected off a triangle with five or six years of stable-looking history understate ultimate losses. The severity events driving corrections are rare enough that a moderate accident-year window looks calm until a handful of verdicts land, at which point the trend reveals itself all at once as adverse development on years that had appeared adequately reserved. A single bad settlement quarter is a plausible explanation; a recurring pattern is the signal that the loss-trend assumption needs resetting rather than the point estimate.

The Offset Depends on a Release That Also Runs Out

The structure here, short-tail redundancy offsetting long-tail strengthening inside one reporting period, is the mechanism by which a carrier reports a stable headline ratio for several quarters while the casualty reserve position weakens beneath it. Business Insurance written premium grew 5% to $4.02 billion, so the top line is real. The underlying ratio, 89.2 against 89.3, did not move.

Workers' compensation has been a dependable source of favorable development industrywide since 2017, as frequency ran below pre-pandemic baselines. Leaning on that release to fund a casualty build is a bet that the redundancy persists long enough, and in sufficient size, to keep covering GL and commercial auto strengthening before net development turns unfavorable outright.

What the market can check against that bet is thin. The earnings release and call give a sequential combined ratio; a reserve opinion needs accident-year-level development on GL and commercial auto specifically, which shows up in the 10-Q's loss reserve discussion and schedule P-style disclosures rather than the press release. Whether the Q2 strengthening concentrates in the most recent two or three accident years, consistent with an emerging severity trend not yet in the loss-trend pick, or spreads across a longer tail, is not observable from outside. That gap is why "improved combined ratio" and "casualty reserve build" are both accurate descriptions of the same quarter.

The reinsurance side is finding the same two lines. Swiss Re has been cutting casualty reinsurance volume even while posting record profitability, citing $15.8 billion of adverse prior-year casualty development across the US industry in 2024 alone, the same pattern social inflation has been driving through loss development factors. General liability and commercial auto are where the market's most sophisticated risk-takers keep finding deficiency, whichever side of the treaty they sit on, and a headline improvement built on the absence of a prior charge does not change which side of that ledger Hartford's casualty book is on.

Further Reading

Sources