The Hartford booked $46 million of general liability prior-year reserve increases in Q2 2026, part of $116 million in total GL adverse development spanning accident years 2017 through 2019 and 2022 through 2023, and management named the cause explicitly: a higher frequency of large losses reaching excess casualty and umbrella attachment points, not a shift in severity trend ("The Hartford Reports Strong Second Quarter 2026 Financial Results," The Hartford, July 24, 2026).

The company also booked $26 million of adverse development in commercial auto liability, tied to accident years 2023 and 2024 and attributed to higher severity from increasing attorney representation and time-limit demands. Both charges landed inside an otherwise strong print: core earnings of $945 million, or $3.42 per diluted share against a $3.21 consensus, a trailing-twelve-month core earnings ROE of 18.7%, and book value per share excluding AOCI of $78.91, up 7% year to date. The stock fell 2.89% to $138.28 in the session after the July 24 call anyway, as investors weighed the reserve commentary against the beat.

What Hartford Actually Named

CFO Beth Costello characterized the $46 million quarterly addition as "a very modest increase when you think about the reserve base," and CEO Christopher Swift framed the review as continuity, not a change in process, telling analysts loss reserving is "both a science and art" that the company has "combined pretty well over a long-term period." Both statements are defensible on the numbers. A $46 million addition against a general liability reserve base running into the billions is not, by itself, evidence of systemic inadequacy.

What is unusual is the specificity of the stated driver. Carriers routinely attribute adverse development to "elevated large-loss activity" without saying where in the tower it originated. Hartford's disclosure went further: the frequency of large losses is what changed, and it changed specifically at the excess casualty and umbrella layers, across accident years old enough (2017 through 2019) and recent enough (2022 through 2023) to rule out a single underwriting-year anomaly. That combination, an explicit frequency attribution spanning six non-contiguous accident years, is not the pattern a pricing actuary would expect from a normal large-loss noise cycle.

Frequency at the Attachment Point Is a Different Assumption Than Severity Trend

Excess and umbrella pricing leans on increased-limit factors and layer-specific loss cost relativities that assume the probability of any given primary claim developing into a loss large enough to pierce the attachment point is low and reasonably stable period to period. A trend load applied at the primary limit, carried up through the ILF curve, is built to absorb inflation in the severity of claims that already reach the layer. It is not built to absorb a change in how often claims reach the layer at all. If the share of primary claims that eventually pierce a $5 million or $10 million attachment point is rising, that is a shift in the shape of the frequency-severity split feeding the excess book, and no amount of severity trend correctly re-prices it, because the trend load was calibrated against the old piercing rate.

The mechanism matters because it also contaminates the historical development pattern the desk relies on to set the current-year loss pick. A loss development triangle built on years when large losses pierced the layer at one rate, and is now developing losses that pierce at a higher rate, will show adverse development in the older diagonals precisely as Hartford's did, across 2017 through 2019 as well as 2022 and 2023. That is a signal the loss development factors themselves, not just the point estimate for the current accident year, may be running behind a frequency regime that has already shifted. A tail factor selected from a stable-frequency era understates ultimate losses in a rising-frequency era even before severity trend is applied on top.

The commercial auto charge is the feeder mechanism worth watching alongside it. Auto liability remains the leading source of catastrophic claims reaching umbrella and excess towers, particularly multi-vehicle and pedestrian-severity events in metro exposure. Hartford's $26 million auto charge was framed as a severity story, higher attorney representation and time-limit demand activity pushing individual claim costs up, but a primary auto claim that grows large enough to trigger a time-limit demand is also a claim more likely to pierce an umbrella attachment point sitting above it. A severity problem at the primary layer and a frequency problem at the excess layer are not two separate stories on the same balance sheet. They can be the same claims, observed from two different altitudes in the tower.

The Renewal Backdrop This Lands Into

Hartford's disclosure arrives as excess and umbrella pricing is already running hard relative to the rest of commercial lines. WTW's Insurance Marketplace Realities 2026 report on casualty puts umbrella liability rate increases at 10% to 20% and excess liability at 8% to 18% for 2026, with lead umbrella and excess programs averaging rate increases above 12% (WTW, Insurance Marketplace Realities 2026, Casualty). More than a quarter of lead umbrella programs are being restructured on capacity constraints rather than simply repriced, which means the primary lever at renewal for a growing share of accounts is attachment point and limit deployment, not rate per million alone (WTW, 2026).

That firming is happening against a broader commercial market that is giving rate back everywhere else. Alliant's mid-year 2026 marketplace report, covered by The Insurer on July 20, 2026, found property and cyber softening for U.S. buyers even as casualty and transportation lines stay under pressure from litigation costs and large verdicts (Alliant, Insurance Marketplace Insights and Observations, mid-year 2026). The verdict data underneath that pressure kept climbing through 2025: 149 nuclear verdicts exceeding $10 million totaled $25.1 billion across 28 states, and the median nuclear verdict has risen from $21.5 million in 2020 to more than $44 million by 2025 (Litigation Sentinel; U.S. Chamber Institute for Legal Reform, 2025). Hartford's frequency framing, more claims reaching the layer rather than each claim costing more once it arrives, is a distinct mechanism from the median-severity story that verdict trackers usually lead with, and it argues the 2026 lead umbrella rate increases are being earned by loss experience rather than pushed defensively ahead of it.

One Carrier's Signal, or a Sector Pattern

The open question for a reserving or pricing desk reading Hartford's print is whether this is a persistent frequency regime across the excess casualty market or accident-year noise specific to Hartford's book. The evidence cuts both ways. CNA booked $106 million of casualty reserve strengthening in Q1 2026, with $56 million of it specifically in excess casualty alongside $50 million in professional E&O, a charge this site covered in detail at the time (see CNA's Q1 2026 Casualty Reserve Charge Flags Soft-Cycle Risk). Two national carriers naming excess casualty specifically within a single quarter of each other is a harder pattern to dismiss as an isolated underwriting-year problem at one shop.

Travelers cuts the other way. On its own Q2 2026 call five days after Hartford's, on July 17, 2026, Travelers reported $578 million of favorable prior-year development and CFO commentary that the company was seeing no pressure in casualty lines including umbrella or commercial auto during the quarter (Travelers, Q2 2026 earnings call, July 17, 2026). That is not proof the frequency shift is contained to Hartford and CNA; portfolio mix, attachment point selection, and claims-handling philosophy differ enough across carriers that one book can show a frequency signal while another, with a different limits profile or a more aggressive claims resolution posture, does not. But it means the desk should treat Hartford's disclosure as a book-specific data point that demands verification against the account's own large-loss experience, not as an industry-wide repricing signal to be applied uniformly across a renewal book.

Where the Next Print Lands

The next confirming or disconfirming data point is Hartford's Q3 2026 10-Q, due in late October, and whichever peers report GL or excess casualty reserve actions alongside it, particularly Chubb and CNA's next casualty-specific commentary. A repeat frequency-driven charge in Q3, especially one that again names excess casualty and umbrella specifically rather than GL in aggregate, would be much harder to read as accident-year noise. A quarter with no comparable disclosure would support Costello's "modest" framing.

In the meantime, at a lead umbrella or excess renewal between now and year end, the question worth putting to the broker or the ceding carrier is not the aggregate loss ratio on the layer but the large-loss count by attachment point over the last three to five accident years, separated from the average severity of those losses. If the frequency of claims piercing $5 million or $10 million has been climbing on a specific account even while severity per claim looks stable, that account's ILF curve and tail factor selection deserve a re-derivation against current piercing rates rather than another year of trend rolled forward on an assumption that stopped holding somewhere back in 2022.

Sources

  1. The Hartford, "The Hartford Reports Strong Second Quarter 2026 Financial Results," July 24, 2026 - sec.gov
  2. The Hartford Insurance Group (HIG) Q2 2026 earnings call transcript, July 24, 2026 - investing.com
  3. WTW, "Insurance Marketplace Realities 2026 - Casualty," 2026 - wtwco.com
  4. The Insurer, "US insurance buyers see relief in property, cyber but casualty costs keep climbing," citing Alliant's mid-year 2026 report, July 20, 2026 - theinsurer.com
  5. Insurance Business America, "CNA's Q1 earnings hit by casualty reserve action and weaker underwriting," 2026 - insurancebusinessmag.com
  6. Insurance Journal, "Q2 Net Income at Travelers Soars 46% on Less Catastrophes, Favorable Reserves," July 17, 2026 - insurancejournal.com
  7. Litigation Sentinel, Nuclear Verdicts Dashboard, 2026 tracking; U.S. Chamber of Commerce Institute for Legal Reform, median verdict data, 2025

Further Reading