Arch's 83.5% Combined Ratio Blends a Primary Loss-Pick Slip

Arch Capital Group's headline 83.5% group combined ratio for the second quarter of 2026 blends two segments moving in opposite directions: reinsurance ran at 77.5% while the insurance segment rose to 98.5% from 93.4% a year earlier (Arch Capital Q2 2026 earnings release, July 28, 2026). Strip out catastrophes and prior-year reserve releases and the group's underlying accident-year loss ratio still widened, to 82.5% from 80.9%. That is the figure a premium-weighted average averages away.

Key Takeaways

• 83.5% group combined ratio blends a 77.5% reinsurance segment against a 98.5% insurance segment, the latter up more than five points from 93.4% a year earlier.
• 82.5% is the group accident-year loss ratio excluding catastrophes and prior-year development, up from 80.9%. On $3.985 billion of net premium earned that is roughly $64 million of incurred loss.
• 12 points separate the two segments on the noise-stripped basis: 91.6% for insurance against 79.9% for reinsurance.
• $165 million of favorable prior-year development, distributed as 15.7 points of loss-ratio benefit to mortgage, 5.3 to reinsurance and 1.4 to insurance. The cushion is thinnest where the deterioration is.
• Net premiums written fell 6.9% to $4.05 billion while gross was roughly flat at $6.13 billion. The retreat sits in what Arch retained, not what it wrote.

Decomposing the 83.5%: Two Segments, One Number

Arch reported on July 28 in its earnings release and financial supplement, filed with the SEC as exhibit 99.1 to a Form 8-K. Net income available to common shareholders came to $1.047 billion, or $3.00 per share, an 18.0% annualized return on average common equity, with book value per share up 2.8% from the first quarter to $68.04.

The segments did not move together. The insurance segment, Arch's primary commercial and specialty book, posted a 98.5% combined ratio, with underwriting income falling to $27 million from $129 million. Reinsurance ran at 77.5% on $410 million of underwriting income against $451 million a year earlier, and mortgage posted its usual structurally low 22.8% on $220 million (Insurance Business America, July 2026).

Catastrophes explain part of the insurance deterioration. Cat activity added 7.6 points to that segment's loss ratio against 2.9 points a year earlier, a 4.7-point swing that alone covers most of the year-over-year gap. Group-wide pretax cat losses net of reinsurance and reinstatement premiums totalled $201 million, with 8.0 points of the load in insurance against 2.8 in reinsurance. The segment's expense ratio also rose to 35.5% from 33.6%, partly on transitional integration costs from the 2024 acquisition of the U.S. MidCorp and Entertainment businesses from Allianz.

The Number That Survives Cat and Reserve Noise

Cat losses are volatile by nature and prior-year development is about business already written. Neither speaks to whether the current accident year is priced adequately. The metric that does is the accident-year loss ratio excluding both, and at group level it moved from 80.9% to 82.5%, a 1.6-point widening. On $3.985 billion of net premium earned that is roughly $64 million of incurred loss a cat-free, development-free comparison did not carry a year earlier.

The segment split is where it becomes actionable. Arch's financial supplement puts the insurance segment's combined ratio excluding catastrophes and prior-year development at 91.6% against 79.9% for reinsurance. A 12-point gap on the noise-stripped basis is not a mix effect. It points to a primary book absorbing loss trend and casualty severity faster than its 2026 rate actions have offset, while a reinsurance book priced off cedents' own experience and repriced annually at renewal has kept pace.

Favorable prior-year development added $165 million to the quarter group-wide, and its distribution matters more than its size. As a loss-ratio benefit it ran 15.7 points in mortgage, 5.3 points in reinsurance and 1.4 points in insurance. Mortgage releases are a recurring feature of a line with short, well-modelled emergence. Reinsurance's 5.3 points are the consequential entry, because they mean a meaningful share of that segment's 77.5% is a favorable revision to past accident years rather than current-year underwriting profit.

Net of that release, reinsurance's 79.9% is the figure comparable to insurance's 91.6%: still better, though not by the margin a 77.5% against 98.5% comparison implies. The insurance segment's own 1.4 points barely offset its underlying deterioration, which is itself the finding. A primary book with slipping current-year picks is not being carried by its own back book the way reinsurance is.

Segment	Combined ratio, Q2 2026	Ex-cat, ex-dev combined ratio	Cat load (points)	Favorable PYD (points)
Insurance	98.5%	91.6%	7.6	1.4
Reinsurance	77.5%	79.9%	2.8	5.3
Mortgage	22.8%	39.8%	—	15.7

The pattern is not company-specific. RenaissanceRe's Q2 2026 showed a comparable divergence between its property and casualty retro books, covered in the site's read of its 72.8% combined ratio, and Chubb's release showed the same casualty-versus-property split, examined in its 83.8% combined ratio against a softening casualty book.

The Cushion Is Thinnest Where It Is Needed

The asymmetry in that $165 million is structural rather than incidental, and it constrains what the next few quarters can absorb. A reinsurer with a large back book of short-tail property business can release from a benign recent cat year without that release saying anything about how the current treaty year is priced. A primary casualty writer has less of that available in the years immediately after a soft-market vintage, because casualty development is slow enough that adverse trend in a recent accident year has often not emerged by the time the reserve is tested.

That is exactly the shape of Arch's numbers: 5.3 points of favorable development in reinsurance against 1.4 in insurance, on a primary segment already 12 points worse on the noise-stripped basis. An actuary reviewing insurance-segment reserves should treat that 1.4 points as thin cover rather than confirmation that current picks are conservative, and should read the 98.5% headline as flattered in the opposite direction by a 7.6-point cat load that will not repeat every quarter.

The premium line suggests Arch's own actuaries reached the same conclusion. Net premiums written fell 6.9% group-wide to $4.05 billion and net premiums earned 8.1% to $3.985 billion, while gross premiums written were roughly flat, down 1.1% to $6.13 billion. The decline is concentrated in what Arch chose to retain and cede, not in what it wrote at the top line. Inside the insurance segment, net written fell 5.1%, or closer to 1.8% excluding non-renewals of certain acquired programs.

A diversified underwriter pulling back on net retained primary premium while gross volume holds is consistent with declining rate-inadequate business rather than losing share, the tell flagged in the site's soft-market reserve adequacy playbook. CEO Nicolas Papadopoulo called it "a strong quarter, driven by solid underwriting performance across our three segments." That reads cleanly at group, reinsurance and mortgage level. At the insurance segment, a 91.6% underlying combined ratio and a shrinking release is the part of disciplined execution still catching up.