Berkshire's $869M Reserve Release Hides a Casualty Build

Berkshire Hathaway's reinsurance division booked $869 million of favorable prior-year reserve development in the first six months of 2026, up from $506 million a year earlier (Berkshire Hathaway 10-Q, August 2026). That release is a blend: lower property loss estimates funded the number, while casualty loss estimates rose underneath it, the same long-tail pressure showing up at Everest, CNA, and AIG this earnings season.

$869M

H1 2026 Reinsurance Reserve Release

53.6%

P&C Reinsurance Loss Ratio, Down From 61.4%

$483M

Tokio Marine Assumed Premium

3.4%

Gross Claim Liabilities Ceded to Retrocessionaires

The Segment Note Behind the Headline Number

Berkshire's insurance operations run through three segments: GEICO, Berkshire Hathaway Primary Group (BH Primary), and Berkshire Hathaway Reinsurance Group (BHRG), which writes excess-of-loss and quota-share property and casualty treaties through the National Indemnity Company (NICO), General Re, and TransRe groups. The $869 million figure sits inside BHRG's property/casualty line, disclosed in the property/casualty reinsurance underwriting table of the August 2026 10-Q. Berkshire's own words describe the mechanics precisely: "changes in prior accident years ultimate loss estimates reduced losses and LAE by $609 million in the second quarter and $869 million in the first six months of 2026... The reductions in each period were mostly attributable to lower-than-expected property losses" (Berkshire Hathaway 10-Q, August 2026). A separate passage in the same filing completes the sentence: the property-driven reductions came "partially offset by increases in estimated losses for casualty exposures."

The rest of the property/casualty reinsurance table confirms how much of the quarter's strength traces to that single line. Pre-tax underwriting earnings for the segment rose to $1.775 billion in the first half of 2026 from $1.113 billion a year earlier, a 59.5% increase, while the loss ratio improved to 53.6% from 61.4%, a 7.8-point swing. Premiums earned actually declined, from $10.343 billion to $9.941 billion, so none of the earnings gain came from top-line growth. It came from the loss side of the ledger, and the reserve note explains where on the loss side that improvement originated.

BH Primary's own reserve release, disclosed three paragraphs earlier in the same filing, is the useful counterpoint. BH Primary reduced its prior accident year ultimate loss estimates by $444 million in the first half of 2026, reversing a $401 million increase booked in the same period of 2025, and Berkshire attributes that reduction "primarily to lower-than-expected property losses and, to a lesser extent, casualty losses" (Berkshire Hathaway 10-Q, August 2026). Both loss categories moved in the same direction for BH Primary. Only BHRG's release is a net figure built from two components pulling against each other, and that distinction does not survive the trip from the 10-Q's footnotes to a headline earnings recap.

The Catastrophe-Free Half That Flatters the Combined Ratio

Reserve development was not the only tailwind. BHRG recorded no losses from catastrophe events exceeding the $150 million reporting threshold in the first half of 2026, compared with $760 million from California wildfires in the first quarter of 2025 (Berkshire Hathaway 10-Q, August 2026). A segment posting a 53.6% loss ratio against a $760 million cat-loss comparison base the year before is clearing a low bar twice: no large events, and a reserve release nearly $363 million larger than the one it booked in 2025.

Consolidated results show the same pattern from a different angle. GEICO's pre-tax underwriting earnings fell to $994 million in the second quarter of 2026 from $1.821 billion a year earlier, a 45% decline, while BHRG's second-quarter earnings rose to $913 million from $650 million, a 40% increase (Berkshire Hathaway 10-Q, August 2026). Total pre-tax insurance underwriting earnings fell 14% to $2.18 billion in the quarter (The Insurer, August 2026), even as BHRG's own number improved sharply. Two segments inside one filing moved in opposite directions for opposite reasons: GEICO's decline traces to rising claims costs, and BHRG's gain traces to a reserve release that a catastrophe-quiet half made easier to book. Reading BHRG's improvement as evidence of stronger underlying reinsurance pricing, rather than the absence of large losses plus a favorable reserve mix, overstates what the quarter actually shows.

GEICO's Bodily Injury Severity Inside the Same Filing

The casualty pressure BHRG cites as a partial offset to its property release is not an abstraction elsewhere in the same 10-Q. GEICO's loss ratio rose to 75.3% in the first half of 2026 from 70.4% a year earlier, a 4.9-point deterioration, which the filing attributes to "higher claims frequencies and average severities." Private passenger auto bodily injury claims frequency increased in a five-to-seven percent range and average severity increased ten to twelve percent, while property damage and collision frequency rose three to five percent (Berkshire Hathaway 10-Q, August 2026). GEICO's own loss ratio deterioration is not the same claims population as BHRG's assumed casualty treaties, but it is a same-company, same-quarter data point confirming that the severity trend BHRG's reserve actuaries cite as a partial offset is visible inside Berkshire's own primary auto book, not just in ceded treaty data from third-party cedants.

That internal consistency matters for how much weight to put on the word "partially" in BHRG's reserve language. A carrier whose captive auto insurer is absorbing double-digit bodily injury severity increases in the same reporting period has direct visibility into casualty severity acceleration, which raises the bar for treating its own reinsurance casualty offset as conservative rather than as a lagging estimate still catching up to trend.

National Indemnity's Tokio Marine Assumption Into a Falling-Rate Market

BHRG's premium volume tells its own story about market conditions. Excluding a new arrangement, reinsurance premiums written declined 5.6% in the second quarter and 3.8% in the first half of 2026, "primarily due to lower property volumes" (Berkshire Hathaway 10-Q, August 2026). What kept total premiums roughly flat was a new whole-account quota-share agreement with subsidiaries of Tokio Marine Holdings, which commenced April 1, 2026 and contributed $483 million of premiums written in the first half. Under the ten-year agreement, NICO assumes a portion of Tokio Marine's non-life premiums, losses, and expenses on a quota-share basis, alongside a separate transaction in which NICO acquired a 2.5% equity stake in Tokio Marine (Reinsurance News, 2026).

The timing is not incidental. Risk-adjusted global property-catastrophe reinsurance rates-on-line fell an average of 14.7% at the January 1, 2026 renewals, the steepest year-on-year decline since 2014 and an acceleration from an 8% decline in 2025 (Howden, January 2026). National Indemnity is filling the gap left by falling organic property pricing with an assumed book acquired through a strategic transaction rather than through renewal terms it negotiated line by line. A quota share assumed at the start of a rate-declining cycle inherits whatever loss ratio assumptions Tokio Marine and NICO agreed to at inception, and because the treaty only began April 1, 2026, none of its experience has had time to emerge into a prior-year development figure yet. The $869 million release describes business already on the books; the Tokio Marine book is the marginal premium keeping BHRG's top line from shrinking, and its loss experience is a 2027 and 2028 question, not a first-half-2026 one.

The Recoverable Line: How Little Berkshire Cedes Away

The casualty component buried inside BHRG's reserve release does not get distributed to third-party retrocessionaires the way it would at a reinsurer running an active outward retro program. Berkshire's consolidated balance sheet at June 30, 2026 carried $122.875 billion in gross property and casualty claim liabilities against $4.235 billion recoverable from other reinsurers, meaning just 3.4% of Berkshire's gross P&C liabilities is ceded away, down from 3.9% a year earlier when recoverables stood at $4.687 billion against $118.788 billion gross (Berkshire Hathaway 10-Q, August 2026). That is the quantitative signature of a company built to retain risk rather than transfer it: National Indemnity and General Re write treaties and hold the resulting liabilities on Berkshire's own balance sheet, funded by the float those premiums generate rather than laid off through a retrocession panel.

The contrast with peers actively reducing casualty exposure through cession and volume cuts is direct. Swiss Re cut June and July 2026 casualty treaty renewal volume 5.9% from "continued restructuring of liability lines," and Munich Re executed a 61% reduction in its retrocession program at the April 2026 renewal, describing the move as declining to renew business that did not meet return expectations (see actuary.info's coverage of Swiss Re's H1 2026 casualty restructuring). Those reinsurers are actively narrowing their casualty exposure at the point of underwriting. Berkshire's posture is the opposite: it retains substantially all of what it writes, and the 3.4% recoverable ratio means whatever the ultimate casualty severity trend turns out to be inside the current book, Berkshire's own balance sheet absorbs nearly all of it rather than sharing the outcome with a retrocession panel.

Casualty Reserve Actions Across the Same Earnings Season

Berkshire's mixed reserve note is one data point inside a broader pattern of casualty reserve activity that surfaced across P&C and reinsurance filings in the same reporting window.

Company	Q2/H1 2026 Casualty Reserve Action	Context
Berkshire (BHRG)	Casualty losses increased, offset within a net $869M favorable release	53.6% H1 loss ratio, no cat losses over $150M
Everest	~$200M North America casualty treaty reserve strengthening	88.5% Reinsurance Treaty combined ratio
AIG	$145M net favorable development, excess casualty saw slight strengthening	88.1% combined ratio
CNA	$77M mass-tort reserve charge	Same quarter as Chubb and Travelers releases
Chubb	$283M favorable prior-year development	Q2 2026
Travelers	$578M favorable prior-year development	Q2 2026

Everest strengthened North American casualty treaty reserves by just under $200 million in the second quarter, choosing to react to emerging loss trends immediately rather than wait for its third-quarter annual reserve study (Bermuda Re, August 2026). Everest CEO Jim Williamson was explicit about the driver: the U.S. tort environment "is corrosive to a well-functioning economy and is putting pressure on industry reserves" (Bermuda Re, August 2026). That is a direct, named-executive statement about the same casualty severity trend that BHRG's reserve note describes only as a partial offset buried inside a net favorable figure. Details on how AIG, Chubb, Travelers, and CNA split their own quarters between casualty strengthening and releases are covered in actuary.info's Q2 2026 casualty reserve divergence tracking and its AIG Q2 2026 reserve development analysis.

Reserve and Capital Implications of a Retained Long-Tail Book

For a reserve actuary reviewing Berkshire as a counterparty, a cedant, or a comparable, the practical implication is that the $869 million figure needs to be read as two components, not one net number. The property component reflects genuinely favorable loss experience; the casualty component reflects the same severity trend showing up across Everest, CNA, and GEICO's own auto book in the identical reporting period. Netting them together in a single headline obscures the direction of the underlying trend even though Berkshire's own footnotes disclose both halves explicitly. An analyst benchmarking BHRG's loss development factors against industry casualty triangles, the kind of exercise actuary.info has walked through in its casualty triangle adverse-development methodology piece, would find the property offset masking exactly the signal that methodology is designed to isolate.

The capital implication follows from the retention data. Because only 3.4% of Berkshire's gross P&C liabilities are ceded, any future adverse development on the casualty book embedded inside the current release falls almost entirely on Berkshire's own reserves and, ultimately, its float. That is a different risk profile than a reinsurer running an active retrocession program, where a casualty miss is partially absorbed by retrocessionaires. Berkshire's insurance investment income allocated across its three underwriting segments totaled $7.008 billion in the first half of 2026, down from $7.624 billion a year earlier (Berkshire Hathaway 10-Q, August 2026); a retained long-tail casualty book that develops adversely in future periods would weigh on reported underwriting income directly, with no cession to diffuse the impact, at the same time the float generating that investment income has itself contracted.

The Tokio Marine transaction adds a second layer to that exposure. A quota share written into a market where risk-adjusted property-cat pricing fell 14.7% at the start of the year is a book whose adequacy will not be testable in reserve reviews for several years, since its first loss development will not surface until 2027 estimates are set against 2026 experience. Reserve actuaries evaluating Berkshire's future filings have a specific item to watch: how the Tokio Marine book's own prior-year development reads once it has enough history to produce one, and whether the pattern that shows up resembles the property-favorable, casualty-adverse mix BHRG is already booking on its legacy treaty portfolio.