Reinsurers Carry 40% of Cat Exposure but Pay Under 25% of Losses

Howden Re's September report puts reinsurers' average share of natural-catastrophe exposure at 40% since 2021, yet their share of 2025 paid catastrophe losses stayed below 25% for a fourth year (Howden Re, September 2026). High attachment points left the frequent loss layer with primary carriers while reinsurers carried the tail.

David Flandro, Howden Re's head of industry analysis, described a market where "profitability is strong, capital is abundant and reinsurance pricing continues to soften" (Howden Re, September 2026). The loss-share wedge shows which balance sheet earned that profitability and which one funded the quiet years.

Key Takeaways

• 15 percentage points separate reinsurers' 40% average share of catastrophe exposure since 2021 from their sub-25% share of 2025 losses. Contract attachment, rather than appetite, explains the gap.
• Under $26.8 billion of Swiss Re's $107 billion in 2025 insured natural-catastrophe losses reached reinsurers on Howden Re's share estimate, leaving primary carriers with more than $80.2 billion.
• 16 points is the 2022 backtest miss: Howden Re's model produced a +21% pricing signal while its observed full-year risk-adjusted property-catastrophe rate-on-line index rose 37%.
• $89.5 billion of dedicated capital disappears in the severe scenario, from $505 billion to $415.5 billion, even though $76.7 billion of investment income offsets part of the underwriting and bond shock.
• Near zero economic value added by mid-2026 sits beside AM Best's 16.3% median 2025 return on equity, exposing how accounting profit and the return above all capital costs can diverge.

The 15-Point Loss-Share Wedge

The two percentages measure different things. The 40% is Howden Re's estimated ceded share of total insured natural-catastrophe exposure, averaged since 2021. The loss share is what contracts actually paid after events met occurrence definitions, deductibles and layer attachments. An exposure allocation describes potential participation across the distribution; a paid loss share records where the realized events landed.

The 2025 event mix landed low in the tower. Swiss Re Institute counted $107 billion of insured natural-catastrophe losses, with secondary perils producing a record 92%. Severe convective storms contributed $51 billion and the Los Angeles wildfires $40 billion (Swiss Re, March 2026). Repeated medium-sized claims can exhaust primary catastrophe budgets without breaching occurrence treaties designed around hurricanes and earthquakes.

Applying Howden Re's sub-25% share to Swiss Re's total gives an upper bound of $26.75 billion paid by reinsurers and more than $80.25 billion retained by insurers. That arithmetic does not allege failed protection. It shows the selected protection responding as written after the 2022 hard market moved retentions upward.

Those retentions survived the price turn. Howden's January renewal report said global property-catastrophe rates fell 14.7%, while retrocession fell 16.5%, yet attachments remained higher than in the previous soft market (Howden, January 2026). By July, Guy Carpenter put the global property-catastrophe rate-on-line decline at 16% (Guy Carpenter, July 2026).

For a primary pricing actuary, the observed reinsurance recovery ratio is therefore a poor substitute for a prospective ceded loss cost. The rate filing still carries the frequency layer below attachment, even if the treaty premium falls. As actuary.info's earlier attachment-point analysis showed, a cheaper unchanged tower lowers ceded expense while leaving retained catastrophe volatility in place.

The Cycle Model and Its 2022 Backtest

Howden Re's model starts with premium and dedicated capital, subtracts attritional, catastrophe and prior-year losses, adds investment income, then applies interest-rate-driven bond marks directly to capital. Combined ratio and capital growth produce a directional pricing signal. The report calls the signal directional rather than statistically calibrated, and its only fully observed test exposes the size of that qualification.

Scenario	Insured cat loss	Rate shock	Combined ratio	Ending capital	Pricing signal
2022 backtest	$126bn	+288 bps	98.8%	$355.0bn	+21.0% model; +37% actual
Current-trend base	$100bn	No shock	92.4%	$559.2bn	-15.4%
Severe convergence	$200bn	+300 bps	110.2%	$415.5bn	+7.1%

The 2022 backtest begins with $421 billion of capital and ends at the observed $355 billion after a $90.7 billion mark-to-market bond loss. It generates a 21% pricing response against the actual 37% increase. Howden Re attributes the difference to its blended all-lines portfolio, which smooths the sharper property-catastrophe move. That is a 16-point miss on the variable this model is meant to signal.

The base and stress ledgers also reveal how the cession assumption behaves. Doubling insured catastrophe losses from $100 billion to $200 billion raises net reinsurance catastrophe losses from $32.6 billion to $69.1 billion, a $36.5 billion increment. The effective recovery share rises from 32.6% to 34.6% as more loss penetrates the tower, still below the 40% exposure share.

The larger capital movement comes from assets. The stress case records a $106.6 billion bond mark, more than twice the $45.3 billion underwriting loss, while $76.7 billion of investment income partly rebuilds capital. A cat-loss shock turns the pricing signal only after the model adds a nearly equal 300-basis-point rate shock and adverse casualty development.

Capital benchmarks use different perimeters. Howden's scenario starts from $505 billion of dedicated capital, while Aon measured total year-end 2025 reinsurer capital at $785 billion, including $649 billion of traditional equity and $136 billion of third-party capital (Aon, April 2026). Both measures describe abundant supply; neither repairs a pricing signal that averaged away the property-cat turn.

Accounting Profit and Economic Value

Howden Re's economic value-added spread was near break-even by mid-2026 even while reported returns stayed positive. Its measure subtracts weighted average cost of capital from after-tax return on invested capital and treats retrocession as contingent capital alongside debt and equity. A low combined ratio can coexist with thin value creation once the price of all three capital sources is charged.

AM Best provides the apparent counterexample: a 16.3% median return on equity for 2025 against a 9.6% cost of equity, while weighted average cost of capital rose to 8.63% in the first quarter of 2026 (AM Best, August 2026). The periods, composites and return denominators differ, but the contrast establishes why reported ROE cannot settle the capacity question.

The distinction reaches the treaty layer. Capacity that clears an accounting profit after four years of sub-25% catastrophe loss shares may still destroy value at the margin when the rate-on-line falls 16%, retrocession has an explicit cost and a primary-peril loss reaches the layers that have stayed untouched. actuary.info's cost-of-equity analysis found the hurdle rising during the same quarter that property-cat pricing fell.

The model leaves out equity-market shocks, represents reinsurance as one blended portfolio and calculates catastrophe losses separately from asset-sale liquidity. Howden Re acknowledges that the separation can understate a genuine convergence, because a large loss may force securities to be sold while bond values are depressed. Each omission weakens the connection between abundant starting capital and the price response.

The 2022 backtest matched ending capital at $355 billion and missed property-catastrophe pricing by 16 points. The balance sheet was calibrated; the market reaction was not.

Further Reading on actuary.info

• Reinsurers' Cost of Capital Rose 96 Basis Points While Property Cat Rates Fell 16%
• Property Cat Reinsurance Down 14%: How to Recalculate Your Cat Load
• Property Cat at -23% from Peak: Reinsurer ROE and the 2027 Cost-of-Capital Horizon
• Swiss Re: Secondary Perils Drove 92% of 2025 Nat Cat Losses