42%, 45% or 52%: Swiss Re's Quiet Half Has Three Insured Shares

Swiss Re Institute put global insured natural catastrophe losses at $42 billion for the first half of 2026, 16% below the ten-year average and the lowest first half since 2020 (Swiss Re Institute preliminary sigma estimates, August 11, 2026). Economic losses came in at $100 billion against $152 billion a year earlier.

The figure that matters to a catastrophe pricing desk is neither of those. Insurance absorbed 42% of the economic loss, against a 30-year average of 33%. The half was cheap in total and adverse in mix.

Key Takeaways

• Swiss Re's own two below-average figures imply a ten-year baseline of 45% (computed: $50bn insured against $111bn economic). Measured against that, this half's 42% ran three points low, not nine points high.
• Severe convective storm supplied $28 billion of the $42 billion, about two thirds of the insured total. It is the peril class most concentrated in the most heavily insured market on earth.
• The Venezuela earthquake caused $20 billion of economic loss with no reliable insured estimate. It enters the denominator of the insured share and barely touches the numerator, which means it pushed the 42% down, not up.
• Excluding that one event lifts the share to about 52% (computed: $42bn against $80bn, treating its insured loss as zero). That is the only reading on which the half looks genuinely mix-adverse, and it rests on one unquantified event.
• The first half is the cheap half by construction. North Atlantic hurricane and the European wildfire season both load into the second, so a 16%-below-average January-to-June total constrains the full year far less than it appears to.

What the $42 Billion Counted

The composition is narrow. Severe convective storm accounted for $28 billion of the $42 billion insured total, and Swiss Re notes that United States storm activity ran above average while relatively few of the highest-impact events struck Texas, the Southern Plains and the Southeast. The peril was busy; it simply missed the most expensive postcodes.

Wildfire contributed only a small share of first-half insured losses despite being, on Swiss Re's reading, the fastest-growing weather peril globally. European insured wildfire losses have risen 8% to 11% a year in real terms since 1970.

Measure, H1 2026	Amount	Comparison
Insured natural catastrophe losses	$42bn	16% below the ten-year average; lowest first half since 2020
Economic losses	$100bn	10% below the ten-year average; $152bn in H1 2025
Insured share of economic loss	42%	30-year average of 33%
Severe convective storm, insured	$28bn	Largest driver; about two thirds of the insured total
Venezuela earthquake, economic	$20bn	No reliable insured estimate; low insurance penetration

Source: Swiss Re Institute preliminary sigma estimates, August 11, 2026.

Why the Share Moved, and What It Does to a Cat Load

Two different things push the insured share above its 30-year average, and only one of them is about this half.

The first is secular. A 30-year baseline spans decades of lower insurance penetration across most of the world, so any recent period will tend to print above it. Reading 42% against 33% as a pure mix signal overstates the case, and the comparison should carry that caveat.

The second is genuine mix, and Swiss Re attributes the elevated share to exactly that: the concentration of damage in highly insured markets and across widely covered perils. The arithmetic supports it. Strip out the Venezuela earthquake, $20 billion of economic loss with no reliable insured estimate, and the remaining $80 billion of economic loss carried $42 billion of insured loss, a share near 52%.

That arithmetic treats the quake's insured loss as zero, which Swiss Re does not say. It says the figure is not reliably estimable and that low penetration will keep it small. Any insured loss it did produce pulls the 52% down. A single near-uninsured event of that size normally drags the ratio down hard. This one did, and the ratio still landed nine points high.

The pricing consequence sits in how a catastrophe load gets on-levelled. A load selected from a multi-year industry insured-loss history implicitly assumes the relationship between economic damage and insured damage is stable. It is not, and this half moved it by nine points against a long-run baseline. A desk converting modelled ground-up loss to an insured take using a historical share anchored anywhere near 33% understates its own exposure on a half that looked, on the headline number, unusually quiet.

The same distortion runs through the reverse test. An underwriting result that came in favourably against a $42 billion industry half is not evidence of a good loss pick, because the $42 billion was extracted from a smaller economic base than usual. The correct comparator is loss relative to exposure inside the insured footprint, and on that basis the half was closer to ordinary than the aggregate suggests.

Which Baseline Decides the Answer

The report gives two below-average percentages, and they are enough to back out the baseline it does not print. Insured losses of $42 billion running 16% below the ten-year average implies a ten-year average near $50 billion. Economic losses of $100 billion running 10% below implies a ten-year average near $111 billion. The ten-year average insured share is therefore about 45%.

Against that number, this half did not run high. It ran three points low.

The 33% figure the report cites is a thirty-year average, and thirty years reaches back through a period when far less of the world's exposure was insured at all. Comparing a 2026 half to it measures three decades of penetration growth as much as it measures this half's mix. The nine-point gap is real and mostly secular.

What survives is narrower and rests on one event. On the ten-year baseline the half was ordinary; excluding the Venezuela earthquake it was about seven points favourable to insurers. Which of those is the true reading depends entirely on a $20 billion loss whose insured component Swiss Re declines to estimate, and the honest answer is that the mix signal is somewhere between neutral and modestly adverse, not the nine points the headline comparison suggests.

That is the practical caution in the protection-gap statistic. It is quoted as though the baseline were settled, and the baseline moves the sign. A cat load rebuilt off a 33% long-run share would carry a nine-point error against recent experience; rebuilt off this half's 42% without adjusting for a single near-uninsured quake, it would carry a smaller error the other way.

The seasonal point stands regardless of baseline. Wildfire contributed almost nothing to the first half despite growing 8% to 11% a year in real terms in Europe since 1970, and Swiss Re flags June's record heat and dry conditions as the setup for an active season. North Atlantic hurricane loads into the second half too. Balz Grollimund, Head Catastrophe Perils at Swiss Re, put it plainly: "A less costly first half of the year does not mean the risk has gone away."

Further Reading on actuary.info

• H1 Cat Losses Hit $46B: The Under-Spent Budget Behind Soft January 2027 Renewal Terms - the broker estimate of the same half, and the renewal pricing read built on it.
• US Severe Convective Storm Losses Top $35B as the August Derecho Outruns 2026 Cat Budgets - the peril that supplied two thirds of the insured total, carried past the first half.
• AM Best: Big Four Kept Cat Appetite by Raising Attachments and Exiting Aggregate Covers - who keeps the losses that a quiet half leaves below the attachment point.
• Moody's Flood Gap and Property Cat Tail Risk - the protection gap measured on a single peril rather than a global ratio.
• Wildfire Cat Bond Issuance Hits $5.18B, Nears 2025's Full-Year Record - the capital market's read on the peril that stayed out of the first half.
• Mexico's M7.3 Near Miss and the Basis Risk in Parametric Cat Bonds - another earthquake where economic loss and recovered loss parted company.
• NAIC's Flood Insurance Blueprint for a Private Market - the domestic version of closing a gap between economic and insured loss.