The Credibility Report
AI-curated actuarial intelligence — Edition 4
AM Best just shifted global reinsurance's outlook from positive to stable—and that one word change tells the whole story of January's renewals. Property pricing softened faster than expected, the buyer's market is real, and even the cautious optimists are sounding notes of warning about where ILS capital goes next.
Meanwhile, $88 billion in global insured losses from wildfires, floods, and severe storms made 2025 a reminder that the machinery of risk transfer exists for a reason.
📊 Headlines
⛈️ Winter Storms in Focus: Scale, Duration, and Rising Losses
Swiss Re's sigma research highlights a concerning trend: winter storms are becoming more severe, longer-lasting, and costlier. The analysis points to rising insured losses driven by climate factors and growing exposure in vulnerable regions. Pricing assumptions built on historical storm data may systematically underestimate forward-looking risk.
Read more → Swiss Re sigma📉 AM Best Shifts Reinsurance Outlook to Stable as Softening Accelerates
In a significant move, AM Best revised its global reinsurance outlook from positive to stable, citing accelerated softening in property pricing at January renewals. The rating agency warns that ILS capital pressure could increase as investors seek returns in a compressing spread environment. This is the first outlook downgrade since the hard market began.
Read more → AM Best📈 US P&C Industry Set for Lowest Net Combined Ratio in Over a Decade
Triple-I/Milliman forecasts the US P&C industry's lowest net combined ratio in 10+ years for 2025. Despite devastating California wildfires in Q1, an Atlantic hurricane season without US landfall provided relief. Personal auto improved to 94.4 points while homeowners sits at 99.6 points.
Read more → Triple-I💰 Allstate Returns with $500m Dual Cat Bond Issuance
Allstate is back in the market targeting $500 million across Sanders Re III and Sanders Re IV—its 24th and 25th deals in the series. The four tranches provide multi-peril per-occurrence coverage for US named storms, earthquakes, wildfires, and severe weather (ex-Florida), attaching at $4.75 billion.
Read more → Swiss Re Capital Markets💊 Aetna Secures $250m from Vitality Re XVII Health Cat Bond
CVS Health's Aetna unit closed its seventeenth Vitality Re issuance at the $250 million target. All three tranches priced at or above initial guidance—Class A at 2%, Class B at 2.4%, Class C at 4%—reflecting investor minimum return requirements even for remote-risk health ILS.
Read more → Swiss Re Capital Markets🛒 RLI Trims Cat Reinsurance by $150m in 'Buyer's Market'
RLI reduced its catastrophe reinsurance limit by $150 million at January renewal, securing 15-20% rate reductions in what COO Jen Klobnak describes as a "buyer's market." Casualty rates dropped roughly 5%. Q4 combined ratio came in at 82.6% versus 94.4% prior year.
Read more → RLI Corp📊 US Insured Catastrophe Losses 12% Above 10-Year Average
Gallagher Re estimates US insured catastrophe losses at $100 billion for 2025—12% above the 10-year average. Severe convective storms and wildfires drove the toll, with Aon separately estimating global insured losses at $127 billion. Secondary perils continue to dominate the loss landscape.
Read more → Gallagher Re💹 Private ILS Offers Better Spreads Than Cat Bonds
Despite spread compression in the cat bond market, private ILS continues to offer better returns, according to Stephan Ruoff of Schroders Capital. The Co-Head of Private Debt & Credit Alternatives notes the market remains attractive in absolute terms, with investors moving up the complexity curve to find value.
Read more → Schroders Capital📖 Research Spotlight
Paper of the Month: A Deep Learning-Copula Framework for Climate-Related Home Insurance Risk
This paper tackles one of the thorniest problems in actuarial climate modelling: how do you capture the complex, non-linear relationships between extreme precipitation and home insurance claims?
The authors propose a two-step framework. First, deep neural networks model the marginal relationship between weather variables and claim frequency/severity. Then, copula-based multivariate analysis captures the dependency structure between these modelled outputs—preserving tail dependence that simpler correlation measures would miss.
The case study uses Canadian Prairies data from 2002-2011, a region with significant hail and flood exposure. The key insight: traditional linear models systematically underestimate joint extremes. When multiple perils compound (heavy rain plus saturated ground plus failing drainage), the interaction effects matter enormously.
💡 Why actuaries should care: Climate risk isn't additive. As warming drives more volatile precipitation patterns, understanding the dependence structure of compound events becomes critical for pricing and reserving. This framework offers a tractable approach that respects both the complexity of the climate signal and the requirements of actuarial practice.
Additional Papers
New methodology for extracting interpretable representations from SOMs—relevant for portfolio segmentation and risk clustering applications.
arxiv.org/abs/2601.13851v1Critical for model validation: how deep learning models degrade under biologically and technically induced distribution shifts.
arxiv.org/abs/2601.14969v1Efficient neural architecture for emulating climate models at scale. Enables rapid scenario generation for cat modelling without full GCM runs.
arxiv.org/abs/2601.15102v1Critical examination of synthetic data quality for financial ML. Important reading as insurers increasingly use generated data for model validation.
arxiv.org/abs/2601.12990v1When should you build one big model versus many small ones? Practical guidance for sparse count data—think low-frequency, high-severity claims.
arxiv.org/abs/2601.14031v1Sharp conditions for when SD exceeds GMD (and vice versa)—directly relevant for understanding heavy-tailed distributions in actuarial practice.
arxiv.org/abs/2601.12414v1🔬 Deep Dive: The Soft Market Signal
AM Best's outlook revision from positive to stable isn't just a rating action—it's a timestamp marking the end of the hard market's expansion phase.
The headline number matters: accelerated softening in property pricing. But the subtext matters more. AM Best explicitly warns that "ILS capital pressure could increase" as the market matures. Translation: three years of strong returns have attracted capital that may not respect underwriting fundamentals when spreads compress further.
The buyer's market evidence is everywhere. RLI secured 15-20% rate reductions and trimmed $150 million in limit. Travelers "meaningfully improved" their reinsurance coverage at favourable terms. Cedents have leverage—and they're using it.
Yet the loss environment hasn't softened. Gallagher Re's $100 billion US estimate sits 12% above the 10-year average. Severe convective storms now dominate global insured losses, per Aon. Winter storms are intensifying in scale and duration, per Swiss Re sigma.
⚠️ The paradox: Profitability is at decade-highs precisely as pricing softens and losses remain elevated. This is the classic cycle setup—and the rating agencies are starting to notice. Don't mistake rate reductions for risk reductions.
🎯 Practical Takeaways
- • Buyer's market means leverage—but 12% above-average losses mean trends are concerning.
- • Private ILS spreads > cat bonds. Consider complexity premium in placements.
- • Winter storm severity increasing—review duration and intensity assumptions.
- • SCS now dominates global insured losses; secondary peril weightings need updates.
- • US P&C decade-best CR masks line-of-business divergence; homeowners still at 99.6.
- • $88bn global insured losses demands fresh IBNR for exposed books.
- • AM Best stable outlook = cycle turning point; stress capacity assumptions.
- • ILS capital pressure building—watch for underwriting discipline erosion.
📄 From the arXiv
| Paper | Score | Focus | Link |
|---|---|---|---|
| A Deep Learning-Copula Framework for Climate-Related Home Insurance Risk | 32 | Climate/Insurance ML | arxiv |
| Inverting Self-Organizing Maps: A Unified Activation-Based Framework | 28 | Manifold Learning | arxiv |
| Robust ML for Regulatory Sequence Modeling under Distribution Shifts | 26 | Model Robustness | arxiv |
| Field-Space Autoencoder for Scalable Climate Emulators | 23 | Climate Modeling | arxiv |
| Beyond Visual Realism: Reliable Financial Time Series Generation | 22 | Financial ML | arxiv |
| Intermittent Time Series Forecasting: Local vs Global Models | 22 | Forecasting | arxiv |
| On the Order Between SD and Gini Mean Difference | 21 | Risk Measures | arxiv |
| Calibrated UQ for Prosumer Flexibility Aggregation | 20 | UQ/Energy | arxiv |
🔮 What We're Watching
AM Best's stable outlook marks the end of expansion. Watching for discipline erosion as capital chases returns in a softening market.
Swiss Re sigma flags increasing scale, duration, and losses. This isn't a one-year blip—structural assumptions may need revision.
Schroders sees better value in private ILS. If institutional money follows, cat bond spreads face further compression pressure.
Severe convective storms and wildfires now drive global insured losses. Primary hurricane pricing may be mispriced relative to actual loss drivers.
Until next time—stay credible.
— The Credibility Report