Edition 28: Reinsurance Renewals, ILS Appetite, and Rare-Event Model Governance

Opening Bell

The mid-year reinsurance story is not “the soft market is new”; prior editions have already covered that. The fresh signal is that abundant capital, record profitability, and renewed ILS appetite are now showing up in concrete April-to-June renewal evidence, transaction pricing, and capital-allocation choices. On the research side, insurance ML papers are moving from generic predictive performance toward model failure modes — class imbalance, probability quality, and fairness — which is exactly where actuarial governance lives.

Key Metrics

Headlines

1. Mid-year reinsurance renewals are a follow-up story, not a rerun

Gallagher Re’s FY2025 report puts dedicated reinsurance capital at a record $648bn, giving cedants more leverage heading into mid-year renewals while keeping reinsurer discipline in the frame. The actuarial question is not “are prices softer?” but “how much ceded-cost relief survives once attachment, coverage, secondary-peril, and retained-volatility assumptions are re-run?” Gallagher Re

Decision Delta - Signal: Capital abundance is now visible in renewal economics. - New vs last issue: Follow-up; generic softening was covered before, the capital-and-mid-year evidence is the new angle. - Functions affected: Reinsurance buying, pricing, capital, planning. - Direction: Lower ceded cost, but unchanged need for technical adequacy. - Horizon: June/July 2026 renewals. - Confidence: High. - Source quality: Broker primary report. - Actuary action: Update ceded-cost allocations together with retained-volatility and attachment-point diagnostics.

2. Kin’s $335m cat bond says ILS capacity is still open for the right risks

Kin secured $335m through Hestia Re 2026-1, its largest and most geographically expansive catastrophe bond. For actuaries, the useful signal is not only the size; it is that capital markets remain willing to take named-storm risk where sponsor data, structure, and expected return line up. Kin / PR Newswire

Decision Delta - Signal: ILS capacity remains receptive. - New vs last issue: Fresh transaction. - Functions affected: Cat modelling, reinsurance, capital allocation. - Direction: Lower marginal cost of well-structured cat risk transfer. - Horizon: 2026 renewal and cat-bond issuance window. - Confidence: Medium-high. - Source quality: Issuer / press release. - Actuary action: Revisit net-cat capital cost assumptions for programs with credible data and clean structure.

3. ADB’s disaster relief bonds widen the cat-risk-transfer playbook

The Asian Development Bank issued its first Disaster Relief Bond offerings, designed to provide rapid liquidity to the Kyrgyz Republic and Tajikistan after severe earthquake or flood events. This matters because parametric cat structures are moving from private portfolio optimization into sovereign resilience finance, where basis risk, trigger design, and liquidity timing become the actuarial centre of gravity. ADB

Decision Delta - Signal: Public-sector catastrophe risk transfer is maturing. - New vs last issue: Fresh transaction. - Functions affected: Cat modelling, ERM, public-sector insurance, development finance. - Direction: More parametric and liquidity-first structures. - Horizon: 2026–2028. - Confidence: High. - Source quality: Multilateral development bank primary source. - Actuary action: Treat trigger design and basis-risk governance as first-class actuarial work, not structuring footnotes.

4. Cyber pricing gets a severity warning light

Triple-I, summarizing Chubb’s cyber claims work, reports that large-company cyber claim frequency fell 34% in 2025 while average severity doubled to more than $4.4m. That is the classic actuarial trap: a frequency improvement that can be eaten alive by severity, accumulation, litigation, and vendor/supply-chain concentration. Triple-I

Decision Delta - Signal: Cyber adequacy is severity-led. - New vs last issue: Follow-up with fresh numeric detail. - Functions affected: Cyber pricing, reserving, accumulation management. - Direction: Higher tail and aggregation concern despite lower count frequency. - Horizon: Current pricing year. - Confidence: Medium-high. - Source quality: Industry research summary of insurer claims data. - Actuary action: Stress severity distributions and correlated vendor events before celebrating frequency improvements.

5. NAIC capital work is getting more peril-specific

NAIC’s P/C Risk-Based Capital Working Group and Catastrophe Risk Subgroup have a May 12 agenda covering Proposal 2026-08-CR, underwriting-risk factors, premium/loss concentration, severe convective storm analysis, and climate impact disclosures. The regulatory direction is clear: generic catastrophe load language is becoming less persuasive than peril-specific, model-to-capital evidence. NAIC P/C RBC WG and NAIC Catastrophe Risk Subgroup

Decision Delta - Signal: Cat capital evidence is becoming more granular. - New vs last issue: Fresh agenda detail. - Functions affected: Capital modelling, regulatory reporting, cat model governance. - Direction: More documentation burden and peril split discipline. - Horizon: 2026 filings and 2027 planning. - Confidence: High. - Source quality: Regulator primary source. - Actuary action: Map hurricane, earthquake, wildfire, SCS, concentration, and climate outputs explicitly to RBC documentation.

6. Carrier earnings look strong — after cat normalization

Chubb reported an 84.0% Q1 P&C combined ratio, W.R. Berkley reported 90.7%, and Markel’s insurance combined ratio was about 93%. These are strong underwriting signals, but peer benchmarking should separate accident-year performance, catastrophes, prior-year development, mix, and reinsurance effects before using them as rate-adequacy evidence. Chubb IR, W.R. Berkley IR, Markel IR

Decision Delta - Signal: Strong underwriting, but normalization required. - New vs last issue: Fresh Q1 update. - Functions affected: Pricing, reserving, competitor benchmarking. - Direction: Favorable results with catastrophe-comparison caveats. - Horizon: Q1/Q2 2026 earnings cycle. - Confidence: Medium-high. - Source quality: Carrier IR primary sources. - Actuary action: Build peer CR bridges before changing trend or rate-need assumptions.

7. ASSA’s 2026 research work is unusually practical

ASSA’s Research Newsletter highlights RAF reform, funeral insurance, the Climate Change Index Phase 1, NHI framework work, and an AI essay competition. That is a useful reminder that actuarial research does not only live in pricing models; it also lives in compensation systems, health financing, climate metrics, and public-interest design. ASSA

Decision Delta - Signal: South African actuarial research is moving into delivery domains. - New vs last issue: Current update. - Functions affected: Public policy, health, social insurance, climate analytics. - Direction: More applied actuarial involvement. - Horizon: 2026. - Confidence: High. - Source quality: Professional body primary source. - Actuary action: Watch the climate-index and compensation-system work for transferable methodology.

8. Emerging-risk work now has to include AI and cyber as operating risks

The SOA/CAS Emerging Risks Survey puts financial volatility and geoeconomic shifts near the top of short-term concern, while AI adverse outcomes and cyber events dominate the longer-horizon technology-risk discussion. This is not just an ERM-register update; it should affect scenario design, model validation, cyber aggregation, and governance of AI-assisted actuarial work. SOA Research Institute

Decision Delta - Signal: AI/cyber have crossed into mainstream actuarial emerging-risk governance. - New vs last issue: Ongoing with current survey evidence. - Functions affected: ERM, cyber pricing, model risk, capital. - Direction: Higher governance and scenario-analysis expectations. - Horizon: 1–3 years. - Confidence: High. - Source quality: Professional body research. - Actuary action: Link emerging-risk scenarios to pricing, capital, and model-validation controls.

Deep Dive — Market

The renewal market is giving cedants relief, but the actuarial interpretation needs to be more careful than “reinsurance got cheaper.” Gallagher Re’s FY2025 market report puts dedicated reinsurance capital at $648bn, up materially, which helps explain why buyers have more leverage. Kin’s $335m Hestia Re 2026-1 transaction shows ILS appetite remains strong for clean, structured cat risk. ADB’s first Disaster Relief Bonds show that cat-risk capital is also spreading into sovereign resilience and disaster-liquidity structures.

The technical implication is that ceded-cost loads should move, but not alone. If a pricing team lowers gross indications mechanically because reinsurance is cheaper, it risks smuggling capital-market competition into the loss cost. The right update is paired: ceded premium, attachment, reinstatement, coverage, basis risk, retained volatility, and peril-level tail assumptions all move through the model together. In other words: cheaper reinsurance is an input, not a conclusion.

Decision Delta - Signal: Mid-year renewal economics are softening through capital abundance and transaction appetite. - New vs last issue: Follow-up with fresh capital and transaction evidence. - Functions affected: Reinsurance buying, pricing, cat modelling, ERM. - Direction: Ceded-cost relief; retained-risk discipline still required. - Horizon: June/July 2026 renewals. - Confidence: High. - Source quality: Broker report + primary transaction sources. - Actuary action: Re-run net indications and capital metrics as one model, not two spreadsheets stapled together.

Deep Dive — Research

The most useful research item this week is “Class imbalance in insurance fraud detection models”. Fraud models are exactly where attractive headline metrics can mislead: the base rate is low, the cost of false positives is operationally real, and the cost of false negatives is financial leakage. A model that improves aggregate accuracy can still be useless if it learns the majority class too well and treats the minority class as actuarial confetti.

For pricing and underwriting actuaries, the lesson generalizes. Insurance modelling is full of rare-but-important regions: large claims, fraud, cyber catastrophe, litigation explosion, nonstandard risks, thin cells, and protected-class fairness diagnostics. The research direction is therefore less about “which classifier wins?” and more about decision quality under skewed data: thresholds, calibration, lift stability, sampling design, operational triage, and governance evidence. This also connects naturally to recent probability-quality work such as the Manokhin Probability Matrix and calibration diagnostics.

Decision Delta - Signal: Insurance ML evaluation is moving toward failure modes and decision quality. - New vs last issue: Fresh paper candidate. - Functions affected: Claims analytics, fraud triage, underwriting, model validation. - Direction: More emphasis on calibration, thresholds, and minority-class utility. - Horizon: Immediate model governance. - Confidence: Medium-high. - Source quality: Recent insurance journal feed. - Actuary action: Evaluate rare-event models with cost-sensitive, calibrated, decision-based metrics — not accuracy theatre.

Practical Takeaways

From the Actuarial Societies

• SOA / CAS — Emerging Risks Survey (research): The 19th survey highlights financial volatility, geoeconomic shifts, AI adverse outcomes, and cyber events. It is a useful bridge from ERM theory to scenario design, cyber aggregation, and model-risk governance. SOA Research Institute
• ASSA — Research Newsletter 2026 (research / public policy): ASSA’s current research update covers RAF reform, funeral insurance, a Climate Change Index, NHI frameworks, and AI essay activity. The practical value is the spread across compensation, health, climate, and AI rather than one narrow technical theme. ASSA
• IFoA — Sustainability Hub (policy / practice resource): IFoA’s sustainability material keeps climate risk in the practical actuarial toolkit. This remains more of an implementation signal than a fresh market event, but it belongs in the governance watchlist. IFoA
• AAA — ACA premium-rate policy analysis (policy): The American Academy’s health-rate analysis is outside the P&C centreline, but it is a useful reminder of how actuarial rate-setting interacts with policy uncertainty and affordability constraints. AAA

Societies covered this week: SOA, CAS, ASSA, IFoA, AAA.

From the NAIC

• P/C RBC Working Group / Catastrophe Risk Subgroup agenda: The May 12 call is set to consider Proposal 2026-08-CR, expose underwriting-risk and concentration proposals, hear severe convective storm impact analysis updates, and discuss climate impact disclosures. Actuarial relevance: capital models need to become more peril-specific and auditable. NAIC P/C RBC WG
• Catastrophe Risk Subgroup: The subgroup’s work keeps hurricane, earthquake, wildfire, severe convective storm, and climate disclosure in the solvency conversation. Actuarial relevance: internal cat model outputs should be traceable to the regulatory risk taxonomy. NAIC Catastrophe Risk Subgroup
• Flood insurance consumer/regulatory material: NAIC’s flood resources remain a useful public-facing signal of continued regulator attention to protection gaps. Actuarial relevance: flood pricing, mitigation credits, and affordability remain politically sensitive. NAIC Flood Insurance

From Carrier IR

• Chubb: Q1 P&C combined ratio was 84.0%, with ex-cat/prior-period-development at 82.1% and P&C cat losses of $500m versus $1.64bn in the prior-year quarter. The actuarial read is strong underwriting, but comparisons need catastrophe normalization. Chubb IR
• W.R. Berkley: Q1 combined ratio was 90.7%, or 88.3% before 2.4 catastrophe points. This supports the view that specialty underwriting remains profitable even while property-cat reinsurance pricing softens. W.R. Berkley IR
• Markel: Q1 insurance combined ratio was about 93%, with cat losses of $35m / 2 points versus $66m / 3 points in the prior-year period. Actuaries should adjust for catastrophe load, business mix, and reinsurance-exit effects before using the figure as a cycle indicator. Markel IR

From the Journals

• Class imbalance in insurance fraud detection models (published online 5 May 2026): A directly insurance-specific ML paper on rare-event modelling and evaluation pitfalls. It is this week’s best research deep-dive candidate because it connects model performance to operational decision quality. Springer
• Index insurance under demand and solvency constraints (published online 8 May 2026): Useful for thinking about parametric products, affordability, solvency, and basis-risk design. It pairs naturally with the ADB disaster-bond story. Springer
• Selection in car insurance when claims are heterogeneous (Version of Record online 4 May 2026): A fresh Journal of Risk and Insurance Early View item on heterogeneous claim selection, useful for pricing and underwriting segmentation diagnostics. Journal of Risk and Insurance

Source note: the journal feed itself did not expose complete publication-date metadata, so the included papers were checked against publisher/search metadata before keeping them in the “last 7 days” journal section.

From arXiv

The direct actuarial arXiv scrape returned zero viable items because arXiv responded with HTTP 429, so this section uses recent Supabase arXiv/general-ML fallback items. Treat these as method-watch items, not actuarial papers.

• The Manokhin Probability Matrix: A Diagnostic Framework for Classifier Probability Quality: Relevant to calibration, probability quality, and model QA for pricing/underwriting systems. arXiv
• Gradient Regularized Newton Boosting Trees with Global Convergence: A tree-boosting optimization paper; useful if it improves stability or convergence diagnostics in tabular insurance models. arXiv
• SHIFT: Robust Double Machine Learning for Average Dose-Response Functions under Heavy-Tailed Contamination: Relevant to causal pricing, treatment effects, and robustness where claims/cost data are heavy-tailed. arXiv
• Adaptive Norm-Based Regularization for Neural Networks: A general regularization item; watch for applications to stable tabular neural nets and actuarial foundation-model training. arXiv

Coming Up (Events — from events tracker)

• NAIC Joint Meeting of the P/C RBC Working Group and Catastrophe Risk Subgroup — 12 May 2026, public Webex: The call covers Proposal 2026-08-CR, underwriting-risk factors, premium/loss concentration, severe convective storm analysis, and climate impact disclosures. This is the clearest near-term regulatory event for P&C capital modelling. NAIC P/C RBC WG
• ASSA Banking Seminar — 18 June 2026: ASSA’s Banking Committee is focusing on non-financial risks, AI/digitisation, climate risk, and regulatory shifts. It is not insurance-specific, but the risk-management themes map directly into actuarial ERM and capital work. ASSA
• ASSA 53rd Convention — 7–8 October 2026, CTICC Cape Town: ASSA’s annual flagship convention is the major South African actuarial gathering on modelling and measurable risk. Worth tracking for climate, AI, public-policy, and insurance-practice content. ASSA Convention

What We're Watching

• 🌀 Mid-year renewals without sloppy optimism — ceded costs may fall, but secondary perils and retained volatility are not cancelled.
• 🧱 Peril-specific capital evidence — NAIC’s RBC/cat work is nudging models toward more explicit hurricane, earthquake, wildfire, SCS, and climate documentation.
• 🧪 Rare-event model evaluation — fraud, cyber, and large-loss models need calibrated decision metrics, not accuracy confetti.
• 🌍 Public-sector cat finance — ADB’s disaster relief bonds show catastrophe modelling moving further into sovereign resilience and liquidity design.