The Credibility Report — Edition 32

May 29, 2026

AI-curated actuarial intelligence, designed by actuaries, for actuaries.

This Week's Headlines

1. SCOR secures USD 75m of multi-year natural-catastrophe protection

SCOR sponsored Atlas Capital DAC Series 2026-1, adding USD 75m of multi-year natural-catastrophe protection. The actuarial read is simple: capital markets capacity is still showing up for named peril and retrocession structures, but ceded-cost savings need to be separated from attachment, term, and peril-mix changes.

2. Korean Re adds targeted multi-peril retro protection through Solomon Re

Korean Re secured targeted USD 75m multi-peril retrocession protection through its second Solomon Re catastrophe bond. This reinforces the same signal as SCOR's deal: ILS is increasingly part of ordinary reinsurance architecture, not a side-market curiosity.

3. Mid-year renewals are reportedly down 15-20%+, with cat bonds competing harder

Aeolus's Andrew Dutt says mid-year property-cat renewals are seeing 15-20%+ reductions, with cat bonds becoming a sharper competitive threat. Actuaries should treat this as a decomposition problem: price, structure, reinsurer share, model loss cost, and basis risk are all moving at once.

4. BMO sees property-cat pricing trending down mid-teens

BMO Capital Markets expects mid-year property-cat pricing to be down in the mid-teens. The useful cross-check is that broker, investor, and ILS signals are now directionally aligned, though not identical in magnitude.

5. Allianz Research says the global P&C market is normalising

Allianz Research describes global P&C markets as moving away from aggressive rate hikes toward stabilisation. That does not make rate adequacy easy; it means the pricing conversation shifts from broad hard-market lift to line-specific trend, inflation, cat load, and cycle discipline.

Research Spotlight

Paper of the Month: insurance pricing as off-policy evaluation

Insurance Pricing Optimization via Off-Policy Evaluation reframes pricing as a decision problem where price sensitivity and counterfactual outcomes matter alongside risk-based fairness. It is directly relevant to tariff work because observed portfolio outcomes are policy-contaminated: the rates offered determine which risks bind, lapse, or never arrive.

Survival diffusion models for time-to-event data

SDPM: Survival Diffusion Probabilistic Model for Continuous- applies diffusion modelling ideas to censored survival analysis. For life, health, and claims analytics, the watch item is whether generative time-to-event models can improve uncertainty representation without losing calibration discipline.

Identifiable Bayesian deep generative copulas

Identifiable Bayesian Deep Generative Copulas with Unknown L targets a persistent weakness in flexible dependence modelling: black-box copulas can fit well while being hard to identify or explain. This is worth tracking for capital, aggregation, and multi-line dependency work.

KAPLAN for prognostic survival modelling

KAPLAN: Kolmogorov-Arnold Prognostic Learnable Activation Ne is another survival-analysis item, using learnable activation structure to model covariate-time effects. The actuarial relevance is less the acronym and more the pressure toward flexible but governable survival curves.

Proxy-based Shapley and Banzhaf interactions

Proxy-Based Approximation of Shapley and Banzhaf Interaction proposes cheaper interaction estimates for complex machine-learning models. If stable, this kind of method can help pricing teams move beyond one-way feature attributions toward interaction governance.

Practical Takeaways

• Reinsurance: do not treat mid-teens cat softening as a one-line ceded-cost scalar; bridge price, attachment, limit, peril scope, reinstatements, and model loss cost.
• Pricing: off-policy evaluation is becoming a serious actuarial pricing lens because historical prices shape observed demand and loss outcomes.
• Life and health: survival modelling research is moving quickly, but calibration under censoring remains the gatekeeper.
• Model governance: interaction explainability matters where tariff factors, dependence assumptions, or fairness diagnostics are not additive.

What We're Watching

• Whether July property-cat renewals confirm the reported 15-20%+ softening or reveal a sharper split by geography, attachment, and secondary peril.
• Whether cat bonds keep competing directly with traditional retrocession for clean, modelled peak-peril risk.
• Whether off-policy pricing methods begin moving from research into actuarial pricing workflows.
• Whether survival and dependency papers can pair flexibility with auditability, not just better benchmark scores.

Edition 32 • May 29, 2026