The Credibility Report
Edition 25: Funded Reinsurance, Geopolitical Risk, and Dependence Modelling
EIOPA insurance risk dashboard stability and natural-catastrophe research cooperation, Swiss Re on fragmented insurance markets, natural-catastrophe trend, and resilience gaps, plus research on Cape Cod reserving MSEP, competing risks, cure copulas, heatwave attribution, and tail-dependence diagnostics.
Welcome to Edition 25. The centre of gravity this week is resilience under stress: prudential regulators are tightening funded-reinsurance expectations, EIOPA is calling the sector stable but geopolitically exposed, and the Baltimore Bridge settlement is a live case study in infrastructure-scale marine accumulation.
The research thread is equally actuarial: reserve uncertainty, survival dependence, competing risks, climate attribution, and tail-dependence diagnostics. Less sparkle, more plumbing — which is usually where the actual model risk lives.
— Ron Richman, Founder, InsureAI
This Week’s Headlines
Source note: every headline link in this section goes directly to a primary source — EIOPA or Swiss Re Institute.
EIOPA: stable insurance risk dashboard, but geopolitics shape the outlook
EIOPA’s latest insurance risk dashboard keeps the overall signal stable, while highlighting geopolitical uncertainty as a forward-looking pressure point. Solvency ratios can look comfortable while scenario uncertainty, spread volatility, and liquidity stresses are quietly re-pricing the tails.
EIOPADecision Delta
- Signal: The base case is stable; the risk budget is being consumed by uncertainty rather than current deterioration.
- Functions affected: ERM, stress testing, investment risk, capital steering.
- Actuary action: Separate current solvency comfort from forward-looking scenario vulnerability.
EIOPA and JRC deepen natural-catastrophe risk research cooperation
EIOPA and the European Commission’s Joint Research Centre are deepening cooperation on natural-catastrophe risk research. For actuaries, the important signal is methodological: regulatory climate-risk work is becoming more data-heavy, model-driven, and explicitly linked to protection-gap analysis.
EIOPADecision Delta
- Signal: Climate-risk supervision is moving toward shared research infrastructure.
- Functions affected: Cat modelling, ORSA climate scenarios, protection-gap analysis, public-policy pricing.
- Actuary action: Align internal climate-risk diagnostics with regulator-facing data and attribution methods.
Swiss Re sigma: world insurance in a riskier, more fragmented order
Swiss Re Institute’s world-insurance outlook frames the market against a more fragmented macro and geopolitical backdrop. The read-through for insurers is practical: growth, capital, reinsurance demand, and protection gaps need to be interpreted through volatility and fragmentation, not just premium growth.
Swiss Re InstituteDecision Delta
- Signal: Insurance growth is increasingly conditional on fragmentation, inflation, and capital-market uncertainty.
- Functions affected: Strategy, capital allocation, reinsurance planning, market-entry assumptions.
- Actuary action: Use scenario-weighted market outlooks rather than single-path premium-growth assumptions.
Swiss Re sigma: natural-catastrophe insured losses remain on trend to USD 145bn
Swiss Re Institute’s natural-catastrophe sigma is a useful anchor for pricing and capital discussions: the relevant issue is not whether 2025 is a single bad year, but whether the industry is managing a persistent trend in secondary perils, exposure growth, and protection gaps.
Swiss Re InstituteDecision Delta
- Signal: Nat-cat loss trend is becoming an attritional capital problem, not only a peak-peril story.
- Functions affected: Property pricing, reinsurance purchase, cat capital, portfolio steering.
- Actuary action: Split loss-trend diagnostics between hazard, exposure, vulnerability, and social inflation components.
Swiss Re Resilience Index: gains are encouraging, but protection gaps remain
Swiss Re Institute’s Resilience Index keeps the focus on the insurance protection gap. For actuarial teams, the useful framing is not merely “more insurance penetration”; it is which perils, geographies, and affordability constraints are keeping risk off balance sheets until after the event.
Swiss Re InstituteDecision Delta
- Signal: Resilience is measurable, but the remaining gap is concentrated and structurally hard.
- Functions affected: Public-private pools, affordability design, parametric cover, catastrophe protection.
- Actuary action: Pair technical premium adequacy with affordability and take-up diagnostics.
From arXiv
Lead paper
A Note on the Generalized Cape Cod Reserving Method
This is the week’s cleanest actuarial paper. The authors close a practical gap by deriving an analytical mean squared error of prediction formula for the generalized Cape Cod reserving method. Chain-ladder and Bornhuetter-Ferguson have long had stochastic uncertainty machinery; this paper gives GCC the same kind of uncertainty footing.
That matters because GCC is exactly the kind of method actuaries use when pure development experience is thin and exposure-level information carries credibility. A deterministic point estimate is not enough when the method is feeding capital, range-setting, or reserve governance.
Read the paper →Bayesian nonparametric competing risks
A principled prediction-curve approach for competing risks in survival analysis, explicitly relevant to actuarial sciences, reliability, and cause-of-death modelling.
Read on arXiv →Bivariate cure copula with zero-inflated frailty
A survival-dependence model for paired event times with cure fractions. Useful read-through for morbidity, longevity, and dependent decrement modelling.
Read on arXiv →Spatio-temporal heatwave attribution
A climate-attribution framework that models daily temperature fields, tails, and spatio-temporal dependence. Cat and health actuaries should care about the persistence layer, not only the marginal return period.
Read on arXiv →Signed multivariate tail-dependence compatibility
Technical, but relevant to portfolio capital and copula governance: when specified tail-dependence assumptions can actually coexist in a multivariate loss model.
Read on arXiv →From the Journals and Scholar Alerts
Embeddings and attention in predictive modelling
Variance paper on embeddings and attention for insurance claim severity, including transfer of learned categorical structure into GLMs and TabTransformer-style contextual embeddings on NFIP data.
Read the article →Interpretable churn prediction in motor liability insurance
A motor-liability retention paper comparing ensemble methods with SHAP explanations. Useful for pricing teams where churn, renewal price elasticity, and high-cardinality rating variables collide.
Read the paper →Deep-testing for dependence detection
A neural approach to hypothesis testing, demonstrated on independence testing. It is worth watching for actuarial model validation where dependence misspecification is often the hidden failure mode.
Read on arXiv →Third-party litigation funding reform
Triple-I tracks state-level reform momentum. For casualty actuaries, TPLF is not just a legal story; it is a severity-trend, settlement-duration, and reserve-uncertainty variable.
Read the analysis →Practical Takeaways
If GCC appears in the reserve toolkit, add explicit MSEP analysis and show how it compares with chain-ladder and Bornhuetter-Ferguson uncertainty ranges.
Stress funded-reinsurance recapture, collateral, liquidity, and asset-liability channels together; treating them separately understates balance-sheet feedback.
Move beyond marginal event severity: heatwaves and marine infrastructure losses both require spatio-temporal and network accumulation views.
Track tort reform and litigation-funding changes as structural breaks in severity, settlement speed, renewal adequacy, and reserve development.
What We’re Watching
- Whether the PRA funded-reinsurance proposals become a template for other solvency regimes.
- How geopolitical uncertainty migrates from dashboard commentary into insurer capital and liquidity stress tests.
- Whether the Baltimore Bridge loss pushes marine reinsurers toward more explicit infrastructure-network accumulation models.
- Whether actuarial ML papers keep moving from generic predictive lift toward uncertainty, dependence, and governance-ready diagnostics.