The Credibility Report
Actuarial Intelligence for Insurance Professionals
What’s in this edition
Liability trends and Australian property risk, plus four actuarial preprints on pricing, mortality and catastrophe-bond valuation. Research summaries are based on abstracts; results are not independently replicated. Scores indicate keyword-based relevance, not research quality.
📰 Headlines (primary sources)
Liability trend: tort costs grow faster than the economy
Triple-I reports that US tort costs rose from $299.7 billion in 2014 to $579.7 billion in 2023. Its analysis includes uninsured and self-insured costs: this is not an insured-loss series. For casualty pricing and reserving, separate legal-system pressure from economic inflation rather than importing an aggregate tort-cost growth rate directly into a loss-trend assumption.
Read source → • Insurance Information Institute • October 30, 2026El Niño is not a guarantee of a dry claims season
The ICA warns that wet and stormy conditions remain possible despite the declared El Niño. It urges property preparation, cover reviews and updated sums insured. Actuarial implication: retain flood and severe-storm stress scenarios alongside bushfire scenarios; a climate-phase label alone is not a portfolio loss forecast.
Read source → • Insurance Council of Australia • October 29, 2026Victoria adopts mandatory flood-risk planning provisions
The ICA welcomes mandatory flood planning rules and a Flood Risk Rating system linking development and building controls to hazard. The industry body calls for consistent council implementation and updated flood mapping to reach insurers. Underwriting implication: distinguish changes to future development from the flood exposure already on the book.
Read source → • Insurance Council of Australia • October 22, 2026🔬 Research Spotlight (arXiv)
Joint Distributional Copula Regression for Insurance Claim Frequency and Severity via Boosting
arXiv • Score: 36 • 2026-10-27
Banghee So and Na Zhang jointly model frequency and severity using boosted, structured additive distributional regression with a copula. Covariates can affect marginal distribution parameters and dependence, not only expected loss. Simulation and real-data results are reported. Actuarial use: benchmark conditional tails and dependence against separate frequency–severity models before assuming independence is adequate.
Open paper →Generalized survival energy models with exogenous covariates for mortality prediction
arXiv • Score: 26 • 2026-10-21
Yasutaka Shimizu models death as a latent survival-energy process reaching zero, with social and economic covariates influencing its dynamics. Japanese mortality results are compared with Cairns–Blake–Dowd benchmarks conditional on realized environmental paths. Important limitation: this is not evidence that future covariates are known, or that covariate effects are causal. Use it as a mortality-scenario challenger, not an automatic replacement for a production basis.
Open paper →Risk Adequacy and Anti-Discrimination in Insurance Pricing: Classification, Aggregation, and Transfers
arXiv • Score: 25 • 2026-10-23
Wiebke Hansen, Stefan Weber and Mario Wüthrich distinguish individual risk adequacy, portfolio coverage and anti-discrimination criteria. Their classification–aggregation–transfer framework describes tariff-design trade-offs and illustrates them with motor insurance. Actuarial use: report protected-group premium effects and capital implications alongside predictive accuracy; a single fairness statistic cannot describe every pricing objective.
Open paper →Catastrophe-bond valuation: separating frequency and severity risk premia
arXiv • Score: 15 • 2026-10-27
Saeid Safarveisi uses financial entropy and geometric optimization to select pricing measures in an incomplete CAT-bond market. In the model, spreads do not separately identify frequency and severity risk premia; the entropy criterion selects a decomposition. The worked example is a hypothetical earthquake bond calibrated to observed spreads. Treat the decomposition as assumption-dependent, not as separately observed market prices of risk.
Open paper →✅ Practical Takeaways
- Casualty pricing and reserving: split economic inflation and legal-system trend assumptions; stress development-tail and excess-layer sensitivity without treating total tort costs as insured losses.
- Property catastrophe modelling: retain flood, severe-storm and bushfire accumulation scenarios; reconcile changed flood maps to geocoded exposures and distinguish new-development controls from existing-stock risk.
- Pricing model validation: fit a dependence-aware frequency–severity challenger and compare out-of-time joint-tail calibration against the independent benchmark, not just average prediction error.
- Life and annuity valuation: benchmark survival-energy mortality scenarios against the current basis and a CBD challenger, testing forecast covariates rather than relying only on realized paths.
- Tariff governance: document individual risk adequacy, protected-group premium effects and portfolio capital outcomes separately before approving class aggregation or premium transfers.
- ILS valuation: hold the observed spread fit fixed and vary the frequency–severity risk-premium decomposition to expose non-identification; record the selection criterion in model documentation.
Until next time—stay credible.
— The Credibility Report
Edition 046 | Published November 1, 2026 (UTC)