The Global Risks Alliance (GRA)
Catastrophe models inform capital, price and rating decisions, so they must be shown fit for purpose. Validation tests each module against observations, compares models, challenges assumptions and documents limits. Supervisors, rating agencies and professional standards set expectations. This page sets out how catastrophe model validation works and where it falls short, the evidence it uses, and who decides.
GRA · In development · service level R0
Catastrophe model validation is the process of showing that a model is fit for its intended use by testing modules against observations, comparing models, challenging assumptions and documenting limits. Validation functions, boards, supervisors and public commissions decide whether a model may inform capital and rates.
Which places drive loss in a severe year, and how far it reaches into reinsurance. Illustrative design values.
Uncertainty, sensitivity and benchmarks. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Hazard is compared with observed event statistics, vulnerability with claims, and the full model with historical losses. Each test covers part of the model, not all of it.
Rare events cannot be validated against long loss histories. Validation relies on scientific plausibility and expert judgement for the tail.
Validation by the team that adopted the model is weaker than independent challenge. Governance separates development, adoption and validation.
Vendor updates can shift results sharply. Change management assesses why results moved and whether the new view is adopted.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Model validation | Independent testing that a model performs as intended and is fit for a stated use. |
| Change analysis | Explaining how and why a new model version changes results. |
| Independent challenge | Review of a model by people not involved in developing or adopting it. |
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Modelled versus historical loss for past events | ratio | per validation | Validation reports | MODELLED |
| Change in results from model version updates | percent change in key metrics | per update | Change analysis | MODELLED |
| Independent validation completed | yes or no, by model | annual | Governance records | DECLARED |
| Internal models approved by supervisors | count | annual | Supervisor records | OFFICIAL |
| Validation findings outstanding | count by severity | periodic | Validation records | DECLARED |
| Public model reviews completed | count | annual | Public commissions | OFFICIAL |
The topic as a computational graph: named sources feed the methods, an evidence review gates the outputs, and the decisions stay with the authorities that hold them.
Fig. 02Nodes carry their type, two metric rows and a state; connectors show what feeds what; dashed frames name the owners; the timeline is the cadence. Design of the pipeline, not a record of a run.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| Florida Commission on Hurricane Loss Projection Methodology standards and reports | Public standards and acceptance reviews for hurricane and flood models | Public records |
| EIOPA and PRA guidance on internal model use and validation | Supervisory expectations for model validation | Free, EU and Bank of England terms |
| Lloyd's catastrophe modelling guidance and minimum standards | Validation and governance expectations for managing agents | Free to read, Lloyd's terms |
| Actuarial Standards Board ASOP on catastrophe modeling | Professional standards for using models | Free, ASB terms |
| Oasis LMF model documentation | Open model documentation enabling comparison | Open source, terms per model |
Reinsurance and Catastrophe prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Model validation functions | Whether a model is fit for each intended use | Test results, comparisons and documented limits |
| Boards and risk committees | Whether to adopt a model or version as the company view | Validation findings and change analysis |
| Supervisors | Whether to approve internal models for capital | Validation and governance evidence |
| Public model commissions | Whether models meet standards for rate filings | Model submissions and review findings |
The events that matter most are the rarest, so there are few observations against which to test tail estimates. A model may match recent losses well but misestimate the tail. Validation must therefore test components separately, against hazard records, engineering studies and claims, as well as the model as a whole.
Sensitivity analysis shows which assumptions matter. Small changes in vulnerability functions, event frequencies or exposure data can change loss estimates significantly. Understanding these sensitivities is essential for using models responsibly.
Comparing models helps reveal structural uncertainty: if independently built models disagree, the disagreement indicates how uncertain the estimate is.
Testing modules against independent evidence, using uncertainty quantification and model validation.
Comparing modelled and observed losses for past events. Exposure changes complicate comparison.
Varying key assumptions systematically. Interactions between assumptions can be complex.
Comparing outputs of different models on common portfolios. Requires shared, consistent inputs.
Validation reduces but cannot remove model uncertainty. It is an ongoing process, repeated as models and science change.
The current service level is R0: information and enquiries. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.