Models and professional methods
Every estimate The Global Risks Alliance (GRA) publishes depends on a method, a set of assumptions and a review. This page sets out how GRA builds risk estimates and scenarios, how it states assumptions and uncertainty, and how methods are reviewed before results are released. Methods are documented so that a professional reader can examine, reproduce or challenge them. The canonical description of each method sits in the STEM Commons stewarded by The Global Centre for Risk and Innovation (GCRI). This page belongs to research and intelligence.
- 01InstitutionThe Global Risks Alliance (GRA)
- 02Service levelR0: information and enquiries
- 03In briefThe short answer
- 04Method familiesFour families run through GRA research
- 05ValidationModel validation, under the 2026 guidance
- 06Spatial methodsHazard, exposure and vulnerability as separate records
- 07DocumentationWhat every method record states
GRA · Institution · service level R0
The short answer.
GRA models and professional methods are documented so that a professional reader can examine, reproduce or challenge every estimate: how risk estimates and scenarios are built, how assumptions and uncertainty are stated and how methods are reviewed before results are released. A GRA estimate is never a forecast of loss, a rating or a capital requirement.
Draft to published version.
Fig.How a finding is reviewed before it is published. Design, not a record.
Financial centres and urban exposure in the United Kingdom and Ireland.
The cities where people, assets and financial activity concentrate, shown for the United Kingdom and Ireland as an example geography for this subject.
Four families run through GRA research.
Actuarial and quantitative
Probability, actuarial mathematics and extreme value theory for rare, severe losses. Reference: actuarial mathematics.
Contagion
How losses travel through exposures, funding and shared providers. Reference: network science and contagion.
Scenario mathematics
Compound physical, operational and economic shocks compared across plausible futures. Reference: scenario design and stress testing.
Model risk
Validation and uncertainty for the models others rely on. Reference: uncertainty quantification and model validation.
Model validation, under the 2026 guidance.
GRA method records follow the validation logic regulators expect of the industry: conceptual soundness, ongoing monitoring and outcomes analysis. In the United States that logic is now set out in the April 2026 revised interagency guidance (SR 26-2), which superseded SR 11-7; in the UK in PRA SS1/23. GRA applies those principles to its own research methods; it does not validate any firm's model.
| Check | How it is done |
|---|---|
| Conceptual soundness | Assumptions and their sources are written down and reviewed by a second analyst |
| Benchmarking | Results are compared with an independent model or published estimate |
| Sensitivity | Key parameters are varied to show which drive the result |
| Back-testing where possible | Modelled losses are compared with reported losses for past events |
Hazard, exposure and vulnerability as separate records.
Spatial estimates keep the hazard layer, the exposure layer and the vulnerability function as separate versioned records, so each can be replaced without hiding the change. Method descriptions sit at GCRI uncertainty quantification and GCRI spatial statistics.
What every method record states.
- The question the method answers and the decision context it is suited to
- Data inputs, their sources, licences and evidence labels
- Assumptions, including those that drive the result most
- Uncertainty and sensitivities, reported as ranges rather than single figures
- What was not tested, and the conditions under which the method should not be used
How a method and its results are checked.
- 01
Independent reviewer
A reviewer not responsible for the work checks the method, inputs and labels before release.
- 02
Reproducibility
Where rights allow, enough detail is published for a competent reader to reproduce the result.
- 03
Challenge
Anyone may challenge a method or result at integrity@globalriskalliance.com, without membership or payment.
- 04
Correction
Errors lead to a recorded new version, listed on corrections and updates.
What a GRA estimate is not.
A GRA estimate or scenario is evidence for a reader's own judgement. It is never a forecast of loss, a rating, a supervisory test, a capital requirement or investment advice, and no institution should substitute it for its own validated models. Related pages: financial data and rights, financial stability analysis and scenarios and stress testing.
Terms used on this page.
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. |
| Uncertainty quantification | Characterising and propagating the uncertainties in a model to state confidence in its outputs. |
| Back-testing | Comparing a model's predictions with realised outcomes to check its accuracy. |
| STEM Commons | The shared reference of science, technology, engineering and mathematics methods for the Nexus network, hosted by GCRI, with one canonical page per method and a named steward platform. |
Where this page's references lead.
The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.
Questions and answers.
Can I reuse a GRA method?
Why does GRA report ranges instead of single numbers?
Are these methods standards?
How do methods relate to the STEM Commons?
Around GRA.
Start with an enquiry.
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.