Physical climate scenarios in insurance capital models using NGFS pathways

Insurers increasingly need to convert climate pathways into capital-relevant evidence rather than treating them as separate sustainability analysis. NGFS scenarios provide a common reference point, but they are not insurance capital models. The practical task is to map pathway variables, hazard changes and macro-financial assumptions into underwriting, catastrophe, reserving, asset and management action modules, while retaining clear governance over uncertainty, use limitations and validation. For ngfs climate insurance capital work, the main control question is not whether a scenario has been run, but whether the translation from climate pathway to solvency impact is decision-useful and reproducible.

Why NGFS pathways matter for insurance capital

The Network for Greening the Financial System provides climate scenarios designed to support financial risk analysis. They combine transition pathways, physical risk assumptions and macroeconomic outputs. For insurers, their relevance lies in creating a structured basis for exploring how future climate states may alter risk profiles over horizons that are longer than most one-year solvency calculations.

Insurance capital models are usually calibrated to a specified confidence level over a defined time horizon. Climate scenarios, by contrast, describe conditional pathways under assumptions about emissions, policy, technology and physical climate response. This difference is fundamental. A pathway is not a probability distribution for annual losses. It can, however, inform stresses, parameter shifts, exposure assumptions, underwriting strategies and ORSA narratives.

Physical climate risk affects insurers through acute hazards such as windstorm, flood, wildfire, hail and tropical cyclone, and through chronic changes such as sea-level rise, heat stress and precipitation pattern changes. These effects can influence:

• frequency and severity distributions in catastrophe models;
• insured values and vulnerability functions;
• geographical accumulation and diversification assumptions;
• claims inflation, repair costs and demand surge;
• reinsurance cost, availability and attachment structures;
• lapses, morbidity or mortality assumptions for some life and health portfolios;
• asset valuations where physical risk affects collateral, infrastructure or real estate.

The practical value of NGFS pathways is that they create a repeatable scenario taxonomy. This can improve consistency across risk appetite, ORSA, pricing, reinsurance and capital planning. It also supports stronger governance when connected to documented model ownership and escalation routes, as described in risk ownership and accountability and insurance enterprise risk management.

From pathway variables to insurance loss drivers

The first implementation issue is granularity. NGFS outputs are not usually at the same spatial, temporal or peril-specific resolution as insurance catastrophe models. Capital teams therefore need a translation layer.

Identify the relevant pathway dimensions

For physical risk, the most relevant pathway dimensions normally include:

• temperature outcomes and time horizon;
• precipitation change and hydrological stress;
• sea-level rise assumptions;
• chronic heat indicators;
• regional hazard indicators where available;
• macroeconomic variables affecting inflation, asset values and exposure growth.

The insurer should document which dimensions are used directly, which are proxied through external hazard science, and which are excluded because they are not material or not model-ready. Exclusions require evidence. For example, a monoline motor insurer may justify limited use of sea-level variables, while a household insurer with coastal exposure may not.

Translate climate signals into peril adjustments

Translation is usually performed through peril-specific adjustments rather than a single climate multiplier. Examples include:

• shifting event frequency for flood in regions where precipitation intensity is expected to increase;
• applying sea-level rise to coastal flood depth distributions;
• modifying wildfire ignition, spread or suppression assumptions using heat and drought indicators;
• changing vulnerability curves where building performance is affected by repeated heat or flood exposure;
• adjusting demand surge after correlated regional events.

The mapping should distinguish hazard, exposure and vulnerability. Combining them into one scalar can be useful for an early stress test, but it reduces diagnostic value and makes validation more difficult. A more robust approach is to document whether each change affects event occurrence, event intensity, insured value, damage ratio, claims settlement cost or reinsurance recovery.

Preserve consistency with capital model architecture

A climate pathway should not be inserted as an undocumented override. If the internal model or standard formula capital process has defined modules, dependencies and management actions, climate adjustments should be assigned to those components. This is particularly important for groups using catastrophe vendor models alongside internal aggregation and reinsurance modules. Related governance issues overlap with solvency and capital modelling and catastrophe risk.

Framework

A workable framework has five stages: materiality, scenario selection, translation, capital integration and governance review.

1. Materiality assessment

The insurer should assess whether physical climate risk is material by portfolio, peril, geography and time horizon. Materiality should be considered both on a current balance sheet basis and under plausible strategic plans. A low current exposure does not necessarily imply low future risk if growth, acquisition or distribution strategy increases concentration in climate-sensitive regions.

Typical evidence includes exposure maps, historical catastrophe loss experience, peril model sensitivity, claims trend analysis, reinsurance recoveries, asset location data and underwriting plans. The output should be a ranked view of material climate-sensitive portfolios and a clear statement of where scenario analysis is required.

2. Scenario selection

The insurer should select NGFS pathways that are relevant to its risk question. A physical risk stress may focus on higher-warming pathways, while a combined solvency scenario may include both physical and transition effects. The selected time horizons should be linked to business use. For example:

• one to three years for business planning and reinsurance purchasing;
• three to five years for underwriting portfolio steering;
• ten years or more for strategic asset allocation, product design and ORSA narrative;
• longer horizons for chronic physical risk analysis where liabilities or assets are long duration.

Capital teams should avoid implying precision at horizons where model and exposure uncertainty are high. Scenario selection should be approved through the same governance route used for material stress testing, with challenge from the risk function and, where material, board-level visibility consistent with board risk reporting.

3. Translation and calibration

The translation layer should define the mathematical relationship between pathway indicators and insurance model parameters. Methods may include expert judgement, catastrophe vendor climate-conditioned views, academic hazard studies, statistical trend analysis, engineering models or internal claims experience. Each method should be ranked by evidential strength and relevance.

Where expert judgement is material, it should be structured, challenged and documented. The governance expectations are similar to other judgement-heavy capital processes, including clear rationale, alternatives considered and sensitivity to the judgement. This aligns with principles discussed in governance of expert judgement.

4. Capital model integration

The integration approach depends on the model type. In a full internal model, scenario-conditioned parameters may be run through the stochastic engine. In a standard formula environment, the insurer may use scenario analysis to assess whether the formula remains adequate for its risk profile or whether an additional management buffer is needed. For ORSA, both approaches may be relevant.

Important design questions include:

• whether the climate scenario is a deterministic stress or a shifted probability distribution;
• whether dependencies between perils change under the pathway;
• whether reinsurance terms are assumed constant or re-priced;
• whether management actions are credible under correlated market and catastrophe stress;
• how climate effects interact with inflation and claims handling capacity.

5. Governance review

The final stage is governance review. The risk committee should understand the scenario purpose, material assumptions, key sensitivities, impact on solvency position and actions under consideration. Outputs should be linked to underwriting appetite, reinsurance strategy, capital planning and risk limits. Scenario analysis that does not influence decisions should be reviewed for relevance.

Worked numerical illustration: coastal flood capital stress

The following example is illustrative and does not represent a regulatory calibration or empirical estimate.

Assume a non-life insurer writes residential property in three coastal regions. Its current catastrophe model estimates a one-in-200 annual gross flood loss of 500 million before reinsurance. Net loss after an excess-of-loss programme is 220 million. The insurer uses an NGFS higher-physical-risk pathway to examine a 2035 balance sheet under unchanged underwriting strategy.

The translation layer applies three adjustments:

1. Exposure growth: insured values increase by 15% based on the insurer's business plan and assumed claims inflation. 2. Hazard shift: coastal flood depth distributions are adjusted so that modelled event severity increases by 10% for affected zones. 3. Vulnerability shift: damage ratios increase by 5% for older properties where flood resilience measures are not assumed.

If applied multiplicatively to the gross one-in-200 loss, the adjusted gross loss is:

500 million × 1.15 × 1.10 × 1.05 = 664.1 million.

Reinsurance is then recalculated rather than scaled mechanically. Suppose the current programme has a 100 million retention and 300 million cover. Under the adjusted event, the net loss is:

Retention of 100 million + loss above exhausted cover of 264.1 million = 364.1 million.

The gross loss increases by about 33%, but the net loss increases from 220 million to 364.1 million, a larger relative increase because the reinsurance layer is exhausted. This illustrates why climate translation should be run through the reinsurance module. Scaling net losses directly would understate the capital effect if covers, reinstatements or aggregate limits bind.

The insurer then runs sensitivities:

• no exposure growth: gross loss 577.5 million;
• lower hazard shift of 5%: gross loss 633.9 million;
• additional demand surge of 8%: gross loss 717.2 million;
• revised reinsurance purchase with higher limit: net loss reduced but premium and counterparty exposure increase.

The scenario output is not a new best estimate of required capital. It is evidence for decisions: whether to revise coastal underwriting limits, increase reinsurance protection, strengthen property resilience incentives, hold additional capital, or change the strategic plan. The key governance point is that management should see both the numerical outcome and the assumptions that drive it.

Validation and control considerations

Climate scenario integration should sit within the model risk management framework. It is not sufficient to state that uncertainty is high. High uncertainty increases the need for transparent controls.

Validation checklist

A validation review should consider:

• Purpose: Is the scenario designed for ORSA, pricing, reinsurance, risk appetite or capital adequacy assessment?
• Scope: Are all material portfolios, perils and geographies included or explicitly excluded?
• Data lineage: Can pathway data, exposure data, model versions and adjustments be traced?
• Translation logic: Are hazard, exposure and vulnerability effects separated?
• Calibration evidence: Are adjustments supported by credible internal or external evidence?
• Expert judgement: Are judgements documented, challenged and approved?
• Model integration: Are reinsurance, aggregation and dependency effects recalculated rather than approximated without justification?
• Sensitivity testing: Are results tested against alternative assumptions and pathways?
• Back-testing: Where possible, are near-term assumptions compared with emerging claims and hazard indicators?
• Use test: Are outputs linked to decisions, limits or capital planning?
• Limitations: Are uncertainty, granularity gaps and non-modelled perils clearly stated?

Independent review should be proportionate to materiality. For a material capital use, validation standards should be comparable to other significant model changes. Useful reference points include independent model validation standards, model risk management frameworks and model limitations and compensating controls.

Monitoring after implementation

Climate model assumptions should not be static. Monitoring indicators may include exposure accumulation in high-risk zones, changes in catastrophe model vendor views, claims severity trends, reinsurance market feedback, flood defence investment, building code changes and observed hazard frequency. The monitoring process should define escalation thresholds. For example, a material increase in coastal accumulation or a change in reinsurance attachment affordability may trigger an updated scenario run before the next ORSA cycle.

ORSA, solvency and management action implications

Regulatory expectations increasingly emphasise that insurers should assess climate-related risks in their own risk and solvency assessment where those risks are material. The ORSA lens is important because physical climate risk may emerge over horizons that exceed the one-year capital measure, while strategic decisions taken now can alter future exposure.

A credible ORSA treatment should explain:

• why selected pathways are relevant;
• how materiality was assessed;
• how climate variables were translated into insurance risk drivers;
• how impacts affect solvency, liquidity, reinsurance and business strategy;
• what management actions are available and under what triggers;
• why the board is comfortable, or not comfortable, with the residual risk.

Management actions require careful scrutiny. Restricting new business in high-risk zones may reduce future exposure but may not reduce existing liabilities immediately. Reinsurance purchase may reduce net capital strain but could be limited by market capacity or price. Premium increases may be constrained by competition, conduct requirements or public policy. Asset sales may be less effective in a system-wide climate stress if market liquidity deteriorates.

Physical climate scenario analysis should therefore be connected to the wider stress testing programme. It should not operate as a separate reporting exercise. Integration with stress testing programmes, climate scenario analysis and orsa governance for strategic decisions helps ensure that scenario results inform capital and strategy discussions.

Limitations

Several limitations should be explicit in any NGFS-based capital analysis.

First, NGFS pathways are not insurer-specific catastrophe models. They provide structured scenario assumptions but do not remove the need for peril-specific hazard science, exposure data and claims modelling.

Second, spatial resolution can be a constraint. Insurance losses are highly location-specific, especially for flood, wildfire and convective storm. Translating regional climate variables to address-level exposure requires assumptions that may dominate the result.

Third, the probability of a pathway is not usually specified. A scenario impact should not be presented as a percentile capital requirement unless the insurer has separately justified a probabilistic interpretation.

Fourth, model uncertainty is material. Climate models, catastrophe models and insurance financial models each contain uncertainty. Combining them can create false precision if outputs are reported without ranges, sensitivities and narrative context.

Fifth, non-stationarity weakens reliance on historical claims alone. Historical experience remains important for validation, but it may not fully represent future hazard regimes, exposure development or adaptation measures.

Sixth, adaptation and public policy are difficult to model. Flood defences, building codes, land-use planning, insurance availability and government compensation schemes can materially alter insured losses. These factors should be considered through scenario variants where material.

Frequently asked questions

Should NGFS scenarios be used directly as capital model inputs?

Usually not without translation. NGFS pathways provide climate and macro-financial scenario information, but insurance capital models require peril-specific frequency, severity, exposure, vulnerability, dependency and reinsurance assumptions. The insurer should document the conversion from pathway variables to model parameters.

Is a physical climate scenario the same as a one-in-200 capital event?

No. A physical climate pathway describes a conditional future state or trajectory. A one-in-200 capital event is a probabilistic severity measure within a defined model horizon. A pathway can be used to condition a catastrophe distribution or to define a stress, but the distinction should remain clear.

How should insurers handle uncertainty in long-term climate projections?

They should use ranges, sensitivities and transparent limitations rather than single-point precision. Material assumptions should be challenged and monitored. Where uncertainty is high and the exposure is material, management may need risk appetite limits, underwriting controls or capital buffers even if exact quantification is not possible.

What evidence should boards expect from management?

Boards should expect a concise explanation of material exposures, selected scenarios, key assumptions, solvency impact, management actions and limitations. They should also see whether the analysis has changed underwriting, reinsurance, capital planning or strategic decisions.

Professional disclaimer: This article is for technical risk governance information only and does not constitute actuarial, legal, regulatory or investment advice.