Internal Models

Material risks


Article 121(4) of the Directive requires that internal models cover ‘all material risks’ to which firms are exposed.

Events not in data


The concept of ENID also applies to the data used to set the parameters for the internal model. Firms should not assume that parameterising the internal model using only historical data will take into account all quantifiable risks, unless an unadjusted distribution has been shown to capture the full range of possible future events, for example by way of stress and scenario testing.


For example, for liability lines, data sets covering recent years may not include sufficient examples of liability catastrophes, which can significantly increase the dependency between policies, and, as a result, the volatility. Parameterising the internal model using such a data set alone would omit the possibility of future liability catastrophes, failing to cover all material risks.

Risks covered by third party models


Where firms use third party models, firms should take particular care to demonstrate that the model covers all material risks in their own risk profile. This is consistent with the expected requirement in the Delegated Acts to monitor limitations arising from the use of external models. For example, where firms have used a third party model for earthquake exposure, they should ensure that the internal model also covers related risks, such as corresponding tsunami exposure.

Consistency with technical provisions


Article 121(2) requires the methods used in the internal model to be ‘consistent with the methods used to calculate technical provisions’.


In order to calculate the movement in basic own funds over one year, it is necessary to calculate technical provisions in the internal model. When selecting a method for this purpose, firms should ensure that the method produces similar results to a full technical provisions calculation throughout the probability distribution forecast, and not just in benign circumstances.


For example, firms should not use an automated re-reserving (‘actuary-in-the-box’) method with a basic chain ladder where this would fail to capture the significant judgement – such as a change in reserving basis – that reserving practitioners would apply following a severe deterioration in claims incurred or the emergence of new information such as a legal judgement.

Assumptions and techniques


Article 121(2) requires firms to base the internal model on ‘adequate, applicable and relevant techniques’ and ‘realistic assumptions’, and to ‘justify the assumptions underlying the model’ to the PRA.

Uncertainty around parameters


Firms should allow for estimation error where this is material and it is practicable to do so, in line with the expected Delegated Acts.


For example, where there is significant uncertainty around a sensitive parameter, so that the correct value could lie anywhere in a range, firms should seek to reflect the parameter uncertainty in their choice of parameter value unless they have otherwise quantified and allowed for this estimation error in the model.

Calendar year effects


Calendar year effects, such as claims inflation, can have a significant impact on the volatility of future reserve development. Firms should only use methods that do not capture calendar year effects explicitly if they have shown that the resulting distribution appropriately reflects the volatility introduced by these effects, or if such volatility is captured elsewhere in the model.

Improvements in performance


Firms should not assume an improvement in performance relative to that seen in the past unless such an improvement has been clearly justified, in line with the expected Delegated Acts. For example, it would not be realistic to base the internal model on a business plan which assumes improved underwriting results unless the measures taken have been shown to be effective.

One-year emergence of risk


Firms should not assume that insurance risk emerges simply according to a historical paid or incurred development pattern. Where firms use an emergence factor method (where one-year risk is assumed to be a proportion of ultimate risk), firms should not base the emergence factor purely on the incurred or paid pattern.


Where historical paid or incurred patterns are used in the model, firms should not assume that these will be repeated in future, unless the firm has shown that this is a realistic assumption throughout the probability distribution forecast.

Industry standards


While, in line with expected requirements in the Delegated Acts, firms should ensure that the internal model reflects progress in generally accepted market practice, assumptions cannot be justified solely on the grounds that they are ‘industry standard’ or ‘established good practice’. Firms should justify assumptions on the basis of their own specific risk profile.

Default options


When justifying the assumptions underlying an external model, it is not sufficient to justify the assumptions on the grounds that they are selected by default. Firms should justify all assumptions on the basis of their own specific risk profile.


For example, where a catastrophe model is set by default not to allow for clustering of storms, firms should demonstrate that this assumption is appropriate for their risk profile, and cannot justify this assumption on the grounds that it is selected by default.



Article 121(3) requires that ‘data used for the internal model’ to be accurate, complete and appropriate.

Data used


Any data that can have an impact on the outputs of the internal model should be considered to be ‘used for the internal model’, and must therefore be accurate, complete and appropriate. For example, where a firm has material natural catastrophe risk, the exposure data input into the catastrophe model should be accurate, complete and appropriate.

Risk mitigation


Article 121(6) allows firms to take into account risk-mitigation techniques in the internal model, as long as the risks arising from the techniques are ‘properly reflected’.

Reinsurance exhaustion


The most common risk mitigation technique is the modelling of purchased reinsurance. Where firms model reinsurance, they should allow for the possibility of reinsurance exhaustion in order to ensure that the risks arising from the risk mitigation techniques are properly reflected.

Management actions


Article 121(8) allows firms to take into account ‘management actions that they would reasonably expect to carry out in specific circumstances’.

Renewal of reinsurance


Firms should treat the renewal of reinsurance in the model as a future management action unless it has been shown that the renewal will not rely on a decision made by the firm.

Validation standards


Article 124(1) requires firms to have a regular cycle of validation to ‘review the ongoing appropriateness’ of the internal model.

Specific validation


In order to review the ongoing appropriateness of the internal model, firms should perform validation that relates specifically to their own risk profile. For example, it is not satisfactory to review the appropriateness of a third party model purely on the basis of generic validation performed by the model vendor.

External models and data


Article 126 of the Directive requires firms to apply Articles 120 to 125 to external models and data.

Data from third party models


Firms often use data output from a third party model. Where the assumptions and methods the third party uses to produce the data could have a material impact on the outputs of the firm’s internal model, firms should demonstrate that the external model itself satisfies Articles 120 to 125, and not the data alone.


For example, where firms are provided with catastrophe risk event loss tables by a third party, Articles 120 to 125 should be applied to the model that produced the tables, and not to the tables alone.