Quick Search
|
Internal Models
The alternative to the standard formula approach is very different to the current regime - the opportunity for firms to develop internal models of the business and use these models to calculate their regulatory capital requirements (SCR).
An internal model will allow a much more bespoke assessment of the particular business and the unique set of risks that it faces, which can never be replicated in a standard formula, and has the potential to be a useful management tool. Internal models are already in common use within the larger and more sophisticated firms in Europe, and the directive is clearly endorsing the use of the these models for risk management and regulatory purposes.
The directive also allows firms the option of using a partial model, with some components of the standard formula SCR replaced by results from an internal model. However, such firms will be scrutinised to ensure that they are not cherry-picking those elements that will provide capital relief.
An interesting point arising from the directive is that some firms may be required to develop a (full or partial) internal model if the supervisor considers that their risk profile deviates significantly from that assumed when calibrating the standard formula SCR. It is also likely that some firms may come under “peer pressure” to develop internal models from other similar firms that have a more sophisticated approach.
It should be noted that supervisors will have the power to impose a capital add-on, regardless of the approach, if the risk profile deviates significantly from the assumptions underlying the SCR, or if there are concerns over the governance standards within the firm. Ideally this will be a temporary measure, and firms will be advised on how to adapt their approach to have the add-on removed.
Approval of Internal Models
Firms will have to apply to their supervisor to get agreement for the use of an internal model. The directive states that the supervisor will decide on the application within six months of receipt. This sounds like a tall order, particularly if many applications are received in a short period of time.
As part of the initial approval process of their internal models, firms shall agree with the supervisory authorities on a policy for changing the model and may then change their internal model only in accordance with that policy. It should include a specification of minor and major changes, and any major changes to the model, as well as changes to the policy, shall always be subject to prior supervisory approval. It will be interesting to see how this works in practice.
In order for their model to be approved, a firm must demonstrate that it passes a “use test”, meets particular statistical quality standards, and is capable of being calibrated to the “one-year 99.5%” level specified in the SCR.
The “use test” requires that the model is embedded within the system of governance and is a key tool in the decision-making processes. There must also be processes in place to ensure that the model continues to reflect the risk profile of the firm concerned.
Firms will need to justify the assumptions underlying their models. They may take into account future management actions that they would carry out in specific circumstances, as long as the model allows for the time necessary to implement them. Supervisors may require firms to run their internal models on relevant benchmark portfolios and use assumptions based on external rather than internal data in order to verify the calibration and to check that it is line with accepted market practice. Models will need to be sufficiently flexible to allow this to be performed.
Regular model validation should be performed, including a comparison of results against actual experience. This validation process should allow the firm to demonstrate to their supervisory authorities that the resulting capital requirements are appropriate. In particular, sensitivity of the results to changes in key underlying assumptions should be tested.
The model should be thoroughly documented, including an outline of the design of the model and details of the methodology, data and assumptions used. Major changes to the previous version of a model should also be documented.
|
|
|
