Banks re-estimating their credit risk models for usage in the Internal Rating Based (IRB) approach are facing vast and complex challenges, investing considerable time, money and management focus during the lifetime of the project. The ambitions for the institutions vary and where one bank may align existing models to new regulatory requirements and expectations, another (bank) may try to address structural deficiencies in the earlier generation of models. To be successful in its modelling initiatives it is critical for the IRB-programme to understand how the risk models align to the objectives and interests of the end-user groups / stakeholders.
Simplified, the dynamic and challenges of an IRB-programme is still well captured by listing the model motivations for the end-user groups, and understanding and accepting the motivations and rationale is critical in IRB-programmes since one way of measuring the success of the programme is to evaluate the stakeholders’ updated capacity (with the new set of models) to perform the activities for which they are responsible.
As example: The business side will find the success of an IRB-project highly dependent on the degree of business orientation that is encapsulated by the models. Business orientation in this context for instance could be models very risk sensitive to type-, sub-type of collateral. Or strong model recognition to the expert based derivation of likelihood for restructuring in the event of a default. From the centralised risk point-of-view, these model features may be in conflict with e.g. achieving robust and stable model parameters, derived from observed data and applied to a less complex model structure resulting in low MoC:s (Margins of Conservatism). These conflicting views of the ambition of the models will result in both sides having to make concessions during the project and where the management of these concessions constitute one of many critical success factors.
While the regulations and guidelines in many areas are clear and prescriptive, leaving no room for interpretation, other, less prescriptive, areas may provide modelling opportunities which could be explored, albeit at the cost of higher initial MoC:s during a defined transitional period (until the set of assumptions and / or data can be confirmed).
Thus, it is important to acknowledge that a need for regulatory interpretation does not solely arise as a consequence of discussions driven by the business side or any other stakeholder. Even the risk function will face a need to interpret and make regulations operational during the course of modelling (and this will eventually form the basis for the next generation of unexplained REA variability, however this time at lower levels). This process of operationalisation is important and where the regulatory objectives must be considered carefully to ensure that model / risk measurement simplifications are not conflicting with the regulatory objectives. I.e. that the institution(s) in all of their activities shall measure risk accurately so that informed business decisions can be taken, supported with an appropriate (conservative) amount of capital.
Parallel to the measurements for regulatory capital purposes runs the internal business equity models which, by means of additional risk measurement capacity, are fundamental for the capital distribution- and transaction pricing processes. In the general case, the more an IRB-programme tilts towards a strict regulatory capital orientation (e.g. minimising MoC:s), the stronger the need to support the IRB- models with enhanced business equity models (still to be based on the IRB-models as part of the use-test requirements). And the other way around, the more complexity that is built in to cater for business requirements, the less the need for additional adjustments in a business equity context, but the higher the initial capital add-on to the regulatory capital, in addition to any potential increase in authorisation risk.
Concluding the description of the complexity and challenges that are inherent in IRB modelling projects, FCG would point out the following success factors.
- Wide and engaged participation: Senior managers and executives from management, risk- and the business side should participate in the modelling programme to ensure correct input with respect to both historical- and forward looking strategic directions which may have an impact for the risk modelling stream (e.g. in the interpretation of the available internal data).
- Deep Understanding: While Deep Learning may be deployed from a technical modelling perspective, it is even more critical to apply a “Deep Understanding”-approach throughout the organisation during the lifetime of the project. Recurring education sessions for senior management and the board, combined with coherent and encompassing impact analysis’ during the design-, modelling- and implementation phase, support resource allocation initiatives during the programme. It also provides input to stakeholders on the impact on tools and processes, enabling appropriate recalibration to these in due time
- Consistent governance: Ensure consistency between the internal project- and the ongoing governance structures with respect to risk measurement and model risk management. Inconsistent governance may result in conflicting programme objectives which eventually may lead to a misunderstanding of the project output and its capacity to support risk measurement-, capital- and pricing strategies
- Manage trade-offs and model limitations: Identifying, deciding, accepting and documenting trade-offs is essential during the project. Correctly identified and managed model limitations can be appropriately addressed immediately or in relation to the project-phase-II planning (e.g. activities such as changes / updates to risk appetite, limit control, business equity models or even planned updates to the newly approved IRB-models)
- Prudent regulatory operationalisation: IRB authorisation risk is often associated with aggressive assumptions on validity of data points, estimation techniques etc. and where the introduction of more assumption and model complexity will decrease the likelihood of getting regulatory approval. This holds true and is an essential component in the modelling strategy decisions. But it is also important to acknowledge that too much simplification, leading to non-appropriate risk measurement tools, may also constitute an authorisation risk. If the objective of the regulation is to enhance risk measurement capabilities and support transactions with sufficient capital, this should also be dominating features in the models.
FCG has long experience in managing strategic- as well as methodological decisions related to IRB modelling programmes and where we can support the institution to reach the overall objective of the programme. To ensure that all stakeholders gets updated tools and processes that align to their respective areas of responsibility.