16.7 Quality of Models

Good Model:

• Show the balance between risk and reward for a range of different strategies

  (e.g. changing asset mix or reinsurance program or LoB to grow)

• Recognizes and reflects its own imperfection

  (Model only approximates reality and the parameters are uncertain)

Weak Model:

• Over(under)state risks \(\leadsto\) Overly cautious (aggressive) choices

Factors that affect the suitability and usefulness of the model

• Data quality

• Choices on assumptions and mathematical method

• Variety of different risk elements and how they are represented

Key elements that differentiate model quality:

• Model:

  1. Reflects relative importance of various risks to business decisions

  2. Includes mathematical techniques to reflect dependencies

• Modelers:

  1. Have a deep knowledge of the fundamentals of those risks

  2. Have a trusted relationship with senior management

Model should take into account the uncertainty from other models (e.g. CAT, ESG, credit risk) that are used as inputs

Risks less amenable to detailed representation in enterprise risk model (Non-quantifiable risk):

• e.g. Op risk (IT exposures, pension inadequacy, key man risk, rogue traers, frade)

• They are extremely important to the success of a business

• Requires specialized management process and difficult to incorporate into an overall risk model

• Can be modeled in bulk using informed judgments (but with high degree of uncertainty)

• Use model to manage the risk for which the modeling process is effective

• While some risks are at best weakly represented in such a model and requires other management methods