Model Risk

A model risk refers to the risk of loss due to using a financial model with fundamental errors or incorrect use of a model for decision-making. The entities can assess and reduce the risk using an effective validation process, evaluating fundamental correctness and output analysis.

Companies or investors use models for a wide range of objectives. Its application helps make better decisions, save money, and reduce various risks. However, the inefficient use of models can cause negative effects like financial loss, poor strategic decision-making, damage to reputation, and noncompliance with laws and regulations.

• Model risk emerges when a financial model with fundamental inaccuracy is applied or incorrect use of a model for business or strategic decision-making happens.
• The main types are specification risk, implementation risk, and model application risk.
• Model Risk Management (MRM) controls risks indicated by the possible adverse effects of choices made using flawed or inappropriate models.
• The entities can assess and reduce the risk using an effective validation process, evaluating fundamental correctness and output analysis.

Model risk refers to the threat that arises from the failure of a financial model and in scenarios where it is not performing as expected and is flawed due to many errors and mistakes. In modern organizations, different models exist that are designed to help them avoid any contingencies and uncalled consequences and make loss averse so that they can make effective decisions for the management and operations of the company.

With different models exists the issue of model risk. Therefore, organizations need to participate in managing such risks at different levels of operations and work. Every company has a structure designed to mitigate such risks by regularly introducing and practicing risk audits to improve the accuracy of making decisions, avert the risk of financial loss and damage to the system, and grow their productivity.

It also exemplifies the vulnerability to technical mistakes, data issues, bugs, and other internal and external factors. Most such systems are quantitative, so every company has a risk analyst to evaluate them, which is a long, tedious, and complex process; therefore, the whole process of model risk assessment is very significant. 

Let's look into some significant types:

• Specification Risk: The specifications of a model play a vital role. A model with the correct specifications will implement correctly and bring results with a sense of usability; otherwise, the model will turn inapplicable. It is typically associated with the design of the model. The risk exists even after designing it in such a type and may need periodic modifications.
• Implementation Risk: It refers to the flawed implementation of a model. The data used for the input may differ, the structure may be inappropriate, and the usage of inaccurate numerical approximations also can happen. In these cases, risk often indicates that the outputs will get disengaged, complex, and mislead the organization in a different direction. The risk can occur in an information technology environment due to programming and technical errors.
• Application Risk: It reflects the unsuitability of a model regarding the field. When a model is misinterpreted, it only produces false data with no reliability. Any actions based on it will only bring chaos and dysfunction to the organization. Therefore, it is best to have a suitable model for the subject, and only then are actions advisable.

Let's look into some examples for a better understanding of the concept:

Example #1

A business evaluates its workforce using a third-party software program making judgments about remuneration and career progression. The application includes features that enable a citizen data scientist to create and apply AI models for hiring, evaluating, and promoting personnel.

The risk management and compliance teams are not involved in the deployment or operation of the model. A worker passed over for a promotion complains that the AI model is biased, files a lawsuit against the firm for unfair labor practices, and challenges it to demonstrate that the model is impartial.

Example #2

In 2013, it was disclosed that JP Morgan Chase & Co. (JPM), whose trading loss of over $6.2 billion in 2012, was caused by adopting a formula or model that undervalued risks. They use a specific formula to study the risk of its credit derivatives position, and they have used many models since January 2012.

Bruno Iksil, sometimes referred to as the London Whale because of the size of his bets, constructed the portfolio. In addition, in the first quarter of 2013, the bank modified how it calculated value at risk (VaR) for the credit derivatives book to improve uniformity with related products within the corporate and investment bank.

MRM controls risks indicated by the possible adverse effects of choices made using flawed or inappropriate models. Itprimaryin purpose is to employ strategies, behavior, and practices that measure and mitigate such risks. Companies design an MRM framework to mitigate model risk and improve decision-making. Some essential concepts surrounding the management include the following:

• Governance: Governance comes with policies, activities, and guidelines regulating the execution of MRM. The U.S. Federal Reserve Bank has established a regulatory standard called SR 11-7 that guides MRM. In addition, the Federal Reserve, the Federal Deposit Insurance Corporation, and the Office of the Comptroller of the Currency (collectively, the agencies) have explained how the risk management principles outlined in the agencies' "Supervisory Guidance on Model Risk Management" (also known as "model risk management guidance" or MRMG) relate to the systems or models used by banks to help them act by the requirements of Bank Secrecy Act laws and regulations.
• Validation: It confirms the accuracy of a model and validates its purpose and efficiency by testing it in various scenarios and with data inputs to ensure that the model is predictive with the right conditions. The model risk validation also helps validate by comparing different experimental data results. Furthermore, the relevance and quality of data are equally considered.

This article has been a guide to what is Model Risk & its definition. Here, we explain the topic in detail, including its types, examples, and management. You may also find some useful articles here -

• Markowitz Model
• Enterprise Risk Management
• Credit Risk Management