What Is Model Risk?

Model risk is the potential for adverse consequences resulting from decisions based on incorrect or misused models. The use of models for valuing financial instruments, measuring risks, or making crucial business decisions is ubiquitous, and model risk emerges from potential losses if these financial models are inaccurate or misused. The consequences of model risk have been highlighted in financial events such as the London Whale incident, the collapse of the Long-Term Capital Management (LTCM) hedge fund, and the subprime mortgage crisis of 2008–2009. For instance, during the subprime mortgage crisis, the housing market collapsed because many of the financial instruments that backed or relied on subprime mortgages turned out to contain lower-value properties than their ratings indicated. This became a problem when the models used did not accurately reflect the dynamics and risks of the market, leading to widespread financial instability. Without accurate models, financial institutions are unable to properly assess risk, which can result in poor decision-making and significant economic consequences.

From the perspective of model risk, Modelscape is instrumental in reinforcing the management of such risks by equipping organizations with advanced tools for the entire life cycle of financial models—ranging from their creation to deployment. This comprehensive approach not only ensures models adhere to stringent regulatory standards but also fortifies the governance processes, pivotal in curtailing potential financial losses arising from model failures.

What Causes Model Risk?

Model risk can arise from a variety of sources, including:

• Incorrect data: The most sophisticated model is vulnerable to garbage-in/garbage-out scenarios. Incorrect or outdated data can skew the model’s output, leading to erroneous decisions. For example, an algorithmic trading model fed with outdated stock prices due to a data feed glitch could erroneously execute large purchases, leading to substantial financial losses.
• Inaccurate models: When financial models contain flaws in their design, the decisions based on them can lead to significant financial losses. For example, a bank suffered major losses on mortgage-backed securities after its risk assessment model underestimated the likelihood of a nationwide housing market collapse.
• Inappropriate use: Using a model beyond its intended purpose, such as applying a home loan model for assessing the probability of default in a different context, can yield misleading results.

Mitigating Model Risk

Given the reputation and financial impact of model risk, financial institutions deploy a comprehensive suite of strategies to mitigate it. These strategies include:

• Model documentation: Keeping detailed records of model design, assumptions, and operational procedures
• Model validation and monitoring: Regularly reviewing models to ensure their accuracy and relevance
• Scenario analysis and stress testing: Evaluating models under extreme but plausible scenarios to assess their resilience
• Benchmarking and challenging models with machine learning: Using advanced techniques to test models against benchmarks and challenge their assumptions
• Model risk reporting: Keeping stakeholders informed about model risks and the measures taken to mitigate them

Modelscape, MATLAB^®, Statistics and Machine Learning Toolbox™, Risk Management Toolbox™, MATLAB Report Generator™, and MATLAB Production Server™ are tools for modeling and mitigating risk.