modelrisk

Model risk refers to the potential for adverse outcomes arising from the use of incorrect, misapplied, or poorly understood models. A model is any formal representation—statistical, financial, mathematical, machine learning, or expert judgment—that maps inputs to outputs such as prices, probabilities, forecasts, or decisions. Model risk emerges when the model is wrong for the task, when it is used outside its intended scope, or when it is implemented or maintained poorly.

Common sources of model risk include specification errors (invalid functional form or omitted variables), data quality issues (biased samples, missing or non-stationary data), estimation uncertainty (uncertainty in parameter values), implementation mistakes (coding or system integration errors), and misuse (overreliance on models, inappropriate backtesting). Other factors include changing underlying conditions that render a model obsolete, feature drift, and governance gaps such as weak documentation or inadequate model monitoring.

Model risk management is the discipline that identifies, measures, monitors, and mitigates these risks. It typically features an independent validation function, formal policies and controls, a risk appetite for model use, and a comprehensive model inventory. Key activities include rigorous development and testing standards, documentation of assumptions and limitations, change management, and ongoing performance monitoring. Validation often covers conceptual soundness, data adequacy, implementation fidelity, and out-of-time and stress testing to assess robustness.

Applications and regulation vary by sector, with particularly prominent emphasis in banking and financial services due to pricing, risk measurement, and regulatory capital requirements. Regulators and international bodies encourage disciplined model governance, periodic review, and transparent reporting to reduce the likelihood and impact of model risk across industries.