Model risk

from Wikipedia, the free encyclopedia

The model risk is the risk that

  • a model has been designed incorrectly or is unsuitable for the selected application or
  • a model is used improperly or
  • incorrect input data are used for a model or
  • a model is no longer up-to-date or inconsistent

The concept of model risk is used in finance , especially in risk management .

The model risk can be referred to as meta-risk, since it applies to other risks that are measured using models.

Types of model risk

According to Crouhy, Galai & Mark, a distinction is made:

  1. Incorrect / incorrect models and incorrect specifications in the model
  2. Error in the analytical solution
  3. Incorrect specification of the underlying stochastic processes
  4. Missing risk factors in the model
  5. Missing elements in the model
  6. Incorrect classifications or identification of examined parameters
  7. Incorrect implementation of models
  8. Incorrect model calibration
  9. Incorrect processing of market data
  10. Incorrect model application

Model risk as part of operational risk

The model risk can be understood as part of the operational risk . If models are used in the company (particularly for risk assessment), a model risk arises (of an unknown amount). For example, banks have to deposit this as an operational risk with equity under Basel II .

Example: model risk in the valuation of derivatives

The model risk describes the risk or the uncertainty that arises from the fact that models are used to value / hedge derivatives. However, these models only represent a representation (a model ) of reality. This model can be incorrect or incomplete. Conclusions from the model (such as an option valuation according to the Black-Scholes model ) do not have to be valid in reality. Model statements are i. A. Met about a future reality. However, since it is unclear how large the deviation of the model is, model risk arises.

In practice, the risk manifests itself through incorrect valuations (derivatives are sold too expensive / too cheap by banks) and hedge errors (systematic errors in hedging transactions ). Model risk occurs primarily in derivatives valuation and less so in equity valuation, as the model is more dependent on the model.

Reduction of the model risk

The following measures are proposed to reduce the model risk:

  • Control and verification of the models by a person who did not develop the models
  • Clear responsibilities for model testing
  • Clear documentation of the models and their application
  • Ongoing validation and review
  • Risk analysis without a model as described in the article Risk Analysis .

swell

  • Michel Crouhy; Dan Galai; Robert Mark: Model Risk , in: Journal of Financial Engineering , (1998); Vol. 7 (3/4), pp. 267-288, reprinted in Model Risk: Concepts, Calibration and Pricing, (ed. R. Gibson), Risk Book, 2000, pp. Xvii-xxxi.