Browse By

Internal Capital Adequacy Assessment Process (ICAAP): Model Risks

Using internal models as part of the Internal Capital Adequacy and Assessment Process (ICAAP) exposes the bank to model risk. This post is based on the paper “Model Risk” by Emanuel Derman (Quantitative Strategies Research Notes – April 1996).

Model Risks

  • The model selected may not be applicable to the process. As mentioned earlier it is important to fully understand how variables work and interconnect before assigning a mathematical process to explain that phenomena.
  • The model chosen may be incorrect. This may be because all of the important variables have not been considered in the formulation of the model; or variables may be assigned a deterministic process where as they may in fact follow a more stochastic process; or the processes defining the variables may be incorrect; or the assumptions used to relate one variable to another may be wrong; or the current business and market cycle may invalidate the use of the model which was based on historical data that is significantly different; or the model is based on an ideal situation rather than on reality; or that the model is based on a long run average and therefore ignores short run factors; or the parameters of the model have been wrongly estimated; etc.
  • The model may be correct, but the solution from the model may be incorrect due to carelessness or misunderstandings.
  • The model may be correct but it may be put to use in situations where it was never intended in the first place, i.e. the model is applied beyond its intended scope.
  • The model may be correct but the degree of accuracy for obtaining a numerical solution may be limited because of the methodology employed.
  • Implementation of the model into the system could lead to errors due to programming, code optimization processes, revisions by new developers who are not the original writers of the model, etc.
  • Model parameters may be based on historical data which may not be representative of the future and which may vary considerably depending on the sampling window chosen.

We have looked at some model risks that the bank could be exposed to by using models to assess, quantify and stress test its material risks. In the next section we will consider ways in which model risks may be avoided or reduced.

Comodo SSL