A model expectation is a condition implied by a model specification that has to be satisfied.
As such it is a translation of the model specifications in terms of model/data features which basically measure the strength of model/data traits. When we encounter an instance where the expectation is not valid, we have found a model issue.
As an example, a valuation model has as one particular specification, the fact that the NPV has to move smoothly with varying model inputs. A model expectation will express this condition quantitatively, such as e.g. requiring the norm of the curvature of the NPV expressed as a function of model inputs is not larger than a given threshold.
Scale your model risk management
Learn how technology can transform how you and your team manage model risk.
