Sensitivity Analysis
Definition
Sensitivity Analysis denotes a quantitative technique which (e.g. in a Model Validation context) can establish the robustness of a given Risk Model. In practise, if small variations to inputs lead to large variations in outcomes the sensitivity analysis suggests that the model / framework requires additional attention.
Context
We can broadly distinguish inputs as those that normally remain constant during model operation (Model Parameters) and these inputs that vary (market observables, client data etc.). Model sensitivity can be defined with respect to both classes. In the case of variable inputs the sensitivity indicates the range of variation that can be expected in the normal operation of the model. In the case of non-varying model parameters the sensitivity analysis can reveal the variation of outcomes hidden in possibly hard to fix parameters.
Issues and Challenges
- Sensitivity analysis will not identify vulnerabilities linked to a Risk Factor and phenomena that has been ignored in the formulation of the Model Assumptions