The risk of using a flawed model, yielding incorrect outputs. Model risk applies wherever a model, prototype, or other human construct is used. Models are widely used to help explain everything from life expectancy, to disease progression, to aerodynamics, to behavioral responses No model is perfect. Self-respecting scientists and practitioners understand this. Model risk doesn’t refer to the known deficiencies of a model, but rather to the potential that some unknown aspect of the model is imperfect to the extent that use of its outputs may lead to loss or damage.
Example:
An insurance company uses a model to calcuate mortality rates, and then designs life insurance products based on the model’s outputs.
Downside Scenario:
The model contains hidden flaws that lead to poorly-priced life insurance products. More specifically, the model underestimates mortality rates. Over time, as the true rates manifest, the insurance company incurs great losses.
Upside Scenario:
The model properly captures mortality rates, with even greater precision than the models used by competing companies. This greater precision allows the firm to provide its products at a lower cost, enabling it to win market share from competitors.