Harris KhanModel risk is a type of risk that occurs when a financial model is used to measure quantitative information such as a firm’s market risks or value transactions, and the model fails or performs inadequately and leads to adverse outcomes for the firm.
For example, suppose a bank has a model to estimate the loss on default (Loan Loss Allowance) in its loan portfolio. This model may fail in calculating the losses and lead to restructuring of the loan to generate higher losses. Or, if a firm uses a model to price interest sensitive securities like bonds, the model may show a higher price than the actual price on the market.
What is the risk associated with these kind of failures of models?
Let us understand this with an example. A bank has a model which historically has worked very well in measuring the losses in the loan portfolio. Now, consider that the bank lends all the way up to the maximum allowed limits in their past financial performance and at the same time, every borrower may have already breached the terms in the loan agreement on utilization. Now, if the borrowers could not give collateral for the loan or if the counter-parties in the derivatives fails to give collateral, the bank may find its lending capacity to be significantly eroded. Describing the scenario above, model risk is the risk that the model may fail when the business environment changes or when the features of the model change.
Model risk has two components:
- Model Risk : Modeling error which can lead to an inaccurate calculation of the expected loss
- Model Risk Factors : Different situations that can lead to failure of models
Risk factors or Model Risk Factors can be broadly categorised into:
- Model Robustness
- Model Condition
- Model Structure
- Model Consistency
- Model Robustness
The robustness of the model can be examined by introducing changes into the variables or parameters in the model to test how the model responds to different conditions. There are different types of model robustness like:
Regression Model Robustness
Introducing different inputs to the regression models like unexpected loss given default in the loan portfolio will help to examine the robustness of the model
- Forecasting Model Robustness
Instruments that have a similar valuation like the mortgages with similar loans may be evaluated by introducing ‘sudden, drastic, unanticipated macroeconomic changes’ to see if the model also reacts well as the changes were not anticipated.
2. Stress Test Robustness
Models of stress test models simulates the possibility of a set of extreme scenario that may occur in the future. By introducing the extreme scenario to the model, the model should not generate large losses
Model Condition
Model condition is the possibility that specific environment does not correspond to what actually happened in the reality and this is again divided into two subcategories:
- Macro Condition
The macro condition is the possibility of unexpected changes in the economy or political changes. For example, if there is a sudden global financial crisis that decreases the net interest margin from 2% to -1.5% (as in the latest financial crisis), the models may not be able to handle such a change
- Micro Condition
When the conditions are expected to change, the model should be able to handle them. For example, if the credit risk has gone up, the models should be able to take this into account by either increasing the haircuts or reduce the lending coverage ratios
Model Structure
Model structure is the possibility that the structures in the models are not suited to the end-user requirements or may not be design to handle the business requirements. For example, if a model needs to be used for valuing a floating rate, the model should be designed in such a way that the exposures and sensitivities will not rise and fall simultaneously. Or, in the case of a loan portfolio, the models may be designed in such a way that the sensitivity of loan portfolio with respect to the index is not restricted to a certain level of loss
Model Consistency
Model consistency is the possibility that the models are capable of remaining consistent throughout the life of a transaction. How does one examine the model consistency? Normally, model consistency is examined by observing the model output after making changes in the assumptions. For instance, if there is a break in the dependence structures in the models, then this is a possibility that the model will not be consistent. This will show up as unpredictable errors in the output
Model risk can be broadly classified into three types:
- Model Risk for Interest Rate
The model risk for interest rate arises if the interest rates change. The same changes may lead to model sensitivity changes while calculating the risk of interest rate sensitivity.
- Model Risk for Valuation
The model risk for valuation arises is if the valuation parameters change or the valuation method changes.
- Model Risk for Liquidity Risk
The model risk for liquidity risk arises if the liquidity resource varies.
The above classification is given on the premise that the defects in the models may be caused by one or more of the Model Risk Factors. The issue which is important in the analysis of model risk is answering the question whether the models are in-sync with the business environment or not.
The final identification of where the possible defects in the model are lies on the Final Risk Assessment to see if the model is working as per the objective that it was designed to handle or not.
With all the information provided above, it is fair to say that risk is a major concern especially for a financial institution and model risk is one of the most serious problems in banking. With the plethora of models that are implemented in the day-to-day activities in banking, it is important for every bank to have a model risk management mechanism in place. A mis-estimation of losses can cause huge losses to the bank. It has been seen a number of times in the past when the models were not able to handle the changing market conditions and led to the huge losses in the banks and investment companies. So when it comes to the issues of risk, model risk should be given the highest priority.
