In the realm of financial institutions, particularly banks, the accuracy and reliability of market risk models are typically governed by an independent risk department that either does the validation themselves or hires external validators from renowned consultancy firms. Banks are required to employ a rigorous process to assess the performance and validity of these models. This process shares several similarities with the model validation framework we discussed in broad strokes in the previous article, as it involves comprehensive assessments to ensure that market risk models are fit for purpose. The following are the key processes involved in the ‘gold standard’ of assessing a bank’s market risk models.
Data Validation: The Foundation of Market Risk Models
Before delving into the complexities of pricing models, banks begin with a solid foundation – the validation of market data. The integrity, completeness, and appropriateness of the market data used are thoroughly reviewed. Key aspects of data validation include:
Integrity of Market Data Sources: This step involves verifying the accuracy and appropriateness of data sources. For example, banks ensure that Bloomberg tickers and Reuters RICs are correctly specified and accurate.
Completeness of Data Inputs: Banks assess the completeness of various market data inputs, ranging from interest rates and foreign exchange rates to volatilities and spreads across different currencies and term structures.
Curve Construction and Adjustment: Proper setup and implementation of curve construction, including bootstrapping, interpolation, and extrapolation, are examined. Currency basis and LIBOR basis adjustment in discounting curves are reviewed for consistency.
Data Consistency: Consistency across all valuation platforms is crucial. This step ensures that market data remains consistent in various contexts.
Mapping and Specification: Banks confirm that market data files and objects are accurately specified and mapped.
Identifying and Enhancing Data: Any deficiencies in the data are identified, and suggestions for possible enhancements are made.
Once the mental gesticulation of which curves are sexiest and applies the most to the purpose of the models being assessed the validators would then move on to the actual pricing models used.
Pricing Model Validation: Digging Deeper
Market risk models may involve a complex web of pricing models put together to measure a specific outcome. To unravel the puzzle that a bank’s transaction ledger can be the bank’s model validators may opt to select a representative portfolio of sample transactions to assess each deal type within their derivatives and borrowings portfolio instead of the whole portfolio. Given that the sample was done via ‘risk-based sampling’. The process of pricing model validation includes:
Model Selection: The suitability of valuation methodologies and numerical methods is studied to ensure they align with the objective.
Model Assumptions and Parameters: The implementation of the model via assumptions and parameters, such as mean reversion and path characteristics, is inspected.
Model Calibration: The accuracy and appropriateness of calibration instruments and strategies are ascertained.
Model Implementation: The valuation setup is reviewed, including the formulas used for structured coupons, yields, forward rates, discount factors, and spreads.
Credit Spread Determination: The methodology for deriving credit spreads for valuing borrowings is reviewed. Suggestions for methodology improvements are offered.
Independent Valuation: Consistency with other pricing systems is evaluated, and the bank’s valuations are validated using independent and comparable pricing systems.
Identification and Recommendations: Valuation issues are identified, and relevant measures for improvement are recommended.
Now, as overly simplified as the pricing model validation process sounds, it is indeed a very useful tool in understanding the ‘black box’ that are the usual models built into Bloomberg terminals or other tools used by market risk modelers in banks. Since as far as models go, there really aren’t that many that the market uses, some usual suspects include Hull-White 1 Factor, Hull-White 2 Factor, oh and some even use 3 factors. The point being is that there are only a handful of ‘industry standard’ models used by bank’s that usually rely on the alternative model validation methods over the quantitative/qualitative methods, as each step above is usually remarked on by an ‘expert’ in the bank or by an out of sample testing done on the models.
Governance, Process, and Systems: The Organizational Framework
The last bit of model validation that no one really talks about is the assessment of a bank’s organizational governance framework. This portion of model validation plays a crucial role in ensuring that market risk models remain accurate, reliable, and is also downright boring. Here are its key aspects in case you want to look it up: Reporting Structure, Roles and Responsibilities, Model Usage and Business Embedding, Process Enhancement, and System Enhancements.