Treasury Pricing, CVA, PFE and VaR Model Validation
Regional bank
The departure of a key quantitative resource left the bank exposed on continuity and regulatory readiness for its Bloomberg-sourced treasury pricing, counterparty risk and VaR models. The objective was to formalise documentation, benchmarking evidence and governance for the existing framework, with no model development or system replacement required.
Scope of work
Model discovery confirming the inventory along with each model's purpose, regulatory classification and Bloomberg dependencies.
Scope definition across three model families:
- Treasury pricing: interest rate and profit rate swaps, Wa'ad-based derivatives, options and volatility models, and OIS, multi-curve and cross-currency curve construction
- Counterparty risk: exposure simulation across EE, EPE and ENE, netting set aggregation, CSA and collateral modelling, unilateral and bilateral CVA, PFE percentile and time-bucket aggregation, and initial and variation margin with margin period of risk
- Market risk: historical simulation and parametric and Monte Carlo VaR engines, risk factor mapping, back-testing through Kupiec and Christoffersen tests, and stress and scenario VaR
Data lineage mapping, documenting Bloomberg data lineage, market data conventions, credit and exposure inputs and risk factor mapping in full.
Independent replication built in Python using QuantLib or equivalent to evidence conceptual soundness, with outputs reconciled against reference analytics under defined tolerance and materiality thresholds.
Benchmarking and sensitivity testing run across the model set, with variance analysis against the reference outputs.
Governance and handover, producing conceptual documentation, assumptions and methodology write-ups, and embedding model lifecycle controls.
Outcome and value
- Regulator-ready submission packages across treasury pricing, CVA and PFE, and VaR.
- Auditable documentation of methodologies, assumptions, data lineage and limitations.
- Independent benchmarking evidence covering prices and Greeks, exposure measures and VaR metrics.
- Key-person dependency reduced, with model governance strengthened.