OCC 2011-12 and Collateral Valuation: When Your System Becomes a Model

OCC 2011-12 and Collateral Valuation: When Your System Becomes a Model
Financial institutions using pledged-asset lending regularly ask: does our valuation system qualify as a model under OCC 2011-12? The answer matters. If yes, you need model inventory, development documentation, independent validation, and ongoing monitoring. If no, lighter controls may suffice.
The OCC definition is straightforward: a model is "a quantitative method, system, or approach that applies statistical, economic, financial, or mathematical theories, techniques, and assumptions to process input data into quantitative estimates."
Three elements: quantitative method, processes input data, produces quantitative estimates. Many collateral valuation approaches meet this definition.
Does Your Valuation Qualify?
A pure market price lookup probably isn't a model. You retrieve real-time prices from exchanges, return them to users. No method applied. But if the system aggregates multiple sources, detects outliers, applies volume weighting, or makes adjustments, you've added methodology and probably crossed into model territory.
Comparable transaction analysis is almost certainly a model. You're valuing an asset by reference to similar assets, applying adjustments for differences. The adjustment methodology involves assumptions and calculations. The selection criteria for "comparable" involves systematic judgment.
Discounted cash flow is definitely a model. DCF applies financial theory, statistical assumptions about growth and discount rates, and mathematical techniques to produce present value estimates.
NAV-based valuation depends on complexity. Simple lookup may not qualify. NAV with systematic adjustments for staleness, liquidity, or secondary market factors probably does.
Machine learning or algorithmic approaches are obviously models. They qualify. The variable is governance intensity.
The Core Requirements
For systems that qualify, OCC 2011-12 establishes expectations across several areas.
Model inventory means maintaining a comprehensive list of all models, including name, purpose, owner, development date, last validation date, and risk tier. For collateral valuation, each distinct methodology should be a separate entry. A liquid securities model, a private company model, and a cryptocurrency model are three inventory items, not one.
Development documentation explains how the model works. What's the theoretical basis? Why is that theory appropriate? What calculations are performed? What data inputs are required? What assumptions are embedded? What are the model's limitations? What testing was done before implementation? The documentation should be complete enough that someone else could understand and replicate the approach.
Independent validation means assessment by parties independent of development. Validation examines conceptual soundness, ongoing performance, and outcomes analysis. It should occur before initial use and periodically thereafter. Annual validation is common for high-risk models. The key word is independent: having the developer validate their own model doesn't satisfy the requirement.
Governance covers ownership, approval authority, change management, exception handling, and reporting. Each model should have a clear owner accountable for performance. Material changes require appropriate approval. Overrides should be documented and approved. Model risk should be reported to appropriate committees.
Documentation Deficiencies
Examination findings frequently cite documentation problems.
Incomplete development documents mean models in production without comprehensive methodology, assumption, and limitation documentation. The fix is to document before deployment, not after examiners ask.
Missing validation reports mean models that have never been independently validated or where reports can't be located. Schedule validation, retain reports, know where they are.
Undocumented changes mean modifications without written rationale or approval. Build change management into your process. Log everything.
Override records missing means valuations were adjusted without documentation of why. If someone overrides model output, capture the reason and the approval.
Proportionality
OCC 2011-12 emphasizes that model risk management should match model risk. High-risk models need rigorous oversight. Low-risk models can have lighter controls.
For collateral valuation, high risk means models affecting large exposures, involving illiquid or novel assets, or where errors could materially affect credit decisions. These need comprehensive documentation, annual validation, and active monitoring.
Medium risk involves moderate exposures with established methodologies and observable inputs. Standard documentation and periodic validation apply.
Lower risk covers simple valuations for small exposures using direct market prices. Documentation can be lighter. Validation less frequent.
The point is proportionality, not exemption. Even low-risk models need some governance. The calibration should be honest about actual risk, not optimized to minimize compliance effort.
Vendor Models
Institutions frequently assume vendor models are the vendor's problem. They're not. You remain responsible for understanding methodology, ensuring appropriate use, and monitoring performance.
When you use a vendor for collateral valuation, you still need to document how the model works, validate that it performs appropriately for your use case, and monitor ongoing performance. The vendor can provide inputs to this process, but the accountability sits with you.
Examiners will ask how you evaluated the vendor's methodology, how you confirmed it's appropriate for your purposes, and how you monitor performance. Having the vendor's documentation on file is necessary but not sufficient.
What Triggers Revalidation
Certain events should trigger model revalidation before the scheduled cycle: material methodology changes, significant performance deterioration, expansion to new asset classes or use cases, material market condition changes affecting assumptions, or regulatory feedback.
Don't wait for the annual cycle if something material has changed. Early revalidation demonstrates responsive risk management.
Examination Preparation
When examiners review your pledged-asset lending program, they'll request model inventory, development documentation, validation reports, governance evidence, performance data, and override analysis.
Preparation means having these materials current and retrievable. Documentation should be up to date. Validation should be within cycle. Governance evidence should demonstrate active oversight. Staff should be able to explain methodology and limitations.
The time to assemble examination materials is before examination notification, not after.
For current OCC guidance, consult Bulletin 2011-12 directly. This article provides general information and does not constitute legal or regulatory advice.
