Model Risk, Valuation Risk and XVA

Introduction and context

Investment banks depend on quantitative models to price securities, calculate sensitivities, determine collateral and margin, measure counterparty exposure, allocate capital, manage liquidity and report financial results. Models translate assumptions about markets, counterparties, legal agreements and institutional behaviour into decisions with financial consequences. Their usefulness is accompanied by model risk: the possibility that a model is unsuitable, incorrectly implemented, misunderstood or applied outside the conditions for which it was designed.

Valuation risk is the broader risk that the recorded or quoted value of a position differs materially from an economically supportable exit value. It includes uncertainty in market inputs, model assumptions, liquidity, funding, credit, legal enforceability and valuation adjustments. XVA is a collective term for valuation adjustments that modify a risk-free or clean derivative value to reflect such effects. The most familiar adjustments are CVA, DVA and FVA, although the precise list varies among institutions and regulatory regimes.

The central governance challenge is therefore not simply to build a sophisticated model. It is to ensure that the complete model lifecycle—from conception to retirement—is controlled, independently challenged, transparently documented and continuously monitored.

Learning objectives

By the end of this chapter, the reader should be able to:

1. Explain model risk, valuation uncertainty and the relationship between clean valuation and XVA.
2. Design an effective model lifecycle covering development, validation, implementation, monitoring and inventory management.
3. Distinguish market-data uncertainty, model uncertainty, credit risk, funding effects and prudent valuation.
4. Explain and calculate simplified CVA, DVA and FVA measures, while identifying their assumptions and limitations.
5. Assess governance, accounting, regulatory, ethical and managerial implications of valuation models.
6. Interpret a practical model-risk case and recommend proportionate controls.

1. Model risk as an end-to-end governance problem

A model is a quantitative method, system or approach that applies assumptions, data and mathematical or statistical techniques to process inputs into quantitative estimates. In an investment bank, models include option-pricing engines, yield-curve construction tools, exposure simulations, stress-testing models, liquidity forecasts, capital models and machine-learning systems used in surveillance or credit assessment.

Model risk normally arises through five channels:

• Conceptual weakness: the model does not represent the relevant economic mechanism.
• Input weakness: market data are stale, illiquid, inconsistent, missing or incorrectly transformed.
• Parameter and calibration weakness: parameters are estimated from insufficient or non-representative observations.
• Implementation error: code, data interfaces, numerical methods or configuration do not match the approved methodology.
• Use and interpretation weakness: users apply the model beyond its scope, misunderstand outputs or fail to incorporate expert judgement appropriately.

A model can be mathematically correct yet inappropriate. For example, a Black–Scholes implementation may be coded accurately but unsuitable for an illiquid option with a discontinuous payoff, stochastic volatility and material counterparty exposure. Conversely, a simple model may be acceptable if its limitations are known, its use is restricted and compensating reserves are established.

The three lines of responsibility

A robust control environment usually separates responsibilities:

Independence does not mean that validators need no technical expertise. It means that approval, challenge and escalation should not be controlled solely by the model’s developers or profit beneficiaries.

2. The model lifecycle

2.1 Model conception and materiality assessment

The lifecycle begins before coding. The owner should define the decision supported by the model, population of trades, valuation purpose, users, frequency, materiality and consequences of error. Materiality is not limited to profit and loss. A model may be material because it affects regulatory capital, client pricing, liquidity, accounting disclosures, risk limits or senior management decisions.

A proportional assessment considers:

• potential financial loss from model error;
• complexity and novelty of methodology;
• degree of reliance on unobservable inputs;
• volume and liquidity of positions;
• dependence on other models;
• regulatory or accounting significance;
• reversibility of decisions based on the output; and
• quality and availability of benchmarking data.

The result should be a documented model-risk rating, such as low, medium or high, linked to validation frequency, approval authority, monitoring intensity and escalation thresholds.

2.2 Development and documentation

Development documentation should explain the economic rationale, mathematical formulation, assumptions, calibration, data lineage, numerical methods, limitations, intended use and prohibited use. It should identify whether the output is a price, fair-value estimate, exposure distribution, risk measure or decision score.

For a derivative, a clean value may be represented generally as the discounted expected value of future contractual cash flows under a pricing measure:

Here, V₀ is the value at time 0, E^Q is expectation under the chosen pricing measure Q, CF_i is the cash flow at time t_i, and D(0,t_i) is the discount factor from t_i to time 0. The expression assumes that the cash-flow definition, discounting framework, measure, market data and legal terms are consistently specified. It is not a guarantee that the value is observable or unique.

Documentation must distinguish assumptions from facts. “The counterparty will post daily variation margin” is a contractual or operational assumption; “the counterparty posted margin on every day in the sample” is an empirical observation. Both matter, but they have different evidential status.

2.3 Independent validation

Validation should be sufficiently independent, technically competent and proportionate to materiality. It commonly covers:

1. Conceptual soundness: whether the methodology captures relevant risk drivers.
2. Data and calibration: quality, representativeness, transformations and parameter stability.
3. Outcome analysis: comparison with prices, benchmarks, back-testing or realised outcomes.
4. Implementation: code review, numerical accuracy, version control and interface testing.
5. Use and limitations: whether users, controls and documentation match the approved scope.
6. Governance: approval, issue tracking, change control and monitoring.

Validation should not be reduced to reproducing the developer’s spreadsheet. Useful challenger methods include alternative closed-form models, finite-difference or Monte Carlo methods, independent data sources, sensitivity analysis and reverse stress testing.

A validation conclusion should state scope, evidence, unresolved issues, compensating controls, limitations and a clear approval recommendation. “Validated” is not synonymous with “true”; it means that the model has passed specified tests subject to recorded conditions.

2.4 Implementation controls

A model approved on paper can fail in production. Essential implementation controls include:

• controlled source code and peer review;
• segregation of development, testing and production environments;
• reproducible builds and version identifiers;
• automated data-quality checks;
• reconciliation between front-office, risk and finance systems;
• access control and privileged-user monitoring;
• independent replication of material calculations;
• exception handling and break management;
• change approval and rollback capability; and
• secure retention of code, inputs, outputs and approvals.

A particularly important control is the production-to-approval comparison. Each production release should demonstrate that the deployed methodology, parameters and configuration correspond to the approved version. Manual overrides should be logged, justified and independently reviewable.

2.5 Performance monitoring

Monitoring should assess whether the model remains fit for purpose. Relevant indicators include pricing differences against independent sources, realised-versus-forecast exposure, parameter drift, sensitivity instability, unexplained profit and loss, data-quality exceptions, limit breaches and frequency of manual adjustments.

A generic forecast-error measure is:

where y_t is the observed outcome at time t, ŷ_t is the model forecast and e_t is the error in the same units as the outcome. A mean absolute error over T observations is:

This measure is useful for monitoring, but it is not a complete validation test. It can conceal offsetting errors, regime dependence and economically important tail failures. In market risk, a model may have a modest average error but materially understate losses in stressed conditions.

Thresholds should trigger investigation rather than automatic acceptance or rejection. Escalation may lead to recalibration, restricted use, additional reserves, temporary conservative overlays, redevelopment or retirement.

2.6 Inventory and retirement

A model inventory is the authoritative register of models and material model components. It should record owner, purpose, risk tier, systems, products, dependencies, validation status, approval date, limitations, issues, compensating controls, regulatory use and retirement status.

Inventory completeness is itself a control objective. Unregistered spreadsheets, vendor models, pricing libraries and data transformations can create “shadow models.” Retirement should include removal from production, archival of evidence, communication to users, replacement mapping and confirmation that no downstream report still relies on the retired model.

3. Valuation risk and uncertainty

Valuation risk is multidimensional. A position’s value may be uncertain because the underlying market price is uncertain, because the position cannot be liquidated without a price concession, or because the institution’s model does not uniquely determine the value.

Sources of valuation uncertainty

A useful distinction is between uncertainty and error. Uncertainty reflects incomplete knowledge even when processes operate correctly. Error is a failure relative to the intended process, such as a wrong currency or missing coupon. Both can affect valuation, but error requires remediation while uncertainty may require a reserve or prudent adjustment.

Fair value and prudent valuation

Fair value is generally an exit-price concept: the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date. The concept does not imply that the bank intends to sell immediately or that the value is directly observable.

Prudent valuation adds caution where fair-value estimates contain uncertainty. Regulatory prudent valuation frameworks commonly address items such as market-price uncertainty, close-out costs, model risk, concentrated positions, administrative costs and unearned credit spreads, depending on the applicable regime. A prudent valuation adjustment is not a substitute for correcting a known booking or implementation error.

Reserves and valuation adjustments should have clear ownership, methodology, release rules and accounting or regulatory treatment. They should not be used as an opaque mechanism to smooth earnings or conceal weaknesses in the primary model.

4. Clean valuation and XVA

A clean or risk-free derivative value abstracts from some counterparty, own-credit and funding effects. XVA modifies that value to reflect economically relevant features of the transaction and the institution’s valuation policy. The adjustments are not always additive in a strict theoretical sense because credit, collateral, funding and close-out interact. Nevertheless, additive presentation is widely used for management communication:

The signs shown are a common convention for a bank’s asset-side valuation: CVA is a deduction for expected counterparty loss, DVA can increase the value of the bank’s liability to reflect its own credit, and FVA reflects funding effects. Sign conventions differ by position and institution; the policy must define them explicitly. KVA denotes a capital valuation adjustment and MVA a margin valuation adjustment.

4.1 Exposure concepts

Let V_t be the future mark-to-market value of a portfolio from the bank’s perspective. Positive exposure is:

where C_t is collateral held by the bank, measured consistently with the contractual collateral convention. If collateral exceeds exposure, positive exposure is zero in this simplified expression. Netting sets, margin period of risk, collateral disputes, thresholds and minimum transfer amounts can materially alter the calculation.

Expected positive exposure at time t is:

Potential future exposure is a high-quantile measure of the exposure distribution rather than an expected value. Confusing expected exposure with potential future exposure is a common analytical error.

4.2 Credit Valuation Adjustment (CVA)

A simplified unilateral CVA for a discretely monitored exposure can be written as:

Here, R_C is the counterparty recovery rate, DF_i is the discount factor to time 0, EE_ti is expected positive exposure at time t_i, and PD_i^marginal is the probability of counterparty default during the interval ending at t_i. The formula assumes unilateral counterparty credit risk, deterministic recovery, consistent netting and collateral treatment, and a simplified separation between exposure and default. More advanced calculations model wrong-way risk, stochastic recovery, collateral, discounting and dependence between exposure and default.

The factor 1 − R_C is loss given default. CVA is not simply exposure multiplied by the counterparty’s current credit spread. Credit spreads inform default probabilities, but the relationship requires a credit-curve and recovery framework.

4.3 Debit Valuation Adjustment (DVA)

DVA reflects the effect of the bank’s own default risk on the value of its liabilities. In a simplified form:

where R_B is the bank’s recovery rate, E_t^- = max(-V_t + C_t, 0) is expected negative exposure from the bank’s perspective, and PD_B,i^marginal is the bank’s marginal default probability. DVA is conceptually controversial because a deterioration in the bank’s own credit may create an accounting gain even while its economic condition worsens. It can also be difficult to monetise because the bank cannot generally realise the benefit by choosing to default.

4.4 Funding Valuation Adjustment (FVA)

FVA represents funding costs or benefits associated with financing uncollateralised or imperfectly collateralised exposures. A stylised representation is:

where F_t is the relevant funding requirement, s_F,i is the incremental funding spread for interval i, and Δt_i is the interval length in years. The sign depends on whether F_t represents a funding need or surplus and on the institution’s convention. FVA is not universal: its treatment depends on collateralisation, treasury transfer pricing, accounting policy, replication assumptions and whether funding costs are already embedded in discount curves.

Double counting is a major risk. For example, if a valuation already uses a funding-adjusted discount curve, a separate FVA may duplicate the same effect. XVA governance therefore requires a single documented valuation architecture and reconciliation among front office, treasury, risk and finance.

Worked example: simplified CVA

An uncollateralised one-year transaction is monitored at two future dates. Assume the following:

Assume counterparty recovery R_C = 40%, so loss given default is 60%. Applying the simplified formula:

The first discounted expected-loss component is $147,000, and the second is $152,000. Therefore:

The result is an illustrative valuation deduction, not a complete production CVA. It ignores wrong-way risk, exposure-default dependence, collateral disputes, cure periods, close-out conventions, stochastic rates and recovery uncertainty.

5. Integrated XVA governance

XVA models often depend on multiple linked models: yield curves, volatility surfaces, credit curves, trade cash flows, collateral simulation, exposure profiles, netting and default processes. Model risk therefore propagates through dependencies. A small error in a curve or collateral flag may affect thousands of trades.

An effective XVA framework should establish:

• a documented definition of clean value and each adjustment;
• legal agreement and netting-set data ownership;
• consistent market-data hierarchies and credit-curve construction;
• exposure simulation methodology and time-grid controls;
• treatment of collateral, initial margin and variation margin;
• explicit wrong-way-risk policy;
• independent price verification and reserve governance;
• reconciliation from trade-level values to desk, legal-entity and group totals;
• sensitivity, attribution and stress reporting; and
• controls against double counting among CVA, FVA, KVA and reserves.

XVA desks must also coordinate with treasury, collateral management, legal, finance, regulatory reporting and front-office pricing. A model that is correct in isolation may be wrong for the institution if its outputs conflict with contractual terms or internal transfer-pricing policy.

6. Illustrative case: an interest-rate swap portfolio

This is a fictional illustrative case.

Background

Northbridge Investment Bank holds a portfolio of long-dated interest-rate swaps with a corporate counterparty. The front office reports a positive clean value of $42 million. The portfolio is partly collateralised, but a recently amended credit-support annex has not been correctly linked to the exposure engine. The counterparty’s credit spread has widened, and the bank’s treasury department has introduced a new funding curve.

Problem

The daily XVA report shows a CVA of $1.8 million, while an independent review estimates $3.1 million. FVA is also lower in the production system than in treasury’s transfer-pricing analysis. A management report still uses the old values because the discrepancy has been treated as a data issue rather than a model-risk event.

Analysis

The review identifies four interacting weaknesses:

1. The collateral system classifies the amended agreement as uncollateralised for part of the portfolio.
2. The exposure engine uses a stale counterparty credit curve.
3. Production uses a different discounting convention from treasury.
4. The model inventory does not list the collateral mapping spreadsheet as a material model component.

The issue is not solved by selecting the larger CVA. The bank must determine which legal terms, data and valuation conventions are correct, quantify the period affected and assess whether reported financial or regulatory measures require correction.

Approach and solution

The bank establishes a controlled incident team involving front office, XVA, collateral operations, legal, treasury, finance and model risk. It freezes non-essential changes, preserves production evidence and performs a trade-level reconciliation. The collateral agreement is legally confirmed, the mapping is corrected, and the credit curve is independently sourced and approved. Treasury and XVA agree a documented funding convention that avoids double counting.

Pending full redevelopment, the bank applies a conservative interim reserve based on an independently reproduced exposure range. The model is then subjected to challenger testing, including stressed credit spreads, delayed collateral, partial default and wrong-way-risk scenarios. The collateral mapping tool is entered into the model inventory, assigned an owner and placed under change control.

Outcome

After correction, the CVA is $2.7 million and FVA is $0.9 million, compared with the original values of $1.8 million and $0.5 million. The clean value remains $42 million, but the all-in value is reduced by the additional adjustments under the bank’s stated sign convention. The incident produces a formal model-risk finding, revised monitoring thresholds and a requirement for daily reconciliation between legal agreements, collateral records and exposure engines.

Learning points

The case demonstrates that XVA is not merely a pricing calculation. It is an integrated governance process involving law, data, systems, market risk, credit risk, treasury and accounting. Inventory completeness, independent challenge and production controls are as important as mathematical sophistication.

7. Challenges, limitations and emerging practice

Procyclicality and feedback effects

Credit spreads, exposures and funding spreads can widen together during stress. A model calibrated in tranquil markets may understate joint deterioration. Applying stressed inputs without understanding their dependence can also overstate risk. Institutions should use scenario analysis and explain correlations rather than rely on mechanical overlays.

Wrong-way risk

Wrong-way risk occurs when exposure increases as the counterparty becomes more likely to default. A commodity producer may owe more when commodity prices fall, or a financial institution’s exposure may rise as its sector deteriorates. Assuming independence between exposure and default can materially understate CVA.

Numerical and computational limitations

Monte Carlo simulation involves sampling error, discretisation choices and computational constraints. A coarse time grid may miss margin calls or default-sensitive events. Variance-reduction methods can improve efficiency but require validation. Parallel computing and cloud infrastructure introduce additional controls concerning reproducibility, access, data residency and vendor dependence.

AI and machine learning

Machine learning can assist with anomaly detection, proxy modelling, calibration and data-quality surveillance. It should not bypass governance. Explainability, training-data representativeness, drift, leakage, adversarial manipulation and human accountability are particularly important in valuation and risk applications. A surrogate model used to accelerate XVA remains a model and requires scope definition, benchmarking and monitoring against the primary engine.

Ethical and regulatory considerations

Valuation models influence client prices, capital allocation, compensation and reported earnings. Deliberately selecting assumptions to produce a desired reserve or profit outcome violates the principle of faithful measurement and undermines market confidence. Governance should require traceability, conflicts-of-interest controls, escalation without retaliation and clear communication of uncertainty.

Regulatory expectations differ by jurisdiction, but common themes include sound model governance, independent validation, documentation, data quality, conservative treatment of uncertainty, stress testing, auditability and timely remediation. Institutions must apply the rules relevant to their legal entity, reporting basis and products rather than assume that one global methodology satisfies every requirement.

8. Best-practice control framework

A mature framework can be organised around the following principles:

1. Purpose before mathematics: define the decision, user and materiality before choosing a technique.
2. Ownership: assign accountable owners for models, data, legal terms and production systems.
3. Proportional independence: match validation depth and frequency to risk, with genuine challenge.
4. Traceability: retain data lineage, code versions, assumptions, approvals and output evidence.
5. Controlled implementation: reconcile approved methodology to production behaviour.
6. Continuous monitoring: use outcome analysis, sensitivity, stress and issue indicators.
7. Transparent uncertainty: distinguish observable inputs, assumptions, overlays and reserves.
8. No double counting: reconcile clean value, XVA, prudent valuation and accounting adjustments.
9. Inventory discipline: include spreadsheets, vendor components, data mappings and dependencies.
10. Decisive remediation: restrict, replace or retire models that no longer meet their purpose.

Chapter Summary

Model risk is the risk that a model or its use produces an unsuitable or materially inaccurate result. Effective governance covers the entire lifecycle: conception, materiality assessment, development, independent validation, controlled implementation, performance monitoring, inventory management and retirement. Validation should test conceptual soundness, data, calibration, outcomes, code and use limitations.

Valuation risk arises from uncertain market data, liquidity, model structure, parameters, legal terms and operations. Fair-value estimates therefore require independent price verification, sensitivity analysis, reserves and, where applicable, prudent valuation adjustments. Known errors must be corrected rather than hidden in reserves.

XVA modifies clean derivative values for counterparty credit, own credit, funding, capital, margin and related effects. Simplified CVA uses discounted expected positive exposure, loss given default and marginal default probabilities; DVA uses the bank’s own negative exposure and default probabilities; FVA reflects funding requirements and funding spreads. These measures depend critically on collateral, netting, legal enforceability, dependence between exposure and default and the institution’s valuation conventions.

The most important professional lesson is integrative: reliable valuation requires coordination among quantitative developers, validators, front office, treasury, collateral operations, legal, finance, technology and senior management. Sophisticated mathematics cannot compensate for weak data, unclear ownership or uncontrolled implementation.

Key Terms

Model risk: Risk of adverse consequences from incorrect models, implementation, data or use.

Model inventory: Authoritative register of models, components, owners, status and dependencies.

Model owner: Person or function accountable for a model’s purpose, operation and controls.

Independent validation: Technical and governance assessment performed independently of development or profit ownership.

Conceptual soundness: Whether a model’s economic and mathematical structure is appropriate.

Model limitation: A documented condition under which a model may be inaccurate or unsuitable.

Valuation risk: Risk that a recorded or quoted value differs from an economically supportable value.

Fair value: Exit-price estimate in an orderly transaction between market participants.

Prudent valuation: Conservative treatment of uncertainty in fair-value measurement for applicable purposes.

Reserve: Amount held to reflect valuation uncertainty or a known valuation-related adjustment.

XVA: Collective term for valuation adjustments applied to a clean derivative value.

CVA: Credit Valuation Adjustment for expected loss from counterparty default.

DVA: Debit Valuation Adjustment reflecting the bank’s own credit risk on liabilities.

FVA: Funding Valuation Adjustment for funding costs or benefits.

Expected positive exposure: Expected value of positive future exposure after relevant collateral treatment.

Wrong-way risk: Dependence in which exposure increases as default likelihood increases.

Netting set: Transactions covered by an enforceable netting agreement.

Collateral: Assets or cash posted to mitigate counterparty exposure.

Challenger model: Independent alternative used to test a primary model.

Model overlay: Controlled adjustment applied to a model output to address a documented limitation.

Review Questions

Short Answer Questions

1. What are the principal sources of model risk in an investment bank?
2. Why must model validation be independent of model development?
3. Distinguish valuation uncertainty from a known implementation error.
4. What is the role of a model inventory?
5. Define CVA and identify its principal components.

Long Answer Questions

1. Explain the complete model lifecycle and the controls required at each stage.
2. Discuss the sources of valuation uncertainty and the role of reserves and prudent valuation.
3. Explain the economic meaning, assumptions and limitations of CVA, DVA and FVA.
4. Analyse why collateral and legal netting agreements are central to XVA measurement.
5. Evaluate the governance challenges of using machine learning in valuation and risk management.

Analytical / Application Questions

1. A portfolio has expected positive exposures of $12 million and $9 million at two dates, discount factors of 0.97 and 0.94, marginal default probabilities of 1% and 2%, and recovery of 40%. Calculate simplified CVA and state at least three limitations of the calculation.
2. A production XVA engine reports a lower CVA than an independent challenger after a collateral agreement amendment. Design an investigation plan covering legal, data, model, implementation and reporting controls.
3. A bank proposes adding FVA to a valuation that already uses a funding-adjusted discount curve. Explain the double-counting risk and specify the governance evidence required before approval.

Multiple Choice Questions with Answers

Question 1

Which activity is primarily associated with independent model validation?

A. Setting the trading desk’s daily profit target

B. Testing conceptual soundness, implementation and limitations

C. Negotiating collateral terms with a client

D. Selecting the bank’s external auditor

Correct Answer: B. Testing conceptual soundness, implementation and limitations

Question 2

Which item should normally be recorded in a complete model inventory?

A. Only models that produce accounting entries

B. Only models developed by external vendors

C. Models, material components, owners, dependencies and validation status

D. Only models used for regulatory capital

Correct Answer: C. Models, material components, owners, dependencies and validation status

Question 3

In the simplified CVA formula, the factor 1 − R_C represents:

A. Probability of collateralisation

B. Loss given counterparty default

C. The bank’s funding spread

D. Potential future exposure

Correct Answer: B. Loss given counterparty default

Question 4

Expected positive exposure is best described as:

A. The maximum possible portfolio value

B. The expected value of positive future exposure after specified collateral treatment

C. The counterparty’s total notional amount

D. The bank’s own default probability

Correct Answer: B. The expected value of positive future exposure after specified collateral treatment

Question 5

Which situation illustrates wrong-way risk?

A. Exposure and default probability are unrelated

B. Exposure falls when the counterparty’s credit quality deteriorates

C. Exposure rises as the counterparty becomes more likely to default

D. Collateral is posted daily under a margin agreement

Correct Answer: C. Exposure rises as the counterparty becomes more likely to default

Question 6

Why can DVA be controversial?

A. It ignores the bank’s own credit risk by definition

B. A deterioration in the bank’s credit can create an accounting gain

C. It is always larger than CVA

D. It applies only to collateralised assets

Correct Answer: B. A deterioration in the bank’s credit can create an accounting gain

Question 7

Using the chapter’s simplified CVA approach, exposure is $5 million, discount factor is 0.96, marginal default probability is 2%, and recovery is 25%. What is the contribution to CVA?

A. $24,000

B. $48,000

C. $72,000

D. $96,000

Correct Answer: C. $72,000

Question 8

Which control most directly helps demonstrate that production matches the approved model?

A. A production-to-approval comparison using controlled versions and parameters

B. Increasing the number of traders using the model

C. Removing all manual exception logs

D. Replacing validation with user familiarity

Correct Answer: A. A production-to-approval comparison using controlled versions and parameters

Question 9

What is a major risk when adding FVA to a valuation that already uses funding-adjusted discounting?

A. Loss of all counterparty exposure information

B. Double counting of funding effects

C. Automatic elimination of model risk

D. Conversion of CVA into DVA

Correct Answer: B. Double counting of funding effects

Question 10

Which statement best describes a model overlay?

A. An undocumented adjustment used to smooth earnings

B. A controlled adjustment addressing a documented model limitation

C. A replacement for correcting a known booking error

D. A guarantee that the primary model is unbiased

Correct Answer: B. A controlled adjustment addressing a documented model limitation