1. What Is A Financial Modelling Error?

A model is a simplified mathematical representation used to make a decision: pricing, valuation, VaR, credit scoring, ECL provisioning, stress testing, LTV.

A modelling error is any of these 5:

1. Input Error: Wrong data, stale data, manual copy-paste.
2. Logic / Formula Error: Dividing by sum instead of average, wrong discount rate, wrong correlation.
3. Implementation Error: Model is correct on paper, wrong in Excel / Python / Flexcube.
4. Assumption Error: Assuming house prices always go up, correlations stay low, liquidity always exists.
5. Misuse Error: Using a model outside its approved purpose - e.g., using a VaR model designed for small hedges for a $100bn proprietary position.

Under Basel II/III, Operational Risk is defined as risk of loss resulting from inadequate or failed internal processes, people, and systems, or from external events . Model Risk is a sub-set of Operational Risk.

Under US Fed SR 11-7 / OCC 2011-12: "Model risk can lead to financial loss, poor business and strategic decision making, or damage to a bank's reputation." Banks must have independent validation, effective challenge, documentation, ongoing monitoring.

In India, RBI has adopted the same principle in its Master Directions on Model Risk - Board approved framework, independent validation, no model can go live without sign-off.

2. The 4 Routes To A Banking Loss Claim

When a model fails, the bank does not just absorb it. It pursues 4 claims in parallel:

Most big cases fail on Route C and succeed on Route B.

3. Case Laws - The Core Cases You Must Know

CASE 1: JPMorgan "London Whale" - The Classic Model Error Case (2012)

Loss: $6.2 Billion

Trader Bruno Iksil in London built a huge Credit Default Swap position as a "hedge". The bank's risk was measured by a new VaR model.

The JP Morgan Task Force Report (Jan 2013) revealed:

The model "operated through a series of Excel spreadsheets, which had to be completed manually, by a process of copying and pasting data from one spreadsheet to another"

And the formula error:

"The spreadsheet divided by their sum instead of their average, as the modeler has intended... It also remains unclear when this error was introduced"

Effect: It muted volatility by a factor of two and lowered VaR by ∼50%. Management thought risk was $67m, actual risk was $129m+.

Legal Outcome:

• FCA (UK) fined JPMorgan £137,610,000 as a result of large losses incurred in a trading portfolio. The losses were stated to have amounted to US $6.2 billion by end of 2012 and were reported as the "London Whale" trades.
• US SEC + OCC + CFTC fined another ∼$1bn.
• Civil Claim: JPMorgan Chase agreed to pay $150 million to settle investor claims that it hid $6.2 billion in losses.

Principle: A model error does not excuse the bank. The bank is liable for failing to validate the model. The Model Review Group had flagged the Excel process but approved it anyway.

CASE 2: Citibank / Revlon - The $900 Million Model / System Error (2020)

In re Citibank August 11, 2020 Wire Transfers, 520 F. Supp. 3d 390 (S.D.N.Y. 2021)

Citibank was administrative agent for Revlon's syndicated loan. It intended to send $7.8 million interest to lenders. Instead, due to Flexcube software, it sent ∼$900 million - the full principal - from its own funds.

Court found:

Citibank accidentally paid off the facility in full, including the entire principal amount owing, a total of almost US$900 million

The payments were not caused by a technical error, but by human error: the failure to select the FRONT and FUND fields when inputting the default override instructions in Flexcube.

Judgment: Judge Jesse Furman held for the defendant lenders. Under New York law of mistake + discharge-for-value, if a lender receives money in good faith and in satisfaction of a real debt, it does NOT have to return it, even if bank made a mistake.

Citibank lost ∼$500m net. It then sued its vendor Wipro and recovered partially.

Principle for Modelling Claims: If your model/system pays wrong amount, you cannot automatically claw it back. You must prove recipient had notice of mistake. This led to the "Revlon Clause" now in every loan agreement.

CASE 3: Barclays Bank v Christie Owen & Davies [2016] EWHC 2351 (Ch) - The Valuation Model Error Case

This is the leading UK case for banks claiming for a bad valuation model.

Barclays lent against 3 gaming arcades in Great Yarmouth based on Christie's valuation of £2.1m and £1m. Barclays claimed Christie negligently overvalued.

Court held:

"Barclays now seeks to recover its losses... by proceedings for professional negligence against the Defendant. The claim is based on the provision to Barclays of two valuation reports"

Judge found Christie was negligent in not adopting the EBITDA approach, using Turnover Multiplier instead. Barclays was entitled to damages, but reduced by 40% for contributory negligence - because Barclays itself fell below standard of reasonably competent bank.

Principle: Even if model/valuer is wrong, bank's own failure to challenge the model (Effective Challenge under SR 11-7) reduces recovery. Contributory negligence applies to banks.

CASE 4: SAAMCO Principle - How Much Can Bank Recover?

South Australia Asset Management Corp v York Montague [1997] AC 191

Foundation case. If valuer gives negligent valuation, bank can only recover loss attributable to overvaluation, not loss from general market fall or borrower default.

Example: Property valued at £2m, true value £1m, bank lends £1.5m. Market falls 30%. Bank can recover only £1m (overvaluation gap), not full loan loss. This is called the SAAMCO Cap.

Barclays in the Christie case argued its claim "is less than the limit to its loss arising from the decision in SAAMCO, which would be £1.1m".

Every modelling loss claim in UK/Commonwealth is capped by SAAMCO.

CASE 5: Titan Europe 2006-3 plc v Colliers International [2014] - CMBS Valuation Model

Valuer negligently valued property that was collateral for securitised loan. High Court held valuer liable for €32m. Court said CMBS investors can claim even though they were not direct client, because valuer knew valuation would be relied on for securitisation model.

Principle: Duty of care extends to foreseeable users of the model, not just the bank who commissioned it.

CASE 6: Lloyds Bank v McBains Cooper [2018] - Surveyor Model Error

Bank claimed about £1.4 million, being total amount advanced less recoveries, after McBains gave advice that was "unquestionably negligent". Bank sustained loss by advancing money in reliance on information provided. Court allowed claim, subject to SAAMCO cap and contributory negligence.

CASE 7: J.P. Morgan Securities v Vigilant Insurance Co. (NY Court of Appeals, 2013) - Insurance Claim Fails

Bank tried to claim insurance for disgorgement paid to SEC for facilitating illegal trading. Court held:

The disgorgement payment to the SEC in settlement of charges that plaintiffs' predecessors wilfully facilitated illegal mutual fund trading practices does not constitute an insurable loss under the primary professional liability policy

Principle for Banking Loss Insurance: Insurers will argue:

1. Model error loss is not "fortuitous" - it was internal negligence
2. Disgorgement / regulatory fine is not "Loss" - it's restitution, uninsurable under public policy
3. Bank failed to disclose model risk

This is why most banks fail to recover model losses from insurers.

4. How To Prove A Financial Modelling Error Claim - 5 Elements

A bank (or its liquidator/insurer) must prove:

1. Duty: Modeller/valuer owed duty of care to bank. Test: Bolam - would a reasonably competent professional have done same? And Bratt v Jones - valuation must be outside permissible margin of error AND breach professional standards.

2. Breach: Model failed validation, no back-testing, no independent review, Excel without controls.

3. Reliance: Bank actually relied on model to lend / trade / price. If bank ignored model, no claim.

4. Causation: Loss would not have happened but for model error. Need to separate market loss vs model loss - SAAMCO.

5. Quantum: Loss = Difference between modelled value and true value, minus bank's own contributory negligence (Barclays 40% rule), minus recoveries.

Defences that usually succeed:

• Contributory negligence - bank did not challenge model
• No reliance - bank lent anyway
• Market downturn - SAAMCO cap
• Limitation - 6 years from date of valuation/model report, not date of loss

5. Indian Context

India has no separate "Model Risk Act" but claims use:

• Indian Contract Act, 1872 - Sec 73: Compensation for breach
• Tort - Professional Negligence: Against valuer/auditor under ICAI / IBBI rules
• RBI Master Direction on Fraud: Model manipulation = fraud, must be reported
• SEBI (PFUTP) Regulations: If model used to mislead investors
• Insurance: Bankers Indemnity Policy - but Indian insurers often exclude "error in model / computer system" unless specifically added

Recent Indian examples: IL&FS, DHFL - credit rating models gave AAA despite stress, banks claimed against rating agencies for negligent modelling. Cases still in NCLT / Bombay HC, but principle is same as Barclays v Christie.

Summary For Exam / Claim Note

A modelling error is not a market loss. It is an operational risk failure. The bank can claim if it can show (1) a duty, (2) a model outside professional tolerance, (3) reliance, (4) SAAMCO-limited causation. The bank will lose 30-40% for its own failure to validate, and will almost always lose its insurance claim for regulatory fines.