LIBOR made benchmark submissions and trader influence a market-integrity accountability test

Summary

• The Barclays-specific record established attempted manipulation, false reporting, trader efforts to influence submitters, reputation-related submissions and serious systems-and-controls failures within the respondents, benchmarks and periods stated by the relevant authorities.
• Regulatory findings, a corporate non-prosecution agreement, individual charges, guilty pleas and jury verdicts have different legal meanings; none proves that every fixing was distorted or every linked contract suffered a measurable loss.
• Reform and transition reduced reliance on panel-bank judgment, but successor rates are not automatically manipulation-proof or economically identical to LIBOR; integrity still depends on observable inputs, governance, surveillance and precise legacy-contract controls.

The Barclays LIBOR record is an accountability case about control over a number that appeared public, technical and impersonal but was assembled from private judgments. For years, panel banks answered a daily question about the rates at which they believed they could borrow. The administrator discarded outliers and averaged the remainder. The resulting fixings were embedded in derivatives, corporate and consumer loans, municipal financing, investment products and internal valuation systems. A submission moved by only a fraction could therefore matter far beyond the desk that supplied it.

Yet the information, incentives and communications behind that submission were largely inside the contributing bank.

That design created several distinct accountability questions. Who owned the methodology? Who supervised the submitters? Who prevented derivatives traders from shaping inputs that affected their positions? Who challenged implausible submissions during stress? Who aggregated communications and escalated recurring requests? Who could identify harm to a particular contract rather than merely infer it from misconduct affecting a benchmark process? And once the original design was judged unreliable, who had authority to reform, replace and ultimately cease it without destabilising contracts that had assumed the benchmark would continue?

The enforcement record gives concrete answers for specified respondents, currencies, products and periods. It does not support the broader proposition that every LIBOR fixing was false, that every trader at a panel bank participated, or that every contract referencing the benchmark suffered a measurable loss. Those distinctions are not qualifications at the margins. They are the difference between institutional accountability and a story that converts documented conduct into unsupported universal claims.

The US Commodity Futures Trading Commission's June 2012 announcement stated that Barclays would pay a $200 million civil monetary penalty to settle charges of attempted manipulation and false reporting concerning LIBOR and Euribor. It described derivatives traders asking submitters to adjust submissions to benefit positions, and lower submissions made to avoid negative perceptions about Barclays during the financial crisis. Those were findings in an administrative settlement covering the conduct, benchmarks and periods identified by the order; they were not a damages calculation for every instrument linked to a published rate. https://www.cftc.gov/PressRoom/PressReleases/6289-12

The UK regulator imposed a £59.5 million penalty and described significant failings relating to LIBOR and Euribor. The announcement is especially important because it connects communications and incentives to systems and controls: Barclays was responsible not only for isolated messages but for the arrangements that were supposed to manage conflicts and ensure submissions were properly made. The settled outcome establishes the regulatory basis stated in the notice. It should not be read as an adjudicated finding against every employee, customer or counterparty. https://www.fca.org.uk/news/press-releases/barclays-fined-%C2%A3595-million-significant-failings-relation-libor-and-euribor

A benchmark built from judgment

LIBOR was not a single market price observed on an exchange. It was a family of rates by currency and tenor, produced from submissions and a calculation rule. That matters because “the benchmark” can refer to at least four things: the definition of the question, the bank's internal process for answering it, the administrator's collection and calculation, and the published rate used in contracts. Responsibility was distributed across all four, while the strongest position-specific incentives sat inside trading businesses.

A judgment-based contribution need not be dishonest simply because it is not tied to a completed transaction. In a market with sparse borrowing at a particular tenor, an expert estimate can be informative. But discretion creates a verification burden. A robust system needs an explicit hierarchy of transaction data, executable quotes and other evidence; records explaining departures from that hierarchy; separation between submitters and traders who benefit from a fixing; surveillance of communications and patterns; independent review; and escalation when the answer reflects reputation management rather than the benchmark definition.

The detailed FSA final notice supplies the central Barclays chronology. It records requests by derivatives traders, communications involving submitters, concerns associated with external perceptions of the bank during the crisis, and breaches of regulatory Principles. The notice must be used with its own legal character intact: it is a final notice resolving regulatory action against Barclays, not a criminal verdict on each person mentioned and not proof of loss on a particular loan or swap. https://www.fca.org.uk/publication/final-notices/barclays-jun12.pdf

That chronology exposes the weakness of treating a submission as a narrow operations task. A submitter could receive information from money-market activity that was relevant to borrowing conditions. The same person could also encounter traders whose profit and loss moved with the fix. An institution that did not define which contact was legitimate, preserve it, review it and challenge its influence effectively left the boundary to informal practice. The problem was therefore both conduct and architecture: a person could make an improper request, and the institution could fail to make such requests visible and costly.

The administrator's averaging process reduced the effect of any one input but did not eliminate incentives. A trimmed mean is a resilience device, not an integrity guarantee. A contribution can be moved without becoming an outlier, several contributions can lean in the same direction, and a request can be attempted even if it does not ultimately change the published number. This is why attempted manipulation, false reporting and a proven transaction-specific economic effect are separate propositions. Enforcement can establish the first two without quantifying the third for every contract.

Two motives that must not be collapsed

The Barclays record includes at least two analytically different motives. One involved derivatives positions: traders wanted a benchmark outcome that could help their books. The other involved perceived creditworthiness during market stress: higher borrowing submissions could signal that a panel bank was under pressure, creating an incentive to submit lower rates. Both could pull a contribution away from the benchmark question, but they generated different evidence, control failures and chains of responsibility.

Position-driven requests are conflict-management problems. The relevant evidence includes the trader's exposure, the direction and timing of the request, the submitter's response, the bank's submission, the other panel contributions and the final calculation. A message asking for a higher or lower submission may be strong evidence of intent or attempt, yet it is not by itself proof that the fixing changed or that a named counterparty lost money. A full causal claim needs the rest of the chain.

Reputation-driven submissions are governance problems of another order. In a stressed market, the bank may fear that an honest estimate will be interpreted as a public distress signal. If management directs or encourages a contribution based on how it will look, the benchmark input ceases to answer the defined borrowing question. Accountability then reaches beyond a trader-submitters barrier. It asks what senior managers understood, what instructions were transmitted, whether compliance and risk functions recognised the change, and whether the board received information about a practice affecting a systemically important benchmark.

The US Department of Justice's 2012 resolution with Barclays used a non-prosecution agreement and a $160 million penalty. The department described Barclays's admission that misconduct occurred in connection with submissions, including requests relating to derivatives positions and submissions influenced by reputational concerns. A corporate admission in that resolution is meaningful evidence against the entity within the agreement's scope. It is not interchangeable with a guilty plea by the corporate entity, nor does it adjudicate every individual's state of mind. https://www.justice.gov/opa/pr/barclays-bank-plc-admits-misconduct-related-submissions-london-interbank-offered-rate-and

This distinction also explains why “who ordered what” should not be reduced to a single dramatic message. Institutions communicate through meetings, reporting lines, interpreted instructions and shared assumptions. A responsible account identifies the evidence actually cited by an authority and the conclusion that authority drew. It does not upgrade ambiguity to an express command, or infer board knowledge merely because seniority appears somewhere in a chain.

Institutional controls were the decisive layer

The strongest institutional lesson is that conflicts cannot be managed only by telling employees to behave properly. A bank contributing to a benchmark while trading instruments linked to it has an embedded conflict. Its control framework must assume that requests will occur and create evidence capable of detecting them. Necessary elements include formally appointed submitters, documented inputs, restricted communication channels, pre- or post-submission review, electronic surveillance, analysis of rate patterns against market evidence, periodic compliance testing, disciplinary consequences and reporting to accountable executives.

Surveillance must combine content and context. Keyword searches may find an explicit request but miss coded or casual language. Statistical alerts may identify a submission that differs from peers but cannot establish why. Trade data can show the desk's incentive but not whether it influenced the submitter. The useful unit is a joined record: communication, exposure, submission, market evidence, comparator panel data, review decision and escalation. Fragmented systems make each fact appear explainable in isolation.

LIBOR enforcement extended across institutions and cases. That wider record is useful for understanding a market-wide control problem, but it creates a recurring attribution risk. Findings against another bank or trader cannot be imported into the Barclays chronology. Each order has its own respondent, products, time period, admissions or findings and procedural posture.

Board oversight belongs in the chain because a panel bank was supplying infrastructure used across markets. The board did not need to approve daily numbers, but it needed reliable information about the nature of the conflict, control exceptions, regulatory contacts, repeated requests and anomalies. An accountability map should name the management owner of submissions, the independent function testing the process, the committee receiving escalations and the board body monitoring remediation. “The bank” is otherwise too imprecise to support prevention.

Compensation design matters as well. A derivatives trader rewarded for short-term desk revenue has a direct interest in the fixing. A submitter may face cultural pressure to serve the desk or protect the firm's reputation even without a payment tied to the rate. Control design therefore must look beyond formal reporting lines. Performance evaluation, promotion, status and informal access can defeat an organisational chart that appears independent on paper.

Subsequent UK debate about banking standards examined culture and accountability against the background of LIBOR and other failures. Such policy analysis is not a Barclays enforcement decision and should not be used to add factual findings to the 2012 notices. Its institutional value lies in explaining why governance reform moved toward clearer individual responsibilities, stronger challenge and consequences rather than reliance on diffuse collective ownership.

Enforcement was coordinated, but legal outcomes were not identical

The June 2012 resolutions are often compressed into one global “fine.” That loses information. The CFTC applied US commodities law through an administrative order. The UK regulator acted under its statutory regime and Principles. The Department of Justice used a non-prosecution agreement. Monetary amounts went to different authorities under different legal instruments. The facts overlapped, but the legal standards, respondents, procedural rights and consequences were not identical.

This matters for both severity and fairness. A civil or administrative settlement can contain detailed findings and remediation obligations without being a criminal conviction. A non-prosecution agreement can contain corporate admissions without a guilty verdict after trial. An individual prosecution requires proof against the charged person under criminal law; its outcome cannot be predicted from the corporate settlement alone. Conversely, a later acquittal or reversal for an individual does not erase an institution's separate regulatory resolution.

The UK's Serious Fraud Office announced criminal charges in 2014 against three former Barclays employees concerning US-dollar LIBOR. The announcement shows the start and scope of that prosecution, not guilt. Charging decisions are allegations until resolved, and any account of the individuals must follow the later plea, trial, appeal or acquittal record person by person. https://www.sfo.gov.uk/2014/02/17/criminal-proceedings-commenced-three-former-barclays-employees-libor-offences/

US prosecutions similarly demonstrate why role and outcome must remain specific. The Justice Department reported that former Barclays trader Peter Johnson pleaded guilty to conspiring to manipulate LIBOR. A plea is an admission by that defendant to the offence specified; it does not establish guilt for people who did not plead and does not expand the temporal or product scope beyond the case. https://www.justice.gov/opa/pr/former-barclays-trader-pleads-guilty-conspiracy-manipulate-libor

In a different proceeding, the Justice Department announced a jury conviction of two former Barclays traders in 2016. A verdict is a trial outcome against those defendants on the charged counts, subject to post-trial and appellate processes. It is more than an allegation, but it still cannot be converted into a universal finding about all Barclays submissions, all currencies or all colleagues. https://www.justice.gov/opa/pr/two-former-barclays-traders-convicted-libor-manipulation

An accurate institutional account therefore needs a status table even when the prose is concise. The categories are: regulatory finding or settled notice; corporate admission in an agreement; civil settlement without admission where applicable; criminal charge; guilty plea; trial verdict; appellate disposition; and discontinued or acquitted count. Each category answers a different question. Blurring them can exaggerate culpability in one direction and understate formally established conduct in the other.

Coordination among authorities also creates a responsibility question. Parallel cases can share evidence and avoid inconsistent remediation, but every authority must remain clear about its own basis. Institutions need protection against duplicative punishment and unclear credit for cooperation, while the public needs a consolidated explanation of total sanctions and continuing obligations. Transparency is best served by cross-referencing related resolutions and stating whether a payment is additional to or credited against another penalty.

Reform changed the governance of benchmark production

The Wheatley Review was the pivotal UK diagnosis and reform design. It concluded that LIBOR should be comprehensively reformed rather than immediately abolished, recommending statutory regulation, a new administrator selected through tender, a code of conduct for submitting banks, stronger governance and scrutiny, and changes to currencies and tenors. The report is a policy review, not an enforcement judgment. Its recommendations must be distinguished from rules and institutional changes that came later. https://www.gov.uk/government/publications/the-wheatley-review

The reform logic addressed several failures at once. Statutory regulation made benchmark administration and submission subject to direct oversight. A code could define evidence and recordkeeping. Governance could separate commercial interests from calculation. Reduced panels and tenors could focus the benchmark on markets with stronger support. Enforcement powers could attach consequences to dishonest submissions. Yet reform did not manufacture transactions in illiquid unsecured term funding markets. The underlying representativeness problem remained.

The Financial Services Act 2012 created the UK legislative route for regulating benchmark-related activities and introduced criminal provisions concerning misleading statements in relation to benchmarks. Legislation establishes powers and offences prospectively according to its commencement and scope; it does not retroactively convert prior regulatory findings into criminal convictions. https://www.legislation.gov.uk/ukpga/2012/21/contents

International principles then sought to make benchmark governance more consistent. IOSCO's Principles for Financial Benchmarks cover governance, benchmark quality, methodology and accountability, including conflicts, control frameworks, transparency, audit trails and complaints. Principles are standards for design and oversight, not proof that a particular administrator or contributor complied in every period. https://www.iosco.org/library/pubdocs/pdf/IOSCOPD415.pdf

The European Union's Benchmarks Regulation added a binding regional framework for administrators, contributors and benchmark use, including requirements tied to governance, methodology and critical benchmarks. The consolidated legal text is essential for determining obligations at a given date, but it should not be projected backward onto the Barclays conduct resolved in 2012. Nor does legal authorisation alone prove the economic representativeness of every individual fixing. https://eur-lex.europa.eu/eli/reg/2016/1011/oj

These reforms redistributed control. The administrator became a regulated entity with formal methodology and oversight duties. Panel banks received clearer contribution obligations. Supervisors gained direct tools. Users were expected to understand fallback provisions and benchmark risk. But the chain remained interdependent: an administrator could challenge a contribution but relied on input evidence; a supervisor could mandate remediation but not create an active funding market; contract parties could adopt fallbacks but had divergent economic interests.

The Financial Stability Board's 2014 report framed the durable solution as a multiple-rate approach: strengthen existing interbank offered rates where possible while developing nearly risk-free rates. It identified the need to anchor benchmarks in actual transactions and to use rates suited to different purposes. This was a policy roadmap, not a guarantee that transition would be frictionless or that one replacement would replicate every feature of LIBOR. https://www.fsb.org/2014/07/r_140722/

Why replacement became necessary

After the crisis, unsecured wholesale term borrowing by banks declined. That made it harder to ground multiple currencies and tenors in a deep set of transactions. A benchmark can have excellent controls and still be fragile if expert judgment supplies much of the answer. The accountability problem shifted from “are submissions being influenced?” to “is there enough underlying market activity to sustain the rate, and are users prepared for cessation?”

The UK Financial Conduct Authority's 2017 speech announcing that it would not compel or persuade panel banks to submit to LIBOR after the end of 2021 changed expectations. It gave markets a horizon for transition without itself terminating every setting on that day. The speech was a supervisory signal and policy position; the precise cessation or non-representativeness status of each currency-tenor setting depended on later announcements. https://www.fca.org.uk/news/speeches/the-future-of-libor

Replacement rates were deliberately different. Sterling markets adopted SONIA, an overnight rate administered by the Bank of England and based on eligible sterling overnight unsecured transactions. An overnight, transaction-based rate avoids the same panel-bank term-credit judgment, but it is not economically identical to term LIBOR. Compounding conventions, payment timing, credit spread and operational calendars all affect conversion. https://www.bankofengland.co.uk/markets/sonia-benchmark

US-dollar markets adopted SOFR, produced by the Federal Reserve Bank of New York from transactions in the Treasury repurchase market. SOFR is secured and overnight, whereas dollar LIBOR incorporated bank credit and term elements. Its large transaction base changes the manipulation surface, but no benchmark is “manipulation-proof” merely because it is transaction-based. Data quality, market concentration, methodology, operational resilience and administrator governance still require monitoring. https://www.newyorkfed.org/markets/reference-rates/sofr

The Alternative Reference Rates Committee developed recommended fallback language and transition conventions for cash products and derivatives in the United States. Its work helped coordinate a market in which individual contracts otherwise might choose incompatible solutions. Committee recommendations are not legislation and do not rewrite a contract by themselves; their effect depends on adoption, governing law and any applicable statutory solution.

ISDA's IBOR Fallbacks Supplement and Protocol addressed a vast derivatives population through standardized contractual amendments, using adjusted risk-free rates and spread adjustments following defined trigger events. Adherence could create consistency among participating counterparties. It did not automatically amend every non-adhering contract or every cash product, and the adjusted fallback was designed for continuity rather than perfect replication of the original bargain in every market state. https://www.isda.org/2020/10/23/isda-launches-ibor-fallbacks-supplement-and-protocol/

Cessation did not end accountability

The FCA's March 2021 announcement set out future cessation and loss-of-representativeness dates for 35 LIBOR settings. That announcement is the correct type of evidence for setting-specific timing. It should replace shorthand claims that “LIBOR ended” on a single universal date. Some settings ceased earlier; some continued temporarily under changed conditions; synthetic settings were permitted for limited legacy use. https://www.fca.org.uk/news/press-releases/announcements-end-libor

Synthetic LIBOR illustrates why legal continuity and economic representativeness must be separated. A regulator-directed synthetic methodology could provide a bridge for certain tough legacy contracts that could not feasibly be amended. It was not the continuation of panel-bank LIBOR in its prior form, and permission to use it was restricted. A synthetic rate reduced cliff-edge disruption; it did not establish that affected parties received precisely the same economics they expected when contracting.

Legacy contracts generated allocation questions. Who pays when the replacement has a different credit profile? Which spread adjustment is fair? Does a fallback activate on cessation, non-representativeness or another event? Can an agent select a rate? Does statutory override impair contractual rights? The answer can vary by instrument, governing law and drafting.

An institution's transition programme therefore needed a contract inventory, hierarchy of fallback quality, client communications, conduct review, model validation, accounting and tax analysis, and controls against using transition to improve the bank's position at a customer's expense.

The UK critical-benchmarks framework gave the FCA powers concerning designation, methodology changes and legacy use. The FCA's synthetic LIBOR information explains the limited bridge and setting-specific decisions. Regulatory permission is not a blanket safe harbour from private claims, nor is the eventual end of a synthetic setting evidence that every affected contract transitioned without dispute. https://www.fca.org.uk/markets/libor/synthetic-libor

Accountability after cessation also includes operational evidence. Firms should be able to show which contracts were identified, what fallback applied, how spreads and compounding were calculated, when customers were notified, which exceptions were escalated and how disputes were resolved. A dashboard stating that 99 percent of exposure transitioned can conceal the riskiest one percent. Notional value, customer vulnerability, legal uncertainty and economic sensitivity all matter.

Redress requires transaction-specific proof

The scale of LIBOR-linked markets makes aggregate rhetoric tempting. But the path from a problematic submission to recoverable loss is contract-specific. A claimant may need to establish the applicable legal duty or representation, actionable conduct, reliance or causation where required, the fixing used by the contract, the counterfactual benchmark, the direction of payment, limitation rules and a defensible quantum. Different causes of action impose different elements.

Even a demonstrated movement in a fixing does not imply that all users lost. A lower rate might benefit a floating-rate borrower and disadvantage a lender; on a derivative, the effect depends on payment direction, reset date, notional and offsets. A portfolio can contain positions on both sides. Some misconduct involved attempts that may not have moved the published rate. Some periods involved reputation-motivated lowering rather than position-specific requests. These facts resist a single loss multiplier.

Regulatory penalties serve public enforcement objectives and generally are not a claimant-by-claimant compensation fund unless the resolution expressly creates redress. Private settlements may resolve claims without admissions and often contain releases. Criminal restitution, where ordered, has its own statutory basis. Analysts should identify the instrument before describing a payment as “compensation.” Fine, disgorgement, settlement amount, restitution and damages are not synonyms.

Civil litigation can also produce rulings on standing, limitation, antitrust injury, contract and causation that differ by jurisdiction and procedural stage. A complaint contains allegations. A motion-to-dismiss ruling tests legal sufficiency under an assumed or specified standard. Class certification addresses commonality and procedure, not final liability. A settlement resolves claims on agreed terms without necessarily deciding contested facts. Responsible reporting labels each stage.

Authorities monitored LIBOR transition as a financial-stability issue, emphasizing reduction of legacy exposures and operational readiness. That macroprudential lens is useful for system risk, not for proving individual customer loss. Aggregate exposure measures the scale of dependency; it does not determine liability or damages in a particular instrument.

An accountability map for benchmark integrity

The first owner is the benchmark administrator. It controls the definition, methodology, contributor criteria, calculation, publication, correction policy, oversight committee and challenge process. It should publish enough information for users to understand what the rate measures and retain sufficient evidence for audit. It must manage its own commercial conflicts and monitor whether the underlying market still supports the benchmark.

The second owner is the panel bank. Its treasury or money-market function may possess the relevant funding information, but the institution must govern who can submit and on what evidence. Trading desks need clear prohibitions on requests intended to benefit positions. Compliance needs access to communications, exposures and submission records. Internal audit must test whether controls work in practice. Senior management must receive anomalies and certify remediation based on evidence, not reassurance.

The third owner is the supervisor. It authorises or oversees administrators and contributors, investigates misconduct, coordinates across borders and decides when a critical benchmark needs intervention. It also must communicate dates and legal effects precisely. Vague transition signals can create the very disorder that oversight seeks to prevent.

The fourth owner is the contract ecosystem: banks, asset managers, corporates, public bodies, clearing houses, trade associations, calculation agents, lawyers and technology providers. Each must inventory exposure and implement fallbacks. Standardisation can lower collective-action costs, but users retain responsibility for understanding basis risk and customer outcomes.

The fifth owner is adjudication and redress. Courts, prosecutors, regulators and settlement administrators determine different forms of responsibility under different standards. Their decisions should be recorded without collapsing allegation into finding or institution into individual. Data linking resolution, respondent, period, benchmark, legal status and payment type is essential public infrastructure.

What proof of remediation should look like

Evidence of remediation begins with inputs. For every contribution, the bank should retain the data hierarchy, transactions considered, adjustments, judgment rationale, submitter identity, reviewer and timestamp. Exceptions should be machine-readable and sampled. A control that produces no reconstructable record cannot show whether a later anomaly was reasonable.

It continues with conflicts. Surveillance should map relevant trader positions and communications to contribution windows, while respecting legal limits on employee monitoring and data transfers. Alert closure should record who reviewed the evidence and why it was benign or escalated. Repeated low-level alerts involving the same desk, submitter or direction should aggregate automatically.

Management information should report more than alert volume. Useful measures include unsupported judgments, late submissions, overrides, dispersion from transaction evidence, trader contact, repeat exceptions, time to close investigations, disciplinary outcomes and overdue remediation. Independent validation should test whether thresholds miss plausible manipulation strategies. Regulators should be able to reproduce samples.

For successor rates, evidence changes form. Administrators should publish methodology, eligible transaction volumes, revisions, contingency procedures and governance. Users should test calculation engines, compounding, calendars and fallbacks. Risk functions should measure basis between old and new rates under stress. Conduct teams should sample customer outcomes and communications. A deep transaction base is powerful evidence of representativeness, but concentration and structural market change must still be watched.

Proof also requires adverse testing. Could a concentrated group of transactions influence the rate? What happens if the source market closes or data arrive late? Can an affiliate benefit from the administrator's discretion? Are corrections timely? Does a term rate derived from derivatives reintroduce circularity? A replacement benchmark deserves confidence because these questions are measured and governed, not because the old benchmark failed.

Reconstructing responsibility from the evidence

A defensible investigation begins with time. The investigator should construct a submission-window chronology in a consistent time zone: relevant trades and risk positions before the fixing, communications to or from submitters, the internal draft, approval or override, the contribution sent to the administrator, the panel distribution, the published rate and any subsequent profit-and-loss movement. Time precision prevents a later message from being treated as a prior instruction and exposes whether a position existed when a request was made.

The chronology then needs an evidential hierarchy. Contemporaneous system records generally show what was transmitted and when. Recorded communications can illuminate purpose, but shorthand requires context. Witness accounts can explain practice, though memory and incentives must be evaluated. Policies show the expected process, not necessarily the process followed. Statistical analysis can identify unusual direction or persistence, but it cannot by itself identify a speaker's intent.

Enforcement documents synthesize evidence under a defined legal process; they should be cited for the conclusions they actually make, not used as a substitute for the underlying record in a separate damages case.

Position evidence requires particular care. A desk may hold many instruments whose sensitivities offset. A request in the direction of one trade is not automatically beneficial to the net book. The relevant calculation uses the portfolio's exposure to the exact currency, tenor and fixing, including hedges and nonlinear effects. An investigator should distinguish an anticipated benefit at the time of the request from the profit ultimately realised. Intent can exist even if the position changes, the submission is ignored, the panel average is unaffected or the trade loses money for another reason.

Submission evidence also needs a counterfactual. The fact that a bank ranked low or high among panel members does not establish falsity. Its funding profile could differ from peers, and market conditions could move quickly. A stronger assessment compares the submitted value with eligible transactions, contemporaneous quotes, observable funding, internal transfer prices and the bank's documented methodology. Where judgment was permitted, the question is whether the judgment honestly answered the benchmark definition and was supported, not whether another reasonable analyst could have selected a slightly different number.

Reputation-related conduct demands a different counterfactual: what would Barclays have submitted if external perception had played no role? Evidence might include earlier practice, the spread to peers, internal funding observations, management communications and changes coinciding with concern about media or market interpretation. Even then, the finding should remain tied to identified dates and evidence. A pattern during crisis conditions does not prove the same motive in calm periods.

The final step is attribution. A trader can be accountable for an improper request; a submitter for acting on it or reporting falsely; a manager for directing, tolerating or failing to escalate known practices; control functions for deficient design or execution; and the institution for regulatory breaches within the relevant legal framework. These forms can coexist, but none should be presumed solely from job title. Responsibility needs an act, omission, duty, knowledge standard and procedural status.

Data governance was part of market governance

The LIBOR episode is also a data-governance failure. A daily contribution looked like a small data point, but its provenance was economically material. The submission system needed lineage from source evidence through human adjustments to publication. Without lineage, a reviewer could see the final number but not determine whether it came from a transaction, a quote, an estimate, a trader request or a reputational instruction.

Good lineage records the source type, time, currency, tenor, market and any exclusion. Human judgment should be entered as a structured adjustment with a reason code and narrative, not hidden in a spreadsheet cell. Reviewer changes should preserve both versions. Access logs should show who viewed or altered the record. Retention should cover the period required for supervision, litigation and audit, with legal holds when an investigation begins.

Cross-border operations complicate this design. A trader, submitter, server, administrator and regulator may sit in different jurisdictions. Communications monitoring and evidence transfer must comply with employment, privacy, secrecy and data-locality rules. Those constraints do not excuse blind spots. The institution should map lawful access in advance, segregate sensitive data where necessary and create escalation paths for cross-border production. A control dependent on improvised access after an alert is not operationally credible.

Automation can help but should not become an unreviewed authority. A model can rank anomalous submissions, connect communication graphs or compare position direction with requests. It can also reproduce bad labels, miss new language or over-alert on legitimate treasury contact. Model governance therefore needs documented purpose, training and test data, explainable alert factors, performance by desk and language, drift monitoring, change approval and human review. Closing an alert because a score falls below a threshold is not an adequate factual conclusion.

Data sovereignty has a second dimension: no private institution should be able to control the public account by controlling all underlying records. Regulators require timely access, administrators need contributor audit rights, and firms must preserve material communications. At the same time, publication must protect legitimate confidential and personal information. The answer is controlled evidential access and precise public findings, not either total secrecy or indiscriminate disclosure.

Measuring whether reform worked

Sanctions and new policies are outputs, not proof of effectiveness. The first outcome measure is whether unsupported or conflict-tainted submissions became less likely and more detectable. That can be tested through transaction support rates, independent re-performance, contact surveillance, exception recurrence and regulator sampling. A fall in reported alerts is ambiguous: it can mean better behaviour or weaker detection. Measures need paired indicators and challenge.

The second outcome is institutional learning. Did the bank identify similar conflicts in other submissions, valuations, auctions or indices? Did management change incentives and information flows, or only rewrite the LIBOR procedure? Did internal audit verify the remediation after initial closure? A narrow programme can pass a deadline while leaving the underlying habit—commercial influence over supposedly independent data—untouched.

The third outcome is market resilience. For a successor benchmark, administrators should report transaction volumes, concentration, revisions, outages and contingency use. Supervisors should examine whether derivatives used to create forward-looking term rates are sufficiently deep and whether licensing or data dependencies create operational concentration. Users should know how the benchmark behaves in stress and what happens if its underlying market changes structurally.

The fourth outcome is transition fairness. Firms should compare the economic effect of fallbacks across customer groups, investigate outliers and preserve the rationale for discretionary choices. Complaints and litigation should be analysed for recurring design problems rather than treated only as legal inventory. Where legislation or regulatory action supplies a bridge, authorities should state who can use it, for which contracts, until when and with what review rights.

Finally, success should be expressed modestly. Evidence can show that a successor has broader transaction support, fewer judgmental inputs, stronger governance and tested fallbacks. It cannot prove that manipulation, error or conflict is impossible. A credible institution reports residual risk and contingency readiness. Overclaiming certainty recreates the trust problem that benchmark reform was meant to solve.

The durable lesson

LIBOR's apparent simplicity hid a chain of private discretion. Barclays's resolutions made that chain visible: derivatives requests could reach submitters; reputational concerns could influence contributions; controls and escalation could fail; and authorities operating under different laws could impose distinct consequences. Those conclusions are serious without embellishment. Their force depends on preserving their boundaries.

The case does not justify saying that every Barclays submission was manipulated, every linked contract was harmed, or every individual associated with the process was culpable. It does justify demanding institution-level systems capable of separating market evidence from trading interest and reputation management. It also justifies treating a benchmark contributor as a steward of market infrastructure, not merely a firm answering a daily survey.

Reform improved governance, legal oversight, methodology and auditability. Transition to SONIA, SOFR and other risk-free rates reduced dependence on judgment about unsecured term bank borrowing. But cessation and replacement are not retroactive proof of loss, and transaction-based successors are not immune by definition. Their integrity rests on observable depth, transparent rules, operational resilience, conflict controls and continuing challenge.

The market-integrity accountability test is therefore ongoing. It asks whether every actor can show what it controlled, what evidence it used, what conflict it managed, what warning it escalated and what remedy it supplied. Where those records exist, responsibility can be assigned precisely. Where they do not, a benchmark can again turn dispersed discretion into systemic opacity.