Documentary
The Trader Who Broke a 233-Year-Old Bank
Barings Bank opened its doors in 1762 and financed the Louisiana Purchase. It took one trader, one hidden account and three years to end it. Here is what actually broke, control by control.
Prefer YouTube? Watch this episode or subscribe to the Rofterm channel.
Barings was 233 years old when it died. Founded in 1762, it was the oldest merchant bank in London: banker to the Crown, financier of the Louisiana Purchase, the institution the Duc de Richelieu once called Europe's sixth great power. On 26 February 1995 it was insolvent. The Dutch group ING bought it for £1.
The story is usually told as a story about a person: Nick Leeson, 28 years old, running Barings Futures Singapore. That framing is comfortable, because a single bad actor is a problem you can fire. It is also the wrong lesson. What failed at Barings was not one man's judgement. It was a control environment that let one man be his own auditor for three years, and a reporting chain in which everybody assumed somebody else was watching.
The short version. Leeson ran both the trading desk and the back office that settled its trades. Losses went into an error account, 88888, which had been excluded from the reports sent to London. Cash to fund the margin calls kept flowing from head office because nobody linked the funding requests to the positions. On 17 January 1995 the Kobe earthquake moved the Nikkei against him. Six weeks later the bank was gone, roughly £827 million short, about twice its available trading capital.
What actually happened
Barings Futures Singapore was meant to be a low-risk business. Its job was arbitrage: buy a Nikkei 225 futures contract on one exchange, sell the same contract on another, pocket the small difference. Genuine arbitrage is close to riskless because the two legs offset. It also generates unglamorous, predictable profits.
What Singapore actually reported were profits far larger than that strategy should produce. In 1994 the desk was recorded as one of the group's strongest performers. Nobody in London pressed the obvious question hard enough to get an answer: where is this much money coming from, on a strategy that is supposed to be riskless? Outsized return is a risk indicator. It was read as a performance indicator.
The real positions were not arbitrage. They were directional bets that the Nikkei would stay inside a range, expressed partly through sold options, a structure that pays a steady premium while the market is calm and loses without limit when it is not. When the market moved against those positions, the losses were booked into account 88888 instead of appearing in the numbers going home.
Five control failures, and how each one would be caught today
None of these are exotic. Every one of them appears, in some form, in ISO 27001, SOC 2, COSO and every operational-risk framework written since. Barings is instructive precisely because the failures are so ordinary.
Segregation of duties existed on paper, not in the org chart
Leeson was general manager of the Singapore operation and also ran its settlements function. The person creating the exposure was the person reconciling it and the person reporting it. Once that is true, every downstream control is decorative: the reconciliation will always agree with the trade, because the same hands produced both.
Today: a quarterly access review that lists, per system, every user holding both trade-entry and settlement entitlements, with the exceptions signed off by someone outside the desk.
A suspense account nobody reconciled became a hiding place
Account 88888 was opened as an error account, a legitimate mechanism for parking mis-booked trades until they are corrected. It was then excluded from the file sent to London. An error account is only a control if something forces it back to zero and somebody independent checks that it did.
Today: daily reconciliation coverage stated as a percentage of accounts, with an aged exceptions report. The metric that matters is not "we reconcile". It is "we reconcile 100% of accounts, and here is the evidence for every day of the quarter."
Treasury saw the symptom and nobody connected it to the cause
Funding the margin calls on the hidden positions required very large sums to move from London to Singapore, repeatedly. Those transfers were visible. They were processed as funding requests rather than read as what they were: evidence of an exposure nobody had authorised. The information existed inside the bank. It just never met the risk it belonged to.
Today: escalation thresholds that tie treasury movements to position limits, so funding above a defined level automatically triggers an independent position review rather than a payment.
Accountability was diffuse enough that everyone was covered
The Bank of England's Board of Banking Supervision report into the collapse, published in July 1995, found responsibility for the Singapore operation split across product-line management, regional management and group functions, with no single person owning it end to end. Diffuse ownership is not a gap you notice, because every seat feels occupied. It only shows up afterwards, when you ask who was supposed to check and get four plausible answers.
Today: one named owner per risk and per control, a review date on each, and an overdue list the board actually sees. If a risk has two owners, it has none.
Profit bought immunity from scrutiny
Internal audit had flagged the concentration of duties in Singapore. The finding did not force a change fast enough, in an environment where the desk in question was making money. A control function whose findings get softened when the numbers look good is not independent, whatever the org chart says.
Today: audit and risk findings tracked to closure with dates and evidence, reported to the board independently of the P&L owner, so an open finding stays visible whether or not the business unit is performing.
What Barings looks like on a modern risk register
Stripped of the drama, the collapse reduces to five register lines. Any of them, tested honestly once a quarter, would have surfaced the problem long before Kobe.
| Risk | Control that was missing | Evidence you'd ask for |
|---|---|---|
| Unauthorised trading | Segregation of front and back office | Access review showing no user holds both roles |
| Concealed losses | Daily reconciliation of all accounts, suspense included | Signed daily recon, 100% coverage, aged exceptions |
| Funding masking exposure | Treasury thresholds tied to position limits | Escalation log with independent review sign-off |
| Unclear accountability | Single named owner per risk and control | Register export: owner + review date on every line |
| Findings not closing | Independent tracking of audit actions | Board pack showing open findings and their age |
The lesson that survives
Barings did not fail because its controls were missing. Several of them were written down. It failed because nobody could show, on demand, that they had been tested, and because the one person best placed to answer that question was the one person with a reason not to.
That is still the shape of most operational-risk failures. Not an absent control, but an unowned one; not a missing policy, but a policy with no evidence behind it. The question a board should be able to ask on any Tuesday is small and unforgiving: which of our controls were tested this quarter, by whom, and what did the test produce? An organisation that can answer that in minutes is a different organisation from one that needs a week to find out.
Take these into your next risk review
- List every role where one person both performs an activity and assures it. That list is your Barings list.
- Find your suspense, error and clearing accounts. Confirm someone independent reconciles each of them, and ask to see last month's evidence.
- Check whether any treasury or funding movement can happen without a linked position or limit check.
- Export your risk register and count the lines with no owner or a review date in the past. Those are open, not closed.
- Ask how long an audit finding has stayed open in a business unit that is beating its targets, then ask the same of one that is not.
Frequently asked
Which bank did Nick Leeson bring down?
Barings Bank, founded in 1762 and the oldest merchant bank in London. It failed in February 1995, 233 years after it was founded, and was sold to the Dutch group ING for a nominal £1.
How much did the Barings collapse cost?
Roughly £827 million, about twice the bank's available trading capital at the time. The losses came from unauthorised positions in Nikkei 225 futures and options taken by Barings Futures Singapore.
What was account 88888?
An error account opened at Barings Futures Singapore, nominally for booking genuine trade errors. It was excluded from the reports sent back to London, which turned it into a place where loss-making positions could sit unreconciled and unseen.
What was the root control failure at Barings?
Segregation of duties. Leeson ran both the Singapore trading floor and the back office that settled, reconciled and reported its trades, so the person creating the exposure was also the person assuring it. Every other failure downstream depended on that one.
Could a Barings-style failure still happen today?
The specific mechanics are harder now, but the pattern is not rare: one person owning both an activity and its assurance, a reconciliation nobody actually tests, and a control that two teams each assume the other is running. Those are ordinary findings in operational-risk reviews, which is why control ownership and evidence of testing matter more than the control's existence on paper.