UBS, the Swiss investment bank, has been exposed to a $2.3bn loss from a position apparently built up by a trader, Kweku Adoboli, who allegedly inputted false forward-settling trades into the bank’s systems making the position appear hedged, when it was not.

Sources have suggested the future settlement date may not have triggered trade confirmation, delaying detection of the fictitious trades by the bank’s back office. No client positions were affected.

“The loss resulted from unauthorised speculative trading in various S&P 500, DAX, and EuroStoxx index futures over the last three months,” the bank said. “The positions taken were within the normal business flow of a large global equity trading house as part of a properly hedged portfolio.”

On 14 September, the bank discovered that unauthorised trading had been conducted by a trader in its ‘Delta One’ or Global Synthetic Equity business in London. It is alleged that Adoboli, who has been charged by UK authorities with fraud by abuse of position, input “fictitious, forward-settling, cash ETF positions” in the bank’s systems, according to a statement from UBS.

The non-existent trades covered the fact that the index futures trades violated UBS's risk limits. The trader revealed his unauthorised activity on 14 September when UBS control functions reviewed his positions.

Déjà vu

The case mirrors that of Jérome Kerviel, a trader with French investment bank, Société Générale whose positions lost the bank €4.9bn in 2008. The strategy of booking non-existent trades onto an investment bank’s systems to cover the unauthorised exposure was used by Kerviel as well, who also worked in the Delta One function of the bank, dealing with plain vanilla futures hedging.

Kerviel, who has an appeal pending, was found guilty by a court of breach of trust, fraudulent introduction of data into an automated processing system, forgery and use of forgeries.

In the investigation by French magistrates, it was found that that to conceal his positions and results, Kerviel entered, or had his assistant trader enter, fictitious transactions in the ‘Eliot’ computerised database.

“It is true,” Kerviel said during the investigation, “that I entered hundreds of multiple fictitious deals in the Eliot system in order to conceal my positions and results.”

“Afterwards,” he added, “I systematically cancelled these transactions, most often within two to three weeks. I could leave them for one or two months. When I cancelled them, they were erased from the various databases. […] I simply erased them from the Eliot database.”

The bank alleged that Kerviel also made use of the ability to record positive or negative provisions, modifying the valuation calculated by the front office system. This was a process normally limited to trading assistants for the purpose of correcting modelling bias, but on computer systems from which traders were not barred.

Carlos Goncalves, now chief information officer at Société Générale, later testified that it took 800 man-hours to find the false trades, as the system was not designed to find deleted information.

Forward-setting

Sources familiar with the case at UBS say that entering information into the middle- and back-office systems may not have been necessary to avoid back office supervision with forward-settling cash ETF trading; a convention reportedly exists within the bank that such trades don’t have to be confirmed until they are settled.

As the settlement date can be rolled forward, the point at which the back office sees the trades can effectively be delayed as long as they are unconfirmed. Adoboli had worked in the bank’s back office and would have been aware of this convention.

If forward-settling assets did obscure the bank’s exposure, real-time position keeping systems might have been used to provide clarity.

“If a bank is waiting until a forward point to calculate where it really is, it’s driving using the rear-view mirror,” said Dan Hubscher industry marketing manager, capital markets, at Progress Software, a supplier of event-driven market surveillance and monitoring software. “It runs the risk that the wall it thinks it might hit, was one it already hit a few months ago,”

“Firms have approached us about feeding data through to their systems in real time, to give an indication of what the position might be now even though confirmations don’t come though until later,” he adds.

Mat Newman, head of product management, Adaptiv, in SunGard’s capital markets business, says: “I have some sympathy for the view that if bad data is entered into a system, the output will necessarily be incorrect.

“But sometimes people are blinded by the quantitative analytical side of risk management; they forget to employ sanity checks.”

He gives the example banks’ credit divisions which will often run checks beyond the fully netted and offset position, to examine the gross, uncollateralised and un-netted positions as a form of stress test.

“A similar thing could be done for value at risk,” he asserts. “It would allow you to see very big positions building up and the very big hedges that you would have in place against them. Checking them would be valuable, even if it was just for mismatched timings.”

Features