According to this blog post written by someone who knows a DEAI trader, Jérôme Kerviel was an equity futures trader, essentially selling and buying to customers bets on the future value of the European Stoxx 50 index.

His job was to make sure the SG position wasn't getting either short or long : as any good bookmaker, the SG didn't want him to actually bet on the value of the Stoxx 50, and thus Kerviel had to sell roughly as many puts and calls on the index.

What happened was that Kerviel indeed made a bet, that the market would go up. The problem is, he was not allowed to do that by the bank ; so he had to hide it. The way he did it was by creating false bets, with false clients, going the other way ; thus his books appeared balanced.

How could he do it ? as we can see in his resume, he formerly worked in the DEAI middle office - the branch of the bank that is in charge of overlooking the traders, making sure they only do what they were allowed to do. Thus he knew how he was being overseed ; apparently he hadn't taken his vacation time for two years, as vacations prevent hiding one's tracks.

Not only did he work in the middle office, but he worked on the SG referential database - that means part of his job, for two years, had been creating counterparts - clients - for use in the bank's computer system. He knew how to do it.

But - he shouldn't have been able to do it. Supposedly, a basic precept of computer security is that accounts and passwords have to be changed regularly ; that they have to be kept secret, certainly from those who could abuse them. When I worked in the SG, someone doing a job similar to mine, got the boot when it was found his internet access account was used by a whole trading desk. Then, in the beginning of 2007 there had been an information program to make sure we were up to date with the security measures - our passwords were to be unwritten and personal, our desks clean each evening, etc...

Of course, the reality of work in a stressful environment got in the way : we had to give our passwords to the team when going on vacations, for example. I can imagine similar things happening in the Kerviel case : nobody bothering to purge his write-enabled account on the referential database sounds possible. When development has to happen fast, some things are forgotten ; and password management often didn't come in first on the bugfix priority lists.

Many commentators claim that a single trader couldn't possibly lose that much money, and that the 50 billion euros position he must have held is unbelievable. Well, the volume of trades on financial markets is unbelievably high these days ; Kerviel was a market maker, according to his resume, which means SG warranted they'd accept transactions at quoted prices - they made a lot of transactions on his products.

A more intriguing question is why Kerviel did this ; his job didn't include millions in bonuses : he was pay a "measly" 100 000 euros in 2007. This blog post attempts to answer

Our society has glamorized the job of traders ; many young workers in a large bank want to become one. But those that make it realize that they will never get the glamorous jobs in a French bank : those are reserved to people coming from the best "Grandes Ecoles" : Polytechnique, Mines, Centrale Paris. So do the extremely generous compensations that come with these positions : the head trader at DEAI is the best paid wage worker in France. Kerviel wasn't one of them. When he became trader he understood he would only do a pretty routine job, with tight limits on what he could do - and that he'd never really "make money", enough for the instant retirement at 35-40 he may have sought. And he may have become jealous. Maybe he wanted to prove he could make money for the bank. So he did what he wasn't allowed, and started to bet.

When in the middle of January the Stoxx 50 went down a bit too much, it appeared  that the SG was about to inexplicably lose money, and that margin calls were happening when they shouldn't ; after all the SG didn't bet. So all the hiding that Kerviel was doing couldn't go on, and the bank had to sell out of its position. The loss was possibly increased by the fact that it happened on a US bank holiday : the smaller market wasn't able to absorb as easily the massive selling, and what was a billion-and-a-half loss on paper became a five billion euros loss as the markets plummeted.

Thankfully for the SG customers, workers and owners, the bank was large enough to absorb this huge hit. But problems remain : the Société Générale is making huge profits from its retail banking activities ; why should those profits subsidise the risky investment banking activity, and in effect guarantee it?

Will the reputation of the bank - before the scandal, one of the highest in investment banking - hold in face of a clear oversight failure ?

And possibly, is a system where a single man can take such actions that cause risk for a systemic crisis - as may have happened if the SG had indeed gone under - all that stable and viable ?

Update [2008-1-27 14:15:32 by linca]:

The Société Générale has explained what it claims happened. The PDF press release.

It mostly agrees with what I wrote. Except the trader was in arbitrage ; i.e., he wasn't selling bets but profiting of market inefficiencies : if two similar products had different prices despite being supposed to have the same price, he'd make profit on the different and thus make the market more efficient. This implies very large volumes of trading.

SG claim the position unwinding was kept at under 10% of trade volume, i.e. it says it didn't influence the market ; and the market crash started in Asia before the position unwinding began.

Oh, and that first link has really insightful comments. (In French).