Financial Cryptography

SocGen - the FC solution, the core failure, and some short term hacks...

Everyone is talking about Société Générale and how they managed to mislay EUR 4.7bn. The current public line is that a rogue trader threw it all away on the market, but some of the more canny people in the business don't buy it.

One superficial question is how to avoid this dilemma?

That's a question for financial cryptographers, I say. If we imagine a hard payment system is used for the various derivative trades, we would have to model the trades as two or more back-to-back payments. As they are positions that have to be made then unwound, or cancelled off against each other, this means that each trader is an issuer of subsidiary instruments that are combined into a package that simulates the intent of the trade (theoretical market specialists will recall the zero-coupon bond concept as the basic building block).

So, Monsieur Kerviel would have to issue his part in the trades, and match them to the issued instruments of his counterparty (whos name we would dearly love to know!). The two issued instruments can be made dependent on each other, an implementation detail we can gloss over today.

Which brings us to the first part: fraudulent trades to cover other trades would not be possible with proper FC because it is not possible to forge the counterparty's position under triple-entry systems (that being the special magic of triple-entry).

Higher layer issues are harder, because they are less core rights issues and more human constructs, so they aren't as yet as amenable to cryptographic techniques, but we can use higher layer governance tricks. For example, the size of the position, the alarms and limits, and the creation of accounts (secret or bogus customers). The backoffice people can see into the systems because it is they who manage the issuance servers (ok, that's a presumption). Given the ability to tie down every transaction, we are simply left with the difficult job of correctly analysing every deviation. But, it is at least easier because a whole class of errors is removed.

Which brings us to the underlying FC question: why not? It was apparent through history, and there are now enough cases to form a pattern, that the reason for the failure of FC was fundamentally that the banks did not want it. If anything, they'd rather you dropped dead on the spot than suggest something that might improve their lives.

Which leads us to the very troubling question of why banks hate to do it properly. There are many answers, all speculation, and as far as I know, nobody has done research into why banks do not employ the stuff they should if they responded to events as other markets do. Here are some speculative suggestions:

• banks love complexity
• more money is made in complexity because the customer pays more, and the margins are higher for higher payments
• complexity works as a barrier to entry
• complexity hides funny business, which works as well for naughty banks, tricky managers, and rogue traders. It creates jobs, makes staffs look bigger. Indeed it works well for everyone, except outsiders.
• compliance helps increase complexity, which helps everything else, so compliance is fine as long as all have to suffer the same fate.
• banks have a tendency to adopt one compatible solution across the board, and cartels are slow to change
• nobody is rewarded for taking a management risk (only a trading risk)
• banks are not entrepreneurial or experimental
• HR processes are steam-age, so there aren't the people to do it even if they wanted to.

Every one of those reasons is a completely standard malaise which strikes every company, but not other industries. The difference is competition; in every other industry, the competition would eat up the poorer players, but in banking, it keeps the poorer players alive. So the #1 fundamental reason why rogue traders will continue to eat up banks, one by one, is lack of competitive pressures to do any better.

And of course, all these issues feed into each other. Given all that, it is hard to see how FC will ever make a difference from inside; the only way is from outside, to the extent that challengers find an end-run around the rules for non-competition in banking.

What then would we propose to the bank to solve the SocGen dilemma as a short term hack? There are two possibilities that might be explored.

1. Insurance for rogue traders. Employ an external insurer and underwriter to provide a 10bn policy on such events. Then, let the insurer dictate systems & controls. As more knowledge of how to stop the event comes in, the premiums will drop to reward those who have the better protection.

   This works because it is an independent and financially motivated check. It also helps to start the inevitable shift of moving parts of regulation from the current broken 20th century structure over to a free market governance mechanism. That is, it is aligned with the eventual future economic structure.



2. Separate board charged with governance of risky (banking) assets. As the current board structure of banking is that the directors cannot and will not see into the real positions, due to all the above and more, it seems that as time goes on, more and more systematic and systemic conditions will build up. Managing these is more than a full time job, and more than an ordinary board can do.

   So outsource the whole lot of risk governance to specialists in a separate board-level structure. This structure should have visibility of all accounts, all SPEs, all positions, and should also be the main conduit to the regulator. It has to be equal to the business board, because it has to have the power to make it happen.

   The existing board maintains the business side: HR, markets, products, etc. This would nicely divide into two the "special" area of banking from the "general" area of business. Then, when things go wrong, it is much easier to identify who to sack, which improves the feedback to the point where it can be useful. It also puts into more clear focus the specialness of banks, and their packaged franchises, regulatory costs and other things.

Why or how these work is beyond scope of a blog. Indeed, whether they work is a difficult experiment to run, and given the Competition finding above, it might be that we do all this, and still fail. But, I'd still suggest them, as both those ideas can be rolled out in a year, and the current central banking structure has at least another decade to run, and probably two, before the penny drops, and people realise that the regulation is the problem, not the solution.

(PS: Jim invented the second one!)

Posted by iang at February 2, 2008 06:31 PM | TrackBack