Today, SocGen’s Internal Audit report seems to indicate that contrary to my own speculations in a previous post, Jerome K. did not break into the computer systems, and did not use other people’s ID to transact. So how did he manage ? The autopsy of the string of treasury and cash transfers from the front office black box might give an answer, but now we are dealing with other parameters as well: is the truth worth the mangling of one of France’s most performing banks? worth endangering thousands of more competent - and innocent - people’s jobs?
In Social economics calculus ( shades of Jeremy Bentham) may lie the first steps to solve the squaring of the circle.

In this kind of situation, where reports raise more new questions than answer previous issues, more speculation is inept, so I will stop. But it will still be useful, in an Operational Risk ennvironment, to investigate the consequences of a real intrusion into the controlling and transaction systems of a Front Office. If Jerome K. did not do it, somebody someday will…

Copulae are becoming more and more popular these days.

At least, let us say that there are more and more people who know that these strange contraptions are used to model risks. Not a lot of them though, would feel at ease calculating the compounded, “global” risk of a bank using copulae. Even less wouyld know how to compute the VAR (value at risk) of a copula.

Clearly the rogue trader at SocGen did not just overshoot his trading limits, and thus qualify for Basel Op Risk category 6 “Execution Delivery and Process Management” risk. He also usurpated his fellow traders identity, thereby causing Category 1:”Internal Fraud” risk exposure. And he hacked into Middle and Back Office servers, using what is nicely dubbed “social engineering”, creating breaches in Workplace security (category 3). We are looking here at an aggregated, correlated risk. A case for copula.

In order for each of these separate categories of risk to hatch a 99.9% tail event, one must wait for a long, long time - indeed, as a risk manager from Fortis once told me at the Committee of European Bank Supervisor 3 years ago, it’s Star Trek lifespans. But, when compounded, the global risk being calculated as a copula on these marginal individual risks, probabilities rapidly arrive at astounding levels easily. Properties of Copulae force the Iso-Value-at-risk hypersurface to be concave, and thus, each individual risk is less rare than what is expected from its frequency if it were alone.

Our different Projet de Fin d’Etudes on copulae have been the topic of studies and programming for 4 groups of 6 students each on the average, at ECE. Mortgage Credit risk and Operational risk have been studied through copulae. C++ and C# programs have been compiled into DLL’s called by Excel. You guys have all sweated through January and you have all loved them, be they called Franck’s or Student’s, be they Archimedean or just plain Normal.

A new way to view risk and chance, hazard and mishap, a little less qualitatively, a little more quantitatively. Thanks for your burning the midnight oil, guys ! Well done !