In early August 2008, Citigroup agreed to pay $100 million in fines to US state and federal regulators in a settlement covering claims related to the now-defunct Auction Rate Securities market.
While not admitting guilt, Citigroup also agreed to buy-back some $7.5 billion of these securities from investors at par value thus taking any mark-to-market losses onto their books. Sensing the direction of the wind, Merrill Lynch and UBS almost immediately announced that they too would buy back billions of dollars of similar debt that they had sold to their clients. Legal actions continue in other jurisdictions against the same and other banks.
For investment bankers, who do not know their ARS (Auction Rate Securities) from their LBOs (Leveraged Buy Outs), an ARS is a hybrid security that combines attractive features of both long-term bonds and short-term money market instruments. These innovative securities were re-priced at short-term intervals (7-35 days) using a complex auction process organized by a small group of broker/dealers, including Citigroup, Merrill Lynch etc.
Like a game of pass the parcel at an ADHD clinic, the ARS market shuffled some $200 billion of debt around frantic investors until the music was stopped dead by the Credit Crunch in 2007. Stamping their Gucci loafers in disgust, the Wall Street market makers just stopped making a market, leaving investors holding worthless parcels of illiquid securities. The investors claim that they were assured that ARS securities were highly liquid and that the broker/dealers would always buy any securities not sold in an auction - whoops! Bad losers that they are, investors went crying to the New York Attorney General, who, smelling blood and no longer having to seek votes, turned upon Wall Street.
The fines paid by Citigroup in the ARS fiasco clearly constitute operational risk losses under Basel II, specifically "Improper business or market practices" under the 'loss event type category', which is "Clients, products, and business practices".
Furthermore, the losses incurred by taking their clients' ARS debt onto their books at par, rather than market value, must also be added to these fines to calculate the total 'loss amount' for this event.
Under the Advanced Measurement Approach (AMA) described in Basel II, Citigroup, and the other offenders, must set aside 'Operational Risk Capital' (ORC) to cover potential re-occurrences of such 'loss events'.
Unfortunately, it will take some time for these banks to crystallize the full extent of these loss amounts, as the securities must be repackaged and sold at a lower value onto a skeptical market.
Prior to Basel II, the precise value of the losses incurred would not have meant a hill of bean counters to anyone other than the firms involved but now, under Basel II, all AMA banks MUST incorporate these losses into their estimations of Operational Risk Capital, as part of so-called 'relevant external data'. [Note the deliberate emphasis on MUST.]
Until this point, debate on 'external data' in the Basel II community centered on the thorny issue of what percentage of a particular well-publicized loss event was 'relevant' to a specific firm. Since the original loss amount was generally known with a reasonable level of certainty, calculating ORC became a matter of multiplying the chosen percentage by the known amount and adding this (and other relevant external data events) to the firm's selected, but not fully validated, 'loss distribution'.
Since we are unlikely ever to know the full extent of the total value of ARS losses (simply because the firms who are not fined but tow the line are unlikely to want to publish the embarrassing details)we are now in the ridiculous situation that in order to calculate Operational Risk Capital, we must multiply a number we don't know, by a percentage we don't know, to add to a distribution we don't know, for, increasingly, a reason we don't know!
Where is Millard Fillmore* when we need him?
The shambles that is 'external data' in Basel II also appeared in another 'relevant' context recently.
In August 2008, the, heretofore relatively silent body, CEBS (Committee of Banking Supervisors) issued the results of a 'stock-take' of responses by European banks to the losses at Societe Generale incurred by our favourite musketeer M. Jerome Kerviel. Now why this was a stock-take rather than a survey is not disclosed but in any case, CEBS asked banks throughout Europe 'could the same thing happen in your institution?' In a result that would only be more predictable if 100 pre-school children were asked "should all kittens be strangled at birth?' the banks responded unanimously that "No - it couldn't happen to us" and "If it did, any loss would be minimal anyway". They would of course say that - wouldn't they?
Hold on a minute though. The banks surveyed were mainly AMA banks that had been accredited just recently by the very supervisors that were asking the (stupid?) question. Who better to know the answer to this question than the person who had just reviewed and approved the Operational Risk Management systems of the banks surveyed - i.e. the members of the CEBS themselves?
This really is the blind leading the bland.
The survey was, however, not completely useless. The supervisors also asked the banks to provide an estimate of how much their Operational Risk Capital would be changed by incorporating the 'relevant' impact of the $7 Billion loss at SOCGEN into their internal ORC models [which, note, they MUST do under Basel II]. The survey (whoops - stock take) found that there would be "an estimated increase in the regulatory capital charge of up to 20 per cent" - Yes - 20%!
This raises the specter of some very interesting meetings in banks around the world in 2009.
Envisage a meeting, next year in some financial capital, between the Global Operational Risk Manager (GORM) of a large firm and the head of a major business line to review the allocation of Operational Risk Capital for that specific business.
The GORM might start with "First, the good news. After all the hard work, time and effort spent by your business line last year, we have calculated that your prior year's allocated ORC should be cut back by some 10% - may I congratulate you and your staff?"
The GORM continues, "On the other hand, we must note that, because of the Societe Generale affair in France, we are unfortunately required, by regulators, to increase your allocation of ORC by 20%, sorry".
**** *** *** *******. I leave the reader to insert their own form and content of the business head's response. I predict that, in many firms, there will be at least one GORM less after such discussions.
In effect, Operational Risk Managers are being set up to fail. They (and their business clients) may do an outstanding job in reducing operational risk in their own firms but be undermined by some silly oaf doing something stupid somewhere else. This is clearly untenable in the long term, but we are sleepwalking towards many such conversations over the next few years.
Each week, we encounter more examples of how the Basel II rules on the calculation of Operational Risk capital are impossible/farcical to implement in practice. How long will it be until someone applies the coup to grace to this (load of) bull?
Raising the tone, somewhat, I pose a serious question on the use of 'external data' to regulators and the operational risk theorists amongst us.
Due to the credit crisis, major banks are hemorrhaging capital like blood from a recently beheaded chicken. Over the past year, having written assets down at different rates, the 'relative' asset values of firms have changed, sometimes considerably. For that matter, so has 'relative income' as some banks have incurred much bigger losses than others.
Now it has often been hypothesized, but never demonstrated**, that operational risk losses have some direct relationship to asset size and/or income size and that factors based on these relative values should be used to 'scale' external data for ORC purposes. [Come to think of it that is the logic underlying the Standardized Approach to calculating ORC in Basel II].
Now if such a relationship does indeed exist, then surely firms must go back and incorporate the changed relativities into their assumptions about the impact of 'relevant' external data. In practice, this would mean that firms that were 'good' in the credit crisis would have to take on more capital for operational risk (as a relative function of asset size) and those that were 'bad' would get capital relief. This seems counter-intuitive to me.
A trickier technical question arises, which asset or income figure should we use for scaling? Should it be the latest value applied to historical losses? Or should it be the relative asset size at the time of each recorded loss? If it is an historical value, should it be the value at the time that the loss was discovered or when it was first hidden, which may be some time before? Or should it be some average over the period? Likewise would it be the relative asset sizes at the time that write-downs were announced, or when they should have been announced, which was sometime earlier?
For the very high 99.9 percentile confidence interval mandated by Basel II, the choice of scaling factor could turn out to be very significant, potentially dwarfing the losses made by the firm itself!
Whichever values are chosen, it would be interesting to be a fly on the wall of the GORM/business head meeting discussing this anomaly. "And by the way, because a particular bank [insert name here] lost $X billion and they have written down $Y billion because of the credit crunch, the allocation of operational risk capital to your business line must be increased by 30% - sorry". End of conversation I suspect?
Footnotes
* In 1856, Millard Fillmore was the US presidential candidate of the so-called 'Know Nothing' party, gaining almost 25% of the popular vote. If he were still alive, he would appear to be a shoe-in to succeed Mr. Nout Wellink, as next chairman of the Basel Committee.
** On the contrary, academics, such as Dahen & Dionne from HEC Montreal, using publicly available data have concluded that "the size effect is quite weak compared to the other scaling factors". In short we know little about what 'causes' large operational risk losses.
