QIS 4 reawakens concerns over correlation rule

Complaints about the Basel Committee on Banking Supervision's proposed capital rules for correlation trading have resumed following the conclusion on January 15 of tests designed to measure the impact on individual banks.

The rules would allow banks to use their own models to calculate how much capital to hold, but much of the ire is directed at a floor in the proposals that limits the capital reduction they can achieve. As a result of the floor, a risk manager at one large European bank says the capital consumed by its correlation business would account for roughly one-third of total trading book capital - far higher than is warranted by the risk, she argues.

"Not only did they make the modelling really difficult but on top of that, the eventual capital charge ends up being the floor level anyway," she says. "Whatever your model, whatever the quality, whatever the investment you made in quants and the system, at the end of the day you are still floored by this really, really stupid method."

Tests at other banks produced capital numbers that were higher than the floor, but the head of the Basel programme at one of these institutions objects on principle - if banks are going to be allowed to model the risk in their correlation books, regulators shouldn't insist on an arbitrary minimum level, he says.

At this point, nothing about the floor is set in stone, insists one member of the Basel Committee's trading book group, which is drafting the rules. The impact tests - known as QIS 4 - are the precursor to discussions that will look at how the floor should be calculated, what level it should be struck at and, more fundamentally, whether it should exist at all. There are critics of the floor within the Basel Committee, he says, but their voice is countered by others wary of allowing capital to be set by model results alone: "There are members at the committee whose faith in models has been greatly shaken by what happened during the crisis."

Whatever your model, whatever the investment you made, you are still floored by this really, really stupid method

Regulators introduced the comprehensive risk measure (CRM) last July, conceding ground to banks that had argued a prior set of proposals covering securitisations would have killed off correlation trading by burdening it with too much capital. As a quid pro quo, regulators attached strict criteria to the CRM and ruled capital would not be allowed to drop below 8% of the amount generated by the non-modelled, standardised approach. That might seem generous - and banks concede the new approach is an improvement on what went before - but only because the previous option was unworkable, say dealers.

"The original approach was ludicrous as it produced capital charges way beyond the total maximum loss you could have on the book. Even the CRM requires you to hold more capital against hedged positions than unhedged ones, which is a bit ridiculous," says the European bank's risk manager.

The trading book group will now collate results from around the industry, before deciding whether to implement the proposals or make further tweaks. But dealers warn the numbers generated in the tests may not be comparable. The risk manager says banks are using differing interpretations of the proposals, which is why some capital numbers are above the floor level and others well below, and adds that some banks have simply opted to provide estimates rather than using a fully modelled approach.

Regulators counter divergent capital numbers could simply be the result of different underlying portfolios, and therefore reflect the kind of risk sensitivity the proposals are supposed to produce. In addition, banks have been asked to respond to a detailed qualitative questionnaire as part of QIS 4, which will give regulators some insight into how the capital numbers were produced, says a second member of the trading book group. "We're getting a good amount of background information on the underlying calculations that can help us sift through the richness in the results," he says.

One reason for the tension over correlation trading is dealers' suspicion the rules are designed to kill the business by making it economically unattractive - indeed, some regulators have publicly questioned its value to the economy and society as a whole. But the first trading book group member tries to reassure dealers: "The point of QIS 4 is not to produce a measure of the social welfare value of this kind of activity, but to provide some sort of estimate of the underlying risk from a safety and soundness perspective."

And the second group member criticises the idea of regulators sitting in judgement on issues such as social and economic worth: "Yeah, the topic has come up - the idea correlation trading might be socially pernicious - but I don't know why we're especially well equipped to judge how, in the end, this activity may or may not benefit mankind."