Who will be the dummy in CCP crash-tests?

Any crash test needs a dummy. There are various candidates in the suddenly divisive debate about how to ensure the robustness of central counterparties (CCPs).

If the outcome to that debate is a standardised test of some sort, then the CCPs will be the ones putting on the seatbelt. But the design of the test will be critical, because their current practices – a closely guarded secret for each clearing house – vary dramatically.

One major clearing house is said to run its stress tests at a 99.9% confidence level. A rival CCP calibrates its stress tests to seven standard deviations, which translates to a confidence level of 99.9999999997440% – but this CCP assumes a normal distribution of returns, while the first does not.

The number of scenarios used in CCP stress tests also varies, from less than five to more than 100. And the lookback periods range from 10 years to 30 years.

A single, standardised test applied on a uniform basis might end up a long way from the practices in use at any individual CCP, potentially producing very different results and making it look less robust than its peers.

Some banks are currently reconsidering their earlier support for a standardised test, arguing it will produce homogeneous risk management

There are plenty of other, granular differences, and also one big, obvious one. At LCH.Clearnet, each of the markets the CCP clears has its own default fund, while other big CCPs have funds that cover multiple markets – the latter approach raising questions about the extent to which cross-product offsets are recognised in the context of a stress test. The UK-based clearing house is an advocate of a standardised test. It also argues such a test should strictly limit the extent to which cross-product correlations can be assumed – a stance its critics in the industry see as an attempt to undermine the competition.

Others may find themselves thrust into the dummy's role. Some banks are currently reconsidering their earlier support for a standardised test, arguing it will produce homogeneous risk management – with default fund and margin requirements all responding in the same way, at the same time, to the same impetus. CCPs would also have exactly the same blind spots.

"If you think it through, a standardised stress test is just not the right solution," says a senior clearing executive at one European bank – and a former supporter of standard testing.

Regulators may also end up being cast as the dummy. With banks divided over the topic, as well as CCPs, whatever stance the officials take will delight some and dismay others – so the tendency might be to look for some kind of acceptable middle-ground, winning support but failing to go far enough. This kind of compromise increasingly seems to have hobbled the only existing international standards on CCP risk management - the three-year-old Principles for financial market infrastructures.

So, it may be a divisive debate. But regulators should act boldly if they think it is required – the stakes are too high for anything else.