This is the second part of a 2-part post on LCH.Clearnet’s just published “Stress This House, A Framework for the Standardised Stress Testing of CCPs”, an analysis of risk management at CCPs. We look at the 5 elements required for a CCP stress test.
Element 1: Segregating Clearing Member Assets
“…As a first step, a CCP portfolio must be appropriately segregated into default funds prior to the application of any stress test exercise. It is critical that this be performed correctly in order to prevent a potentially serious underestimation of how a portfolio would perform under a real world stress event…”
LCH.Clearnet is suggesting either you allow uncorrelated products in the same default fund and “…do not allow any portfolio margining or offset of default fund contributions between them. This would ensure that each individual component of the default fund is sized to a cover 2 standard on a standalone basis as no inadvertent offsets are allowed…”
Alternatively,
“…Different products can be placed in the same default fund and portfolio margining and offsets can be allowed among them if they meet the following criteria: the products must be highly structurally correlated and there must be an economic rationale to have the products in the same product class. A high correlation in its own right cannot be used to justify the inclusion of multiple products in the same product class. The correlation between the products must also be stable… Essentially, unless the portfolio of products is properly segregated, their individual contribution to the financial assets of the CCP could be understated…”
Rate products, equities, and credit should, according to LCH.Clearnet, not be margined together.
Exactly how one can predict that the correlation between products is going to be stable is a hard one. One is reminded of the Feb 23, 2009 story from Wired Magazine “Recipe for Disaster: The Formula That Killed Wall Street” about the erroneous assumption by many traders that correlation was constant.
Element 2: Constructing Standard Stress Scenarios
“…The segregated asset classes should be subjected to three defined categories of standardised stress scenarios…”
The three scenarios are
a.) Historical stress scenarios: think specific market disruptions like the ERM exit in 1992, the Russian default in 1998, or Lehman’s collapse.
b.) Hypothetical stress scenarios: this involves thinking up the unexpected market moves, including those that go in directions opposite to what has been seen before.
c.) De-correlation stress scenarios: throw any expected correlation out the window and see what happens. To test unforeseen changes in correlation, create all sorts of permutations by moving around many points on a yield curve in different directions. In a pool of equities, the assumption that stocks in the same industry move together is thrown out and idiosyncratic risk is introduced and tested. In credit, assume that “…a single concentrated exposure to the credit of one debt issuer could again run counter to the overall correlated performance of the other names in a portfolio of bonds, thereby skewing the risk profile…”
Element 3: Combining the Stress Scenarios
Test the various scenarios and “…under each scenario, record each member P/L should that scenario happen and calculate the resulting margin erosion and the corresponding usage of default fund (both funded and unfunded) and the usage of skin in the game…”
This is where the rubber hits the road – what exactly happens when put the assets through the stress tests.
Element 4: Stress testing the cover 2 condition
Cover 2 is the concept that the CCP can withstand the default of its two largest clients.
“…for each scenario ask how many counterparties would need to default to exhaust the funded portion of the default fund(s)…The minimum number of counterparties observed under this process across all scenarios would then provide a clear picture of the buffer built into the pre-funded portion of the default fund(s). This would be of immense interest to clearing members who have mutualised the default fund with others, especially if there were plausible scenarios in which the CCP may have to call on the unfunded resources of the default fund…”
Element 5: Auction Risk
“…Having assessed the level of stress necessary to exhaust both the CCP’s own skin in the game and member default contributions, the final element in a stress testing framework is to ascertain whether a defaulted member’s cleared positions could be successfully auctioned without recourse to unfunded resources…”
If everything can be ported from a defaulted member to a non-defaulting one, that negates the auction issue. If not, positions need to be auctioned off. LCH.Clearnet made the point that it is the economic interest of the CCP members to see a successful auction. But there could be mitigating circumstances. For example, a.) if there are only a few non-defaulting members left creating concentration risk or b.) the non-defaulting members are on the same side of the market as those who defaulted and bidders are loath to increase their risk or c.) members who are in a position to absorb the risk have limited capacity. These conditions can and need to be tested for.
Finally, LCH.Clearnet comes back to the underlying premise that it is critical that each CCP is stress tested the way in order to compare apples to apples.
“…Using the methodology outlined in this paper, clearing member risk managers that today struggle with the opacity of non-compatible and non-standardised reports supplied by CCPs, would instead benefit from a single readily understandable framework, detailing in plain language how each of their default funds would endure under the same series of hypothetical stresses at each clearing house…”
The paper goes into more detail than space will allow us here. It is worth a read to understand how to best kick the tires.