In epidemiology, risk of contagion spread is sometimes modeled using a chain model approach. The chain approach involves a continuous chain of risk from a single entity spreading through a network of contacts with various rates of success. Break the chain, break the spread. CDS exposure to netting failures can break the chain of cascading liquidity crisis with a buffer of excess capital.
CDS Notional amounts get a lot of headlines, but few really understand the nature of the risk and why it so large. Netting risk is rare and unusual. Bankhaus herstatt is the only institution I know of to have failed due to a failed netting where the other side was actually good for the deal.
Settlement is boring in terms of risk, most people like to talk about models, VaR, Swans etc. But netting is the plumbing that keeps it all going, failed CDS netting could lead to systemic risk on a significant scale. This risk is heightened due to the following factors:
- Large Notional amounts outstanding
- Low Tier 1 capital in place
- Liquidity challenged environment
- Crude netting, reporting and tracking in some institutions
- Sometimes subjective EOD or delayed agreements of an EOD
- An environment where EOD's are becoming more likely or concentrated
- An environment where a netting counterparty failure may become more likely
- Counterparties offsetting risk elongate settlement chains and therefore could amplify netting risk
Here is a simple graphic model that shows nettting exposure growing as OTC (non exchange traded participants each try to minimize risk by offsetting position with other counterparties. the transactions are identical in terms of reference credit and notional amount.
We can consider this a Chain of settlements that all have to fall in sync assuming "no excess cash buffer". The Chain runs from A to D
Here is an example of exchange traded settlement as a centralized netting agent with the same behaviours of the institutional participants.
As we can see these two settlement systems have very different netting risk exposures in the event of default events. The nature of CDS's which act a bit like extreme barrier options is that these settlement chains rarely get tested or stressed. They haven't evolved robustness from a network perspective. Most organically evolved networks are formed and behave structurally in something called a scale free network topology. Basically this means the number of connections or relationships is logarithmically scaled. Imagine the 80-20% rule in action.
Here are some crude assumptions with absolutely no empirical data behind them.
My general hope is that the efforts of ISDA to accelerate standardized terms for netting and the efforts of the major banks to establish an automated netting facility happen soon. I don't know what the capacity of the financial system is for a failed chain event, but it is certainly cause for concern. We certainly don't want any grad students writing theses with titles such as "CDS netting failure, the day money disappeared."
If anyone has data or a model on the "capacity" for a network to handle these failure it would be interesting to see. Assuming a 0.05% annual failure and clearing rate of 3 business days. with an 80-20% concentration of failure could lead to the system needing to clear $420billlion in a cycle. Assuming the same 80-20 rule for institutional concentration means the top 50 institutions would need a liquidity buffer of $336b.
Again these are made up numbers using the crudest of assumptions, but highlight the risk of failed netting even if the counterparty is good, that or a lot of bank holidays to stretch they cycle thereby increasing the netting capacity of the system.
I won't even touch on the component of an institutional failure being used as the reference credit. I call that recursive failure in the system and it is nasty indeed.