Earlier this month, ISDA published a short paper entitled Counterparty Credit Risk Management in the US Over-the-Counter (OTC) Derivatives Market. The paper used data prepared by the Office of the Comptroller of the Currency (OCC) to examine the extent of losses related to counterparty defaults from 2007 through the first quarter of 2011. During that entire period, the US banking system sustained losses of $2.7 billion on counterparty defaults on OTC derivative contracts. This includes nearly $850 million of losses in the fourth quarter of 2008, which is when we believe losses associated with Lehman were recognized. This leaves less than $2 billion of counterparty losses not related to Lehman.
We were surprised at these relatively small figures as we remembered large synthetic CDOs inflicting very large losses on financial institutions. The OCC data covered just banks so we looked harder. We found very substantial counterparty losses in non-bank affiliates of banks and broker-dealers outside the banking system. These losses were related to mortgage products and they verified our understanding of the market. We were, though, still surprised by the very small counterparty losses sustained by the banking system.
We then decided to model how these exposures and losses might be impacted by the clearing and margining regulations stemming from the Dodd-Frank Act. A portion of the losses were due to corporations that defaulted amid the troubles that followed 2008. These users will be exempt from clearing under most sets of proposed regulation. Another portion must have come from entities, such as hedge funds, that defaulted on mortgage and other complex products that could not be cleared.
The balance of counterparty losses were caused, most likely, by defaults of financial entities that will be required to clear their transactions under the new regulations, and the defaults occurred on these very same transactions. If these transactions had been cleared, then it stands to reason that counterparty losses by US banks would have been lower. This raises an interesting question: how beneficial would clearing and initial margining have been for US banks?
The reason we ask this question is the pure cost of initial margin related to clearing. ISDA estimates that initial margin will amount to anywhere from $200 billion to $500 billion of collateral once clearing is complete. The “cost” of this collateral might be 50 basis points; it might be 100 basis points. That’s a minimum of $1 billion per year and a maximum of $5 billion per year. These estimates are global costs. Perhaps the cost to customers of US banks is $250 million to $2 billion per year. These costs need to be considered when contemplating the benefits of clearing and the need for initial margin.
So let’s consider what this means in light of the counterparty loss experience of US banks. As noted, we have a firm upper bound of $2 billion of costs (the counterparty losses not related to Lehman) that could be attributed to the absence of initial margin. But let’s make some assumptions. First, let’s assume half the losses were attributed to defaults by corporations. That reduces the losses caused by financial institutions to $1 billion. These losses were caused by a combination of:
- no variation margin on clearing eligible products;
- no or insufficient margin on products not clearing eligible; and
- variation margin but no initial margin or insufficient initial margin on clearing eligible products.
It is impossible to quantify these respective losses but common sense indicates that a large majority came from lack of variation margin or from products that are not clearing eligible. We will hazard a guess and say those two causes accounted for 80% of the $1 billion of losses from the defaults of entities that would now be subject to clearing. That leaves $200 million as the benefit of initial margin over a period of four and a quarter years. Our numbers and analysis don’t have to be exact. We shake our heads and wonder: we spend hundreds of millions a year and save tens of millions?
Readers: are we missing something?