US Bank Exposure to Greece, part 3

Tracy Alloway at the FT summarizes some of the push-back against the idea that US banks have substantial exposure to Greece:

Who’s selling Greek CDS, again

There appears to have been a massive amount of misunderstanding about recently published BIS statistics regarding US banks’ Greek exposures. The idea that US banks are on the hook for $32.7bn of CDS written on Greece was first expounded by the Street Light blog, then picked up by us (unfortunately), then spread to Fortune and onwards to markets. This week US bank stocks moved on the concern, apparently.

Anyway, the BIS statistics do not clearly lay out the amount of US banks’ Greek CDS exposure. Anyway, the BIS statistics do not clearly lay out the amount of US banks’ Greek CDS exposure...

If you want a summary version, here’s Nomura’s new banking analyst Glenn Schorr:

The U.S. banks and brokers sold off yesterday [Wednesday] over concerns around exposure to a potential Greek debt restructuring or default. The concern was driven by a BIS report published this month that shows that U.S. banks and brokers have written $32.7bn of credit guarantees (most likely CDS protection) on Greece. Unfortunately, this number is only half the story, as it shows gross protection written but does not show the net exposure of U.S. banks, meaning that any hedges or collateral that the U.S. banks may have in place is not being captured in this number. While there is definitely some Greek exposure in the U.S. system, we think net exposure at the large U.S. banks and brokers is a whole lot less than the $32.7bn.

While I agree that the BIS data does not tell us how much of the exposure is due to CDS contracts, I disagree that it can't be used to tell us the size of the US banking system's exposure to Greek debt. The whole point of the BIS statistics is to measure the ultimate risks that banks face, after all, and so they take great pains to ensure that the statistics reflect the net effective exposure faced by banks. As explained by the BIS in their guide to the international statistics (pdf):

The statistics mainly provide information on international financial claims of domestic bank head offices on a worldwide consolidated basis, ie including the exposures of own foreign offices but excluding inter-office positions. Currently they indicate the nature and extent of foreign claims of banks headquartered in 30 major financial centres. In contrast to the residence or balance of payments principle of the locational statistics, the reporting of consolidated positions offers a more useful measure of the total risk exposure of a reporting country’s banking system.

...Consolidated data based on the residence of the party ultimately responsible for the repayment of an obligation (ultimate risk basis) in addition to total claims based on the residence of the immediate borrower (contractual claims) are more compatible with information produced by banks’ own internal risk measurement systems and are considered a more appropriate measure of country risk exposure.

Regarding the specific question of whether the "guarantees" listed in the BIS data are net or gross exposures, the answer is more complicated than the one provided by the Nomura analyst above. This week I asked BIS for clarification on how they handle CDS data. The BIS has told me that when a bank writes a CDS contract selling default protection, the gross amount appears in the line "guarantees". So Schorr has that part of the story correct -- the guarantees are, in a sense, gross CDS exposures, because they are not offset by the protection purchased (if any) by the bank to hedge the risk it took by selling the protection.

However, that doesn't mean that the offsetting protection purchased by the bank is simply not counted. It will be counted either in the "derivatives contracts" line of the BIS statistics, or the "foreign claims" line of the BIS data. In either case, the protection purchased through ofsetting CDS contracts are counted in the BIS data, so that when you add together the direct + indirect exposures, you have netted out all countervailing CDS positions and have the final, true, net exposure faced by banks.

In other words, the BIS data does indeed represent true net exposure, and in the case of US banks that exposure is about $40 billion to Greece, on par with the exposure faced by banks in France and Germany. No, we don't know exactly how much of that takes the form of CDS contracts, but we do know that the exposure is there.