This Bloomberg article about accounting differences between the US and Europe for derivative-y things comes down pretty squarely on the side of Europe, which is to be expected: European (well, IFRS) standards tend to gross up the size of bank balance sheets, compared to US GAAP standards. Grossing up bank balance sheets makes for bigger numbers and scarier banks, and “US banks are scarier than they seem” is more newsworthy than “European banks are less scary than they seem.” Also intuitively truer. As Bloomberg puts it:
U.S. accounting rules allow banks to record a smaller portion of their derivatives than European peers and keep most mortgage-linked bonds off their books. That can underestimate the risks firms face and affect how much capital they need.
Or it can overestimate the risk European firms face. Or any estimating of risks based on any measure of balance-sheet size is necessarily indeterminate. Risk happens tomorrow, not yesterday.
Anyway though some of these accounting differences are puzzling insofar as they are not accounting differences. Here is the mortgage bond one:
European banks sell covered bonds to finance mortgage originations and aren’t allowed under international accounting rules to move the home loans that back them off their balance sheets. Covered bonds package mortgages like securitizations, and the bonds are sold to investors. In case of bankruptcy, the mortgages that back the covered bonds are walled off from other assets of the bank and can be seized by bondholders.
Buyers of the bonds can demand that banks replace soured mortgages with performing ones, leaving the credit risk with the originator. That’s similar to buyback requests in the U.S. Executives at U.S. banks disagree, saying the securitizations pass mortgage-default risk to the government and investors, while covered bonds don’t.
Now executives at U.S. banks say that because it is true. RMBS securitizations actually do pass default risk on to the buyers (and/or GSE guarantors). You can tell by the fact that buyers and the GSEs lost fuckloads of money on defaults on mortgages backing their mortgage bonds.
This is not a difference in accounting standards; it is a difference in actual instruments and market practices. A covered bond is a recourse obligation of a (European) bank that is collateralized by mortgages (or whatever); the buyer’s cash flows are guaranteed both by the mortgage pool and the bank. A mortgage securitization is a nonrecourse obligation of a shell company that is collateralized by mortgages; the buyer’s cash flows are guaranteed only by the mortgage pool. Credit risk – the risk that a loan will default – really is transferred by a securitization, and really is not transferred by a covered bond. The different facts lead to different accounting, which accurately reflects the facts.
I mean, sort of. Of course there is some truth to Bloomberg’s claim that mortgage putback requests on RMBS securitizations are similar to the recourse nature of covered bonds. Putbacks are not about credit risk; they are about … I was going to say “fraud” but the right term is “violation of reps and warranties.” When a bank sells a mortgage into a securitization, it makes a bunch of promises about the underwriting of the loan, the borrower’s income, the LTV, and so forth. If those representations are wrong, the securitization’s buyers can demand that the bank buy back the loan. In the limit case, if every defaulted mortgage in a securitization also violated its reps and warranties, then the bank ends up retaining 100% of the credit risk, just as it would in a covered bond.
That limit case is pretty close to being true. Bank of America, arguably the fraudiest mortgage originator due to its purchase of Countrywide, has basically infinity billion dollars of putback liability, in part because Countrywide originated over $2 trillion of mortgages that were securitized in the run-up to the financial crisis, and in part because BofA is being sued multiple times over many of those mortgages.1
Should that be reflected on the balance sheet? To some extent it is – the big banks have somewhat nebulously disclosed but large reserves for potential mortgage putback claims – but that is not the same as keeping the entirety of your mortgage securitizations on your balance sheet because they might turn out to be 100% fraudulent and you might end up being on the hook for all of them. That seems pretty drastic. There is an important difference between contractually remaining on the hook for credit risk, which seems like it should be on your balance sheet, and legally remaining liable for your own fraud and misstatements. You’re always liable for your own fraud and misstatements, but you rarely capitalize that liability and put it on your balance sheet. In expectation, it’s supposed to be zero! You’re not supposed to commit fraud!
Is that a hopelessly naïve view? One thing you might ask is why European practice demands that banks remain on the hook for the mortgage bonds they sell, while American practice does not, except in the looser form of reps-and-warranties liability. The answer, I suspect, has a lot to do with structural things like the GSE guarantee program (and SEC disclosure requirements for covered bonds). But I also suspect that the practical results of this difference are more important than the accounting results.
The other accounting difference that Bloomberg mentions, which really is an accounting difference, is the divergence in how GAAP and IFRS treat derivative netting. I am a patriotic American who pledges allegiance to GAAP every morning, so I might be biased, but I can’t escape the fact that reflecting net, not gross, positions for derivatives subject to an enforceable master netting agreement is right, and not doing that is wrong, which is a good argument for doing it the American way.2 European banks agree with me, though they would:
While European banks have long favored the U.S. approach to netting derivatives, they haven’t pushed for change because it wouldn’t have an impact on income statements or capital requirements, which are based on risk-weighting of assets, according to Andrew Spooner, a London-based partner at Deloitte LLP.
Notice there that Basel risk-weighted assets are grossed down for enforceable master netting agreements: the numbers used by international regulators to make actual decisions about risk and capital requirements follow the American approach. The IFRS, or wrong, approach is used only to confuse European bank investors, not to actually measure risks. By that standard, European bank risks really are overstated by gross balance sheet size.3
On the other hand. If you start from a place where you expect a bank’s balance sheet to reflect all of its myriad potential frauds and misstatements, then maybe you’d come to a different place. “Derivatives” are mostly interest-rate swaps.4 Interest-rate swaps are mostly based on Libor. Libor was mostly manipulated. One unfortunate feature of Libor manipulation is that, if you are a Libor-manipulating bank, you stand a decent chance of eventually owing money to any counterparty whose position was hurt by your manipulation, but very little chance of being able to claim money back from any counterparty whose position was helped. Your netting breaks down a bit.5
The U.S. accounting approach essentially relies on contracts – whether for mortgage bonds or for derivatives netting – being respected for what they say. As an accounting matter, this seems more elegant than the European approach. But it does sort of rely on the contracts being right and fraud being the exception rather than the rule. The European approach, whatever else you can say against it, may end up better reflecting economic reality.
1. Much of the history of financial innovation is about squeezing the distribution while fattening the tails: reducing day-to-day volatility and risk, and reducing the probability of loss, while making the less-likely loss far more horrible. From the investors’ perspective, mortgage securitization is a classic example, as mortgages ended up in AAA securities that were almost certain to pay off at par – but, when they didn’t, they all didn’t at once, and the world ended.
But the theory also works on the sellers’ side. Selling covered bonds leaves you liable for a predictable amount of the credit risk in the underlying mortgages: 100%. Selling securitizations leaves you liable, in expectation, for a much lower amount of credit risk: in theory, if your documentation is right and your underwriting is careful and your reps and warranties are true, you’re liable for zero percent of the credit risk. So your risk of losing money on a few mortgage defaults here and there is drastically reduced.
That becomes a little too good to be true, for moral-hazard reasons: if you have no credit risk, why do good underwriting? Why not stretch a bit on the documentation? And so you get things like the recent study finding that ~9% of securitized mortgages contained misrepresentations about either owner-occupancy status or junior liens.
And then people sue you multiple times on every loan, and you end up settling some of those suits for 105 cents on the dollar, and your exposure to the credit risk of the mortgages you securitized is uncertain but seems like it might exceed 100%. Then again sometimes you settle for 0.6 cents on the dollar. On balance it is safe to say that the U.S. banks that did mortgage securitizations will not eat the majority of the default losses, though I guess it’s still too early to tell.
Netting allows banks and trading partners to add up the positions they have with each other and show what would be owed if all contracts had to be settled suddenly. These master agreements are only relevant during bankruptcy and underestimate risk, according to Anat Admati, a finance professor at Stanford University. When a bank’s solvency is in doubt, derivatives partners demand to be paid immediately, causing a run.
There is some sociological truth to this – if your counterparty calls you up and asks to close out all his asset positions while keeping on his liability positions, and you are in trouble, you might be tempted to say yes to keep his business. But there is not a ton of legal truth to it; if anything banks tend to have more robust early termination rights than their clients do.
4. Meh. “Interest rate contracts comprise 80% of total derivatives,” and “Swap contracts continue to represent the bulk of the derivatives market at $136 trillion (60%),” but that’s by notionals. Fair-value or revenue measures would make it closer. Whatever.
5. I mean, presumably you could still net within one client? Maybe? This is all pretty early stages of speculation.