Bloomberg this week had an article about how bespoke synthetic CDOs are coming back in vogue, and various people have fretted about that, because synthetic CDOs are scary, financial crisis, etc. And, sure, it’s certainly possible that the next financial crisis will be exactly like the last, only with more Cyprus.1 But today let’s talk about something tangentially related.
If you require banks to have capital based on risk-weighted assets, and if capital is expensive (at least for bankers), then you’ll have banks who want to lower the risk weights of their assets. There are many ways to do this, including buying safer assets, selling riskier assets, monkeying with models, etc., but one popular way is to buy credit protection against risky assets. The reason that this is popular is because of regulatory discontinuities: if you have $100 worth of stuff with a 200% risk weight, then you have $200 of risk-weighted assets, but if you buy protection against the riskiest $10 of it then you might go from $200 of risk-weighted assets all the way to $6.30, because the safest $90 of it might have only a 7% risk weight.
That’s a big jump. If your aim is to have capital equal to 12% of your risk-weighted assets, then your capital requirements go from $24 to like 75 cents. If your cost of capital is 10%, then that jump saves you $2.32 a year. So you could pay, say, $2 a year to the protection provider and still be up a few cents, versus not buying credit protection – plus, of course, you’ve got credit protection (meaning that you get more money back if there are defaults). And if you pay $2 a year for five years to protect $10 worth of risk, then the protection provider should do that trade all day long: he’s getting paid $10 to take $10 of risk. At worst – if 10% of your stuff, or for that matter all of your stuff, defaults – he breaks even. It’s free money.
That’s oversimplified (time value, counterparty risk, whatever), but it’s kind of a thing. To some extent that thinking underlies things like the glorious Credit Suisse PAF2 trade, where Credit Suisse basically wrote credit protection to itself because doing so saved it so much on risk-weighted assets. But the folks on the Basel Committee on Banking Supervision don’t particularly like it, and so they released a document today yelling at banks about it. It’s worth distinguishing why the Basel people don’t like it – because it is an obvious gaming of the capital rules and leans on the gameability/weakness/arbitrariness/discontinuity of those rules – from why they say they don’t like it, which is that they want “to ensure that the costs, and not just the benefits, of purchased credit protection are appropriately recognised in regulatory capital.”2
The way they ensure that the costs are recognized is by saying: if you do something like that, you basically need to deduct from your capital the entire future cost of that protection, day one.3 That should eliminate the free-money option, and make buying this sort of protection more punitive.
Still not very punitive. Basel provides an example (“Transaction D”) where you buy protection on the first-loss 10% of your position, paying 15% of that tranche per year for 5 years. That is, you pay 7.5% of the total initial notional of the position over five years, without regard to whether there are defaults along the way; your counterparty can make a maximum of 7.5% (if there are no defaults) and lose a maximum of 2.5% (if >=10% of the portfolio defaults). This is not quite as silly as paying 10% to get a maximum of 10% of protection, but it’s still probably a decent trade for the counterparty, depending on what the stuff is.4
Here’s some math on that example:
If you assume – plausibly? whatever – that banks (1) have to have capital equal to at least 10.5% of their risk-weighted assets5 and (2) have a cost of “capital” of 10%,6 then that tranche CDS that pays the counterparty 1.5 units (dozen bitcoins, million dollars, trillion Cypriot rubles, whatever) a year actually costs the bank only 0.39, after you take into account the capital savings. In other words capital regulation “pays for” three-quarters of the cost of the credit protection. That’s after Basel fixed the regulatory arbitrage today, mind you.
By the way you can change that 15%-of-tranche-notional annual price to 20% of tranche notional (2 units a year) and it still only costs the bank 1.2 units a year – the trade is still 40% “subsidized.” The bank effectively pays 6% of assets, the counterparty receives 10% of assets, and the counterparty protects the bank from losses on the first 10% of assets. Great trade for the counterparty, and far less pointless for the bank than it first appears.
If you are a bank you probably should be buying lots of this stuff – and what is “this stuff” if not “bespoke synthetic CDOs”? – because you only pay part of the cost. So feel free to overpay a bit: it’s not your money. If you are a shadow-bank-y thing – a hedge fund, a pension fund, a Blackstone, an AllianceBernstein – you should, similarly, sell a lot of it, since banks are happy to overpay for it and you should be happy to get overpaid.
Oh and yields elsewhere are low blah blah blah:
Synthetic credit, which amplified the financial crisis five years ago, is enticing investors after corporate-bond yields dropped to less than half the 20-year average. By betting on the degree to which a group of companies will default, a CDO may pay relative yields of more than 5 percentage points, four times that of a typical credit-swaps transaction on similar debt.
“That’s a valid strategy for this part of the credit cycle: Don’t stretch on credit quality, but rather leverage your exposure to better-quality credit,” Ashish Shah, the head of global credit investment at New York-based AllianceBernstein LP, which oversees $256 billion in fixed-income assets, said in a telephone interview.
I mean, what do I know about what Ashish Shah is doing? Per Bloomberg, the boom in synthetic CDOs seems to include lots of mezz-y tranches (which don’t work that well for Basel arbitrage), lots of investment-grade portfolios (which attract lower capital requirements to begin with and so are less likely to profit from the above strategy), and small broadly offered trades done off corr desks rather than block trades done by the people in charge of capital needs. So I doubt that Ashish Shah at AllianceBernstein is literally in the business of writing capital-optimization trades.
Still: if banks can pay uneconomically high yields on some synthetic tranches, because those yields are in effect subsidized by capital regulation, that might have a knock-on effect to synthetic CDO yields more broadly, and help spark the comeback in the asset class. The externalities can help make the market: the regulatory subsidy means that the protection-buying bank wins, the protection-selling fund wins, and the spirit of capital regulation loses.
Which means, if you mostly believe in the bank-capital-regulatory mission, that the losers are whoever suffers when banks are inadequately capitalized. That is a somewhat puzzling class to identify these days – depositors? creditors? governments? But it’s reasonable to guess that, whoever they are, they’re more broadly sympathetic than banks and hedge funds are.
Synthetic CDOs Making Comeback as Yields Juiced [Bloomberg]
1. Though Felix Salmon makes an excellent point: “Why does Citigroup, of all banks, have Wall Street’s largest correlation book? And does anybody think that Citigroup has the risk-management chops to ensure that it doesn’t blow up, a la the London whale?”
2. That’s not fair – they also, and prominently, mention concerns about “potential regulatory capital arbitrage.” But that’s harder to fix: the discontinuities are there, and are hard to avoid in any risk-weighted system.
3. More accurately: calculate the entire present value of the premiums you’ll have to pay, and risk-weight those at 1250%. At an 8% capital requirement, 1250% risk-weighting means that you need capital equal to the amount of the exposure.
4. It’s 200% risk-weighted stuff, which might include for instance BB rated bonds, which have about a 1.5% annual default rate recently (and a ~0.9% long-term average). So getting 1.5% a year to insure a pool of BB bonds against default isn’t nuts.
5. For big banks a 10.5% total capital plus capital conservation buffer seems to be the Basel standard. This includes non-Tier 1 preferred-y things. Obviously banks prefer not to run right at the line either.
6. I mean, like, whatever. I will quote a favorite line of mine: “To simply cover its debt expenses and other capital costs, Morgan Stanley must achieve a return on equity closer to 10 percent.” Make of that what you will. Incidentally I’m sort of implicitly assuming a zero cost of non-equity funding here (or, I guess, a 10% spread between cost of equity and cost of debt), but whatever. Rates are low, TBTF banks are subsidized, etc. etc.