Felix Salmon

Paul Krugman has a reasonably good overview

of the mess in the CDO market today (free version here).

But he goes much too far when he tries to impress upon us how big the problem

is:

With the collapse of the $800 billion market in bonds backed by subprime

mortgages — the price of a basket of these bonds has lost almost 40

percent of its value since January — it’s now clear that many

investors who bought these securities didn’t realize what they were

getting into.

This is simply not true. The thing which has lost 40% of its value since January

is the ABX.HE index of BBB-rated subprime credit default swaps. Which is nowhere

near being a basket of subprime-backed bonds. For one thing, it measures not

bond values but CDS prices. Maybe that’s a niggle. But much more importantly,

it measures the value only of the riskiest tranches of the CDOs that Krugman

is so worried about.

As with most tranched securities, CDOs are structured so that the riskiest

portions take the first loss, the next-riskiest portions take the second loss,

and so forth. The bits of CDOs which have dropped in value by 40% since January

are precisely these riskiest portions, which account for maybe 5% of the total

amount of subprime-backed debt outstanding. The vast majority of CDO investments

are triple-A securities which haven’t dropped in value by anything near 40%.

Maybe a few points, tops.

As for Krugman’s bigger argument, he, like James

Hamilton, sees a clean and direct causal relationship between falling house

prices, on the one hand, and falling CDO valuations, on the other. I’m willing

to believe that relationship is there, but so far I have to take it on faith,

because no one has done a good job of explaining how this relatinship has worked

in practice. I can see how falling property prices lead to higher default rates

on subprime loans. But no one has done a good job of explaining how high subprime

default rates feed through into big losses for MBSs in general and CDOs in particular.

People seem to make that step without really thinking about it – and I

suspect that what happens in reality is actually rather more nuanced.

Bloomberg’s Mark Pittman isn’t hedging his bets in a story

headlined "S&P,

Moody’s Mask $200 Billion of Subprime Bond Risk":

Standard & Poor’s, Moody’s Investors Service and Fitch Ratings are masking

burgeoning losses in the market for subprime mortgage bonds by failing to

cut the credit ratings on about $200 billion of securities backed by home

loans.

There’s no attribution to anybody else – these are Pittman’s own numbers,

it would seem.

Almost 65 percent of the bonds in indexes that track subprime mortgage debt

don’t meet the ratings criteria in place when they were sold, according to

data compiled by Bloomberg.

But rather than immediately explain what he means, Pittman then embarks upon

2,000 words of throat-clearing before explaining how he got his numbers. Eventually

he does tell us: he’s basing everything on an old S&P criterion for determining

how well insulated a bond is against default.

S&P abandoned seven-year-old criteria for determining a bond’s protection

against default in February.

Under the old guidelines, S&P said a bond’s “credit support” must be

twice the rolling 90-day average of the sum of value of mortgages delinquent

by three months or in foreclosure plus real estate that has been seized by

the lender.

Credit support for a bond is determined by looking at the number of lower-rated

securities that would have to go bust before it suffered losses, the dollar

amount of mortgages available to pay back the interest and the annualized

interest the mortgages generate in excess of what needs to be paid to bondholders.

The measure was one of four tests used by S&P, said Chris Atkins, a spokesman

for the company, a unit of New York-based McGraw-Hill Cos. A failure to meet

the credit support standard wouldn’t have automatically resulted in a downgrade,

he said.

Pittman applied the now-defunct criterion to $300 billion of mortgage-backed

bonds, and concluded that $200 billion of them didn’t meet it – that’s

where his headline comes from.

Ratings agencies, quite rightly, don’t simply plug a bunch of ratios into a

formula to generate a rating. If they did, they’d add precious little value.

Instead, they have a dynamically-changing set of criteria which they use to

help them make an informed decision as to what their ratings should be.

It seems that between 2000 and early 2007, one of S&P’s four criteria for

rating mortgage-backed bonds was a ratio between that bond’s credit support

and the quantity of impaired loans. In February, S&P decided that wasn’t

a particularly useful a criterion to use, but if it had continued to

use it, then it would have discovered that a lot of mortgage-backed loans would

no longer meet it. Which wouldn’t mean that those bonds should necessarily be

downgraded, of course, but it might be a red flag.

How that means that S&P, let alone Moody’s, is "masking bond risk,"

I have no idea.

In fact, I’m far from clear what the criterion really shows, since Pittman’s

explanation of how it works is very hard to understand. For one thing, the criterion

doesn’t seem to change at all depending on the rating desired, which is weird:

the criterion for an AA-rated bond can’t be the same as that for a BBB-rated

bond.

Pittman goes on to sum up the criterion when he talks about loans which "contain

so many defaulted loans that the credit support is outweighed by potential losses."

But the does the criterion really measure potential losses? Only, it would seem,

if you consider the potential losses on a delinquent mortgage to be the entire

value of the mortgage. Which is crazy: home prices might have fallen, but they

haven’t fallen to zero. And in any case, the criterion said that the credit

support had to be twice the potential losses, using total delinquent

loans as a proxy for potential losses.

One can see why S&P changed its criterion. For one thing, it used this

criterion to rate bonds as they were being issued: that is, when the

underlying mortgages were very young indeed. Over time, any pool of mortgage-backed

bonds is going to see the number of delinquencies in it go up from where it

was at the very beginning. That’s only natural, and shouldn’t be cause for a

downgrade. Somewhere in the criterion, one would think, would be some adjustment

for how old the pool of loans in question is. And if there wasn’t, then the

individual making the rating would in any case take that into account when working

out what the credit ratings should be.

But Pittman did none of that. Instead, he took a single, now-defunct criterion,

applied it to a pool of loans, and came up with a highly contentious headline.

Then, he buried his argument at the bottom of an extremely long article.

The whole thing made

Tanta’s brain explode, which is bad enough in itself. But it also looks

very much as though Pittman is trying to play "gotcha" with the ratings

agencies. And that really isn’t helpful.

The

Epicurean Dealmaker explains why I’m a bit too sanguine about the prospects

for the CDO market: it all comes down to margin calls. All it takes is one CDO

getting liquidated at prices substantially below market, he says, and "all

hell breaks loose":

First, the prime brokers reset their hedge fund clients’ margin requirements

to match the current lower market prices. Bang! First margin call. Second

(or simultaneously), the bank resets the input expected volatility in its

VAR model to reflect the new, more volatile behavior of the securities. Bang!

Second margin call. All of a sudden, a hedge fund that was sitting fat and

happy with a portfolio of nicely behaved CDOs on its books is looking at a

huge margin call and usually no way to meet it without liquidating securities.

Oops. Fire sale, look out below.

All this is a definite possibility, and I’d never say that this kind of vicious

cycle can’t happen. It can, and it might. But then again, it might not.

Why am I not convinced? Four reasons. Firstly, I’ve seen what’s

happened in practice with Bear Stearns’ High-Grade Structured Credit Enhanced

Leverage Fund. We still don’t know the endgame to that story, but we do know

that $7 billion of leverage got brought down pretty painlessly to $1.2 billion

in leverage without any fire sales and without any bailout from Bear itself.

Clearly, there are non-fire-sale solutions to these kind of problems.

Secondly, there are two very different types of leverage employed by hedge

funds and other owners of CDOs. One is the kind of prime-broker leverage that

our anonymous dealmaking blogger is referring to. And then there’s the kind

of leverage that hedge funds and other investors find in the wonderful world

of credit derivatives. Many CDOs and hedge funds are loaded up with CDSs: credit

default swaps which bring in a certain income unless and until underlying credits

start defaulting. And when hedge funds write credit protection, they don’t have

to put up margin in the same way they do when they borrow cash money. For that

reason, CDSs are a popular way of leveraging your exposure to a certain credit,

and they also don’t suffer from the same margin-meltdown risk that normal loans

carry.

Thirdly, I think that the move from public and transparent markets to private

and opaque markets is more than a blip. CDOs are in the middle: they’re public

and opaque. Where do they move from here? One possibility is that they snap

chaotically back to being public and transparent. But the other possibility

is that they move in the other direction, and become private and opaque: that

would involve being snapped up by pools of private capital which don’t mark

to market and which can invest with a long time horizon. Those pools are already

being formed, and they could prove to be very popular.

Finally, I wonder whether banks are quite as rule-bound as the epicurean dealmaker

imagines. Take another look at the attempted liquidations of Bear Stearns’ funds’

assets. Prime brokers seized, them, put them out to bid – and then, in

the end, didn’t sell them. Why not? If the market price is ever and always superior

to the model price, then the prime brokers should just accept the best bid,

and sell as much as they can at that level. But mark-to-market is not always

the best way of valuing something as complex and illiquid as a CDO. If trillions

of dollars of CDOs have been sold in the primary market, does it really make

sense to revalue them all on the basis of two or three small trades in the secondary

market? Besides, all CDOs are different, and the vast majority of them don’t

trade at all, which means that there’s literally no market for them to mark

to. I suspect that if marking CDOs to market becomes a real problem in the prime

brokerage community, then they might end up not marking CDOs to market.

In the 1980s, banks around the world sat on hundreds of billions of dollars’

worth of defaulted sovereign debt, while carefully making sure never to mark

it to market. Banks are good like that: they know when to recognise reality

and when not to. If the A-rated tranches of CDOs start defaulting unexpectedly,

then at that point it makes sense to mark down their prices. But at the moment

default rates are still incredibly low, which should give banks and investors

the ability to muddle through, somehow.