Believe those who are seeking the truth. Doubt those who find it. Andre Gide

Tuesday, August 9, 2011

A Bad Rap for the Bond Raters?

Who's going to listen to a company whose name translates to "average and below average"? Jon Stewart.

In my previous post (Wonderland), I asked what sort of evidence justifies making public sport out of the major bond rating agencies and their role in the recent financial crisis. The type of sentiment I question was repeated the other day by Paul Krugman here; I quote:
And S&P, along with its sister rating agencies, played a major role in causing that crisis, by giving AAA ratings to mortgage-backed assets that have since turned into toxic waste.
The first thing we have to ask, of course, is what does a AAA rating actually mean? The only thing most people know is that AAA is the highest rating that agencies attach to bonds. But that's an ordinal statement; it does not necessarily imply "absolutely free of risk." Apart from death and taxes, there are no perfect guarantees in life.

The second thing we have to ask is what sort of risk are the ratings trying to measure. In a nutshell, they measure the risk that the terms of a contract are not fulfilled. They do not measure the liquidity risk associated possession of the asset (the major problem during the financial crisis). Note: by liquidity risk, I mean the ease with with one can dispose of an asset (or use it as collateral in a loan) over a short period of time, without the asset being ridiculously discounted in the market.

Now, there seems to be no question that in the depths of the crisis, a lot of MBS was treated as if it were toxic (it was heavily discounted). But as I said above, bond ratings do not (I do not think) measure liquidity risk. They measure things like, well, did the MBS actually deliver on the interest and principal that was expected?

It surprising how difficult it is to find out just how much of the outstanding MBS actually did end up as toxic waste. Paul Krugman seems to think it was a lot. So do a lot of other people. But where is the data?

I decided to ask Gary Gorton, who knows more than most about these matters. Here is how he replied to me:
Of the notional principal amount of AAA/Aaa subprime bonds issued in the years 2006, 2007 and 2008 (which is almost $2 trillion), the realized principal loss as of Feb 2011 is 17 basis points – almost nothing, but much higher than AAA/Aaa other stuff.  Yes, I think the agencies are being treated unfairly by uninformed people.  There are many other facts that are similar that are also inconsistent with the popular narrative of the crisis. 
If this is true, it is indeed remarkable. The actual losses on these "toxic" products has been tiny. That is the type of risk that was being evaluated. (I might add that the losses on bank deposits during the great financial panics of the U.S. National Banking Era (1863-1913) were reportedly in the order of 50 basis points.)

Evidently, Gary has a PhD student working on this. I look forward to reading her findings (and reporting them here). Of course, if anyone out there has evidence relating to this question, I'd greatly appreciate hearing from you.

Update: August 10, 2011

Thanks to Jesse for this link to Bond Girl, who makes a lot of the same points (and more).

I also received this note from Don Brown (thanks, Don--I will investigate):

I cannot verify current loss=17bp figure, but it doesn't sound surprising to me.  HOWEVER, that misses the point.  There are significant losses to be taken on 05-07 vintage subprime, due to massive delinquencies.  It will be on the order of 12-15% of the original amount of AAA subprime securities that FNMA/FHLMC bought.  I encourage you to do your own original research to verify this, but as a quick back-of-the-envelope:

I know that agency portfolios were buying wide-window AAAs off subprime, mostly 06-07 vintage.  They bought a lot ($250B if I remember correctly).  Note: portfolios were the hedge funds that FN/FH were running, outside of their traditional business as mortgage guarantor.

Average original subordination was 26-30%.  So actual losses on subprime would have to be around this level to start showing losses.  Currently, actual losses aren't at this level, but it's easy to see that they will increase (barring a miracle).

To date, most subprime deals have taken losses in the 15-20% range (percentage of original balance).  They have serious delinquencies 40-60% of current balance (serious dlq defined as 90+ days dlq, FCL, or REO).
15% current losses leave you with 11-15% subordination (on average).  Currently, subprime loss severity is around 80% (upon liquidation of the house).  For a 50% dlq deal => 50%x80% = 40% additional losses from here (distributed over time).  That would imply actual losses of 25-29% on the current face of the class.  Average factor for these classes are 0.55.  Very roughly, this translates to 12-16% loss on the original investment.

For examples, take a look at OOMLT 07-5 1A1(cusip 68403HAA0) and CWL 06-26 1A (cusip 12668HAA8).  One of the agencies owns these classes (almost all subprime deals have a Group1 / Group2 structure.  Group1 was conforming balance subprime loans that went to the agencies).  You can find the remittance reports at BoNY web site or CTS.  They are publically available.

Update: August 14, 2011

Rebel Economist (see comments below) directs me to this interesting link: Is Blaming AAA Investors Wallstreet Serving PR? 


  1. Hi David,

    The credibility of credit rating agencies (CRAs) has been called into question in another interesting debate that recognizes their potential conflicting roles as informational intermediaries and holders of “regulatory licences” within the financial system. In the case of former, the question is whether CRAs provide unbiased estimates of default risk that improve upon the assessments that individuals can make using publicly available data. As financial intermediaries, CRAs are expected to do a decent job because of their exposure to reputational risk. At the same time, if CRAs are paid by issuers rather than investors, under the commonly adopted issuer-pay model, then their ability to objectively operate as informational intermediaries may be compromised. In the legal context, too, CRAs have often invoked their rights to “free speech” as conveyed in their rating assessments.

    Under the U.S., and more generally, global financial system, credit ratings have been incorporated by regulators and bank supervisors in determining the capital adequacy requirements of banks and other financial institutions. In such an environment, debt debtors may issue bonds to a wide class of regulated institutional investors at a relatively low cost (interest rate) if their securities are being “endorsed” by a favourable letter rating grade (i.e. CRAs provide “regulatory value”). And bond issuers have the incentive to shop around for the most favourable rating assignment in a largely oligopolistic rating industry. In the short term, CRAs may compete with each other for market share by simply providing “regulatory value” at the expense of “informational value” as long as the gains from the former exceed the expected losses associated with foregoing the latter.

    So, as far as this debate goes, the credibility of CRAs is called into question because of this unpleasant trade-off that CRAs must make between providing “regulatory” and “informational” value.

  2. Horatio, thanks for this.

    I wonder about the issuer-pay model. Why does it seem to work well elsewhere? For example, if I want to sell my car, I may want it inspected by a certified inspector (locally, the British Columbia Automobile Association).

    The trade-off you mention between regulatory and informational value is interesting. Of what empirical relevance is it, however? If what Gorton says above is true, then it seems that reputational concerns are paramount--the actual losses on AAA subprime MBS were trivial--not outlandishly inconsistent with the high rating they were accorded.

    I also wonder how freer entry into the business might affect things. Governments frequently prescribe the set of suitable raters (say, for assets to be held in a pension fund). Any thoughts?

  3. So who's to blame, a ratings agency with known blunt knives and conflicts of interest, or the people that download ratings tables into Excel, filter based on "AAA" and buy the basket?

    Sorry if I can't put all the onus on the ratings houses. A dog is a dog, don't make it into anything but.

  4. Here's what Bond Girl thinks of ratings agencies:

  5. Dear Dave,

    Thanks for getting back into the saddle.

    I guess there is good news and bad news in the data that you are beginning to adduce from Gary Gorton. 17 basis points on $2 trillion is a tiny fraction of the value of all houses in the United States. But Gorton himself says that this rate is a lot higher than for all other AAA/Aaa. So, at least by ex post standards, the ratings agencies did not do their jobs.

    Of course the ratings agencies did not rate the CDS's whose values were based on the mortgage backed securities.

    Assume that you own 10,000 mortgages. I take insurance out on the event that your assets go sour in the next year.

    I am not sure what happens now. If 17 of your mortgages are not performing, then the insurance agency AIG (for whom Gary Gorton consulted) might owe me the whole value of the insurance policy I bought. All of a sudden, a small loss in the value of your assets creates a liquidity crunch for the insurance company.

    How wrong is my simplistic conjecture? It seems it would have made sense for AIG to write contracts that paid out depending on what fraction of your mortgages became non-performing, not on the whole package that you held.

    I really enjoyed Lewis's The Big Short, but in the end that book posed more questions than it answered.


  6. I think Gary's talking about a small subset of Aaa junk. I'm told the ABSs died a horrible death.

  7. @Davd Backus: it appears you may be correct. See my update above.

    @Little Fish: Of course the ratings agencies did not rate the CDS's whose values were based on the mortgage backed securities. What are you suggesting by this statement? (Just curious. Oh, and I plan to read the Big Short soon--thanks!)

  8. I think you are asking just the right questions; what does a credit rating mean and how accurate have they proved ex-post. The same questions have occurred to me: so I shall keep watching your progress here.

    Contrary to popular/lazy/cynical conception, credit ratings are not really a measure of credit risk; they are a measure of risk of default (S&P) or expected loss though default (Moody's). In other words, if a triple-A security has a volatile value but does not actually default, don't blame the credit rating agencies! I always suspected that some structured securities (CPDOs?) were made to exploit this difference between perception and reality to the full.

    That said, I think that it it has been true that the credit losses on triple-A CDOs have been worse than their original rating implied: . You might find more comprehensive information in Gillian Tett's book.

    You might also find the following paper interesting:

    Note also that you can see from the S&P definition above why a downgrade of the US was appropriate. Irrespective of debt or deficit to GDP, some influential US politicians had just shown themselves to be willing to drive the US into default in the near term if that is what it takes to get their own way, and the debt ceiling issue was not pushed that far into the future.

  9. Rebeleconomist: "Note also that you can see from the S&P definition above why a downgrade of the US was appropriate. Irrespective of debt or deficit to GDP, some influential US politicians had just shown themselves to be willing to drive the US into default in the near term if that is what it takes to get their own way, and the debt ceiling issue was not pushed that far into the future."

    Under what possible circumstances would the US have defaulted on its debt payments due to the recent debt ceiling impasse?

    I think that the Treasury had the legal authority to determine priorities for paying bills and there was/is plenty of revenue -- many times what would be necessary -- to pay debt payments coming due.

    Now, there was/is not enough incoming revenue to continue Federal spending at current levels, so non-debt service spending would have to be cut by about 40 percent or so. (I am doing this from memory.)

    So who was willing to "drive the US into default?" A limited shutdown was certainly possible. Default? Not so much.

  10. @Chris,

    "Default? No so much."

    OK, sure. Not so much. What does that mean? Is the probability zero? Did the probability increase, even if it remained small? Note that US debt is still rated very highly; just a small notch below AAA. Not so much (prob of default), in other words. No?

  11. "What does that mean? Is the probability zero?"

    No. Not zero. Pretty darn low in the near term.

    "Did the probability increase, even if it remained small?"

    Yes, it is more likely that something would "go wrong" in the short term if the debt ceiling was not raised. There might a "trembling hand" outcome in which the Treasury would miscalculate and run out of money.

    My argument is that the "tea party" people doing the negotiating did not think that their actions made a default inevitable or even much more likely. In fact, they would probably argue that tough negotiating made default less likely in the long run.

    Rebeleconomist seemed to equate the willingness of the "tea party" to not raise the debt ceiling with a willingness to default. I believe that the "tea party" reps did not believe that default would result.

    (Though I cannot read their minds, I go by their public pronouncements.)

    My complaint is that there seems to be a new story going around that the "tea party" reps wanted or were willing to accept default and I don't think that this is true.

    This comment should not be considered an endorsement of either side's position. My own prefered outcome does not correspond neatly to either side's position. I just don't think that default was a real possibility, even in the event of a failure to raise the debt ceiling.

  12. "I just don't think that default was a real

    Well, it happened in 1979.

  13. @ankle biter.

    Read my post again. I did not say that the probability of default was zero -- or is ever zero -- I said that "it is more likely that something would "go wrong" in the short term if the debt ceiling was not raised. There might a "trembling hand" outcome in which the Treasury would miscalculate and run out of money."

    The 1979 incident to which you refer was precisely the sort of trembling hand outcome to which I refer. Thanks for bringing it up.

    By "not a real possibility," I meant that default was still very unlikely, definitely not synonymous with a failure to raise the debt ceiling.

    I should have chosen my words more carefully, knowing that there are always people who choose to misinterpret everything.

  14. The theory applies better to bills of exchange. Here an individual actually created a bill in exchange for some good, and that bill circulated onward upon endorsement of individual community members. But that's still pushing it. buy from china