I thought it might be helpful to summarize some of the background on how we got into our present mortgage mess.
Once upon a time, when you needed to borrow money to buy a house, you went to a bank (in the old days it was often called a “savings and loan”). The bank gave you the money you needed to purchase the house (indicated by the green arrow in the figure below), and you promised to repay the bank, with interest, over a certain length of time (red arrow).
Today, it is quite uncommon for the bank that originally made the loan to you to be the institution that actually receives your interest payments. Instead, those payments are likely going into a pool from which mortgage-backed securities were created by the government sponsored enterprises Fannie Mae or Freddie Mac, or, particularly in the last few years, arranged by large private institutions such as Countrywide Financial or Wells Fargo (the light blue “asset-backed securities issuers” in the graph below).
 |
This process of securitization of household mortgages was associated with a tremendous increase in the total volume of U.S. mortgage debt, which grew three times as fast as GDP over the last decade.
 |
I’ve created a schematic figure below to summarize the major flows that characterize the process whereby your monthly mortgage payments get turned into a security. The money with which you paid for your house came from a “mortgage originator”– ours came from Norwest Mortgage. The originator took your promise to make monthly payments and sold the rights to receive those payments to an “arranger”– probably Fannie or Freddie, if your mortgage qualified as one GSEs would buy, or to a private arranger if not.
But the arranger wasn’t interested in hanging on to the loan, either. In the case of many private arrangers, they in turn set up a separate legal entity (or “trust”) to which they sold the loan. And where did the trust get the money to pay the arranger? That money came from investors in the trust, who could have been anybody– pension funds, banks, or general investment funds. As a technical aside, these investors often let a separate fund manager make the actual decision of where their dollars got invested, a finer point that will make an appearance shortly when we get into what went wrong with this system. But before getting into any further details, you can get the big picture by following the money (green arrows) in the diagram below. When all is said and done, the cash you delivered to the seller of the house ultimately came from an investor at the far end of the chain.
In return, the trust is making payments to the investors not just out of the mortgage payments you make, but pools them with a large number of other borrowers. A given pool of mortgages was divided up into “tranches”. The way in which this is done can be fairly complicated, but the basic idea is pretty simple. Each tranche would make specified payments to investors over time according to a certain schedule, with every tranche meeting all its payments if all of the original mortgage borrowers make their payments on schedule. If some households in the pool default on their mortgage payment, the trust would be unable to make the full payments on all of the securities, and any shortfalls would be borne by the most junior tranches. For example, if the mortgages end up collecting 90% of the payments promised by borrowers, then the buyers of the securities in the top 90% of the tranches would receive 100% of what they were promised and those in the bottom tranche would get nothing.
Federal Reserve Bank of New York economists Adam Ashcraft and Til Schuermann have a very interesting new paper (hat tip: CR) in which they describe this process and what went wrong. Among other contributions, the paper investigates details of the securitization of a pool of about 4,000 subprime mortgage loans whose principal value came to a little under $900 million and which were originated by New Century Financial in the second quarter of 2006, a small part of the $51.6 billion in loans that the company originated in 2006 before declaring bankruptcy in early 2007.
A striking feature of this pool of loans is the magnitude of the increase in monthly payments to which borrowers were agreeing even if there had been no change in the LIBOR rates to which the “adjustable rate” mortgages were keyed. This increase would result from the 2/28 or 3/27 “teaser rate” feature of the vast majority of these mortgage contracts, according to which the borrower would be virtually certain to need to make a huge increase in the monthly payments within two or three years. Ashcraft and Schuermann calculate that the monthly payments that the recipient of the loan is supposed to pay were scheduled to increase by 26-45% (depending on other details) within 2-1/2 years of the loan being issued, even if LIBOR rates held steady at their values at the time the loan was originated, and by which time the total principal owed would have increased substantially relative to the sum that had originally been borrowed. One has to wonder what circumstances one would be counting on to expect such payments to be made on schedule from a pool of borrowers with a history of other credit problems.
A second remarkable feature of this pool is the high credit rating assigned to all but the most junior tranches. Out of the $881 million in original mortgage loans, there were created $699 million (or 79% of the total) in “senior-tranche” mortgage-backed securities that received the highest possible credit rating (AAA from Standard & Poor’s or Aaa from Moody’s). Only $58 million (or 6-1/2% of the total) received a rating as low as BBB or Baa. There is no reason to believe this is unrepresentative of the nearly half trillion dollars in subprime mortgages that were securitized in the U.S. in 2006.
It’s now clear to everybody that most of these loans should never have been made at the terms that they were, and that a good deal of money is going to be lost by a good number of people. As the multiple arrows in the above diagram attest, there were plenty of individuals who could (and did) make some serious mistakes in this whole process. Ashcraft and Schuermann catalog these and inquire how we might prevent these problems in the future. At the top of their list of “informational frictions” that contributed to the subprime debacle is the one that so far has received the most attention from the media and legislators, namely that between the originator and the borrower. To the extent that the originators just resold the loans before the problems came home to roost, the originator had an incentive to misrepresent overly complex instruments to financially unsophisticated borrowers. The authors’ proposed resolution of this problem is “federal, state, and local laws prohibiting certain lending practices, as
well as the recent regulatory guidance on subprime lending”.
Second on their list of the most important frictions is one we have long been emphasizing here, namely that between the investor and the fund manager. To the extent that fund managers are evaluated on the basis of recent performance subject only to ratings guidelines, there is an incentive for managers to invest the funds in the riskiest vehicles that somehow manage to get a AAA rating. Ashcraft and Schuermann recommend that the investment mandates of any managed funds be rewritten to note the distinction between the ratings on corporate debt and those on artificially structured securities.
Ashford and Schuermann discuss a great many other informational frictions in the whole process that have also been widely discussed elsewhere, including inadequate equity stakes on the part of the originator and arranger, and need for different guidelines and procedures for the rating process. The authors nevertheless note:
We suggest some improvements to the existing process, though it is not clear that any
additional regulation is warranted as the market is already taking remedial steps in the
right direction.
Still, it’s useful to have Ashcraft and Schuermann’s careful summary of exactly what wrong, perhaps not so much in order to tell us to close the barn door as to understand just how all those cows got of the barn.
I’m curious about all those arrows in your second, much more complicated, picture. Is it naive to assume that each of those is a transaction with associated transaction costs? It seems like there are so many more middlemen than in the traditional model that the associated fees must be much higher. Is this actually true? If so, who’s incurring the cost? The borrowers or the investors in the mortgage based securities?
That’s an interesting point, ramster. Ex post I think we’d say that all those costs ended up being paid by the investors. Here’s how the details shape up. As an investor, you know that the fund manager is taking some of your yield, but you hope he or she is earning that by exercising better judgment than you would on your own. The fund manager knows that the arranger is taking some of the interest payments, but thinks it’s worth it because by slicing these into tranches the fund manager gets to buy AAA-rated junk. The arranger of course is paying something to the originator, but thinks it’s worth it because creating the trust was so profitable (at least, it was while the game was still on).
But with hindsight we would have to say that not only were each of these costs not worth it to the respective buyers, but there was another “transaction” cost resulting from the whole system that dwarfs them all, which is the fact that some very unwise investments were made as a result of the informational frictions that piled up from these layers of transactions.
Was this AAA rated junk getting sold for the exact same price as real AAA stuff (government bonds or something like that)? I.e. did the purchasers really think it was AAA or was there some degree of discounting given the associated risk. Did people really buy the con that these sliced and diced tranches of someone’s no proof of income ARM mortgage was really AAA material?
You have a graph of mortgage debt to GDP. In his last post, Menzie added a graph of consumer debt to GDP that looks VERY similar. Of most interest, it seems like the slope of the graph increase significantly in ’98 or ’99, on both graphs.
Any chance that you could graph say, the motgage debt % vs. the consumer debt %? Is it a 45% line (perfect correlation) and thus reacting to the same factors?
Would that factor be much faster monetary growth, first in reaction to Y2k, then 9/11, then ???
Serendipitous that this post should come the same day the BOA buys Countrywide. There are those cynics that think maybe this was deal that was not made in heaven but rather in our nation’s capital. What think thee?
Understanding the Securitization of Subprime Mortgage Credit
A post today in Econbrowswer (which in turn was tipped by Calculated Risk) alerted me to a new Fed study: Understanding the Securitization of Subprime Mortgage Credit, which on first examination looks excellent.
I will admit that I have not thoroughly …
The informational friction hypothesis is not plausible. The yields on these deals were high enough that fund investors would not have noticed an added expense of, say, 0.5bp. On a $500M deal, 0.5bp is $25k, which will easily pay for an accountant to visit the seller and spend a few hours seeing if the underlying assets pass the giggle test. The only hard part is finding a mentally-stable accountant willing to spend half their life on an airplane.
And do not tell me in shocked tones that “that just isn’t the way things are done in Big Finance”. That is my point. The information is not a trade secret and not difficult to organize. The fact that it was neither made available nor asked for means that the relationship was not business to business, it was con artist to mark. Any proposed solutions have to take that into account. You can’t keep people from buying the Brooklyn Bridge with any amount of bridge rating reforms.
I’m interested in the phrase “AAA-rated junk”.
According to Fitch’s recent report, “US RMBS and SF CDO Rating Actions”, dated January 10, 2008, 2007 had a total of 19,255 AAA rated securites with a value of $917-billion.
66 of these securities have been downgraded but are still investment grade (according to Fitch), with a value of about $8-billion. An additional 2 of these securities, worth $0.2-billion, have been downgraded to junk status.
The big damage – according to Fitch’s assessment of credit quality – has been done to the mezzanine tranches of RMBS and CDOs (the latter are heavily dependent upon the former).
When the phrase “AAA-rated junk” is used, does “junk” refer to credit quality of the asset described, or its liquidity, or its market price? These are three very different attributes.
Thank of you for getting this sort of information out there. It’s annoying how little the media, especially sources that tend to be more in depth like the Economist, have failed to get some of this basic information out to the public.
I’d be curious as to why we saw a bubble in the mortgage market and mortage-backed bonds but yet the essentially the same system exists for other assests and asset backed bonds. Is it that their aren’t quasi-governmental companies involved in ASBs? The paper appreciation of housing that we saw that just didn’t and isn’t likely to happen with autos, boats and such?
And how is it investors didn’t catch on to what was happening? Were the guarantees that banks were offering were enough to reduce the risk enough?
“For example, if the mortgages end up collecting 90% of the payments promised by borrowers, then the buyers of the securities in the top 90% of the tranches would receive 100% of what they were promised and those in the bottom tranche would get nothing.”
Wild claim, not supported by the paper you are quoting, this is only true if defaults concretize the day after the securitized package is launched.
In a typical subprime securitization it takes on average 6.5% [of original balance] realized losses before the lower junior tranches start to lose money .Mortgage securitizations have vast amount of overcollateralization and earn vast amount excess spread interest,and while you wait for defaults to concretize, you get paid, since those hit first the excess spread and the overcollateralization account.The credit enhancement features for subprime securitizations is good enough that to date none the ABX tranches including the BBB- has experienced a single cent of principal write-downs and only minor interest shortfalls.
Here is what the paper you mentionned says page 40 about the riskiest tranche:
“The face value of the Class X tranche is $12.3 million. To date, this tranche has been paid
excess spread in the amount of $16.1 million. Note that the amount paid to this tranche has
decreased over time as credit losses have reduced excess spread. Interestingly, even if the
owners of this class are not paid another dollar of interest, they will have a return of 30.9% in
just over one year with their principal returned.”
So much for getting nothing…
Let the Wall Streeters take their haircut. I am very disturbed by the tendancy towards bailouts. This is such a neoliberal tendancy. Having Bush in power (and Republicans interested in controlling Congress) is WORSE than if the Democrats were in power. Becuase the REpukes end up pushing for more liberal policy to get elected and allow the Democraps to move further left.
Don’t forget that Rubin wanted to bail out Enron. That the Goldman/Clinton weenies ran up the internet bubble.
ramster, the AAA-rated MBS paid a slightly higher yield than AAA-rated corporate, which is why the fund managers were attracted to it.
Buzzcut, in 2006 total mortgage borrowing was $1 trillion and consumer credit borrowing was $100 billion, so I think the mortgage component is likely going to dwarf the others.
James Hymas, whether it’s legitimate to refer to a significant chunk of the AAA-rated MBS as “junk” may depend on how seriously you take the ABX indexes.
jck, I was abstracting, quite deliberately and I thought accurately warning the reader, from details such as timing assumptions, the burden assumed by originator and servicer, prepayment, and the bottom equity or overcapitalization tranche. You’re certainly correct that if the defaults do not come until several years after issue, investors have received some payments prior to that. With these numbers I was simply intending to communicate quickly and efficiently the concept that losses will be borne by the junior classes first, which was the theoretical basis for thinking the senior tranches were solid.
OK, so in the phrase “AAA-rated junk”, the word “junk” refers to its market price, not to its credit quality or liquidity (at least, not explicitly).
I find it interesting that the market prices of today are considered accurate reflections of value and used to prove that market prices of yesterday were not accurate.
Clearly, one set of prices (at least) is grossly incorrect, but I don’t see a lot of evidence that we may consider today’s prices the epitome of hard nosed analysis. Panic, maybe; analysis, not so much.
Clearly, there has been an increase in the information available over the past year regarding the credit quality of the instruments; an Efficient Markets zealot would have us believe that the increased information is 100% responsible for the change (which implies that buyers last year were acting at their normal 100% efficiency).
Now, I’m not a specialist in sub-prime RMBS, but I have been around the markets for a while. When a story gets as big as sub-prime is now, you know that the headlines have taken over and analysis has gone out the window. The last time I saw such unanimity on a market issue in the press, the story was that the Internet was going to become pervasive and I’d be arranging my cat’s meals by logging on to petfood.com
As with all instances of market hysteria, there is some truth contained in the headlines. As I noted, the mezzanine tranches and their derivatives have lost a lot of value and I think a good chunk of that loss is justified by the fundamentals. But “AAA-junk”? Fortunately, that’s a testable hypothesis and we will, eventually, learn what the market-implied ratings on the senior tranches should have been.
And I’ll bet you a nickel that the answer will be closer to AAA than BB+.
the AAA-rated MBS paid a slightly higher yield than AAA-rated corporate, which is why the fund managers were attracted to it.
Professor, I rise on a point of pedantry!
The Fed research article shows (with disapproval) that the increase in allocation to non-agency MBS in the single institutional portfolio examined came at the expense of the agency MBS allocation. There aren’t too many AAA corporates left in the world, anyway.
Also, it is not necessarily the case that yield pickup was the sole or even the determining factor. Agency MBS are single-tranche pass-throughs while the senior tranche non-Agency MBS have a more complicated system of allocation of principal repayments. From the prospectus of the issue I use as a paradigm:
Unless the certificate principal balances of the Class A Certificates have been reduced to zero, the subordinated certificates will not be entitled to any principal distributions until at least February 2008, or during any period thereafter in which delinquencies or realized losses on the mortgage loans exceed certain levels.
and
Principal Distribution Amount. On each distribution date, the trustee will withdraw from the Distribution Account the Principal Distribution Amount for such distribution date and will apply such amount as follows:
(a) for each distribution date (i) prior to the Stepdown Date or (ii) on which a Trigger Event is in effect:
first, to the Class A Certificates, until the certificate principal balance thereof is reduced to zero;
This provision, for first crack at any principal repayments is obviously most important in terms of credit enhancement; but it also implies a change in the cash flow assumptions that are made in the analysis of the senior tranche.
Effectively, their principal repayment rate will be about 33% higher than the prepayment rate experienced by a simple pass-through, given (and, I admit, this is a big “given”!) identical prepayment behaviour of the underlying pool.
To what extent the expected cash-flow differences were incorporated in buyers’ analyses of the attractiveness of the senior tranches is something I simply don’t know. I do know that the negative convexity that results from varying prepayment rates in the agency market can have massive effects on yield curve shape (higher interest rates implies slower prepayments implies a longer expected term of the paper implies selling [usually of 10-year Treasuries] to rebalance the portfolio).
One has to wonder what circumstances one would be counting on to expect such payments to be made on schedule from a pool of borrowers with a history of other credit problems.
Ashcraft & Schuermann speculate (p23) that “the likely answer is that lenders expected the borrower would be able to refinance before payment reset”.
This sort of jump is often seen in Canadian banks’ Innovative Tier 1 Capital and Subordinated Debt – the idea is to make the step-up severe enough that investors will believe the issue will be redeemed on the step-up date, while being not so severe that the regulator recategorizes the note for capital calculation purposes.
Canadian mortgages are typically five year loans with a 30-year amortization period – the borrower must refinance after five years.
Which makes me curious – was the existence of 30-year sub-prime mortgages a cultural thing, with Americans rejecting any possibility of the balloon payment, or are there legal reasons for lenders to extend such a long term loan as a matter of course? The availability of financing for such long-term debt in the prime market (via agencies) is obviously a contributing factor.
Felix Salmon is now casting more aspersions on the ABX indices than he was previously.
I’m very confused myself as to how you would keep the prices of such indices honest, given that you probably can’t short the underlying RBMS; and in this environment it would be an ordeal in funding horror to go long the RMBS while buying CDS insurance of some kind or another.
Hi –
One of the more disturbing aspects of the second chart you have there regarding the structure of the system is the disconnection between those who finance the mortgages and those who used the mortgages. I think that this is what drove the system to failure: the originators and arrangers of the loans were basically paid simply to sell mortgages and acquire portfolios. They carried, as such, no inherent capital risks, merely a business risk of failing to be paid when they failed to sell mortgages. As such, this is a situation that just begged for fraud and fly-by-night originators who didn’t care if the business of the mortgages was fundamentally sound or not: they got paid one way or the other.
Part of the problem, of course, was the erroneous belief of the banks involved that their investments were actually backed by properties that they could claim if the mortgages didn’t pan out: instead what they got was shares in a pool, and several court cases have shown that the investing banks cannot go to deliquent clients directly and gather their claims, but rather the only one permitted to do so is the pool administrator, who is loath, of course, to actually foreclose or pull out non-performers, since only by virtue of pooling does the whole business work (i.e. non-performers are covered by performing loans up to a certain point…).
In other words, it was a lovely, lovely system doomed to failure due to lack of accountability and enormous opportunities for fraud…
The failure was in
Economics 102: Mortgages and economic stimulus