Briefly, a place for mortgages.

Photographer: Kiyoshi Ota/Bloomberg.

Nomura and RBS Told a Few Mortgage Fibs

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.
Read More.
a | A

The global financial crisis of 2007-2008 has given rise to vast literatures of history, analysis and personal reminiscence. It's also been pretty good for judicial opinions. Yesterday Denise Cote, a federal judge in New York, published a massive history of the crisis in the form of a 361-page opinion in a lawsuit brought by the Federal Housing Finance Agency against Nomura and Royal Bank of Scotland, accusing them of selling shoddy mortgage-backed securities. It gets right to the point, in case you don't have the stomach for the other 360 pages:

This case is complex from almost any angle, but at its core there is a single, simple question. Did defendants accurately describe the home mortgages in the Offering Documents for the securities they sold that were backed by those mortgages? Following trial, the answer to that question is clear. The Offering Documents did not correctly describe the mortgage loans. The magnitude of falsity, conservatively measured, is enormous.

But if you can handle the rest of it, or at least skim judiciously, it is a good, if somewhat conventional, history of the mortgage-securitization problems that led up to the financial crisis. Nomura is the main villain, as the sponsor of seven mortgage-backed securities; RBS, which helped underwrite and market four of them, is a co-defendant. Nomura would buy mortgages from originators, put them in a pot, slice them into junior and senior tranches, get AAA credit ratings on the senior tranches and sell the resulting bonds to investors. The investors included Fannie Mae and Freddie Mac, which lost of a lot of money buying this sort of thing, and so their regulator, the FHFA, sued a bunch of banks for selling it. Most of the banks have settled, for eye-popping amounts; Nomura chose, somewhat foolishly by most accounts, to go to trial. Yesterday it lost. 

It lost because Judge Cote found that:

  1. the loans that it bought were bad,
  2. it knew that, and
  3. it didn't disclose it to people who might be interested, like the ratings agencies and buyers of its securitizations. 

As for the loans being bad, Judge Cote concluded that the originators from whom Nomura bought its loans pretty much ignored their own underwriting standards -- "Measured conservatively, the deviations from originators’ guidelines made anywhere from 45% to 59% of the loans in each SLG materially defective, with underwriting defects that substantially increased the credit risk of the loan" -- and regularly granted loans based on appraisals that did not reflect the true value of the underlying houses. ("SLG" means "supporting loan group," that is, the pool of mortgages that were packaged into a securitization.) With hindsight, no one really disagrees that the loans were bad. Page 267 of the opinion:

Today, defendants do not defend the underwriting practices of their originators. They did not seek at trial to show that the loans within the SLGs were actually underwritten in compliance with their originators’ guidelines. At summation, defense counsel essentially argued that everyone understood back in 2005 to 2007 that the loans were lousy and had not been properly underwritten.

As for Nomura knowing specifically about the defects, Judge Cote describes the actual process of securitization in some detail, explaining how Nomura would bid to buy mortgages from originators, and would then conduct due diligence after winning the auction but before closing the purchase. Her conclusion is that Nomura's diligence was rushed and defective, and overly influenced by the trading desk's desire to buy loans, but still rigorous enough to find lots of defects in the loans the desk had agreed to purchase. But then Nomura largely ignored those defects, telling its due diligence consultants to change the loans' grades from failing to passing, so that Nomura could securitize the loans.

As for its disclosure failings, the main problem is that the results of due diligence never went anywhere. Nomura got "loan tapes" from the originators, data files with the underlying borrower and property information for each loan. Then it did due diligence and, often, found that that information was wrong. But it didn't change the loan tapes, which then fed into things like the ratings agency review of the securitizations, or the summaries of the loans that Nomura wrote in its prospectuses. So the ratings and prospectuses didn't reflect the bad information that due diligence dug up. From page 75 of the opinion:

Despite speculation by a few Nomura witnesses at trial that the Diligence Group may have been consulted in some undefined way during the securitization process, the Diligence Group had no role whatsoever in the securitization process and did not review or approve the information included in the Prospectus Supplements.

Page 76:

In sum, Nomura had no reliable way to extrapolate the results from its due diligence efforts to an SLG, made no effort to do so, and never even thought about doing so. Even if its pre-acquisition due diligence had been adequate, once the link between the trade pool and SLG was broken, there was no way Nomura could reasonably rely upon the results of that pre-acquisition due diligence in making representations in the Prospectus Supplements.

So the prospectuses were false, and issuing a false prospectus creates Securities Act liability. That's the main legal issue here, but Judge Cote goes beyond it to give a textured history of mortgage securitization. She covers the competitive pressures that drove Nomura, which "came late to the RMBS business," to cut corners. Page 67:

The reason for Nomura’s lackluster due diligence program is not hard to find. Nomura was competing against other banks to buy these subprime and Alt-A loans and to securitize them. As its witnesses repeatedly described and as its documents illustrated, Nomura’s goal was to work with the sellers of loans and to do what it could to foster a good relationship with them.

Given this attitude, it is unsurprising that even when there were specific warnings about the risk of working with an originator, those warnings fell on deaf ears.

There's some good anecdotal history of how hard everyone was working to churn out mortgage securitizations. I liked this bit about the real estate appraisers who testified at trial (page 167):

They performed hundreds of appraisals apiece each year during the housing boom, but assured the Court that they never took shortcuts and in fact spent many hours on each and every appraisal. Clagett reported that he performed more than 700 appraisals each year in the period of 2005 to 2008, and took about five to six hours on each of them. Platt performed about 300 to 400 appraisals each year in 2005 and 2006, taking a minimum of four to five hours to perform each one despite the fact that he was also working fulltime as a fireman. For the period of 2004 through 2008, Morris conducted approximately 600 appraisals per year, which is about 12 per week. To justify those numbers, Morris claimed to have worked long hours seven days a week.

Apparently appraisers needed to appraise 70-80 hours a week, every week, for four years, to feed the mortgage securitization beast. 

But Judge Cote doesn't just describe the workings of the mortgage-securitization machine; she also ties those workings to the broader financial crisis. Pages 211-212:

The evidence at trial, including expert testimony, as well as common sense drive a single conclusion. Shoddy origination practices that are at the heart of this lawsuit were part and parcel of the story of the housing bubble and the economic collapse that followed when that bubble burst. While that history is complex, and there were several contributing factors to the decline in housing prices and the recession, it is impossible to disentangle the origination practices that are at the heart of the misrepresentations at issue here from these events. Shoddily underwritten loans were more likely to default, which contributed to the collapse of the housing market, which in turn led to the default of even more shoddily underwritten loans. Thus, the origination and securitization of these defective loans not only contributed to the collapse of the housing market, the very macroeconomic factor that defendants say caused the losses, but once that collapse started, improperly underwritten loans were hit hardest and drove the collapse even further. The evidence at trial confirms the obvious: Badly written loans perform badly. 

Again: This is all a pretty conventional description of the rise of mortgage securitization, its influence on underwriting standards, the resulting defaults of poorly underwritten loans, and the effect of those defaults on the housing market and the broader economy. But if you want a good, detailed, somewhat technical version of that story, Judge Cote's opinion is a good place to start.

But I also want to talk about a few more surprising elements of this case. The standard narrative above looks like one of deception and agency cost: Nomura knew that it was packaging and selling bad loans to unwitting victims, but did it anyway, because the money was good. Nomura offered to poke some holes in this story, but Judge Cote didn't let it. Pages 287-291 of her opinion cover "Excluded Evidence," and it's a fascinating section. There are two kinds of evidence that Judge Cote refused to consider. First:

Defendants sought to introduce evidence that Nomura retained a residual interest in the lowest subordinate tranche of each Securitization and that it lost money as a result. This evidence, they contended, attests to Nomura’s stake in the Securitizations, its confidence in its due diligence process, and its motivation to perform adequate due diligence.

Judge Cote ruled that "Nomura’s motivations in creating and running its due diligence program are of such minimal relevance to the issue of falsity" that she could ignore this evidence. As a legal matter, that strikes me as fair enough: The issue here is whether Nomura lied about these mortgages, not, like, how that worked out for Nomura.


Defendants also sought to offer evidence of Fannie Mae and Freddie Mac’s due diligence processes performed in the part of their businesses in which they purchased whole loans. As expressed in Nomura’s Offer of Proof, defendants sought to present evidence of the GSEs’ due diligence and the findings from that program, including the exceptions they made to deviations from originators’ underwriting guidelines, the GSEs’ disclosures in their own securitizations, and the GSEs’ communications with and evaluations of originators. With this evidence, defendants say, they would have shown that the GSEs employed due diligence processes similar to those employed by Nomura, that the GSEs regularly “waived in” loans graded 3 by their due diligence vendors, and that the GSEs’ disclosures in their Offering Documents for securitizations were similar to those made by Nomura.

Judge Cote found that this evidence would not be relevant to prove industry standards of due diligence, and "would have been highly prejudicial, wasteful, and misleading" because the sorts of loans that Fannie and Freddie bought and securitized in their own whole-loan businesses were very different from the sort that they bought from Nomura in already-securitized form. This strikes me as slightly less fair but, you know, fine.

But if you're reading her opinion not for its law but for its history, these points both seem pretty relevant. They add some nuance to the story of evil Nomura tricking innocent Fannie and Freddie into buying its garbage. Nomura kept the worst of the garbage, "a residual interest in the lowest subordinate tranche of each Securitization," and lost money on that interest.  If it had done a better job of due diligence, and had bought and securitized better loans, it would have lost less money. And Fannie and Freddie, at the same time they were buying Nomura's garbage, were also building their own securitizations with similarly lax standards. So they can't have been entirely surprised, or fooled, by the laxity of Nomura's standards.

This excluded evidence suggests that this case may not be the story of a bad guy lying to an innocent victim. It's the story of everyone lying to each other, and to themselves. The delusion was shared broadly, and led everyone to relax their standards, buy and sell terrible mortgages and blow each other up.

The other surprising thing here is that while of course the financial crisis was bad, the harm in this case looks relatively mild. The FHFA's lawsuit entitles it to rescissionary damages: Fannie and Freddie can hand back the mortgage-backed securities to Nomura, and get back the money they paid for them (minus the repayments they've already received).  That number could mean something like $500 million of repayment. But that doesn't mean $500 million of damages; Nomura would get back the bonds in exchange for its money. Peter Eavis notes:

The banks’ lawyers could argue on appeal that even though Fannie Mae and Freddie Mac took steep paper losses on the bonds, their actual losses have been small. All payments on the roughly $2 billion in bonds have been made, except for about $25 million, according a member of Nomura’s legal team, who spoke on the condition of anonymity.

The seven mortgage-backed security pools at issue here all issue periodic statements; I've linked them below.  The statements include a column for "Cumulative Realized Losses," that is, the amount of principal that has been written off of each tranche because the mortgages have stopped paying. Lots of the mortgages have stopped paying, and there are significant realized losses for the junior tranches of the mortgage-backed securities.

But the way these securitizations work is that the junior tranches absorb the first losses, protecting the senior tranches from losses even if there are a lot of defaults. And Fannie and Freddie bought senior tranches, and most of the senior tranches have seen no losses at all. Fannie bought one "senior tranche of Securitization NAA 2005-AR6," which has had no losses.  Freddie bought senior tranches in the other six securitizations; five of those deals have had no losses at all in their senior tranches;  the sixth had some losses, including $25.3 million in the tranche that Freddie bought.  That $25.3 million seems to be Fannie and Freddie's only realized loss; that is, the only money that they've lost because these shoddily underwritten mortgages defaulted. In total, Fannie and Freddie bought just over $2 billion worth of defective mortgage bonds from Nomura, and have had cumulative realized losses of 1.25 percent. 

One point two five percent! Over eight to ten years! That is not terrible! The 10-year Treasury lost 1.16 percent of its value yesterday.  

One possible conclusion here is that the securitization process didn't work that badly. Some originators made some mortgage loans, and they were bad. Nomura bought them, and they were bad. Nomura packaged them into mortgage-backed securities, got them rated AAA, and sold them on to Fannie and Freddie. And those AAA mortgage-backed securities were good! Fannie and Freddie bought them, the mortgage market collapsed, and Fannie and Freddie sued, claiming that the AAA securities were full of fraud. And they were full of fraud! "The magnitude of falsity, conservatively measured, is enormous." And yet the AAA tranches that Fannie and Freddie bought barely lost any money.

Obviously this did not work out as well for everyone: If you bought a more junior tranche of these securitizations, you didn't do very well, and if you bought a AAA tranche of a collateralized debt obligation that aggregated and re-sliced those junior tranches, you might have done very poorly indeed. But that 1.25 percent loss rate, if nothing else, gives you a sense of why people liked mortgage securitization so much: Despite the originators' fraud and Nomura's dishonesty and a massive global financial crisis, Fannie and Freddie's AAA rated senior bonds were almost entirely protected from losses. The structure of securitization, of enhancing the credit of mortgages by packaging them into pools and then selling AAA tranches of those pools, worked. And it worked as a substitute for, you know, honesty. Fannie and Freddie didn't need to trust Nomura's due diligence, or its prospectuses, because they were protected by the structure of the bonds. They had no reason to demand honesty, so they didn't get it. From the narrow perspective of the losses on their investments, that worked out fine. From a broader perspective, it was a bit of a disaster. 

  1. And sometimes Nomura even hid the diligence results from people who were expecting them. Pages 68-69:

    In September 2006, Nomura withheld due diligence information from its co-lead underwriter RBS. As Nomura was preparing to send a report to RBS showing the results of AMC’s due diligence review of loans that would be securitized in NHELI 2006-FM2, Nomura discovered that there were 19 loans still rated as having material deviations. In an email with the subject line “HUGE FAVOR,” Nomura’s Spagna requested that AMC act “ASAP” and retroactively re-grade the 19 loans as client overrides since Nomura had decided to buy them “for whatever reason.” Only after the grades had been altered and the report re-run, did Nomura forward the AMC results to RBS. In a conference call with RBS on that same deal, Spagna reported to a Nomura colleague that he “took the liberty to bullshit” RBS, adding “I think it worked.”

  2. I mean, that oversimplifies a lot of discussion. One of Nomura's defenses is that it didn't say the loans were originated in accordance with underwriting standards, but rather that they were "generally" originated in accordance with those standards. That's a big "generally"! Judge Cote doesn't buy it; there's a delightful subsection (pages 273-274) devoted to "The Meaning of 'Generally,'" deciding that 

    the use of the word “generally” can’t bear the weight defendants wish it to bear in this case. It could not and did not convey that roughly half the loans had substantially increased credit risk because they were not originated in compliance with their originators’ guidelines, even after one accounts for exceptions to guidelines justified by compensating factors.

  3. Actually there are others -- this is not the only section titled "Excluded Evidence" -- but these are the interesting ones.

  4. I don't want to overstate this: Nomura seems to have retained only a small slug of residual certificates and may well have made more on fees than it lost on retention. But the retentions weren't trivial, exactly; the latest of the deals, NHELI 2007-3, featured "approximately $54,951,765" of overcollateralization, which was allocated to the non-offered Class X certificates retained by Nomura.

  5. See pages 322-323: 

    The plaintiff is entitled “to recover the consideration paid for [the] security with interest thereon, less the amount of any income received thereon, upon the tender of such security, or for damages if he no longer owns the security.” 15 U.S.C. § 77l(a)(2) (emphasis added). The GSEs retain the Certificates they purchased from defendants, and thus their recovery is measured by the statutory formula. In the case of an RMBS, investors’ original investment is repaid in regular increments. Accordingly, each month, the GSE was to be paid a return of principal on its Certificate and a coupon reflecting an interest payment.

    To calculate damages, FHFA’s expert deducted principal payments received by each GSE on a month-to-month basis to arrive at the amount of pre-judgment interest due on the consideration paid. The proceeds balance at the time of judgment plus the accumulated prejudgment interest make up the statute’s “consideration paid” with “interest thereon.” Finally, to arrive at a damages figure, the expert added all of the coupon interest payments received by the GSEs and subtracted that amount from the sum of the consideration paid and prejudgment interest.

    There are various boring discussions about interest rates, etc., if you like that sort of thing.

  6. Their unfelicitous names are:

    1. NAA 2005-AR6
    2. NHELI 2006-FM1
    3. NHELI 2006-HE3
    4. NHELI 2006-FM2
    5. NHELI 2007-1
    6. NHELI 2007-2
    7. NHELI 2007-3

    See page 40 of the opinion. The statements are from Bloomberg.

  7. Though some of the other senior tranches in that deal (not bought by Fannie) have seen total losses of about $14 million, out of a total original principal of senior tranches of $591 million (2.4 percent). I'm counting tranches I-A, II-A-1, II-A-2, III-A-1, III-A-2, IV-A-1 and IV-A-2 as "senior"; those were the originally AAA-rated tranches. Between them they show $13,977,178.60 of "Cumulative Realized Lossses" on the April 27, 2015 distribution summary.

    Footnote 16 on page 40 of Judge Cote's opinion says which securitizations Fannie and Freddie bought, and the FHFA's complaint lists the tranches; Fannie bought tranche III-A-1, which has had zero dollars of cumulative realized losses. Original principal amounts (and ratings) come from the prospectus

  8. NHELI 2006-FM1, NHELI 2006-HE3NHELI 2006-FM2, NHELI 2007-2, NHELI 2007-3.

  9. That's NHELI 2007-1. Here is the April 2015 distribution summary showing cumulative realized losses, and here is the original prospectus with principal amounts (and seniority). According to the FHFA complaint, Freddie bought tranche II.1.A, which has lost $25,284,611.19 out of an original principal amount of $100,548,000 (25.1 percent). Overall this securitization has had $123 million of losses on senior tranches, out of $949 million of original principal amount of senior tranches (13 percent).

  10. Source: Bloomberg (from 98.6953125 on Friday to 97.5546875 at yesterday's close). Obviously this is an unfair comparison, but, ha.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the author on this story:
Matt Levine at

To contact the editor on this story:
Zara Kessler at