Journal of Law, Finance, and Accounting, XXXX, XX: 1–57

The Nonprime Mortgage Crisis and

Positive Feedback Lending
Bernard S. Black1 , Charles K. Whitehead2 and Jennifer Mitchell
Coupland3∗
1
Nicholas J. Chabraja Professor, Northwestern University, Law School and
Kellogg School of Management, Department of Finance;
[email protected]
2
Myron C. Taylor Alumni Professor of Business Law, Cornell Law School;
[email protected]
3
Northwestern Law School (Class of 2012); [email protected]

ABSTRACT
The “great recession” of 2007–2009 was sparked by a bubble in U.S.
housing prices, driven in turn by a bubble in nonprime mortgage
lending. We collect evidence that the risk of a nonprime housing
bubble (not the certainty, but a meaningful risk) should have been
obvious to the main participants in the markets for nonprime
lending and related mortgage-backed securities (nonprime MBS),
including originators, securitizers, rating agencies, money managers,
and institutional investors. Those who did not see the risk were,
in many cases, willfully blind. We also discuss the strong positive
feedback nature of typical nonprime mortgages. This positive
feedback made it highly likely that, if nonprime housing prices
flattened, let alone fell, they would soon crash and take many
nonprime MBS with them. We discuss regulatory responses that
might limit positive feedback lending, cause the next bubble to be
smaller and less likely, and make the post-bubble aftermath less
painful.

Keywords: Due diligence, Securities offerings, Banks, Auditors, Rating

agencies
JEL Codes: G14, G38, K22
∗ Ms. Coupland worked from 2004–2008 at Countrywide Financial Corporation in their

whole loan purchasing group, which purchased loans from smaller originators, normally to
be securitized with loans originated by Countrywide.
An early version of this article was prepared for the Abraham Pomerantz Lecture at
Brooklyn Law School (March 2009) and a keynote speech at the European Financial

ISSN 2164-5744; DOI 10.1561/104.00000019

©2018 B. S. Black, C. K. Whitehead, and J. M. Coupland

Electronic copy copy

Electronic available at:at:https://ssrn.com/abstract=1973662
available https://ssrn.com/abstract=1973662
2 Bernard S. Black et al.

1 Introduction

Many different explanations have been offered for the “great recession” of
2007–2009. Nevertheless, there is widespread agreement that the spark for the
recession was a “nonprime crisis” – a downturn in U.S. housing prices and an
accompanying surge in default rates on risky mortgages, often called “subprime”
or “Alt-A” (which together we call “nonprime”) loans.1 Those, in turn, led to
falling prices on mortgage-backed securities (MBS), whose value was based on
nonprime loans (nonprime MBS), which sparked a cascade of other events.2
In this article, we discuss how the linked nonprime housing and nonprime
MBS markets became both large enough and fragile enough to provide that
spark. Our article is limited to the nonprime crisis. We do not seek to explain
why the nonprime spark touched off such a large fire.
We focus on the rapid growth of nonprime MBS prior to the nonprime
crisis. Elements of this growth include: the repackaging of the lower, riskier
tranches of direct nonprime MBS into collateralized debt obligations (nonprime
CDOs); the further repackaging of the lower, riskier tranches of nonprime
CDOs into still other structured securities, including CDO-squareds (CDO2 s),
so-called because they were CDOs based mostly on other CDOs; the creation of
“synthetic” CDOs and CDO2 s, whose payoffs were based on the returns of other
structured securities; and the emergence of “structured investment vehicles”
(SIVs) as major holders of those securities. Below, we refer to nonprime MBS
and CDOs as “nonprime MBSall .” 3 What can explain the boom in nonprime

Management Conference on Corporate Governance and Control (Cambridge, England, April

2009). We thank Ronen Avraham, Lynn Baker, Anne Beatty, John Coffee, Amil Dasgupta,
John Griffin, Marc Gross, Jay Hartzell, Todd Henderson, April Klein, and Brian Rider, and
participants in the Pomerantz and European Financial Management lectures in the Centre
for Economic Policy Research, European Summer Symposium in Financial Markets (2009),
and in workshops at the Hong Kong Institute of Chartered Public Accountants, and the
University of Texas, Law School and McCombs School of Business, for comments. We thank
Robert Hicks, Feny Liu, Aashim Usgaonkar and David Sverdlov for their invaluable research
assistance. Our thinking has also benefitted greatly from a related modeling project with
Simon Gervais, which seeks to formally model the incentives that led market participants to
ignore the obvious and build a large bubble.
1 Most of the securitization dollar volume in risky mortgages was for subprime and Alt-A.

Second mortgages (closed-end second mortgages and home equity lines of credit), some
made to prime borrowers and some to nonprime borrowers, often were similar in risk to
nonprime mortgages, and also were part of the picture. In 2005, second mortgages accounted
for approximately 12% of origination volume, but only 4% of securitization volume; in
2006, these loans were 14% of origination volume and 8% of securitization volume. Inside
Mortgage Finance, vol. 1, at 4.
2 See, e.g., Gorton (2009) (“The [2007–2009] credit crisis was sparked by a shock to

fundamentals, housing prices failed to rise.”).

3 There is no standard term that covers the full range of MBS and related CDOs. We will

refer to the full range of MBS and CDOs based largely or entirely on nonprime mortgages as
“nonprime” MBSall and to the full range of CDOs based largely or entirely on nonprime MBS
as “nonprime CDOall ,” but we will omit “nonprime” when the meaning is clear from context.

Electronic copy copy

Electronic available at:at:https://ssrn.com/abstract=1973662
available https://ssrn.com/abstract=1973662
The Nonprime Mortgage Crisis and Positive Feedback Lending 3

MBSall offerings that underlay the boom in the prices of homes financed
through nonprime loans?4 Why was the bust so severe?
For the nonprime crisis, as for the broader financial crisis, a number
of partial explanations have been offered. One is government pressure on
commercial banks to make loans to lower-income borrowers and on government-
sponsored entities (GSEs) to buy and securitize those loans, plus moral hazard
at the GSEs in light of their implicit government-backing.5 This pressure can
partly explain growth in nonprime “purchase” loans that were then securitized
into nonprime MBS. But many nonprime loans were refinancings that had little
to do with making homes affordable for new buyers.6 Moreover, commercial
banks were enthusiastic rather than reluctant participants in the bubble, and
investment banks (who felt no government pressure to make loans) were
equally, if not more, enthusiastic. The GSEs did none of the repackaging of
MBS into CDOs and CDO2 s. And the financial crisis was sparked by losses
incurred by private actors, not the GSEs, which were bailed out by the U.S.
government. For bank behavior, partial explanations include: (i) bank net
capital rules permitting AAA assets to be highly leveraged, which encouraged
banks to create and hold AAA-rated MBS; (ii) auditors failing to question the
banks’ use of SIVs to move structured finance assets off-balance-sheet, only
for the assets to return in the financial crisis; and (iii) the banks’ own risk
management departments’ failing to assess the riskiness of holding nonprime
MBSall , including assets that the securitizers could not readily sell to outsiders.7
Finally, until the introduction of the ABX-HE indices in early 2006, there
was no easy way to short nonprime MBS, which could have helped to sustain
inflated prices.
We offer here two additional, also-partial explanations for why the nonprime
lending and securitization process went far off-track, despite the participation
of sophisticated financial players at every stage of the origination and securi-
tization process, and for why the correction was so painful. First, we argue,
with supporting evidence where available, that it should have been obvious to
market participants that:

1. there was a meaningful risk (not a certainty or even a probability, just

a meaningful risk) that the United States was in a nonprime housing
bubble, which could end with falling prices for homes bought with these
mortgages—nationally rather than just locally, and in both nominal and
real dollars; and
4 For related discussions of MBS lending, see Levitin and Wachter (2012); Eggert (2009);

Engel and McCoy (2007).

5 See, e.g., Morgenson and Rosner (2011); Wallison (2011); Calomiris (2009a).
6 Bear Stearns (2006); see also Mayer et al. (2009) (in 2005, purchase loans accounted

for only 40% of subprime originations).

7 Whitehead (2015) suggests that investment banks were unfamiliar with many of the

risks they assumed during the period prior to the financial crisis.

Electronic copy copy

Electronic available at:at:https://ssrn.com/abstract=1973662
available https://ssrn.com/abstract=1973662
4 Bernard S. Black et al.

2. nonprime mortgages, and MBSall securities based on them, would perform

badly if housing prices fell.

If there was even a 1% risk that we were in a nonprime bubble, then the AAA
ratings given to a large percentage of nonprime MBS (ratings which implied
an annual default rate well under 0.1%) were unjustified, and the AAA ratings
given to nonprime CDOs, which were based heavily on BBB-rated nonprime
MBS, were a fantasy.
Second, we argue that structural features of most nonprime loans created
strong positive feedback. If prices for nonprime-financed houses started to
fall (or even failed to rise), the dynamics of many nonprime loans, especially
adjustable Alt-A loans with low initial teaser rates and punitive reset rates
after two to three years, ensured strong positive feedback in prices and default
rates. That feedback would predictably lead to a crash in nonprime housing
prices, nonprime MBS, and nonprime CDOs. The positive feedback would be
(and, in fact, was) rapid and vicious. A drop (or even a non-rise) in housing
prices would spark loan defaults, which would drive forced sales by owners
and foreclosures by lenders, followed by distressed sales. Those sales would
drive housing prices down further, sparking more defaults, further forced sales,
and additional foreclosures. Falling prices and higher default rates would
trigger tighter lending and securitization standards, which would further limit
purchases and refinancings of houses and reinforce the downward cycle. The
whole nonprime securitization process might (and, in fact, did) shut down,
as it had in the 1990s. This would greatly increase the pressure on housing
prices. A number of core elements of this feedback loop were national and
thus non-diversifiable.
Yet, most market participants assumed away the risk of a housing bubble
and paid little attention to many risks, such as the fraud risk posed by loans
with low or no documentation of a job, income, or assets (“lo-doc” loans), and
widely-reported claims that loan amounts were often based on artificially high
appraisals. Moreover, so far as we can tell, participants entirely ignored the
positive feedback that would result from a drop (or non-rise) in housing prices.
As a result, they significantly underestimated how poorly nonprime MBSall
could perform.
We argue that these risks were gross and obvious. Failure to recognize them
cannot be explained simply as reasonable decisions made by smart investors
that turned out badly in hindsight. Prior analyses have made nonprime MBSall
seem complicated, which they were. But that too easily makes it appear that
the investors’ mistakes were reasonable. They were not. Instead, we argue,
many market participants were willfully blind. That blindness requires an
explanation.
We argue that willful blindness arose because, at every stage of the lending,
structuring, and purchase chain, the key actors had incentives to understate

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 5

and under-investigate MBSall risks:

1. Originators profited from originating loans that they then resold to

banks, which would securitize the loans. So long as they could sell the
loans, the originators were largely insulated from the risk of poor loan
performance.
2. Bankers are often seen as “reputational intermediaries” who engage in a
“due diligence” investigation of issuers and securities and then vouch for
the securities they underwrite. Instead, as the nonprime boom gathered
steam, bankers limited their diligence, maintaining the pretense of doing
so but conducting little real investigation or disregarding the results of
that investigation. Many banks bought loan originators in order to gain
greater access to the assets needed to create nonprime MBSall . They
were unlikely to reject loans from their own affiliates.
3. Rating agencies are also held out as reputational intermediaries, but the
profits from casting an uncritical eye on nonprime MBSall were large
enough to override their concern with reputation. The agencies made
incredible modeling assumptions, and when that was not enough to
justify their high ratings, they added large, unexplained “out-of-model”
adjustments to boost the ratings they bestowed.
4. Money managers failed to press the rating agencies on their assumptions.
Key modeling assumptions remained entirely hidden until the financial
crisis. Nonprime MBSall had a limited history, which counseled caution;
instead, many money managers bought them aggressively to gain a bit of
extra yield. The managers were competing in a tournament for investor
funds and could tell investors they were beating their peers without
taking too much risk—after all, they were investing in highly-rated
securities.
5. Given the limited track record of nonprime MBSall , investors could not
easily assess the true risk level of these securities. This short track
record gave them reason to act with caution, but many did not do so.
In part, investors who placed funds with money managers did not want
to underperform their peers. In part, they were lulled by high ratings.
Top-rated assets are often viewed as “informationally insensitive” and
properly so if the ratings are justified since substantial investor diligence
would not be cost-effective. Nonprime MBSall ratings did not reflect
actual risk, but some investors surely did not understand this.

Banks, rating agencies, money managers, and investors also found safety in
being part of a crowd—if they were wrong, they would not be alone. All three
major rating agencies blessed nonprime MBSall with unjustified ratings; all

Electronic copy available at: https://ssrn.com/abstract=1973662

6 Bernard S. Black et al.

the major banks peddled those securities; many money managers bought them;
and investors happily pocketed the extra returns. Distorted incentives also
existed within banks and rating agencies. The bonuses and careers of MBSall
bankers depended on riding the securitization wave, not sitting it out or raising
warnings within their institutions. So too at the rating agencies—the more
offerings they rated, the more the nonprime MBSall raters would prosper, and
this required that they provide high ratings. For them, the reputational and
financial risks to their employers were externalities.
Securitization is often seen as a way to repackage risk so that it can be spread
more widely and borne by those who are best able to bear it. But securitization
is also dangerous—it makes it easier to hide and concentrate risk. If bankers
are skilled at hiding and concentrating risk, they need to find only a few fools
to buy the “toxic waste” that more knowledgeable investors will shun. The
process of re-securitizing the riskier tranches of nonprime MBS into nonprime
CDOs, and the riskier tranches of nonprime CDOs into nonprime CDO2 s,
allowed a roughly $2 trillion market for nonprime MBS to rest on the bankers’
ability to place only around $20 billion in toxic, low-rated CDO2 tranches.8
We argue that, even though nonprime MBSall were top-rated assets that
were informationally insensitive, which gave money managers less reason to
investigate risk, investors did not conduct even the bare minimum diligence
they sensibly might have undertaken given the very large amounts they invested.
Stepping back from the nonprime crisis, diligence and skepticism are often
weak precisely when they need to be strong—for new financial products or
ideas, or when prices are rising rapidly and markets are bubbly. The nonprime
bubble is an example, albeit a large one, of a recurring problem.
To diagnose the linked problems of willful blindness and positive feedback
lending is not to offer a cure. We have no full cures to offer. We discuss
the regulatory responses to date, which do little to prevent a recurrence, and
sketch reforms that would strengthen the due diligence obligations of bankers,
money managers, and rating agencies, as well as the rating agencies’ disclosure
obligations. Those reforms could limit the size and frequency of future bubbles.
The systemic risk posed by positive feedback lending is a separate problem
that may require direct regulation. For any one lender, positive feedback
features in loan contracts are an externality. Any regulatory response must
be tailored to the nature of the loans. To date, we are unaware of significant
efforts to limit the positive feedback features of mortgage loans. We sketch
below some ways to do so.
This article is part of a vast literature on the financial crisis, too vast to
usefully cite. The closest in spirit to our argument that market participants
were willfully blind is Calomiris (2009b). But Calomiris largely just asserts that
8 There is no single good source on aggregate issuances of nonprime mortgages. The

$2 trillion estimate is reasonable based on the incomplete data we found. The $20 billion
estimate is developed below.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 7

the rating agencies’ loss assumptions were too low (and, therefore, that ratings
were too high), and does not tie the fate of nonprime MBSall to housing market
dynamics. We know of no close predecessor to our second core argument, on
the role of positive feedback lending in inflating the bubble and intensifying
the post-bubble price collapse.
There is also a substantial literature on bubbles and their causes, which we
do not repeat here. Potential causes include boom and bust pricing behavior;9
overuse of leverage in boom times;10 the entry during booms of new, often
inexperienced, market participants;11 cognitive errors;12 hype and media
coverage;13 and simply poor judgment.14 Many of these general causes have
been proposed as explanations for the housing bubble leading to the financial
crisis.15 Other broad explanations for the crisis include large financial flows
from emerging to developed markets, which drove down asset returns and
led to a search for yield;16 overly lax monetary policy;17 a run on collateral
within an interconnected financial system;18 lax regulation of investment banks
and derivative instruments;19 and the bankruptcy super-priority given to
repurchase (repo) transactions.20
This article proceeds as follows. Part II develops our main claim—the
risks posed by nonprime mortgages, including the strong positive feedback
if a downturn began—were obvious had market professionals wanted to see
them. In Part III, we explain how market participants had incentives to
under-investigate, under-disclose, and generally downplay nonprime MBSall
risks. Part IV discusses why post-crisis reforms fall well short of correcting
the errors of the past and sketches some reform proposals.

2 Nonprime Mortgage Risks Were Many and Obvious

We explain here why the risk —not a certainty, only a significant risk—of a
sustained nationwide drop in house prices for homes financed with nonprime
loans should have been obvious to the professionals who created, rated, and
bought those securities, enough so that they had to be willfully blind not to
9 See,
e.g., Kroszner (2005).
10 See,
e.g., Blair (2010) and Caginalp (2000).
11 See, e.g., Haruvy et al. (2007); Dufwenberg et al. (2005); Porter and Smith (1994);

Smith et al. (1988).

12 See, e.g., Shiller (2016) and Partnoy (2000).
13 See, e.g., Shiller (2016); Barber and Odean (2006); Fisch and Sale (2003).
14 See, e.g. Hunt (2011).
15 See, e.g., Lo (2012); Brunnermeier and Julliard (2008); Shiller (2008a).
16 See, e.g., Jagannathan et al. (2009).
17 See, e.g., Taylor (2009).
18 See, e.g., Gorton and Metrick (2012).
19 See, e.g., Rhee (2010) and Pozen (2009).
20 See, e.g., Roe (2011).

Electronic copy available at: https://ssrn.com/abstract=1973662

8 Bernard S. Black et al.

see the risks. A core part of that risk involved the positive feedback nature of
many nonprime loans, which intensified the pre-crash rise in home prices, and
would amplify any downturn. In Part III, we develop a related premise: It was
readily knowable to the professionals who originated, securitized, and bought
nonprime MBSall that a sustained drop in the prices of nonprime-financed
homes would lead to soaring default rates on nonprime mortgages, sufficient
to imperil MBSall , and perhaps a shutdown in nonprime securitization.
To summarize this Part, with details to be provided below:
• It was known that housing prices had experienced an unprecedented rise
in real terms, and it was obvious that this large rise might be followed
by a similar fall.
• It was known that default rates on nonprime mortgage loans would rise
substantially if local housing prices failed to rise and would soar if local
prices fell.
• It was known that real estate appraisers were being pressured by lenders
to inflate home values.
• It was known that nonprime loan terms were becoming ever flakier, so
that prior experience might understate future default rates.
• It was obvious, and readily knowable if anyone had checked, that lo-doc
loans, which were a dominant share of Alt-A loans and a large share of
subprime loans, often involved fictitious claims of income and assets.
• It was known that nonprime loans were soaring in volume, were becoming
an ever-larger fraction of overall mortgage originations, and had become
the dominant loan form in many lower- and middle-income real estate
markets (“nonprime markets”).
• It was known that many nonprime loan types had not been tested in a
serious economic downturn. It, therefore, was obvious that the rating
agencies were guessing about how they would perform in such a downturn.
Finally, it was obvious that many of these aspects contributed to strong positive
feedback in the linked housing, nonprime mortgage, and MBSall markets. That
feedback would reinforce any price downturn, once it started.
The professional participants in the nonprime lending chain did not need to
know each of these elements to understand the risks embedded in the nonprime
mortgage and MBSall markets. They only needed to know the first two—
namely, that nonprime housing prices might fall, and if they did, nonprime
MBSall would perform badly. The remaining factors simply reinforced the risk
of falling home prices and the extent to which loans would perform poorly if
nonprime home prices started to fall.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 9

2.1 Default Rates Would Rise Sharply if Local Housing Prices Fell

Begin with the simple fact, both known and obvious: Default rates on nonprime
mortgages would rise substantially if local housing prices simply failed to rise
and would soar if local prices fell.
Consider first a traditional “prime” mortgage, with a maximum loan-to-
value ratio (LTV) of 80%, made to a creditworthy borrower at a fixed interest
rate. Default and foreclosure rates on those mortgages have historically been
low, and for good reason.21 If the owner had significant equity in the house,
and ran into hard times and could not pay the loan, she would have a strong
incentive to sell the house and repay the loan rather than default on payments
and let the house fall into foreclosure, which normally would lead to a lower
sale price. In a traditional prime mortgage, the lender starts with a healthy
cushion against default. This equity cushion tends to rise over time, because
nominal house prices tend to rise and principal payments reduce the amount
owed.
Moreover, even if the house’s price falls below the loan’s principal (say, by
10% or less), the borrower usually will continue to pay if she can, and usually
she can for a number of reasons.22 First, paying preserves the valuable option
to keep the house if prices rise again. Second, absent job loss, the borrower
can afford the payments—the lender has assured this, with a safety margin,
by limiting the borrower’s “total debt to income” ratio (ratio of mortgage pay-
ments, other loan payments, property taxes, homeowners’ insurance, and other
recurring payments (such as child support) to income).23 Third, creditworthy
borrowers tend to be stable job-keepers. Fourth, most prime loan programs
included a requirement that the borrower have “reserve cash” sufficient to
ensure that the borrower can pay the mortgage loan for several months.24
Fifth, in “recourse” states, the lender can seek a personal payment from the
borrower for any shortfall, and some borrowers will have other assets that can
satisfy the shortfall. And, sixth, prime borrowers often try to avoid defaults
to preserve their credit ratings. Thus, home prices need to decline by 30% or
more before default rates on prime loans will rise significantly.

21 For example, in the first quarter of 2007, before the housing bust took hold, the

“seriously delinquent” rate for prime loans was 0.89%, contrasted with 8.33% for subprime
loans. Mortgage Bankers Association (2010). Post-crisis delinquency peaked in the fourth
quarter of 2009, with prime loans at 7.01% and subprime loans at 30.56%. The seriously
delinquent rate is a non-seasonally adjusted rate that includes loans that are 90 days past
due or are being foreclosed, but excludes already-foreclosed loans. Id.
22 See Guiso et al. (2013).
23 See Lendingtree (2017). Some loans required the borrower to be at or below ratios both

for total debt to income and for “home” debt payments (mortgage payments, other loan
payments, property taxes, and homeowners’ insurance) to income. See Bear Stearns (2006).
24 Prime loan programs typically require cash reserves, ranging from two months to

twenty-four months, depending on the loan amount, property type, and other risk factors.
See, e.g., Lerner (2010); Fannie Mae (2013).

Electronic copy available at: https://ssrn.com/abstract=1973662

10 Bernard S. Black et al.

Nonprime lending both weakens the cushions against default and increases
the likelihood of sharp price drops in a number of ways. First, nonprime
borrowers usually have limited equity, because nonprime loans often involve
high loan-to-value ratios. In 2005, the median combined loan-to-value ratio
(CLTV) for a subprime loan to finance a new home purchase was 100%;
the median for an Alt-A loan in 2006 was 95%.25 Second, many refinanced
subprime loans included a cashout, in which the loan amount exceeded the
remaining principal on the prior loan plus refinancing costs, thus limiting the
equity cushion from prior price appreciation. In 2005, for example, 51% of
subprime adjustable-rate mortgages (ARMs) and 68% of subprime fixed rate
loans were cashout refinancings.26 To be sure, loan-to-value ratios were lower
for refinancings, with the median around 80%.27 Maximums were higher, but
normally were still below those for loans made for original purchases.28
Second, many cashout borrowers used the cash they received to support
lifestyles that were beyond their incomes or to pay mortgages they could not oth-
erwise afford. Those borrowers needed rising property values to pay their mort-
gages; they would have difficulty paying if housing prices were merely flat.29
Third, debt-payments-to-income (DTI) ratios were relaxed. Typical per-
centage limits for prime loans were in the mid-thirties.30 The limits for a
nonprime loan depended on several factors, including the loan-to-value ratio
and whether the loan was full-documentation, but percentage limits often
exceeded 50%.31 Moreover, for ARMs, the ratios were based on the initial
“teaser” rate, not the reset rate. A large percentage of nonprime loans were
ARMs with low initial rates, and much higher reset rates, with the reset usually
occurring in two or three years.32 A typical reset might increase post-teaser
payments by 25%; thus, an initial 40% ratio (already at the upper end of
affordability) would become an often-unaffordable 50% once the reset kicked
in. Teaser-rate ARMs also often included prepayment penalties, which raised
the effective interest rate during the teaser period.33 Some nonprime loans
25 See
Mayer et al. (2009). The “combined” ratio includes both first and second mortgages.
26 BearStearns (2006). For prime loans, the comparable cashout refinancing percentages
were 26% for ARMs and 28% for fixed rate loans.
27 In 2005, the median combined loan-to-value ratio for refinancings was 80% for subprime

loans and 79% for Alt-A loans. Mayer et al. (2009).

28 See, e.g., Sacramento Wholesale Mortgage Alt A Matrix (2007) (95% maximum CLTV

for non-cashout refinancing Alt-A loan; 90% maximum for cashout refinancing).
29 Ryan (2008).
30 See, e.g., SMC Bancorp (2006). Lenders would sometimes permit higher ratios if there

were other mitigating risk factors, such as a low loan-to-value ratio or significant other
assets.
31 See Ryan (2008); Sacramento Wholesale Mortgage Alt A Matrix (2007); Washington

Mutual Bank (2006); Federal Deposit Insurance Corporation (2001).

32 Between 2004 and 2006, ARMs which reset in two or three years (called “2/28s” and

“3/27s”) accounted for approximately 70% of all subprime loans. See Sengupta (2010).
33 Prepayment penalties could be hard or soft. Hard penalties applied to both refinancings

and sales; soft penalties applied only to refinancings. See Document Systems, Inc (2013).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 11

were “interest-only” or even “negative amortization” loans (payment not even

covering interest), with full amortization often starting only after five or ten
years and computation of the debt-to-income ratio based only on the initial
payments.34 Thus, for a large fraction of nonprime loans, unaffordability was
baked in. The borrower had to refinance her home, sell it, or default on the
mortgage loan. As Gary Gorton wrote, “The defining characteristic of a sub-
prime [ARM] is that it is designed to essentially force a refinancing after two
or three years.” 35 This forced refinancing was a core driver of positive feedback
on the downside, of nonprime lending, which had no parallel in prime lending.
Fourth, low- or no-documentation loans were a substantial part of the
nonprime market (see Table 1 below). Those loans were often called “liar loans”
because of the high risk that the borrower’s reported income was inflated. As
we discuss below, those loans deserved this reputation.
Put those risk factors together, and a substantial percentage—perhaps
a majority—of nonprime borrowers would have trouble making their loan
payments if housing prices merely did not rise, by enough to permit refinancing,
over a two-to-three-year period. Many would be both unable to afford their
mortgage loan payments and unable to refinance their loans, due to the lower
loan-to-value limits for a refinancing, with medians around 80% (see Table 1
below), compared to the 95%-100% medians for purchase loans.36 The principal
means to avoid default would be to sell. Nonprime borrowers would be in
even worse trouble if housing prices were flat or fell, because then they could
not sell at a price high enough to repay their loans after transaction costs
and any prepayment penalties. Difficulty in refinancing was still greater for
interest-only or negative amortization loans, which were the dominant form of
Alt-A lending; by 2007, only 18% of Alt-A loans were traditional amortizing
loans. The combination of forced refinancing, and inability to refinance unless
prices rose, was a further core driver of positive feedback on the downside, of
nonprime lending, again with no parallel in prime lending.
The sensitivity of nonprime default rates to home prices was known. As
Stephen Ryan observed:37
Penalties varied. One common penalty was six months’ interest. See, e.g., id. (Sample
Payment Addendum to Note). A second form was a percentage of the remaining principal,
such as 3%. The penalty might decline over time. See Mortgage News Daily (2006).
34 In 2005, interest-only loans accounted for 15% of mortgage origination volume and 32%

of non-agency securitization volume; negative amortization loans, called “Option ARMs,”

accounted for another 8% of origination volume and 12% of non-agency securitization volume.
In 2006, those numbers increased to 18% of originations and 30% of non-agency securitizations
for interest-only loans and 9% of originations and 14% of non-agency securitizations for
Option ARMs. Inside Mortgage Finance, vol. 1, at 6, and vol. 2, at 14.
35 Gorton (2010) (emphasis added). This observation was as true for Alt-A loans as for

subprime loans.
36 Some lenders allowed higher refinancing CLTVs. See, e.g., Sacramento Wholesale

Mortgage Alt A Matrix (2007) (95% maximum CLTV for a purchase or rate and term
refinance Alt-A loan; 90% for a cashout refinancing).
37 Ryan (2008).

Electronic copy available at: https://ssrn.com/abstract=1973662

12 Bernard S. Black et al.

It has been well understood by market participants since the

Russian debt/hedge fund crisis of 1998 . . . [that] the frequency of
defaults on subprime mortgages is low if mortgagors are able to
obtain cash-out refinancing and high otherwise.

Table 1: Nonprime Loan Quality Metrics

2007
Characteristic 2003 2004 2005 2006 (Jan.–June)
Panel A. Subprime loans
Purchase 30 36 40 42 31
Includes “Piggyback” Second 7 15 24 28 15
Mortgage(s)
Median CLTV (Purchase) 90 95 100 100 100
Median CLTV (Refinance) 80 80 80 80 80
Median FICO Score 615 615 618 616 613
Fixed Rate 30 21 17 20 27
Investor Property 8 8 7 7 8
Low or No Documentation 32 34 36 38 34
Interest Only 2 11 21 13 11
Negative Amortization 0 0 0 0 0
Prepayment Penalty 74 73 72 70 69
Penalty Period Longer than 10 8 6 3 2
Teaser Period
Panel B. Alt-A loans
Purchase 46 54 52 49 37
Includes “Piggyback” Second 12 27 35 42 33
Mortgage(s)
Median CLTV (Purchase) 90 90 90 95 95
Median CLTV (Refinance) 74 75 75 79 79
Median FICO Score 710 706 708 701 707
Fixed Rate 69 36 39 39 42
Investor Property 27 23 22 21 22
Low or No Documentation 63 62 69 80 81
Interest Only 16 37 40 44 52
Negative Amortization 2 16 24 26 30
Prepayment Penalty 26 33 39 44 40
Penalty Period Longer than 12 18 30 32 16
Teaser Period
Source: Mayer et al. (2009). Amounts are in percentages.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 13

In addition, nonprime borrowers had personal characteristics that raised

the risk of default. Less creditworthy borrowers are at higher risk of job loss
and other sources of financial distress. Most did not have other significant
assets. Reflecting this reality, nonprime loan programs, unlike their prime
counterparts, usually imposed no reserve requirements. Many nonprime loans
were made in non-recourse states.38 And many nonprime borrowers already
had adverse events—delinquencies, judgments, foreclosures, repossessions,
charge-offs, or bankruptcies—on their credit record.39 Those borrowers may
have been more willing to default if under financial stress or if doing so would
be financially advantageous.

2.2 Positive Feedback Lending

Some of the positive feedback aspects of nonprime lending are already apparent.
Others, with national rather than just local scale, will be developed in later
sections of this article. Consider the upside first. If house prices were rising
and interest rates were stable or declining, most nonprime borrowers could
refinance. Those who could not could resell at a profit to others. Rising prices
also fueled homeowner wealth. Some of that wealth was consumed, but much
went into home buying, which helped to keep prices rising (we provide evidence
on this below).
But home prices could not rise forever. What if prices merely flattened?
Borrowers who had put little money down could not refinance and would need
to sell (or default). The nonprime housing market could presumably handle a
few such forced sales. But typical nonprime ARMs, which combined minimal
down payments with forced refinancing, were the dominant product in many
markets. If prices flattened, and forced sales therefore soared, these sales
would push prices down, which would make refinancing still harder, and force
more sales, which would push prices further down. Declining prices would also
encourage defaults by borrowers with little equity or negative equity. This
would lead to foreclosures and more forced sales. The overhang of houses with
loans in default, but not yet foreclosed on or sold, would discourage buyers.
Local housing markets were already known to be subject to positive feed-
back, with rising prices persisting in the near-term and reversing in the
medium-term.40 The structure of nonprime loans would make that feedback
far stronger. How much stronger—no one knew.

38 De jure or de facto non-recourse states include Arizona, California, and Florida. Some

states are non-recourse de jure, others de facto because they permit a faster foreclosure
process if the lender foregoes a deficiency judgment. Lenders usually prefer the faster,
cheaper process over the limited value of making a borrower with limited assets liable for a
deficiency judgment.
39 Federal Deposit Insurance Corporation (2001).
40 See, e.g., Case and Shiller (1989).

Electronic copy available at: https://ssrn.com/abstract=1973662

14 Bernard S. Black et al.

Does a mortgage product that demands either sale or regular refinancing

make sense for a single lender, over a full market cycle, assuming other lenders
do not behave similarly? We have doubts. Nonprime products were developed
in a boom market; prime mortgages do not have similar features. But positive
feedback loans surely make no sense as a dominant product. The downward
spiral that would happen—that had to happen—if prices merely flattened is
apparent in hindsight. We believe the drivers behind positive feedback should
have been apparent to lenders had they stopped to ask, “What if prices flatten?”
It is tempting for any one lender to ignore the market dynamics that will
result from widespread lending with strong positive feedback. Those dynamics
are largely an externality for one lender – its own positive feedback loans make
the overall reaction to a market flattening or downturn more severe, but much
of the cost will be felt by other lenders. Yet, this begs a central question.
Widespread positive feedback lending created extreme market fragility. Should
not some lenders have pulled back in the face of this fragility, instead of fueling
the cycle with their own loans? Yet, no one did. Moreover, we have found no
contemporaneous discussion of the implications of positive feedback lending
for nonprime default risk.
That is, it is not as if lenders recognized the inherent instability of typical
nonprime mortgages and went ahead anyway, because each lender externalized
much of its own contribution to that instability onto other lenders. If any
recognized the risk, we have found no traces in the public record. That
blindness needs explanation.

2.3 Real Estate Appraisers Were Pressured to Inflate Home Values

Many nonprime mortgages were originated by mortgage brokers who were

paid only for loans that closed and thus had strong incentives to massage
applications to meet the lenders’ approval criteria. During the boom, mortgage
brokers often pressured home appraisers to inflate home values so that loans
would meet CLTV targets. Many appraisers would acquiesce rather than risk
losing business. If enough were willing to do so, brokers could find enough
compliant appraisers and simply bypass the more conservative ones. Thus,
the nominal median CLTV values noted above (100% for subprime purchase
loans; 95% for Alt-A) understated true CLTV ratios to an uncertain extent.
This pressure was widely understood to exist. Some anecdotes: A 2007
Washington Post article reported a survey of 1,200 appraisers across all states
that found that 90% of them felt pressured by mortgage brokers, real estate
agents, lenders, and consumers to inflate their value estimates.41 That number
was up from 55% in a similar survey in 2003. Also in 2007, then-New York At-
torney General Andrew Cuomo sued a major real estate appraiser for assisting

41 Harney (2007).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 15

Washington Mutual (a major nonprime lender) “to hand-pick appraisers who

bring in appraisal values high enough to permit [Washington Mutual’s] loans
to close.” 42
By how much were appraisals inflated? At the time, no one knew. As with
lo-doc loans, no one investigated. Ex post evidence, compiled by Griffin and
Maturana, indicates that typical overstatement was in the 8-10% range.43
Inflated appraisals were an additional reason why nonprime lenders, se-
curitizers, rating agencies, and investors knew, or should have known, that
apparent risk, reflected in reported CLTV ratios, DTI ratios, and the like,
understated true risk. They were also another source of positive feedback from
nonprime lending. As long as prices kept rising, appraisers could be generous,
and their upward shading of value was unlikely to be caught. But if prices
flattened, and houses could not be sold for their appraised values, appraisals
would come under scrutiny. A tightening of appraisal standards would limit
the supply of new loans and reinforce the feedback cycle. The financial crisis
indeed prompted a major tightening of appraisal standards.

2.4 Nonprime Loan Terms Were Becoming Flakier

It was well known that typical nonprime loan characteristics had become
increasingly extreme as the housing boom continued. They needed to do
so—as housing prices rose, and incomes failed to keep pace, borrowers needed
looser terms in order to afford to buy or refinance a house.
Table 1 provides summary information over 2003-1H 2007. By 2006, a
typical subprime purchase loan had a 100% CLTV ratio, was an ARM with
a reset in two-to-three years, and had a 40% DTI ratio based on the teaser
rate.44 Over one-third were low- or no-documentation loans. By 2006, a
typical Alt-A purchase loan had a 95% CLTV ratio, was lo-doc (true for 80%
of Alt-A loans), permitted the borrower to pay interest only or allowed for
negative amortization (70% of loans), and likely involved a 35%–40% DTI
ratio based on amounts paid during the initial period.45
The lower CLTVs for refinancing loans, shown in Table 1, were well known.
If (i) purchase loans were made with minimal equity, and often with no or
minimal principal repayment during the initial years, and (ii) buyers needed
20% equity to refinance their loans (the difference between a 100% median
CLTV ratio for a purchase loan and a 80% median CLTV ratio for a refinancing),
where would that equity come from? For most nonprime borrowers, housing
prices needed to rise substantially in order for a refinancing to be possible.
42 Summons, People v. First American Corp., 2007 N.Y.S. Ct. Pleadings LEXIS 35.
43 Griffin and Maturana (2016a).
44 Jaffee (2008) provides information on DTI ratios.
45 This range reflects the judgment of one of this article’s co-authors, Jennifer Coupland,

who was employed as a buyer of nonprime mortgages for Countrywide in the mid-2000s. We
have not found a data source on DTI ratios for Alt-A loans.

Electronic copy available at: https://ssrn.com/abstract=1973662

16 Bernard S. Black et al.

A second refinancing would be available for a once-refinanced loan only if

home prices had not fallen—and if the borrower did not need a cashout
refinancing to afford the loan payments or the refinancing fees, or to make
other payments. The prepayment penalties that were attached to many ARMs
made the refinancing problem even worse.
The asymmetry between purchase loans and refinancing loans made non-
prime lending inherently unstable. If prices were merely flat, the market would
soon be flooded with involuntary sellers who could neither refinance the loans
nor afford their payments after the interest rate was reset. Those forced sales
and defaults would place downward pressure on home prices, which would
further exacerbate a price decline. Nonprime lending helped fuel the boom in
prices; so too, it would fuel a subsequent bust. The flakier the loan terms, the
more likely it was that borrowers were at or beyond their affordability limits,
and thus the stronger the downward forces would be.

2.5 Many Lo-Doc Loans Were Based on Fictitious Income Claims

It should come as no surprise that some lo- and no-doc “liar loans” were based
on inflated claims of income, assets, or both. One would expect estimating
the extent of fraud to be central to making and securitizing those loans. In
fact, almost no one looked. The data we have been able to find is sparse. Yet,
it is ample to suggest a large problem.
A 2006 survey of 2,140 mortgage brokers who sold lo-doc loans found
that 43% knew that many borrowers lacked the income needed to qualify for
full-documentation loans, in which income would be verified.46 The Mortgage
Bankers Association for Responsible Lending compared 100 stated income loans
with Internal Revenue Service records and found that 90% exaggerated income
by 5% or more and “almost 60% of the stated amounts were exaggerated by more
than 50%.” 47 Liar loans indeed! Fitch Ratings (Fitch) also reached troubling
conclusions from its review of forty-five loan files in 2007. It found that 44% of
the files included employment status or income that was inconsistent with other
portions of the loan application or facially implausible.48 We found no evidence
that the other two main credit rating agencies—Moody’s Investors Service
(Moody’s) and Standard & Poor’s (S&P)—ever assessed the liar loan risk.
Lo-doc loans were a major part of the nonprime market. As Table 1 shows,
by 2006, they comprised 38% of subprime loans and 80% of nonprime loans.49
46 Harney(2006).
47 Mortgage Asset Research Institute (2006).
48 Fitch Ratings (2007). Reasonableness was based on a number of factors. First, the

borrower’s profile was compared to the level of savings or assets in the borrower’s bank
accounts. Second, the borrower’s stated employment and income was compared to income
levels for similar job titles and geographic locations. Id.
49 Other sources provide higher estimates for the proportion of lo-doc subprime loans. A

Fitch report notes that over half of subprime loans were lo-doc. Calhoun (2006). Acharya
et al. (2009) estimate this proportion at 44% in 2006.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 17

Yet no one in the lending chain—not originators, not securitizers, and not
the credit rating agencies—ever assessed fraud risk or its implications for
default risk. (Fitch’s tiny sample, assessed only in 2007 with the bust already
underway, hardly counts.)
The risks posed by liar loans were no secret. Wells Fargo’s chairman, for
example, stated that they were an “open invitation for fraud.” 50 Yet Wells
Fargo still made lo-doc loans—merely less aggressively than other lenders.51
Originators could have checked borrowers’ incomes. Most lo-doc loans required
the borrower to sign IRS Form 4506 or 4506-T, which gave the originator
permission to obtain the borrower’s federal income tax returns.52 Securitizers
could have requested that information. Instead, neither the originators nor
the securitizers chose to look, even on a spot check basis. That is why the
available data are so limited—it comes from one sample of 100 loans, and a
second sample of 45 loans. Willful blindness, indeed!
More generally, early in the bubble period, the banks often hired specialist
firms—Clayton Holdings was the most prominent—to spot check loans and
assess whether they met the banks’ loan purchase criteria. As the bubble
grew and lo-doc lending and other forms of flaky lending grew with it, the
banks steadily cut the percentage of loans that were reviewed. This was not
because the loans consistently met the banks’ standards. Clayton Holdings
later reported to the Financial Crisis Inquiry Commission that, by 2006,
28% of the loans it reviewed were severely substandard.53 Of the loans that
were checked and found wanting, a third stayed in the banks’ mortgage pool.
Presumably, so did the loans that were not checked.

2.6 Nonprime Loan Volume, Housing Prices, and National Positive

Feedback

It was well known that nonprime mortgage loans were soaring in volume and as
a fraction of all mortgage loans. As Table 2 shows, nonprime originations began
to grow in the late 1990s, dipped in the early 2000s due to fallout from the
failure of some nonprime originators and a general decline in originations, and
took off starting in 2003–2004. Between 2000 and 2005, nonprime originations
rose from $125 billion to over $1 trillion, growing from roughly 10% to one-
third of originations. Nonprime lending remained at that level in 2006 before
50 Kovacevich (2009) (“Isn’t it just common sense that especially for subprime borrowers,

their stated income, negative amortization, teaser rates, no-doc and lo-doc mortgage loans
primarily originated by brokers are inappropriate, should never have been offered, and are
an open invitation for fraud?”).
51 Heid (2007) (Wells Fargo executive testifies that Wells Fargo limited the volume of Alt-A

lo-doc loans it extended; no similar statement for subprime loans). Compare Kovacevich
(2009) (claiming that Wells Fargo decided not to originate any lo-doc loans).
52 See Fannie Mae (2013).
53 Morgenson (2010).

Electronic copy available at: https://ssrn.com/abstract=1973662

18 Bernard S. Black et al.

Table 2: Mortgage Loan Origination Percentages, 1990–2009

Gov’t and Jumbo and Nonprime

Year GSEs Home Equity Subprime Alt-A (Subprime + Alt-A)
1990 67.7 23.6 8.1 0.7 8.7
1991 64.4 25.1 9.4 1.1 10.5
1992 67.4 22.6 8.9 1.0 10.0
1993 69.5 21.1 8.3 1.1 9.4
1994 64.3 24.7 9.7 1.3 11.0
1995 61.0 28.0 9.4 1.6 11.0
1996 61.8 26.8 8.9 2.5 11.5
1997 58.7 28.5 9.9 2.9 12.8
1998 58.6 29.7 9.3 2.4 11.7
1999 58.5 28.5 9.9 3.1 13.0
2000 58.2 29.9 9.5 2.4 11.9
2001 65.0 26.0 7.2 1.8 9.0
2002 65.2 25.5 6.9 2.3 9.3
2003 67.9 22.1 7.9 2.2 10.0
2004 46.1 28.9 18.5 6.5 25.0
2005 37.8 30.0 20.0 12.2 32.2
2006 35.9 30.5 20.1 13.4 33.6
2007 52.1 28.7 7.9 11.3 19.2
2008 81.4 14.3 1.5 2.8 4.3
2009 90.1 9.3 0.2 0.3 0.6
Source: Inside Mortgage Finance (2010). Amounts are in percentages of all mortgage loans.

plummeting in 2007 as the financial crisis took hold, and all but disappearing
in 2008. In mid-2007, Merrill Lynch estimated that there were $2.3 trillion in
outstanding nonprime mortgage loans.54
It is a lending truism, for many loan types, that rapid growth in outstanding
loans is a warning sign for declining loan quality and rising risk of future
defaults. To expand volume quickly, lenders must relax lending standards and
find new, formerly-uncreditworthy borrowers. As we saw above, nonprime
lenders did both.
In many parts of the country, especially lower-income areas, nonprime loans
became the principal way in which homes were financed. Easy credit, in turn,
fueled a run-up in housing prices. Figure 1 provides supporting evidence. We
are not aware of a price index that is specific to nonprime-financed houses or to
specific geographic areas in which nonprime mortgage loans were the dominant

54 Merrill Lynch (2007).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 19

180%

160% FL-Miami

CA-Los Angeles
140%
Housing Price Appreciation, 2000-2006 (%)

120% NV-Las Vegas

DC-Washington
AZ-Phoenix FL-Tampa
CA-San Diego
100%
NY-New York

80%
WA-Seattle
CA-San Francisco
60%
IL-Chicago
MN-Minneapolis
MA-Boston
40%

NC-Charlotte GA-Atlanta
20% CO-Denver
TX-Dallas
MI-Detroit
OH-Cleveland
0%
15% 17% 19% 21% 23% 25% 27% 29% 31% 33% 35%
Subprime Loans as % of Total Originations (2006)

Figure 1: MSA-specific Housing Price Appreciation versus Subprime Lending

Note: Figure 1 shows housing price appreciation in 20 major metropolitan areas during 2000–
2006 in relation to the share of subprime loans as a percentage of total mortgage loan orig-
inations in the state in 2006. The correlation coefficient is r = 0.57; the regression line
(% Price Increase) = −0.86 + 6.30 ∗ (% Subprime) (t = 2.75).

source of lending. Nor is there a readily available source for nonprime lending
volume at the local level. But we do know the Case-Shiller housing price
index for 20 major metropolitan areas.55 We plot the percentage increase from
2000–2006 on the y-axis. We also have data on the fraction of subprime loan
originations by state in 2006 from Inside Mortgage Finance.56 We plot this
on the x-axis. The correlation between the two measures is striking given the
crudeness of each measure. The greater the share of subprime loans within
total lending, the larger the housing price rise during the nonprime boom
period (correlation = 0.57; t = 2.75). We lack data on Alt-A lending volume,
but a similar pattern is likely. Alt-A borrowers had higher FICO credit scores
than subprime borrowers, but were often of similar economic status and lived
in the same areas.
The correlation between housing prices and nonprime lending on the way
up (ever-looser lending standards helped fuel the boom) implies that there
would likely be a similar correlation on the way down. Consistent with a
downside correlation, Griffin and Maturana report that a higher proportion of

55 See S&P Corelogic Case-Shiller Home Price Indices, http://us.spindices.com/

index-family/real-estate/sp-corelogic-case-shiller (last visited Jan. 25, 2018).

56 Inside Mortgage Finance (2010), vol. 1, at 245–50.

Electronic copy available at: https://ssrn.com/abstract=1973662

20 Bernard S. Black et al.

misreporting during loan origination (inflated income, for example) predicted

larger price rises during the boom years of 2003–2006, and a larger drop during
2007–2012, at the zip-code level.57
The link between nonprime lending and nonprime home prices was a second,
crucial source of positive feedback. Any tightening in lending standards would
affect home prices and could spark a turn in the housing cycle. Any turn would
be amplified by the positive feedback nature of nonprime lending. Nonprime
lending standards, in turn, depended on what loan terms buyers of nonprime
MBSall would accept. If buyers became concerned over loan terms, and stopped
buying securities based on the flakier loans, those loans would disappear and
this would affect housing prices. The risk that buyers would pull back was more
than hypothetical. As we discuss below, a drop in nonprime securitization
had already occurred in the late 1990s and contributed to an earlier nonprime
mortgage loan crisis.
Any drop in securitization would put pressure on nonprime lending and,
accordingly, on nonprime-supported prices across the country. That is, local
lending depended on the national, indeed international, comfort of nonprime
MBSall buyers with the ever-flakier loans that supported the boom. All local
nonprime housing markets relied on the same group of nonprime MBSall
investors. This strengthened the likely correlation of housing prices in different
nonprime lending markets. The higher that correlation, the worse nonprime
MBSall would perform in a downturn, and the more reason for nonprime
MBSall buyers to be skittish. The worse nonprime MBSall would be likely to
perform in a downturn, the more likely it was that buyers would become still
more cautious in deciding what to buy. That is, the nonprime MBSall market
was also prone to strong positive feedback. A tightening in the nonprime
MBSall market, in turn, would affect loan availability and loan terms, thus
strengthening the positive feedback effect between nonprime mortgage loans
and home prices, and doing so nationally.
A mere flattening in volume could be enough to spark a downward spiral
in the linked nonprime lending and nonprime MBSall markets. The loan terms
that nonprime MBSall investors would accept could not become flakier forever.
If the trend toward investors accepting securities based on ever-flakier loans
merely flattened, loan supply would flatten as well, home prices would likely
flatten, and the downward spiral would begin.

2.7 New Loan Types Had Not Been Tested in a Downturn

MBS based on prime mortgages had existed in the two broad housing downturns
that preceded 2007 (in the 1980s and 1990s). Thus, the rating agencies
had some basis for predicting how they would perform in a downturn. In

57 Griffin and Maturana (2016b).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 21

contrast, many nonprime-loan types were invented, or turned from niche to

mass products, only during the nonprime boom that began in the late 1990s.
They had not been tested during a period of flat home prices, let alone falling
prices. For example, as Table 2 shows, the Alt-A loan category was small until
the boom. The same was true for interest-only and negative amortization loans
in both the subprime and Alt-A loan categories. Lo-doc loans had existed in
the 1990s and were known to be troublesome, but had been a niche product.
The characteristics of a niche product, such as Alt-A or lo-doc loans in the
1990s, might be very different once the loan type became a mass product.
Thus, for a large percentage of nonprime loans, the rating agencies had little
history to help them predict default rates in a downturn. Money managers
and investors knew or should have known this.
The same lack of history applied to many types of nonprime MBSall
securities. Simple MBS had existed for decades, but nonprime CDOs were new,
as were synthetic CDOs and CDO2 s.58 The performance of those securities
partly depended on the performance of the underlying assets, but also and
crucially on the correlation across the returns on different assets. Here, too,
money managers and investors knew or should have known that: (i) the rating
agencies had little to guide them on correlations; (ii) correlations often rise
in a downturn; and (iii) as we describe in more detail below, the agencies’
models assumed near-zero correlation across assets in any given CDO—hence
near-zero correlation across the returns on different nonprime MBSall . How
sharply correlations would rise, and what they would be, no one knew. The
agencies’ actual assumptions were opaque. They said little and what they said
was hard to find and far from clear.59 At the same time, it was obvious that,
as Mason and Rosner wrote in early 2007, “several hundred underlying assets
from the same sector, i.e., [nonprime] RMBS, does not add true diversification
to the pool, leading many in the industry to question the relevance of the
[agencies’ correlation assumptions].” 60
Thus, money managers and investors knew or should have known that the
rating agencies in large measure had little to guide them in assessing MBSall
performance at two levels—how the loans would perform in a serious downturn;
and how nonprime CDOs would perform in light of loan performance. This
should have caused them to probe the rating agencies’ assumptions, assess their
reasonableness, and evaluate how nonprime MBSall securities would perform
58 “Most of the [structured finance] CDOs issued during 1999–2002 were cash flow trans-

actions backed by diversified pools of instruments such as CMBS, manufactured housing

ABS, RMBS, REITs, HY CBOs, and corporate debt.” Hu (2007). But by 2005, subprime
residential mortgages had grown to comprise almost 50% of structured finance CDOs. Id.
59 The very little that the rating agencies said can be found in Moody’s Investors Service

(2009, 2005), and Standard & Poor’s (2002). The only research we know of on how the
assumptions translated into correlations between defaults on various types of MBSall is
Griffin and Nickerson (2017).
60 Mason and Rosner (2007b); see also Crawford (2010) and Hunt (2010).

Electronic copy available at: https://ssrn.com/abstract=1973662

22 Bernard S. Black et al.

under other assumptions. Yet, as best we can tell, few money managers
or investors ever took this basic step. The rating agencies did not disclose
core assumptions about housing price appreciation, and investors did not
ask. The rating agencies minimally disclosed correlation assumptions, but
what those assumptions implied for actual correlations was never stated—and
indeed remains unknown today. Again, money managers and investors did
not ask.
This is astonishing. Money managers and investors knew or should have
known that: (i) nonprime loans would do badly in an economic crisis; (ii) there
was strong positive feedback in nonprime lending and, in turn, home prices
in nonprime markets; and (iii) the rating agencies had little basis from past
history to use in estimating outcomes and correlations in a downturn. Yet,
as we discuss further in Part III, they apparently rarely, if ever, asked what
assumptions the agencies were making.

3 Nonprime MBSall Risks Were Understated by the Rating Agencies But

Obvious to Anyone Who Chose to Look

If housing prices started to fall (or even failed to rise), the dynamics of many
nonprime loans ensured strong positive feedback in prices and default rates,
both locally and nationally, that might well lead to a crash in nonprime housing
prices, nonprime MBSall , and nonprime CDOs. To make the case for willful
blindness, we do not need to claim that a crash was a certainty or even a
probability. We need to claim only that it was a possibility, with chances far
greater than the well below 0.1% annual default risk that AAA-rated securities
were expected to represent. Yet, most market participants paid little attention
to the many risks presented by nonprime MBSall . In particular, they appear to
have ignored the positive feedback that would result from a drop (or non-rise)
in housing prices. As a result, they greatly underestimated how poorly these
securities could perform.
In light of the large and obvious risks discussed in Part II, how did the
rating agencies persuade themselves that a large fraction of MBSall deserved
a high, often AAA, rating? A short answer for MBS: The agencies required
MBS mortgage pools to be geographically diversified. They assumed very
low correlations across home prices in different areas. In particular, they
assumed that national home prices would not decline in nominal dollars. They
understood that local housing price declines could occur, but assumed these
would be offset by gains in other areas. The rating agencies deemed a national
decline to be so unlikely that they did not even test how MBSall securities
would perform if such a decline occurred. In fact, it was obvious that national
housing prices could fall, perhaps substantially, in both nominal and real
dollars. We collect several sources of evidence below.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 23

In theory, the credit rating agencies played a vital role in bridging the
information gap between sponsors and investors in structured financial prod-
ucts, but they failed to do so with MBSall during the period leading up to the
financial crisis.61 The explanations for this failure fall along three general lines.
One emphasizes conflicts of interest rooted in a business model where the
issuer (rather than investors) paid for ratings, which led the rating agencies to
compete with each other for business by lowering their ratings standards.62
Ratings shopping, where an issuer or sponsor would solicit preliminary rat-
ings from competing agencies, placed pressure on the agencies to inflate their
ratings.63 That pressure was exacerbated by the high proportion of revenues
the agencies earned from structured finance deals and the dependence of those
revenues on a few banks that could credibly threaten to move their business
elsewhere if they did not obtain the desired ratings.64 Rating agencies and
sponsors also worked together to develop structured finance products, and
the ratings agencies made their models directly available to investment banks,
which could then design products to take advantage of those models and their
limitations,65 heightening the likelihood of distortion.66
A second argument derives from regulatory use of credit ratings as bench-
marks for risk, in effect, giving the rating agencies the ability to determine
the substantive effect of legal rules.67 Investors had an interest in the rating
agencies underestimating risk in order to take advantage of the resulting regu-
latory benefits.68 And, increasingly, in light of the long history of regulatory
requirements linked to credit ratings, investors may have over-relied on the
value of credit ratings as a measure of risk.69
The third argument highlights failures in the agencies’ models used to assess
nonprime MBSall risk.70 As we noted in Part II, the models were untested
under stress and assumed very low default correlations across assets. Other
assumptions were also suspect. For example, the agencies based their models
on historical data drawn from a boom period and assumed a national home
price decline could never happen.71 We provide details below.
61 See, e.g., Merrill Lynch (2007); Iacobucci and Winter (2005); Partnoy (1999).
62 See, e.g., Hill (2010); see also Lynch (2009).
63 See, e.g., Cohen and Manuszak (2013); Ellis et al. (2012); Coffee (2011a); Riddiough

and Zhu (2009); Skreta and Veldkamp (2009).

64 See He et al. (2011); see also Dennis (2009) (reporting the growing proportion of

Moody’s total revenue that came from rating structured finance over the decade preceding
the housing crisis); Coval et al. (2009) (noting that in 2006 44% of Moody’s revenues was
from structured finance, versus 32% from corporate bonds).
65 See Morgenson and Story (2010).
66 See, e.g., Levitin and Wachter (2012); Cane et al. (2012); Krebs (2009).
67 See, e.g., Lynch (2009); Partnoy (2000); Partnoy (1999).
68 See, e.g., Horton (2013); Hunt (2010); Calomiris (2009b); see also Coffee (2011b).
69 See, e.g., Partnoy (2000); see also Adelino (2009).
70 See, e.g., Levitin and Wachter (2012); McNamara (2012); Merrill Lynch (2007).
71 See, e.g., Levitin and Wachter (2012); Green (2009); Dennis (2009); Hill (2004).

Electronic copy available at: https://ssrn.com/abstract=1973662

24 Bernard S. Black et al.

Nonprime MBSall were complex securities. To what extent can that com-
plexity explain both the rating agencies’ errors, and the willingness of money
managers and investors to rely on those ratings, often with little or no investi-
gation? In our view, complexity alone is an insufficient explanation. Willful
blindness by money managers and their investors needs to be added as a key
part of the explanation. To summarize the key points developed in this Part,
nonprime MBSall impounded multiple known or obvious risks beyond those in
the underlying nonprime mortgage markets because:
• It was knowable, to anyone who asked, that the rating agency models had
(absurdly) assumed away a national housing price decline and, similarly,
had ruled out a national decline in nonprime housing prices.
• It was known or knowable that the rating agencies had applied large
“out-of-model” adjustments to boost nonprime CDO ratings beyond what
their models would justify.72
• It was known that nonprime MBS were the dominant assets within
nonprime CDOs and knowable, to anyone who asked, that the rating
agency models for nonprime CDOs assumed very low correlations across
assets, including very low correlations across nonprime MBS.
• It was known that most originators used an “originate to distribute”
model, and that many originators and securitizers had little or no skin
in the game, which might limit their incentives to make bad loans, as
long as those loans could be resold for a profit.
• It was knowable that, if default rates rose, nonprime securitization might
shut down, as it did in the 1990s.
• It was obvious that, if nonprime securitization shut down, a national
price decline for nonprime-financed homes was a virtual certainty.
• There were warnings, from credible sources, that we might be in a housing
bubble.
Nonprime MBSall had a limited history, which counseled caution. Instead,
many money managers bought them aggressively in order to gain extra yield
and then told investors that they were investing in AA- and AAA-rated
securities. The result was a failure to cast a critical eye on nonprime MBSall
or to question how the rating agencies reached their credit conclusions. The
ratings were unjustified, and money managers and their investors could have
easily known this if they had chosen to inquire.
72 Given the large out-of-model adjustments that the rating agencies applied to CDOs, it

seems likely that they also applied out-of-model adjustments to nonprime MBS, but we are
unaware of evidence of this.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 25

210
200
190
180
170
160
150
140
130
120
110
100
90
80
70
60
50
1953 Q4
1955 Q2
1956 Q4
1958 Q2
1959 Q4
1961 Q2
1962 Q4
1964 Q2
1965 Q4
1967 Q2
1968 Q4
1970 Q2
1971 Q4
1973 Q2
1974 Q4
1976 Q2
1977 Q4
1979 Q2
1980 Q4
1982 Q2
1983 Q4
1985 Q2
1986 Q4
1988 Q2
1989 Q4
1991 Q2
1992 Q4
1994 Q2
1995 Q4
1997 Q2
1998 Q4
2000 Q2
2001 Q4
2003 Q2
2004 Q4
2006 Q2
2007 Q4
2009 Q2
2010 Q4
1890
1896
1902
1908
1914
1920
1926
1932
1938
1944
1950

Figure 2: Case-Shiller Index: Real Home Prices in the Long Run

Note: Case-Shiller Index (inflation adjusted) for 1890–2010.
Source: Shiller (2008b), as updated by the authors using data at http://www.econ.yale.edu/
~shiller/data.htm (last visited Jan. 26, 2018).

Here, as in Part II, the professional participants in the nonprime lending

and securitization chain did not need to know each of these elements in order
to understand the risks of the MBSall market. They only needed to know that
there was a meaningful risk that nonprime housing prices might fall nationwide
and, if they did, nonprime MBSall would perform much worse than the agency
models predicted.

3.1 Nominal Housing Prices Could Fall Nationally

We begin this part with the truly obvious. Step back to 2006. Home prices
had soared nationally and had risen further and faster in nonprime markets
(Figure 1). Incomes had not kept up, especially for moderate income borrowers.
As a result, affordability had plummeted. Builders had responded to rising
prices by building more homes, and the supply of unsold homes was rising.
This set the stage for a flattening or a fall in prices. Any flattening or fall
would be intensified by the positive feedback discussed in Part II.

3.1.1 Data on Real Home Prices

Figure 2 shows national, real home prices from 1890 until 2010 based on the
Case-Shiller Home Price Index. This Index is a standard reference for housing
prices, which uses repeated sales of the same house to track price changes

Electronic copy available at: https://ssrn.com/abstract=1973662

26 Bernard S. Black et al.

for twenty major metropolitan areas.73 One sees immediately that real home
prices can be roughly flat over long periods—indeed real price levels were about
7% lower in 1998, when the pre-financial crisis boom began, than they had been
in 1953, just after the post-World War II boom. One also sees that the last
two times real prices rose, before the pre-financial crisis boom, they soon fell
back to the average post-World War II level of around 110. Given this history,
it was obvious that the pre-financial crisis rise could also be followed by a drop.
The huge run-up in prices over 1998–2006 meant that the drop could be large.
We saw earlier that even a flattening of nonprime-based home prices would
generate intense pressure to push those prices downward, likely leading to
widespread price declines. Thus, by the mid-2000s, a large drop was not
only possible, but likely. As the bubble grew, the likelihood and size of the
subsequent correction grew as well. But our claim is milder: We claim only
that a significant nationwide correction in nonprime-based home prices was
possible, with a probability far above the well under .001 annual default risk
that the rating agencies claimed for AAA-rated debt. We do not think anyone
can look at Figure 2, as it would have appeared in 2005 or 2006, and say
there was not a significant risk that what had just gone way up might come
part-of-the-way back down.
The rating agencies assumed that nominal house prices could not fall, not
real prices. What would a large fall in real housing prices, commensurate with
the 1998–2006 rise, imply for nominal home prices? Figure 3 shows nominal
prices. During the late 1970s’ fall in real prices, inflation was high and nominal
prices continued to rise. During the late 1980s and early 1990s, inflation was
lower; nominal home prices were effectively flat over 1989–1992. So it was
literally true that national home prices had not fallen in nominal dollars since
the period from 1925 through the Great Depression. But this was hardly a
law of nature. Given low inflation rates, if the 1998–2006 rise was followed
by a similar decline, even if only partial, nominal housing prices would also
drop substantially. This risk of a large drop was enhanced because any drop
in nonprime-supported home prices would be strongly self-reinforcing.

3.1.2 Housing Affordability Index

Other sources also warned that trouble might lie ahead. One important
cautionary sign came from the national housing affordability index, published
by the National Association of Realtors (NAR). The index “measures whether
a family earning the median income could qualify for a mortgage loan to buy a
home at the national median price.” 74 Figure 4 shows the affordability of ARM
73 Standard & Poor’s, S&P/Case-Shiller Home Prices, http://www.standardandpoors.com/

indices/sp-case-shiller-home-price-indices/en/us/?indexId=spusa-cashpidff--p-us----.
74 NAR, Methodology of the Housing Affordability Index, https://www.nar.realtor/

research-and-statistics/housing-statistics/housing-affordability-index. An index of 100

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 27

200

180

160

140

120

100

80

60

40

20

0
1953 Q3
1955 Q1
1956 Q3
1958 Q1
1959 Q3
1961 Q1
1962 Q3
1964 Q1
1965 Q3
1967 Q1
1968 Q3
1970 Q1
1971 Q3
1973 Q1
1974 Q3
1976 Q1
1977 Q3
1979 Q1
1980 Q3
1982 Q1
1983 Q3
1985 Q1
1986 Q3
1988 Q1
1989 Q3
1991 Q1
1992 Q3
1994 Q1
1995 Q3
1997 Q1
1998 Q3
2000 Q1
2001 Q3
2003 Q1
2004 Q3
2006 Q1
2007 Q3
2009 Q1
2010 Q3
1895
1901
1907
1913
1919
1925
1931
1937
1943
1949
Date

Figure 3: Case-Shiller Index: Nominal Home Prices in the Long Run

Note: Case-Shiller Index (nominal dollars) for 1890–2010.
Source: Shiller (2008b), as updated by authors using data at http://www.econ.yale.edu/~shiller/
data.htm (last visited Jan. 26, 2018).

loans for 1977–2010, as reported by NAR. Falling interest rates, and growing
use of ARMs with teaser rates, kept apparent affordability roughly constant
in the early boom years. But the composite index dropped in 2004, dropped
sharply in 2005 to 113, marking the first time since 1991 that it had been below
120, and dropped again to 108 in 2006. Even creative mortgage engineers could
no longer bridge the gap between rising prices and flat median family incomes.
For most nonprime loans, the NAR index was also seriously misleading.
Actual affordability was far lower. This was readily knowable. The NAR index
assumed a 20% down payment. Actual nonprime down payments were far
lower. Moreover, for ARMs, the NAR index ignored the jump in payments
that would occur when the initial teaser rate reset.
In Figure 5, we focus on ARMs, which were the dominant nonprime loans,
and on the principal nonprime period of 1995–2007. The top line shows the
“official” NAR housing affordability index, carried over from Figure 4. The
middle line shows a downpayment-adjusted “nonprime affordability index”
that assumes a down payment of 10% in 1995–2003, 5% in 2004, and 0% in
2005–2007.75 The downpayment-adjusted nonprime index never exceeds 120

means that a family with the median income has just enough income to qualify for a
mortgage on a median-priced home, assuming a 20% down payment. The index is based only
on principal and interest payments; it does not reflect taxes or property insurance payments.
75 This is based on median CLTV over 2003–2007. See Table 1. For earlier years, we lack

a data source, so we assume that median CLTV equaled the 2003 level.

Electronic copy available at: https://ssrn.com/abstract=1973662

28 Bernard S. Black et al.

180

160

140

Composite
120
Fixed
ARM

100

80

60
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Figure 4: Housing Affordability Index (as reported by NAR)
Note: The Index is based on the effective rate on purchase loans for existing homes. An afford-
ability index of 100 assumes a 25% ratio of housing expense to gross income and a 20% down
payment.
Source: Housing Affordability Index for 1977–2010, as reported by NAR, at http://www.realtor.
org/topics/housing-affordability-index (last visited Jan. 26, 2018).

and plunges, in the later part of the boom, from 116 in 2003 to 86 in 2006.
Thus, the NAR index, adjusted for a realistic, median downpayment, was
flashing a serious warning of unaffordable prices.
Even the downpayment-adjusted affordability index was a mirage that
would disappear if housing price appreciation slowed. As we discuss above, low-
downpayment purchase loans could be refinanced, once the initial teaser rates
expired, only if prices rose substantially during the teaser period. The bottom
line in Figure 5 provides an estimate of post-reset affordability. To generate
this line, we assumed a reset based on 6-month LIBOR (a common base floating
rate for nonprime ARMs) plus a typical “margin” above this rate.76 A July
2007 Merrill Lynch analysis, citing The Mortgage Investor, reported that those
who could refinance their homes would experience a 17% increase in payments,
and those who could not would see a 28% payment jump.77 Measured this
way, affordability is low but moderately stable during 1995-2004, in the range
of 96 to 101. After that, affordability plummets—to 71 in 2005 and 68 in
2006—before recovering somewhat to 77 in 2007 as home prices start to fall.
These are disastrously low levels, far below any historical experience.
76 We rely on Foote et al. (2008) for initial rates and margins for 2004–2007. For earlier

years, we assume a 25% increase in principal and interest payments on reset. The average
increase on reset for 2004–2007 was 23%.
77 Merrill Lynch (2007).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 29

150

140

130

120

110
ARMs (as reported by NAR)
100
Downpayment Adjusted
90 Downpayment and Reset Adjusted Index

80

70

60

50
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Figure 5: Adjusted Housing Affordability Index for Nonprime ARMs

Note: Reported and Adjusted Housing Affordability Index for Nonprime ARMs for 1995–2010,
as reported by NAR. The source for initial rate and margin: Foote et al. (2008) at 291, 298.
The downpayment-adjusted index is based on median CLTVs from 2004 for Purchase Subprime
Mortgages from Mayer et al. (2009). For 2003 and prior years, the graph assumes a 25% increase
in payments on reset (based on data from 2004 to 2007).

The message from Figure 5 is stark: The typical nonprime borrower was
not close to being able to afford an ARM loan once the reset occurred. Indeed,
many could not afford their loans even before the reset. Lenders used interest-
only and negative-amortization loans to camouflage the underlying affordability
issues. The fraction of nonprime loans that were fully amortizing (payments
sufficient to repay principal over the life of the loan) crashed from around 95%
in 2001 to 35% by 2006.78 The choice for many homeowners became refinance,
sell, or default. And refinancing only worked if prices continued to rise.
Even the bottom line shown in Figure 5 is optimistic, given that many
nonprime borrowers took out lo-doc loans and reported inflated income. Sup-
pose we assume an average overstatement of income of 10%—which seems
reasonable, even conservative, given the proportion of lo-doc loans and the
limited data on true income for lo-doc borrowers (see Part II). Affordability
would have fallen to around 90 or below in the early 2000s as lo-doc loans
became more common (and, likely but unprovably, income exaggeration in-
creased as mortgage brokers learned that no one was spot-checking those loans).
Affordability would have fallen into the low 60s by 2005.

3.1.3 Rising Supply of Unsold Homes

A further warning sign was a rising supply of unsold homes. Figure 6 provides
monthly data from 2000–2010. The number of unsold homes rose from around
1.6 million in January 2000 to around 2.2 million on average during 2004. The
supply then swung up sharply in 2005, rising to 2.5 million in January 2006
78 Lockhart (2007).

Electronic copy available at: https://ssrn.com/abstract=1973662

30 Bernard S. Black et al.

4.00

3.50

3.00

2.50

2.00
Unsold Homes (Millions)

1.50

1.00

.50

.00
Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10

Figure 6: Supply of Unsold Single-Family Homes

Source: NAR, Existing Single-Family Home Sales: Housing Inventory, at https://research.
stlouisfed.org/fred2/series/HSFINVUSM495N{#} (last visited Jan. 26, 2018).

30% 1.60%

20% 1.40%

10% 1.20%

0% 1.00% Housing
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Start Year
Over Year
-10% 0.80% Growth
Population
-20% 0.60% Year Over
Year Growth

-30% 0.40%

-40% 0.20%

-50% 0.00%

Figure 7: Housing Starts v. Population Growth

Source: Source for Housing Start Year over Year Growth: U.S. Census Bureau U.S. Department
of Housing and Urban Development, Housing Starts: Total: New Privately Owned Housing
Units Started, at https://research.stlouisfed.org/fred2/series/HOUST (last visited Jan. 26, 2018).
Source for Population Year Over Year Growth: Population and Housing Unit Estimates, at
http://www.census.gov/popest/ (using mid-year intercensal estimates for 1991–2009) (last visited
Jan. 27, 2018).

and 3.3 million in August 2006. Housing prices continued to rise, which would
further stimulate housing starts. The excess supply implied home prices would
likely need to fall in order for the housing market to regain equilibrium.
As Figure 7 shows, housing starts also exceeded population growth during
this period. This was a further reason to expect the housing market to cool,

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 31

which would put pressure on nonprime home prices due to the positive feedback
structure of nonprime lending. The data presented in Figures 6 and 7 were
readily available. For example, the supply of unsold homes is contained on the
National Association of Realtors website and is updated monthly.

3.1.4 Income Was Not Increasing at the Same Rate as House Appreciation

The declining affordability shown in Figure 5 was readily knowable. Housing

prices were rising rapidly during the boom period, and incomes were not
keeping pace. Figure 8 shows the difference between growth rates in housing
prices and median incomes from 1995 to 2007, in nominal dollars. Beginning
in 2000, growth in the median price for single family homes outpaced growth
in median family income. From 2000 to 2006, the median home price grew by
an average of 6% per year, while median family incomes grew at an average of
3%. This was another source of evidence that housing prices were departing
from underlying fundamentals.

3.2 Rating Agency Assumptions and Investor Ignorance about Them

As the housing boom continued, a correction became more likely, its likely size
grew, and the chances of a significant decline in both real and nominal dollars
rose as well. The actual decline began in the second half of 2006, with the
Case-Shiller national index falling in nominal dollars from 189.93 in 2Q 2006
to 183.17 in 2Q 2007 (a 3.6% drop). By early 2007, one-year futures contracts
on the Case-Shiller index were predicting a further 4% drop over the next year.
It was knowable that the rating agencies assumed away any risk of a
national housing decline in their models. What is remarkable is the failure
of the rating agencies to revisit their assumptions when prices started falling
and the apparent failure of investors to ask what housing price appreciation,
default correlation, and other assumptions the rating agencies were using.
When money managers and investors did ask about the agencies’ assump-
tions for housing price appreciation, the answers were stunning. Mutual fund
manager Robert Rodriguez reported that, when his analyst questioned Fitch
at a press conference about the assumptions underlying its ratings model in
early 2007, Fitch replied that its ratings of subprime mortgages were largely
based on two factors: FICO scores and predicted Home Price Appreciation
(HPA) rates, for which Fitch assumed low- to mid-single-digit appreciation.
Thus, by early 2007, Fitch was imagining a world that no longer existed and
not modeling the world that did exist, even as a stress scenario.
Mr. Rodriguez described Fitch’s response when further pressed: “My
associate then asked, ‘What if HPA was flat for an extended period of time?’
They responded that their model would start to break down. He then asked,
‘What if HPA were to decline 1% to 2% for an extended period of time?’ They

Electronic copy available at: https://ssrn.com/abstract=1973662

32 Bernard S. Black et al.

250,000 100,000

200,000 80,000

150,000 60,000

Median Price: Single Family Home

Family Income

100,000 40,000

50,000 20,000

0 0
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Figure 8: Income and Housing Price Appreciation

Source: Source for Median Price: Single Family Home (left-hand y-axis): NAR, Housing Af-
fordability Index, at www.realtor.org/topics/housing-affordability-index (last visited Jan. 27,
2018). Source for Family Income (right-hand y-axis): Census Bureau American Commu-
nity Survey, as reported by the NAR, Housing Affordability Index, at www.realtor.org/topics/
housing-affordability-index (last visited Jan. 27, 2018).

responded that their models would break down completely.” 79 The limits of
the Fitch model, and the lack of realistic “downside” assumptions, apparently
came as a surprise to money managers and investors.
How about the other two major rating agencies? Investor Steve Eisman
told author Michael Lewis that he had pressed S&P for details regarding the
assumptions underlying its subprime loan models. When he asked S&P what
would happen if real estate prices fell, S&P was unable to tell him. S&P used
a logarithmic model for home prices, which could not accept negative numbers.
S&P literally could not predict default rates if national home prices were flat
or declining.80
For the third agency, Moody’s, we have found no public source on the
assumptions they used in their model, but we were able to confirm informally
from a source at Moody’s that they too assumed steadily rising home prices.
We are not sure which should have been more alarming—the unrealistic
assumption or the fact that Moody’s apparently never needed to disclose, then
or later, what assumptions it was making.
The agencies’ models also used data on nonprime default rates from a
period of generally rising home prices. For example, the Moody’s and S&P
79 Rodrigues (2007).
80 Lewis (2008).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 33

models were based primarily on the performance of subprime mortgage loans

in the mid-1990s, a period of housing price appreciation (see Figure 2) and
far fewer flaky loans.81 The agencies could have used data from the subprime
mortgage loan crisis of the late 1990s, but chose not to do so. Nor did they
update their data as prices rose and loan terms became flakier in the 2000s.
The agencies’ models also ignored many characteristics of the nonprime loans
that underlay the securities they rated. For example, their models had no
way to allow for: the low FICO scores of some individual borrowers (they
instead used only the average FICO scores of the borrowers that formed a
pool of mortgage loans); fraud risk for lo-doc loans; or second mortgage loans
that increased many borrowers’ CLTVs and left these borrowers with little
or no equity and hence little incentive to repay their mortgages.82 Moody’s
model of the correlation across assets in CDOs used a “diversity score” created
when MBS were a small fraction of the loans that comprised CDOs. This was
inappropriate for the nonprime CDOs of the mid-2000s, in which low-rated
nonprime MBS were the dominant assets included in CDOs.83
Nonprime CDO ratings were further distorted by out-of-model adjust-
ments by the rating agencies. Those agencies used different assumptions for
probabilities of borrower default and correlations across mortgage loan assets,
but their outputs—the fraction of securities they rated as AAA, AA, and
so on—were remarkably similar. A principal reason was that the agencies
cheated on their own models by applying large, often undocumented, “out of
model” adjustments. A typical adjustment for nonprime CDOs converted a
BBB rating to AAA!84 The rating agencies inflated their ratings to remain
competitive. In one instance, S&P increased the fraction of nonprime CDOs
that would receive a AAA rating by 4% over its model in order to better match
Moody’s, whose assumptions had produced more AAA-rated securities than
S&P’s model.85
Even when the risks were obvious, as with the fraud risk for lo-doc loans,
no one investigated to assess the effect of relaxed screening on default risk.
As best we can tell, the rating agencies simply assumed that stated income
was true income and that lo-doc loans would perform similarly to full-doc
loans.
The models themselves were influenced by similar pressures. The Statement
of Facts accompanying S&P’s settlement with the Department of Justice made
clear that the risk of losing revenues influenced how S&P constructed a core
81 Senate Report on Financial Crisis (20111) (“the models operated with subprime data

for mortgages that had not been exposed to stagnant or falling housing prices” citing
memorandum from Daniel Nuxoll to Stephen Funaro, “ALLL Modeling at Washington
Mutual,” FDIC_WAMU_000003743-52, at 47).
82 Lewis (2008).
83 Senate Report on Financial Crisis (2011).
84 Griffin and Tang (2012).
85 Griffin et al. (2013).

Electronic copy available at: https://ssrn.com/abstract=1973662

34 Bernard S. Black et al.

CDO model and its later decisions on whether and when to downgrade CDOs
experiencing higher-than-expected delinquency rates and loan losses.86
Even when rating agencies began to adjust their models beginning in
2006 to reflect rising default rates on nonprime loans, investors’ demand for
such products initially remained robust, with $1 trillion in nonprime loan
originations in 2006. Issuance of nonprime CDOs soared.
If one looks, it was knowable that CDO default rates for a given rating,
under the rating agencies’ own models, were far higher than MBS or corporate
bond default rates! Moody’s published data in 2005 showed that default
rates on Baa-rated CDOs were approximately ten times those on Baa-rated
corporate bonds.87 S&P’s target three-year default rate for an AA CDO was
comparable to an A- MBS.88 Perhaps some investors discovered this, but many
surely did not realize that the same rating implied very different default risk
as one moved from corporate bonds, to MBS, to CDOs.

3.3 Originators Had Moved From “Originate to Hold” to “Originate

to Distribute”
It was well known that nonprime loan originators had moved from an “originate
to hold” to an “originate to distribute” model. The institutions involved in
creating subprime products were numerous: the mortgage broker, the loan
originator, the bank that originated and repackaged loans into MBS, and the
perhaps-different bank that engineered CDOs and CDO2 s. Each intermediary
charged a fee and transferred most or all of the risks down the line. Loan
originators and securitizers had little “skin in the game” that might provide
incentives to screen borrowers.89

3.4 Nonprime Securitization Might Shut Down, as It Did in the Late

1990s

Securitization of subprime loans began in the 1990s (Alt-A was not yet in
existence on a meaningful scale).90 Before the 2007–2009 crisis, there had
been a smaller subprime crisis in the late 1990s following the East Asian and
Russian debt crisis and the failure of Long-Term Capital Management. Those
events prompted an investor flight to quality and away from risky assets such
as subprime MBS. Subprime mortgage securitization fell by approximately
18% from 1998 to 1999, and by 2000, subprime loan originations had fallen

86 U.S. Department of Justice (2015).

87 Id.;Tomlinson and Evans (2007).
88 Nomura Securities International (2006).
89 Purnanandam (2011); Acharya et al. (2009).
90 Financial Crisis Inquiry Commission (2011).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 35

by 26% compared to the preceding two years.91 The failure of a few sub-
prime lenders did not threaten major banks or significantly affect the overall
housing market;92 we are not aware of any research on the effect of lower loan
originations on nonprime home prices.
The sharp contraction in subprime securitization in the late 1990s, leading
to a contraction in lending, suggested a major risk for the nonprime boom
of the 2000s. Investors might ignore rising risk for a while, but likely not
forever. An event that sparked a pullback from nonprime MBSall —even, as
in the 1990s, an event not directly related to nonprime lending—could spark
a drop in lending, which would (almost of necessity) affect nonprime home
prices. And even a flattening, let alone a drop, in nonprime home prices would
trigger severe positive feedback, which would reinforce any investor pullback
from nonprime MBSall .
The sensitivity of nonprime securitization to rising risk was higher in the
2000s than in the 1990s. In the 2000s, unlike the 1990s, nonprime MBS offerings
rested on the ability of the securitizers to repackage lower tranches into CDOs.
Even a modest rise in MBS risk, enough to affect the BBB tranches, could
spook CDO investors. As Mason and Rosner stressed in early 2007, “the CDO
market’s willingness to move out of collateral types at a moment’s notice”
posed a major threat to nonprime lending generally.93
Any pullback could be both sudden and sharp—as, in the event, it was.
Thus, the 1990s contraction ought to have, but did not, serve as a “canary in
the coal mine” moment for rating agencies and investors.94 Yet, we have found
no evidence that originators, securitizers, or investors asked, “What is likely to
happen to nonprime home prices if nonprime securitization shuts down (or
significantly contracts)?”

3.5 Warnings about a Bubble

If one wants to diagnose a possible bubble in hindsight, it helps for there

to have been bubble warnings in foresight. We discuss several warnings in
this section. The doom-sayers might have been wrong, but their views surely
counseled in favor of investigating what might happen to nonprime MBSall if
they turned out to be right.
An early warning came from Congressman Ron Paul at a 2003 hearing of
the House Financial Services Committee. He warned of the “day of reckoning”
when the housing “bubble” ended.95 In 2004, economists Robert Shiller and
Karl Case presented evidence of a growing housing bubble. First, the authors

91 Id.
92 McLean and Nocera (2010).
93 Mason and Rosner (2007a).
94 Id.
95 Paul (2003).

Electronic copy available at: https://ssrn.com/abstract=1973662

36 Bernard S. Black et al.

looked at the relationship between housing prices and personal income per
capita from 1985 to 2002. Case and Shiller observed that, in most states, the
rise in housing prices could be largely explained by rising borrower income,
but in eight states, including California, Massachusetts, New York, and New
Jersey, the correlation between income and housing prices was weaker.96 Case
and Shiller concluded that “the pattern of smoothly rising and falling price-to-
income ratios and the consistent pattern of under-forecasting of home prices
during 2000–02 mean that we cannot reject the hypothesis that a bubble exists
in these states.” 97 By 2005, Shiller was more confident of a housing bubble
and published a second edition of his book, Irrational Exuberance, in which he
added the housing bubble to the dot-com bubble that had been at the center
of the first edition.
Beginning in late 2004, The Economist warned that there “were plenty of
symptoms of a [housing] bubble mentality in the United States.” 98 Using the
average price/income ratio from 1975 to 2000 as a baseline, it estimated that
American homes were roughly 30% overvalued. The Economist also noted
concerns over a housing bubble expressed by International Monetary Fund
and Goldman Sachs economists, including a Fund warning over the adverse
impact on the global economy of a downturn in the housing market.
In June 2006, Dean Baker, co-founder of the Center for Economic and
Policy Research, examined ten different economic indicators, including home
sales, vacancy rates, and housing starts. He warned that “[e]vidence [was]
mounting that the housing bubble ha[d] passed its peak.” 99
Thus, from a variety of sources, there were growing signs by the mid-2000s
that the United States might be in for a period of flat or declining housing
prices, and perhaps in an outright housing bubble. Whatever the bubble risk
was, it was surely far above the very low risk levels that AAA-rated nonprime
MBSall promised.
In sum, there were multiple pieces of evidence suggesting there was a
substantial risk that home prices might fall and take many nonprime MBSall
with them. The risks were readily apparent, had the professionals in the
securitization chain chosen to see them.

4 Some Thoughts Toward Reform

To this point, our focus has been on explaining why MBSall risks were both
large and obvious. We have also stressed the positive feedback nature of
nonprime lending. Nonprime lending terms, and the interaction among those
96 Case and Shiller (2004).
97 Id.
98 The Economist (2004).
99 Baker (2006).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 37

terms and home prices and nonprime securitization markets, created strong
positive feedback to the point where a mere flattening in nonprime home prices
would likely cause a crash in those prices and soaring default rates.
In our view, investor oversights cannot be explained as reasonable mistakes
by purchasers of complex securities. We have argued that more was involved—
that market participants were willfully blind at all stages of the lending,
structuring, and purchase chain, from nonprime mortgage originators to the
banks who packaged mortgages into MBSall , the rating agencies who rated
them, the money managers who purchased them, and the institutional investors
who provided funds to these money managers. At all levels, the professionals
involved in bringing MBSall to the market benefited personally—through
paychecks or careers—by promoting nonprime MBSall rather than assessing
their risks. The complex securitization process, with nonprime mortgages
flowing into MBS, lower MBS tranches flowing into CDOs, and often lower CDO
tranches flowing into CDO2 s, made it easier to hide, concentrate, and ignore
those risks. High credit ratings, and a favorable recent past of steadily rising
home prices, gave investors in nonprime MBSall an excuse not to investigate
the risks too closely.

4.1 Regulatory Responses to Date

We now argue that the regulatory responses to date have done little to prevent
a recurrence of willful blindness (perhaps directed at a different asset class),
and nothing to address the risks posed by positive feedback lending. An impor-
tant focus of the Dodd–Frank Wall Street Reform and Consumer Protection
Act (Dodd-Frank Act) has been on improving mortgage quality. Among other
things, the Act created the Consumer Financial Protection Bureau, which
is charged with protecting consumers against misleading financial products
(including mortgages, but excluding securities and other specified assets). The
Dodd-Frank Act also imposes minimum underwriting standards for residential
mortgages. It requires lenders to make “a reasonable and good faith deter-
mination . . . that, at the time the loan is consummated, the consumer has
a reasonable ability to repay the loan, according to its terms.” 100 This, in
effect, bars use of lo-doc loans. But interest-only loans, and indeed negative
amortization loans, are still allowed – for both prime and non-prime borrowers.
The Act does give preferential regulatory treatment to higher-quality “qualified
residential mortgages” by exempting banks that securitize these mortgage
loans from the “skin in the game” requirement discussed below.
Higher-quality mortgages are likely to result in a more stable securitization
process. The cost—reduced access to mortgage loans by borrowers who cannot
document their income or realistically expect to repay their loans based on their
100 Dodd-Frank Act §1411(a)(2), implemented in Federal Reserve Board Regulation Z,

12.F.R. Part 226.

Electronic copy available at: https://ssrn.com/abstract=1973662

38 Bernard S. Black et al.

current income—seems modest. The Dodd-Frank Act also established new

requirements for appraisal independence, implemented by a Federal Reserve
Board rule.101 But much else remains unchanged. There are no limits on
CLTV or loan/income ratios. There are no risk retention requirements for
loan originators. Anecdotally, some aggressive nonprime mortgage structures
have returned to the market.
When we move to other players, the Dodd-Frank Act does much less. For
securitization, it obligates banks that create MBSall to retain at least 5% of
the credit risk. Implementing regulations were adopted in late 2014, but their
practical import remains unclear. We suspect that banks will often succeed in
limiting risk or limiting the period for which they retain risk.
The process by which banks concentrated risk, by repackaging lower-rated
MBS tranches into CDOs and lower-rated CDO tranches into CDO2 s, remains
largely unregulated. Thus, banks can still use elaborate structures to camou-
flage risk and minimize the number of buyers they needed for the resulting
“toxic waste.” The banks remain free of explicit due diligence obligations, as
long as they sell structured securities to institutional investors pursuant to
Rule 144A under the Securities Act of 1933 (described below), rather than to
the public.
The Dodd-Frank Act minimally affects the credit rating agencies. It
modestly increases regulation and supervision, including directing the Securities
and Exchange Commission (SEC) to adopt rules on conflicts of interest between
the agencies and issuers of asset-backed securities (ABS).102 The Act eliminates
direct regulatory reliance on credit ratings. It requires some ratng agency due
diligence, although how much is unclear. But the market pressure to have
securities rated has not changed. And the core problems the crisis unearthed
remain untouched. These include the agencies’ ability to use proprietary
models without disclosing core assumptions (such as the pre-crisis assumption
that nominal housing prices could not decline nationally), the pressure from
banks on agencies to provide generous ratings or risk losing large amounts
of business, the agencies’ ability to add out-of-model adjustments to meet
competitive pressures, the agencies’ ability to assume away important risks
instead of evaluating them, the agencies’ ability to hide their reports from
academic or public scrutiny, and the career incentives that give MBS-raters
within rating agencies incentives to be liberal.
The Dodd-Frank Act does not touch the money managers who invested in
nonprime MBSall while closing their eyes to the risks in order to earn a few
extra basis points and appear to be outperforming their competitors. Nor does
it touch the institutional investors who provided funds to these managers in
search of those few basis points.
101 See Federal Reserve Board Regulation Z, 12 C.F.R. 226.42.
102 See 17 C.F.R. 240.17g-5 (paragraph (b)(9) identifies rating ABS, if paid for by the
issuer, sponsor, or underwriter, as a conflict of interest subject to the regulation).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 39

4.2 The Reform Road Not Taken

The reforms to date miss several key points. First, they do not directly address
the willful blindness of market participants. Loan originators could not have
made the risky mortgages that underlay nonprime MBSall unless they could
sell their loans to banks who ignored the risks (or, worse, partly knew them
and profited from foisting bad assets on unwitting buyers),103 rating agencies
who used ludicrous models to assess credit quality, and money managers (and
their institutional investors) who sacrificed safety for nominal yield. At any
stage, the risks in nonprime mortgages could have been identified, and the
marketplace—the “demand” for nonprime mortgages and MBSall —could have
taken the risks into account in deciding how (if at all) to price them. That
collective blindness may be directed the next time at another asset class. Thus,
the regulatory response cannot be limited to nonprime mortgage lending.
Second, the severe problems with subprime mortgages reflected both bor-
rowers’ credit quality per se (could borrowers repay their loans?), and, of equal
or greater importance, the positive feedback effects that resulted from how
nonprime loans were structured and how nonprime lending, nonprime home
prices, and nonprime MBSall related to each other. Without that feedback, the
nonprime housing bubble would have been smaller, the correction would have
been far less sharp, many nonprime MBS would have performed reasonably
(if not perfectly), and nonprime CDOs would not have become the wasteland
that they became, with losses approaching 100% on many issues and wiping
out both regular and “super-senior” AAA-rated instruments. Positive feedback
lending in part involves an externality—each lender’s actions strengthen the
overall cycle. But the feedback structure also relates to willful blindness—it was
only possible because capital providers turned a blind eye to the risks involved.
Even when Dodd-Frank Act provisions move in the right direction—for
example, the skin-in-the-game requirements for MBSall originators—they do
so in a clumsy way. Banks that create MBSall must retain some risk. But
there is no express requirement in the typical Rule 144A offering that the
banks disclose the remaining risks in the MBSall offering documents. Thus,
the Dodd-Frank Act adds a layer of regulation, but does not reinforce the
market discipline that might also constrain banks to hold some residual risk.
In other words, the Dodd-Frank Act has gotten it mostly backwards. Stan-
dards for underwriting mortgages (supply) are useful, but far from sufficient.
Most of the problem, and hence most of the focus, needed to be on the
process by which capital for new mortgages was raised (demand, fueled by
securitization).
103 Goldman Sachs did so in the notorious Abacus offering, where it allowed John Paulson
to choose the assets in a CDO, so that he could bet against them, and forgot to disclose this
to the buyers. See Morgenson and Story (2009). And Morgan Stanley did so in its STACK
2006-1 and -2 offerings, where Morgan Stanley bet against the equity layers without telling
the buyers. See Weiss (2013).

Electronic copy available at: https://ssrn.com/abstract=1973662

40 Bernard S. Black et al.

4.3 What to Do Instead

4.3.1 Addressing Willful Blindness

Our core approach to addressing willful blindness in the capital-raising process

is a combination of disclosure and explicit due diligence requirements, detailed
below. The core idea: If we rub the noses of participants in the muck, it will
be harder for them not to look or later to claim that they just did not know.
Consider money managers. If they know the main risks (and cannot avoid
knowing them, because rating agencies and securitizers will be required to tell
them), and must disclose those risks to their clients, their incentives should
change. If money managers invest anyway, without good disclosure to their
clients, they will face legal risk for their failed investments. If they do disclose
the main risks, then: (i) some investments will not be made (no one will want to
tell clients, or be questioned ex post, about having bought securities known on
the street as “toxic waste”); and (ii) institutional clients can more easily police
the money managers and choose the level of risk they intend to take. Some
clients will be less sophisticated and may still invest foolishly, knowing the risks
or not really thinking through them. But there is little to be done in that case.
For the rating agencies, their modeling assumptions should be disclosed,
both to investors and publicly. Public disclosure will let researchers assess the
realism of the rating agencies’ assumptions. There is a case for allowing the
agencies to treat modeling details (as opposed to assumptions) as proprietary,
but only a limited one. In practice, major banks will know many of those details,
in part because employees move back and forth; and in part from negotiations
over what assets must be in a pool for the bank to obtain particular ratings
treatment. But for core assumptions, there needs to be explicit disclosure. If
the agencies want to assume that housing prices will never decline nationally
in nominal dollars, they should be required to say so, conduct diligence to
assess whether this assumption is reasonable, and defend that reasonableness in
writing. What they say should be in the offering documents for any offering they
rate, subject (as discussed below) to a recklessness standard for not conducting
due diligence or providing adequate disclosure. It is also absurd that basic
rating agency reports on their nonprime rating process are subscription-only.
The rating agencies make their money by charging for ratings; they do not
need to hide their opinions and assumptions behind a pay wall.104
Likewise, securitizers should be required to disclose what due diligence they
conducted and what interest (if any) they have in the success or failure of an
offering. This will let noteholders assess the extent of any conflicts of interest.
Sunlight remains a good disinfectant. What rating agencies and securitizers
104 An anecdote: In researching this article, we were often unable to obtain rating agency

publications that others had cited (without paying thousands of dollars for them, which we
did not have). Mason and Rosner (2007b) also note that, prior to the crisis, even financial
regulators did not have access to ABS deal documents. They still lack that access.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 41

were willing to do in the gloom of limited disclosure, they might not do if

subject to public access and criticism. Due diligence also creates a paper trail,
and thus liability risk, if their disclosures are incomplete or misleading.
Two advantages of our disclosure and due diligence approach: First, it can
be quite general, and apply to a broad variety of “semi-public” forms of capital-
raising under Rule 144A. It may forestall or weaken the next blindness-induced
crisis, likely involving a different asset class. Second, it relies on enhancing
market discipline for the capital-raising process, rather than substantive regu-
lation (such as skin-in-the-game rules), which can be highly technical and may
be gameable.

4.3.2 Addressing Positive Feedback Lending

Positive feedback has been a recurring source of financial problems. Other

examples include portfolio insurance, which strongly reinforced the 1987 stock
market crash;105 and low margins on repurchase agreements in good times,
which rise in bad times, can spark a liquidity crisis, and helped turn the
nonprime crisis into a broader financial crisis.106 The external costs of positive
feedback are not felt by any one securitizer or investor, and thus may require
substantive regulation or some other mechanism that would cause lenders and
other financial actors to internalize the costs.107 No post-crisis rule directly
addresses positive feedback lending. The post-crisis rule that lenders must
assess mortgage borrowers’ ability to repay their loans will help. But lenders
can still pressure borrowers to refinance or sell in a number of ways. Lenders
can still stretch the boundaries of affordability by allowing high DTI and
CLTV ratios, and offering interest-only or negative amortization loans. They
can still offer teaser rates, with punitive resets, intended to force refinancing.
What else can be done? Disclosure and due diligence requirements can
help. They can include a specific requirement that securitizers and rating
agencies assess and disclose the extent to which structured securities, or the
underlying assets, have positive feedback features that, if widely used, could
affect an overall market. Rating agencies and money managers would then
need to take positive feedback into account in their ratings and investment
processes, instead of ignoring it. Money managers should be required to
disclose investments in positive feedback assets, and the nature of the feedback,
to their clients. Major institutional clients should be highly averse to positive
feedback risk, because it can affect the rest of their portfolio. Once again, some
clients may invest foolishly, knowing the risks or not really assessing them.
But the risk of a crisis sparked by positive feedback will be reduced—perhaps
substantially so.
105 Shiller
(1988).
106 Gorton and Metrick (2012).
107 See Whitehead (2011).

Electronic copy available at: https://ssrn.com/abstract=1973662

42 Bernard S. Black et al.

As a way of focusing this disclosure, securitizers and rating agencies could

be required to assess ABS performance under a range of pre-specified scenarios,
for example, average annual home price appreciation (HPA) (−3%; −2%; −1%;
0%; 1%; 2%; 3%) and future changes in interest rates (no change; +1%; +2%;
+3%; +4%; +5%). The result would be a grid: How would a given loan
perform if most loans (say, 75%) in the market had similar features.
A scenario with −3% HPA and a 5% rise in interest rates is severe, but by
no means impossible. Rates have risen by more than this before (in the 1970s
and early 1980s), and a spike in interest rates would likely induce falling home
prices. Could AAA-rated ABS survive? If not, they should not be seen as a
truly AAA. The scenarios will rub the money managers’ and investors’ noses
in what might happen, even if the rating agencies pretend it never will.
Is more needed? Should the Federal Reserve Board, say, draft substan-
tive limits on positive feedback lending, much as it sets minimum capital
requirements for banks and margin requirements for secured lending against
securities? We are incrementalists, concerned over the unanticipated conse-
quences of such a major change, and so at this point recommend only that
its feasibility be investigated and, perhaps, implemented in staged steps.108
But the special risks posed by positive feedback do provide additional impetus
behind proposals to impose market discipline on banks by requiring them to
issue minimum amounts of subordinated debt, so that the debtholders—and
the publicly-visible interest rate—can provide market discipline.109

4.4 Specific Proposals

There is no way to prevent willful blindness or positive feedback lending,
but there are ways to reduce the likelihood and magnitude of each. Today,
a substantial portion of ABS—and, most of the riskier and more complex
ABS offerings, such as CDOs—are sold under Rule 144A. We believe that the
following regulatory changes are appropriate for Rule 144A offerings, although
we pass over important details, including which agency (if any) has both the
legal authority and expertise to adopt the necessary rules:110
1. Banks and other securitizers should be subject to the same disclosure and
due diligence standards as would apply in public offerings of securities,
plus specific requirements for specific asset types.
2. Banks and other securitizers should be liable to ABS investors for due
diligence or disclosure failures under a standard that is stricter than the
current scienter requirement. That standard might be negligence, as in
108 See Whitehead (2012).
109 See,e.g., Evanoff and Wall (2000).
110 Readers who are expert in securities law will notice various lacunae and imprecisions in

our discussion of current law. These do not meaningfully affect our reform proposals.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 43

public offerings, or perhaps a moderately higher standard such as gross

negligence. One would want to know the views of major institutional
investors on the liability standard, which involves balancing the costs of
a stricter standard against its benefits.

3. Rating agencies should be subject to due diligence and disclosure stan-

dards for any ABS offerings they rate, including disclosure in the offering
documents of significant modeling assumptions, the data that are fed
into the models, and the extent of, and reasons for, any out-of-model
adjustments. The liability standard for failure to meet the due diligence
and disclosure obligations should be stricter than scienter, but perhaps
less strict than for banks and other securitizers. For example, banks
might face a gross negligence standard, while rating agencies face a
recklessness standard.

4. All ABS offering documents should be publicly available, with an ex-

ception for unusual asset classes that are too small to have potential
systemic consequences.

5. There should be specific disclosure requirements for any positive feedback

aspects of lending and securitization structures.

6. Money managers who invest in ABS are fiduciaries. They should be

required to conduct and document their own due diligence and disclose
it to their investors, and should be liable to their investors under the
usual negligence standard for fiduciaries.
We propose both explicit due diligence rules, and rules for documenting
the diligence that is performed. These steps will make it easier for courts
to find liability in practice.

These proposals would narrow the gap between Rule 144A offerings and
traditional registered public offerings. We would allow Rule 144A offerings to
benefit, relative to public offerings, from a more relaxed liability standard if
regulators find, as part of the rulemaking process, that institutional investors
find the cost-benefit calculus to favor a more relaxed standard than the
negligence standard that applies to underwriters in public offerings. But the
basic due diligence and disclosure rules should be similar. We should not
again run the risk that the Rule 144A market, which has become large enough
to spark a major financial crisis, remains vulnerable to willful blindness and
opaque to scrutiny by regulators and academics.
In contrast, the current system encourages—or, at least, does little to
discourage—willful blindness. Relaxed and ambiguous due diligence standards,
coupled with a scienter standard for liability, encourages blindness, because
it is difficult for investor plaintiffs to impose liability on banks for what they

Electronic copy available at: https://ssrn.com/abstract=1973662

44 Bernard S. Black et al.

did not know. Our proposals will also extend due diligence obligations to
rating agencies and money managers, subject the rating agencies to disclosure
obligations, and strengthen the liability standards for each group.
Positive feedback lending—and the systemic risks it poses—is a separate
problem that may require direct regulatory intervention. But our due diligence
and disclosure proposals will help and may instill enough market discipline so
that we can avoid the need for substantive regulation.
We provide selected details on our proposals below.

4.4.1 Enhanced Disclosure in Rule 144A ABS Offerings

SEC Rule 144A is a safe harbor exemption from the registration requirements
of the Securities Act of 1933 for sales of qualifying securities to qualified
institutional buyers. Rule 144A offerings are not subject to the due diligence
and disclosure requirements that apply to registered public offerings. The
liability standard is the relaxed scienter standard under Rule 10b-5, instead of
the essentially strict liability standard for issuers, and the negligence standard
for underwriters, in registered public offerings.
SEC Regulation AB, adopted in 2004, sets out registration, disclosure, and
reporting requirements for public offerings of ABS. In 2014, in response to
concerns raised during the financial crisis, the SEC adopted new disclosure
and reporting requirements (so-called “Regulation AB II”) for registered public
offerings by issuers other than the GSEs. The new rules required enhanced
loan-level disclosure for both residential and commercial MBS, in standardized,
tagged data (XML) format, so that investors can download the information
directly into spreadsheets for analysis and modeling. The required disclosures
apply both at the time of issuance and, in some cases, during the life of the
securitization, and include: (i) basic data about the payment stream of each
asset, such as contractual terms, scheduled payment amounts, interest rate
calculations, and whether the payment terms will change over time; (ii) data
about the collateral supporting the asset, such as property location, value, and
LTV ratio; (iii) historical performance data for the asset class over time, such
as delinquency; (iv) data about any losses that may pass through to investors
and loss mitigation efforts by the ABS servicer; (v) data about mortgage
insurance coverage, lien position, and whether income and employment have
been verified;111 (vi) the loan origination or underwriting criteria, and the
type and amount of assets that do not conform to these criteria; and (vii) the
amount and nature of any continuing interest in the ABS assets retained by
sponsors, servicers, and originators, as well as any material hedges used to

111 As noted above, other new rules effectively bar lo-doc mortgage loans. These rules,

however, do not require lenders to go beyond the documentation to verify the information
that borrowers provide.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 45

offset the credit risk. Finally, Regulation AB II requires disclosure for each
ABS sponsor’s prior securitized pools for the same asset class to show trends
in how groups of assets (“static pools”) perform over time.
Regulation AB II is comprehensive and mostly sensible in what it covers.
But it has a gaping loophole—it does not cover Rule 144A offerings. It should.
Those offerings are too large a part of the ABS market, especially newer, riskier
offerings, to be allowed to fly under the disclosure radar.

4.4.2 Due Diligence in Rule 144A Offerings

For public offerings of stocks or bonds, a number of actors—directors, senior

officers (including the CEO and CFO), underwriters, and accountants—have
explicit due diligence obligations. Much of this diligence falls away in the Rule
144A market. For nonprime MBSall offerings, the originators were not directly
responsible for the offering, the rating agencies disclaimed responsibility to
verify the data on which they based their ratings, and the banks doing the
securitizations had no affirmative due diligence obligations. Minimal diligence
became the norm.
This must change. For ABS, there should be affirmative due diligence
obligations, similar to public offerings. Moreover, diligence cannot be left
to bankers alone. Bankers’ incentives to engage in meaningful due diligence
rest, in part, on legal liability. But reliance on banks alone also rests on the
belief that banks care about maintaining their long-term reputation as credible
gatekeepers to the capital markets. The current banker culture—often driven
by short-term compensation set by demand across the marketplace, rather
than by an individual bank, with limited employee loyalty to the firm or vice
versa112 —is antithetical to careful due diligence and to banks’ willingness
to walk away from a bad deal. We have seen too many examples of “Banks
Behaving Badly, Again” (in the title of a New York Times editorial).113
Instead, diligence obligations should extend to the multiple participants in
the securitization chain, including originators, rating agencies, and the money
managers who buy ABS. Each will have different, although overlapping, roles.
The diligence standard can be similar to that for public offerings. Each should
be responsible for conducting a “reasonable investigation,” with having done so
being an affirmative defense to liability for inadequate disclosure, to be proved
by the defendant rather than something the plaintiff must disprove.
The details of who should be responsible for what are a matter for detailed
regulation and beyond the scope of this article. We expect that, as in public
offerings, the need to establish a due diligence defense will result in an elab-
orate due diligence process involving legal counsel, accountants, and other
professionals.
112 See Sepe and Whitehead (2015).
113 The New York Times (2014).

Electronic copy available at: https://ssrn.com/abstract=1973662

46 Bernard S. Black et al.

4.4.3 Liability Standards in Rule 144A Offerings

In public offerings, the federal securities laws effectively impose strict liability on
the issuer for a material misstatement or omission in the registration statement
at the time of effectiveness, unless the issuer can show that the plaintiff knew
of it when it acquired the security. Other participants to the offering may
avoid liability if they can sustain a due diligence defense—in effect, if they
can show non-negligence. Participants in Rule 144A offerings, in contrast,
are subject to liability under Rule 10b-5 under the Securities Exchange Act
of 1934 for selling securities through fraud, misrepresentation or deceit. The
required mental state is scienter—a term which the courts have interpreted as
severe recklessness, approaching a conscious disregard of material information.
There is no need to establish a due diligence defense. Many securities law
practitioners have counseled issuers and underwriters to conduct a due diligence
review nonetheless, on the basis that failure to do so could be regarded as
reckless. But, in MBSall offerings, this advice was often not heeded. Instead,
as the boom progressed, the banks progressively abandoned the loan-specific
diligence they had once conducted.114 Here, too, the banks were willfully
blind—they appeared to want to not know that many of the securitized loans
failed to meet the bank’s or the originator’s loan-quality guidelines.
The right liability standard is unclear. The Rule 10b-5 standard, with the
burden on the plaintiffs to show the defendants acted with scienter, is too
high a bar. Indeed, it invites participants to not investigate, and hence to
not learn about problems in an offering. A strong position would be to hold
participants in Rule 144A offerings to the same liability standard for material
misstatements or omissions that applies in public offerings—strict liability
for issuers and negligence for the other participants, with the burden on the
defendants to show they conducted due diligence. We propose a potentially
milder approach. We see recklessness as the minimum standard, but would
want regulators to ask institutional investors what liability standard they think
is appropriate and to impose a stricter standard if investors see this as justified.
Major institutional investors are well-placed to judge whether the benefits
from more careful diligence, and a stricter liability standard, outweigh the
associated costs. Whatever the diligence standard, the burden should be on
defendants to show that their diligence, and knowledge after diligence, met this
standard. The defendants’ burden would arise, as in public offerings, once the
plaintiff shows there was a material misstatement or omission in the offering
documents. That burden shift is what forces underwriters to conduct a due
diligence review in public offerings. We believe that an affirmative diligence
obligation, with liability for failing to meet that obligation, would have made
it far less likely that MBSall participants would have ignored what should have
been obvious.
114 Muolo and Padilla (2010) discuss the drop in loan-specific diligence.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 47

4.4.4 Public Disclosure of Rating Agency Models, Data, and Adjustments

For all ABS offerings, we believe the rating agencies must disclose to regulators
and the general public their significant modeling assumptions, the data fed into
the models, and the existence and reasons for any out-of-model adjustments.
A further step to consider is whether the models themselves should be public.
The upside would be greater transparency on what the rating agencies are
actually doing; the downside would be that, since there are only three main
rating agencies (Moody’s, S&P, and Fitch), full disclosure of model details
might foster tacit collusion.
The need for greater rating agency diligence is reflected in the Dodd-Frank
Act, which makes the rating agencies liable to investors if they knowingly or
recklessly fail to investigate the securities they are rating.115 We would go
further, especially for disclosure, but see this diligence obligation and related
liability as an important step in the right direction. The Dodd-Frank Act
also requires the SEC to hold rating agencies to the same standard of “expert
liability” that auditors, lawyers, and others face when they give opinions in
public offerings.116 But when the SEC issued its rating agency rule in 2010,
the rating agencies threatened to withhold ratings. The SEC retreated and
the rule has never been enforced.
In significant part, our approach bypasses the issue of whether the rating
agencies should be liable under a negligence standard (as in public offerings),
something closer to a scienter standard (as they now are under the Dodd-
Frank Act), or somewhere in between. Public disclosure of the rating agencies’
models, data, and out-of-model adjustments will impose market discipline
on the agencies’ assumptions, even without legal liability. If the agencies
publicly rely on fantastic assumptions, such as the ones they made privately
for nonprime MBSall , we expect that regulators, academic researchers, the
business press, and perhaps the institutional investors at the end of the chain,
who hired money managers who invested in nonprime MBSall , will be less
likely to ignore the obvious the next time a bubble starts to inflate.
Our approach also addresses the rating agencies’ threat not to rate. We
would make the agencies liable only for recklessness (unless institutional
investors, in the regulatory process, support a higher standard). That may
not change the standard that the Dodd-Frank Act now imposes on the rating
agencies. We would make the diligence obligations more explicit and require
disclosure of the agencies’ modeling assumptions and the diligence they conduct.
But we would extend liability to everything they rate. The SEC blinked in
the face of the agencies’ threat not to rate public offerings, rather than face
negligence liability. It would be a much harder for the rating agencies to credibly
threaten to rate nothing, and either shut down or find a new business model.
115 Dodd-Frank Act §933.
116 Dodd-Frank Act §939.

Electronic copy available at: https://ssrn.com/abstract=1973662

48 Bernard S. Black et al.

4.4.5 Certification

Sarbanes-Oxley Act §302 requires the CEO and CFO of a public company to
certify that they believe the company’s financial statements and other financial
information fairly present the company’s financial condition and results of
operations, and to have a reasonable basis for that belief. We would sup-
plement the due diligence requirements we propose with similar certification
requirements for appropriate personnel of mortgage loan originators, securitiz-
ers, and credit rating agencies. We propose that the key personnel involved in
a loan securitization verify that they believe, after reasonable investigation,
that: (i) the loans or other assets which are securitized meet the standards
they purport to meet (which nonprime loans often did not in the lead-up to
the financial crisis); (ii) the modeling and other assumptions on which the
securitization is based are reasonable and fairly disclosed; (iii) the models used
to estimate default risk are reasonable, given these assumptions; and (iv) the
data used to test the model are reasonable, given the models, the assumptions,
and the past history of returns on similar assets.
A certification requirement would expose to potential liability those people
in the best position to understand a securitization, its assets, and its rating, if
the assets later default and they were reckless in so certifying. An example
of how certification could help: The originators and securitizers surely knew
(or were reckless in not knowing) that many of the nonprime loans that were
sold to securitizers did not meet the securitizers’ purchase standards. That
degree of willful blindness would no longer be possible without exposing the
individuals involved to personal liability.

4.5 What About the Costs?

Our due diligence and disclosure proposals have costs. Do the benefits exceed
those costs? Our answer is an easy yes. The economic costs of financial crises
are huge. There are also important political costs through loss of confidence in
a market economy. The costs of due diligence and disclosure pale in comparison,
especially because the optimal level of diligence is not zero. A premise of the
relaxed regulation of Rule 144A offerings is that sophisticated participants will
choose optimally what should be investigated and what should be disclosed.
That premise largely failed. Instead, each participant in the MBS securitization
chain had private incentives to ignore the obvious. The result was a massively
suboptimal collective result for the nonprime housing and MBSall markets,
with large negative externalities for financial markets and the world economy.
Moreover, the potential gains from securitization are limited. The promise
of securitization is that risk will be diversified and then transferred to those who
can bear it more cheaply. But the potential gains from complex securitization
are a small fraction of the amount securitized. The concern is that, without

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 49

diligence and disclosure, risk will be hidden, concentrated, and sold to a few
fools who do not understand the risks they are buying. The nonprime bubble
fits the second picture much more closely than the first. “Structured squared
madness” was one term used to describe the MBS to CDO to CDO2 process,
with “Structured Investment Vehicles,” often created by banks, as major buyers
of nonprime CDOs and CDO2 s.117

5 Conclusion

We have argued that a combination of willful blindness and positive feedback

lending underlay the nonprime mortgage crisis, which in turn sparked the
financial crisis of 2007–2009. Willful blindness is hard to prevent. As Upton
Sinclair wrote, “It is difficult to get a man to understand something, when
his salary depends on his not understanding it.” 118 But explicit due diligence
requirements, with the need to disclose what you find, and liability for not
disclosing or not conducting diligence, can help for Rule 144A offerings much as
it helps for public offerings. We propose diligence and disclosure requirements
for Rule 144A offerings covering the major participants in the ABS market.
Positive feedback lending is also hard to prevent. It may seem privately
rational to individual lenders, who fail to see (or see, but ignore) the system-
wide consequences of many lenders behaving similarly. Here, too, due diligence
and disclosure requirements can help. They may prove not to be enough, but
they are an important step in the right direction, and perhaps a sufficient one.

117 Felsenheimer and Gisdakis (2008).

118 Sinclair (1935).

Electronic copy available at: https://ssrn.com/abstract=1973662

50 Bernard S. Black et al.

References

Acharya, V., T. Philippon, M. Richardson, and N. Roubini. 2009. “The Finan-

cial Crisis of 2007: Causes and Remedies”. Financial Markets, Institutions
& Instruments. 18: 89.
Adelino, M. 2009. “Do Investors Rely Only on Ratings? The Case of Mortgage-
Backed Securities”. unpublished paper. url: http : / / fisher . osu . edu /
supplements/10/9861/01132010- Manuel%20Adelino.pdf (accessed on
01/26/2018).
Baker, D. 2006. “Is the Housing Bubble Collapsing? 10 Economic Indicators
to Watch, Center for Economic and Policy Research Issue”. Brief. url:
www.cepr.net/documents/housing_indicators_2006_06.pdf (accessed on
01/25/2018).
Barber, B. M. and T. Odean. 2006. “All That Glitters: The Effect of Attention
and News on the Buying Behavior of Individual and Institutional Investors”.
The Review of Financial Studies. 21: 785.
Bear Stearns. 2006. “Quick Guide to Non-Agency Mortgage-Backed Securities”.
url: https : / / www . scribd . com / document / 24292145 / 19605638 - Bear -
Stearns - Bear - Stearns - Quick - Guide - to - NonAgency - Mortgage - Back -
Securities (accessed on 01/26/2018).
Blair, M. M. 2010. “Financial Innovation, Leverage, Bubbles and the Distribu-
tion of Income”. Review of Banking & Financial Law. 30: 225.
Brunnermeier, M. K. and C. Julliard. 2008. “Money Illusion and Housing
Frenzies”. The Review of Financial Studies. 21: 135.
Calhoun, M. D. 2006. “Testimony Before the Senate Committee on Banking,
Housing and Urban Affairs, Subcommittee on Housing and Transportation,
and Subcommittee on Economic Policy, Hearing on “Calculated Risk:
Assessing Non-Traditional Mortgage Products”. (Citing Fitch Ratings,
Structured Finance: U.S. Subprime RMBS in Structured Finance CDOs).
url: http://www.responsiblelending.org/media- center/press- releases/
archives/Testimony-Calhoun092006.pdf (accessed on 01/26/2018).
Calomiris, C. W. 2009a. “Financial Innovation, Regulation, and Reform”. Cato
Journal. 29(Winter): 65.
Calomiris, C. W. 2009b. “The Subprime Turmoil: What’s Old, What’s New
and What’s Next”. The Journal of Structured Finance. 15(Spring): 6.
Cane, M. B., A. Shamir, and T. Jodar. 2012. “Below Investment Grade and
Above the Law: A Past, Present and Future Look at the Accountability of
Credit Rating Agencies”. Fordham Journal of Corporate & Financial Law.
17: 1063.
Case, K. and R. Shiller. 1989. “The Efficiency of the Market for Single-Family
Homes”. American Economic Review. 79: 125.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 51

Case, K. and R. Shiller. 2004. “Is There a Bubble in the Housing Market?”
url: http://www.econ.yale.edu/~shiller/pubs/p1089.pdf (accessed on
01/25/2018).
Coffee, J. C. 2011a. “Ratings Reform: The Good, the Bad, and the Ugly”.
Harvard Business Law Review. 1: 231.
Coffee, J. C. 2011b. “Systemic Risk After Dodd-Frank: Contingent Capital
and the Need for Regulatory Strategies Beyond Oversight”. Columbia Law
Review. 111: 795.
Cohen, A. and M. Manuszak. 2013. “Ratings Competition in the CMBS
Market”. Journal of Money, Credit and Banking. 45: 93.
Coval, J. D., J. Jurek, and E. Stafford. 2009. “The Economics of Structured
Finance”. Journal of Economic Perspectives. 23: 3.
Crawford, J. 2010. “CDO Ratings and Systemic Instability: Causes and Cure”.
New York University Journal of Law & Business. 7: 1.
Dennis, K. 2009. “The Ratings Game: Explaining Rating Agency Failures in
the Build up to the Financial Crisis”. University of Miami Law Review. 63:
1111–1136.
Document Systems, Inc. 2013. “Prepayment Fees”. url: https://web.archive.
org/web/20121230195038/http://www.docmagic.com:80/compliance/
prepayment-fees/index (accessed on 01/25/2018).
Dufwenberg, M., T. Lindqvist, and E. Moore. 2005. “Bubbles and Experience:
An Experiment”. American Economic Review. 95: 1731.
Eggert, K. 2009. “The Great Collapse: How Securitization Caused the Subprime
Meltdown”. Connecticut Law Review. 41: 1257.
Ellis, N. S., L. M. Fairchild, and F. D’Souza. 2012. “Conflicts of Interest in the
Credit Rating Industry After Dodd-Frank: Continued Business As Usual?”
Virginia Law & Business Review. 7: 1.
Engel, K. C. and P. A. McCoy. 2007. “Turning A Blind Eye: Wall Street
Finance of Predatory Lending”. Fordham Law Review. 75: 2039.
Evanoff, D. D. and L. D. Wall. 2000. “Subordinated debt as bank capital: A
proposal for regulatory reform”. Federal Reserve Bank of Chicago, Eco-
nomic Perspectives. 24(Summer): 40. url: https : / / www . chicagofed .
org / publications / economic - perspectives / 2000 / 2qepart3 (accessed on
01/25/2018).
Fannie Mae. 2013. “Selling Guide Fannie Mae Single Family”. url: https://www.
fanniemae.com/content/guide/sel073013.pdf (accessed on 01/25/2018).
Federal Deposit Insurance Corporation. 2001. “Expanded Guidance for Sub-
prime Lending Programs”. url: www.fdic.gov/news/news/press/2001/
pr0901a.html (accessed on 01/25/2018).
Felsenheimer, J. and P. Gisdakis. 2008. Credit Crises: From Tainted Loans to
a Global Meltdown. Wiley-VCH.
Financial Crisis Inquiry Commission. 2011. “Financial Crisis Inquiry Report”.
url: http://fcic.law.stanford.edu/report (accessed on 01/25/2018).

Electronic copy available at: https://ssrn.com/abstract=1973662

52 Bernard S. Black et al.

Fisch, J. E. and H. A. Sale. 2003. “The Securities Analyst as Agent: Rethinking

the Regulation of Analysts”. Iowa Law Review. 88: 1035.
Fitch Ratings. 2007. “The impact of poor underwriting practices and fraud
in subprime RMBS performance”. url: https : / / www . themaven . net /
economonitor / emerging - markets / editor - s - pick - the - impact - of - poor -
underwriting - practices - and - fraud - in - subprime - rmbs - performance -
f0zkgkbvQEWBzJ7M53CNRw?full=1 (accessed on 01/25/2018).
Foote, C. L., K. Gerardi, L. Goette, and P. S. Willen. 2008. “Just the Facts:
An Initial Analysis of Subprime’s Role in the Housing Crisis”. Journal of
Housing Economics. 17: 291.
Gorton, G. 2009. “Information, Liquidity, and the (Ongoing) Panic of 2007”.
American Economic Review: Papers and Proceedings. 99: 567.
Gorton, G. 2010. Slapped By the Invisible Hand: The Panic of 2007. Oxford
University Press.
Gorton, G. and A. Metrick. 2012. “Securitized Banking and the Run on Repo”.
Journal of Financial Economics. 104: 425.
Green, S. D. 2009. “Disquiet on the Home Front: Disturbing Crisis in the
Nation’s Markets and Institutions”. Pace Law Review. 30: 7.
Griffin, J. and G. Maturana. 2016a. “Who Facilitated Misreporting in Securi-
tized Loans?” The Review of Financial Studies. 29: 384.
Griffin, J. and G. Maturana. 2016b. “Did Dubious Mortgage Origination
Practices Distort House Prices?” The Review of Financial Studies. 29:
1671.
Griffin, J., J. Nickerson, and D. Y. Tang. 2013. “Rating Shopping or Catering?
An Examination of the Response to Competitive Pressure for CDO Credit
Ratings”. The Review of Financial Studies. 26: 2270.
Griffin, J. and D. Y. Tang. 2012. “Did Subjectivity Play a Role in CDO Credit
Ratings”. Journal of Finance. 67: 1293.
Griffin, J. and J. Nickerson. 2017. “Debt Correlations in the Wake of the
Financial Crisis: What are Appropriate Default Correlations for Structured
Products”. Journal of Financial Economics. 125: 454.
Guiso, L., P. Sapienza, and L. Zingales. 2013. “The Determinants of Attitudes
Toward Strategic Default on Mortgages”. Journal of Finance. 68: 1473.
Harney, K. 2006. “The Lowdown on Low-Doc Loans”. Washington Post. url:
http://www.washingtonpost.com/wp-dyn/content/article/2006/11/24/
AR2006112400503.html (accessed on 01/26/2019).
Harney, K. 2007. “Appraisers Under Pressure to Inflate Values”. Washington-
post.com. url: http://www.washingtonpost.com/wp-dyn/content/article/
2007/02/02/AR2007020200712.html (accessed on 01/26/2018).
Haruvy, E., Y. Lahav, and C. N. Noussair. 2007. “Traders’ Expectations in
Asset Markets: Experimental Evidence”. American Economic Review. 97(5):
1901.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 53

He, J., J. Qian, and P. E. Strahan. 2011. “Credit Ratings and the Evolution of
the Mortgage-Backed Securities Market”. The American Economic Review.
101: 131.
Heid, M. 2007. “Letter to the Board of Governors of the Federal Reserve System”.
(Letter submitted by the Division President, Wells Fargo Home Mortgage).
url: http://www.federalreserve.gov/SECRS/2007/August/20070821/OP-
1288/OP-1288_76_1.pdf (accessed on 01/26/2018).
Hill, C. A. 2004. “Regulating the Rating Agencies”. Washington University
Law Quarterly. 82: 43.
Hill, C. A. 2010. “Why Did Rating Agencies Do Such a Bad Job Rating
Subprime Securities”. University of Pittsburgh Law Review. 71: 585.
Horton, B. J. 2013. “Toward a More Perfect Substitute: How Pressure on
the Issuers of Private-Label Mortgage-Backed Securities Can Improve the
Accuracy of Ratings”. Boston University Law Review. 93: 1905.
Hu, J. 2007. “Assessing the Credit Risk of CDOs Backed by Structured Finance
Securities: Rating Analysts’ Challenges and Solutions”. The Journal of
Structured Finance. 13: 43.
Hunt, J. P. 2010. “Rating Dependent Regulation of Insurance”. Connecticut
Insurance Law Journal. 17: 101.
Hunt, J. P. 2011. “Taking Bubbles Seriously in Contract Law”. Case Western
Reserve Law Review. 61: 681.
Iacobucci, E. M. and R. A. Winter. 2005. “Asset Securitization and Asymmetric
Information”. Journal of Legal Studies. 34: 161.
Inside Mortgage Finance. 2010. “The 2010 Mortgage Market Statistical An-
nual”.
Jaffee, D. 2008. “The U.S. Subprime Mortgage Crisis: Issues Raised and
Lessons Learned”. In: Urbanization and Growth.: Commission on Growth
and Development.
Jagannathan, R., M. Kapoor, and E. Schaumberg. 2009. “Causes of the Great
Recession of 2007–2009: The Financial Crisis was the Symptom Not the
Disease!” Journal of Financial Intermediation. 22: 4.
Kovacevich, R. 2009. “Keynote Remarks by Wells Fargo & Company Chairman
Richard Kovacevich”. Stanford Institute for Economic Policy Research
Economic Summit, 2009 Mar. 13. url: https://www.youtube.com/watch?
v=hnSY-3Qwc5k.
Krebs, J. D. 2009. “The Rating Agencies: Where We Have Been and Where
Do We Go from Here”. The Journal of Business, Entrepreneurship & the
Law. 3: 133.
Kroszner, R. 2005. “Asset Price Bubbles, Information, and Public Policy”.
In: Asset Price Bubbles: The Implications for Monetary, Regulatory and
International Policies. Ed. by W. C. Hunter, G. G. Kaufman, and M.
Pomerleano. MIT Press.

Electronic copy available at: https://ssrn.com/abstract=1973662

54 Bernard S. Black et al.

Lendingtree. 2017. “Calculating Your Debt-to-Income Ratio”. url: http://

www . lendingtree . com / mortgage - loans / advice / qualifying - for - a - loan /
calculating-debt-to-income/ (accessed on 01/26/2018).
Lerner, M. 2010. “Qualifying for a conventional mortgage”. Bankrate.com.
url: http://www.bankrate.com/finance/mortgages/qualifying- for- a-
conventional-mortgage.aspx (accessed on 01/26/2018).
Levitin, A. J. and S. M. Wachter. 2012. “Explaining the Housing Bubble”.
Georgetown Law Journal. 100: 1177.
Lewis, M. 2008. “The End”. Portfolio.com. url: http://www-stat.wharton.
upenn . edu / ~steele / Courses / 434 / 434Context / The % 20End % 20by %
20Michael%20Lewis.pdf (accessed on 01/26/2018).
Lo, A. W. 2012. “Reading About the Financial Crisis: A Twenty-One-Book
Review”. Journal of Economic Literature. 50: 151.
Lockhart, III, J. B. 2007. “Housing, Subprime and GSE Reform: Where are
We Headed?” (Address of Director of Office of Federal Housing Enterprise
Oversight at Exchequer Club of Washington, D.C.). url: https://web.
archive.org/web/20130313030127/http://www.fhfa.gov/webfiles/1410/
Exchequer71807.pdf (accessed on 01/26/2018).
Lynch, T. E. 2009. “Deeply and Persistently Conflicted: Credit Rating Agen-
cies in the Current Regulatory Environment”. Case Western Reserve Law
Review. 59: 227.
Mason, J. R. and J. Rosner. 2007a. “How Resilient Are Mortgage Backed Secu-
rities to Collateralized Debt Obligation Market Disruptions?” unpublished
paper. url: http://papers.ssrn.com/sol3/papers.%20cfm?abstract_id=
1027472 (accessed on 01/26/2018).
Mason, J. R. and J. Rosner. 2007b. “Where did the Risk Go? How Misapplied
Bond Ratings Cause Mortgage Backed Securities and Collateralized Debt
Obligation Market Disruptions”. unpublished paper. url: http://papers.
ssrn.com/sol3/papers.cfm?abstract_id=1027475 (accessed on 01/26/2018).
Mayer, C., K. Pence, and S. Sherlund. 2009. “The Rise in Mortgage Defaults”.
Journal of Economic Perspectives. 23: 27.
McLean, B. and J. Nocera. 2010. All the Devils are Here: The Hidden History
of the Financial Crisis. New York: Penguin Books.
McNamara, S. 2012. “Informational Failures in Structured Finance and Dodd-
Frank’s “Improvements to the Regulation of the Credit Rating Agencies””.
Fordham Journal of Corporate & Financial Law. 17: 665.
Merrill Lynch. 2007. “Mortgage Credit Losses: How Much, Where, and When?”
Moody’s Investors Service. 2005. “Moody’s Revisits its Assumptions Regarding
Structured Finance Default (and Asset) Correlations for CDOs”. url:
http://fcic-static.law.stanford.edu/cdn_media/fcic-testimony/2010-0602-
exhibit-finance-default.pdf.
Moody’s Investors Service. 2009. “Moody’s Approach to Rating SF CDOs”.

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 55

Morgenson, G. 2010. “Raters Ignored Proof of Unsafe Loans in Securities,

Panel is Told”. The New York Times. Sept. 27.
Morgenson, G. and J. Rosner. 2011. Reckless Endangerment: How Outsized
Ambition, Greed, and Corruption Led to Economic Armageddon. Henry
Holt and Company.
Morgenson, G. and L. Story. 2009. “Banks Bundled Bad Debt, Bet Against It
and Won”. The New York Times. url: http://www.nytimes.com/2009/
12/24/business/24trading.html?pagewanted=all&_r=0 (accessed on
01/26/2018).
Morgenson, G. and L. Story. 2010. “Rating Agency Data Aided Wall Street
in Deals”. The New York Times. Apr. 23. url: http://www.nytimes.com/
2010/04/24/business/24rating.html?dbk (accessed on 01/26/2018).
Mortgage Asset Research Institute. 2006. “Eighth Periodic Mortgage Fraud
Case Report to Mortgage Bankers Association”. url: https : / / www .
lexisnexis.com/risk/downloads/literature/MBA8thCaseRpt.pdf (accessed
on 01/26/2018).
Mortgage Bankers Association. 2010. “National Delinquency Survey Q1 2010”.
Mortgage News Daily. 2006. “The mortgage prepayment penalty”. url: http:
/ / qna . mortgagenewsdaily. com / questions / the - mortgage - prepayment -
penalty (accessed on 01/26/2018).
Muolo, P. and M. Padilla. 2010. Chain of Blame: How Wall Street Caused the
Mortgage and Credit Crisis. John Wiley & Sons.
Nomura Securities International. 2006. “Bond Rating Confusion”. url: http:
//www.markadelson.com/pubs/Bond_Rating_Confusion.pdf (accessed on
01/26/2018).
Partnoy, F. 1999. “The Siskel and Ebert of Financial Markets?: Two Thumbs
Down for the Credit Rating Agencies”. Washington University Law Quar-
terly. 77: 619.
Partnoy, F. 2000. “Why Markets Crash and What Law Can Do About It”.
University of Pittsburgh Law Review. 61: 741.
Paul, R. 2003. “Fannie Mae and Freddie Mac Subsidies Distort the Housing
Market, Hearing before the House Financial Services Committee”. url:
https://darkpolitics.wordpress.com/2009/12/21/fannie-mae-and-freddie-
mac-subsidies-distort-the-housing-market/ (accessed on 01/26/2018).
Porter, D. P. and V. L. Smith. 1994. “Stock Market Bubbles in the Laboratory”.
Applied Mathematical Finance. 1: 111.
Pozen, R. 2009. Too Big to Save? How to Fix the U.S. Financial System. John
Wiley & Sons.
Purnanandam, A. 2011. “Originate-to-Distribute Model and the Subprime
Mortgage Crisis”. The Review of Financial Studies. 24: 1881.
Rhee, R. 2010. “The Decline of Investment Banking: Preliminary Thoughts on
the Evolution of the Industry 1996–2008”. Journal of Business Law and
Technology. 5: 75.

Electronic copy available at: https://ssrn.com/abstract=1973662

56 Bernard S. Black et al.

Riddiough, T. J. and J. Zhu. 2009. “Shopping, Relationships, and Influence

in the Market for Credit Ratings”. unpublished paper. url: http://www.
hhs.se/df/seminarsworkshops/documents/riddiough.pdf (accessed on
01/26/2018).
Rodrigues, R. 2007. “Absence of Fear, CFA Society of Chicago Speech”. url:
http://www.fpafunds.com/docs/special-commentaries/absence_of_fear.
pdf?sfvrsn=2 (accessed on 01/26/2018).
Roe, M. J. 2011. “The derivatives market’s payment priorities as financial
crisis accelerator”. Stanford Law Review. 63: 539.
Ryan, S. 2008. “Accounting in and for the Subprime Crisis”. The Accounting
Review. 83: 1605.
Sengupta, R. 2010. “Alt-A: The Forgotten Segment of the Mortgage Market”.
Federal Reserve Bank of St. Louis Review. 92: 55.
Sepe, S. M. and C. K. Whitehead. 2015. “Paying for Risk: Bankers, Compen-
sation, and Competition”. Cornell Law Review. 100: 655.
Shiller, R. J. 1988. “Portfolio Insurance and Other Investor Fashions as Factors
in the 1987 Stock Market Crash”. NBER Macroeconomics Annual. 3: 287.
url: http://www.nber.org/chapters/c10958.pdf (accessed on 01/26/2018).
Shiller, R. J. 2008a. “Infectious Exuberance”. The Atlantic. url: https://
www.theatlantic.com/magazine/archive/2008/07/infectious-exuberance/
306839/ (accessed on 01/26/2018).
Shiller, R. J. 2008b. The Subprime Solution. Princeton University Press.
Shiller, R. J. 2016. Irrational Exuberance. Princeton University Press.
Sinclair, U. 1935. I, Candidate for Governor, and How I Got Licked. University
California Press.
Skreta, V. and L. Veldkamp. 2009. “Ratings Shopping and Asset Complexity:
A Theory of Ratings Inflation”. Journal of Monetary Economics. 56: 678.
SMC Bancorp. 2006. “SMC Core-Fixed Rate Conforming Loan Programs”.
url: http://www.rebeccachang.com/smcbancorp/forms/conf_fix.pdf
(accessed on 01/25/2018).
Smith, V. L., G. L. Suchanek, and A. W. Williams. 1988. “Bubbles, Crashes,
and Endogenous Expectations in Experimental Spot Asset Markets”. Econo-
metrica. 56: 1119.
Standard & Poor’s. 2002. “Global Cash Flow and Synthetic CDO Criteria”.
url: www.globalriskguard.com/resources/crderiv/sp_glob_cash_flow.pdf
(accessed on 01/26/2018).
Taylor, J. 2009. Getting Off Track: How Government Actions and Interventions
Caused, Prolonged, and Worsened the Financial Crisis. Hoover Institution
Press Publication.
The Economist. 2004. “Flimsy Foundations”. url: http://www.economist.
com/node/3477796 (accessed on 01/26/2018).

Electronic copy available at: https://ssrn.com/abstract=1973662

The Nonprime Mortgage Crisis and Positive Feedback Lending 57

The New York Times. 2014. “Banks Behaving Badly, Again”. url: http :
/ / www . nytimes . com / 2014 / 11 / 08 / opinion / banks - behaving - badly -
again.html?_r=0 (accessed on 01/26/2018).
Tomlinson, R. and D. Evans. 2007. “CDO Boom Masks Subprime Losses,
Abetted by S&P, Moody’s, Fitch”. Bloomberg Markets. url: http://www.
bloomberg.com/news/articles/2007-05-31/cdo-boom-masks-subprime-
losses-abetted-by-sp-moodys-fitch (accessed on 01/25/2018).
U.S. Department of Justice. 2015. “Justice Department and State Partners
Secure $1.375 Billion Settlement with S&P for Defrauding Investors in
the Lead Up to the Financial Crisis—Settlement Agreement, Annex 1:
Statement of Facts”. url: http://www.justice.gov/file/338706/download
(accessed on 01/26/2018).
U.S. Senate Permanent Subcommittee on Investigations. 2011. “Wall Street
and the financial crisis: anatomy of a financial collapse”. Senate Report on
Financial Crisis.
Wallison, P. 2011. “Dissent from the majority report of the financial crisis
inquiry commission”. American Enterprise Institute.
Washington Mutual Bank. 2006. “Subprime Underwriting Guidelines”. url:
http://www.scribd.com/doc/212796842/WAMU-Subprime-Underwriting-
Guidelines (accessed on 01/26/2018).
Weiss, D. C. 2013. “Morgan Stanley Lawyer Had Joke Names for Toxic Assets
Sold Overseas, Including ‘Subprime Meltdown’”. ABA Journal. url: http:
//www.abajournal.com/news/article/morgan_stanley_lawyer_had_
joke_names_for_toxic_assets_including_subprime_me/ (accessed on
01/26/2018).
Whitehead, C. 2011. “Destructive Coordination”. Cornell Law Review. 96: 323.
Whitehead, C. 2012. “The Goldilocks Approach: Financial Risk and Staged
Regulation”. Cornell Law Review. 97: 1267.
Whitehead, C. 2015. “Size Matters: Commercial Banks and the Capital Mar-
kets”. Ohio State Law Journal. 76: 765.

Electronic copy available at: https://ssrn.com/abstract=1973662