Jump to ContentJump to Main Navigation
A Debtor WorldInterdisciplinary Perspectives on Debt$

Ralph Brubaker, Robert M. Lawless, and Charles J. Tabb

Print publication date: 2012

Print ISBN-13: 9780199873722

Published to Oxford Scholarship Online: January 2013

DOI: 10.1093/acprof:oso/9780199873722.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 04 April 2020

Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Chapter:
(p.87) 4 Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis
Source:
A Debtor World
Author(s):

Amir Sufi

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199873722.003.0004

Abstract and Keywords

This chapter focuses on the tension between the preservation of lender incentives and shedding credit risk. It addresses this issue in three steps. First, it discusses foundational theories of financial intermediation in which a bank's main purpose is to exert screening or the monitoring of loans to reduce borrower moral hazard and adverse selection. One of the primary features of these theories is that banks must retain at least part of the credit risk of the loans in order to preserve their incentives to monitor and screen borrowers. Second, it examines how contractual arrangements attempt to preserve bank incentives while transferring credit risk in a variety of markets, including credit default swaps, loan syndication, and securitization. It then examines the recent subprime mortgage crisis with a particular emphasis on how the conflict between lender incentives and shedding credit risk affected default patterns since 2005. Finally, the chapter explores potential reasons securitization failed to preserve lenders' incentives to screen and monitor borrowers.

Keywords:   financial intermediation, loans, credit risk, bank incentives, subprime mortgage crisis, securitization, borrower screening

Introduction

One of the most important trends in financial intermediation is the advancement in financial technologies that allow debt originators to shed credit risk. These innovations are numerous and significant. For example, Loutskina and Strahan (2009) show that the fraction of home and commercial mortgages securitized since 1980 has increased by 0.50 and 0.30, respectively. Drucker and Puri (2007) document that US secondary corporate loan market volume grew from only $8 billion in 1991 to $176 billion in 2005. Longstaff, Mithal, and Neis (2005) show that the market for credit default swaps, or insurance contracts that protect a creditor against the default risk of a borrower, grew from only $180 billion in 1997 to almost $5 trillion by 2004.

Although diversification of credit risk is desirable for financial stability, an important question is: Can banks simultaneously shed credit risk while retaining the incentive to screen and monitor borrowers? This question has gained increased importance given the recent subprime mortgage crisis and its resulting negative effects on the real economy and financial system. Anecdotal evidence and academic research suggest that the current mortgage crisis is at least partially due (p.88) to the aggressive securitization of subprime mortgages from 2001 to 2006 (Keys, Mukherjee, Vig, and Seru (2009); Mian and Sufi (2009)).

This tension between the preservation of lender incentives and shedding credit risk is the focus of this chapter. I address this issue in three steps. First, I discuss foundational theories of financial intermediation in which a bank's main purpose is to exert screening or monitoring of loans to reduce borrower moral hazard and adverse selection. One of the primary features of these theories is that banks must retain at least part of the credit risk of the loans in order to preserve their incentives to monitor and screen borrowers. Second, I examine how contractual arrangements attempt to preserve bank incentives while transferring credit risk in a variety of markets, including credit default swaps, loan syndication, and securitization. I then examine the recent subprime mortgage crisis with a particular emphasis on how the conflict between lender incentives and shedding credit risk affected default patterns since 2005. Finally, I explore potential reasons securitization failed to preserve lenders’ incentives to screen and monitor borrowers.

Lender Incentives and Financial Intermediation: Theoretical Background

Information frictions and incentive problems are at the root of foundational theories of financial intermediation. Although there are a number of incentive-based theories of financial intermediation (which are discussed below), they generally have the following features: Borrowers have either private information on the value of their projects, or misaligned incentives such as private benefits or excessive risk taking. In the absence of a financial intermediary, private information or misaligned incentives would prevent borrowers from obtaining financing, given that the providers of financing would lose money in expectation. Financial intermediaries have access to certain technologies that allow them to reduce the inefficient outcomes associated with private information or misaligned incentives. The key insight of these theories is that financial intermediaries may also be subject to the same incentive problems as borrowers. The key question then becomes: How can financial arrangements ensure that financial intermediaries such as banks have the incentive to properly screen or monitor borrowers?

The article by Diamond (1984) provides the seminal incentive-based theory of financial intermediation. In his model, true profits realized by borrowers are unobservable to external financiers unless they exert costly monitoring. In such a situation, borrowers have the incentive to “lie” and underreport their true income to avoid paying external financiers. Costly monitoring by a financial intermediary allows the intermediary to see actual profits with some probability. Although this monitoring (p.89) is potentially beneficial, as Diamond notes, “the information production task delegated to the intermediary gives rise to incentive problems for the intermediary” (393). In other words, a financial intermediary with the ability to monitor a loan must be given the incentive to exert effort in costly monitoring if the monitoring technology is to have value to the economy. Diamond argues that diversification of a large number of projects within the intermediary can reduce monitoring costs and therefore make it credible for the financial intermediary to monitor borrowers.

In the model by Holmstrom and Tirole (1997), borrowers face a moral hazard problem. They can pursue a profitable project, but they can also pursue an unprofitable project that provides a private benefit. In the absence of monitoring by a financial intermediary, borrowers would be unable to raise financing given that they cannot commit to taking the profitable project instead of the unprofitable project with a private benefit. In the Holmstrom and Tirole model, a financial intermediary has a costly monitoring technology that allows it to reduce the private benefit from the unprofitable project. The costly monitoring technology is unobservable to other lenders, which gives rise to a moral hazard problem on behalf of the intermediary. As Holmstrom and Tirole note: “While we assume that each intermediary has the physical capacity to monitor an arbitrary number of firms, the moral hazard problem puts a limit on the actual amount of monitoring that will take place” (669).

What is the solution to the intermediary moral hazard problem? Holmstrom and Tirole (1997) argue that an intermediary can commit to monitoring by placing its own capital at risk. This “skin in the game” mechanism is described as follows: “Moral hazard forces intermediaries to inject some of their own capital into the firms they monitor…[other lenders] invest directly in the firm, but only after the monitor has taken a large enough financial interest in the firm that the investors can be assured that the firm will behave diligently” (669). In other words, a financial intermediary can only commit to monitoring a borrower if it has at least some capital at risk if the borrower performs poorly. Although Holmstrom and Tirole focus primarily on the ex post monitoring activity of the intermediary, it is important to emphasize that a financial intermediary with “skin in the game” also commits to the proper screening of the borrower when the loan application is reviewed (see also Gorton and Pennachi (1995), Parlour and Plantin (2008)).

These foundational models of financial intermediation are increasingly relevant given recent trends in credit risk policies of financial intermediaries. When financial intermediaries attempt to shed the credit risk of their lending portfolio, a critical question is whether the transaction can be designed in a manner that preserves the incentives of financial intermediaries to engage in monitoring and screening. In the section below, I address these issues by examining recent trends in syndication, credit default swaps, and securitization.

(p.90) Preserving Lender Incentives: Syndication, Credit Default Swaps, and Securitization

Theoretical research in financial intermediation highlights the tension between the preservation of lender incentives and banks’ attempts to shed credit risk. In this section, I explore how financing arrangements attempt to resolve this tension in syndication, credit default swaps, and securitization.

Loan Syndication

The corporate syndicated loan market is one of the most important sources of finance for US firms, with over $2 trillion of syndicated loans originated annually. In a syndicated loan, multiple creditors lend to a firm through one contractual arrangement. A single bank takes on the lead arranger role. The lead arranger establishes and maintains a relationship with the borrower, and takes on the primary information collection and monitoring responsibilities (Sufi 2007). After signing an initial commitment letter with the firm, the lead arranger syndicates out a part of the loan to participant lenders. Although participant lenders have some information about the firm, they generally rely upon the lead arranger for both screening and monitoring duties. For example, almost all syndicated loans contain financial covenants, and the participant lenders rely on the lead arranger to monitor and enforce these covenants.

The syndication process immediately implies an incentive problem for the lead arranger: The lead arranger is responsible for the lion's share of monitoring and screening duties, but retains only a fraction of the loan and, consequently, the credit risk. As a result, the lead arranger has reduced incentives to properly examine the borrower at origination or enforce covenants going forward.

A large body of empirical research shows that contracting parties understand this incentive problem, and syndicated loan contracts are structured to minimize its effect. For example, Dennis and Mullineaux (2000), Lee and Mullineaux (2004), and Sufi (2007) show that lead arrangers retain a larger fraction of the syndicated loan when monitoring and screening are likely to be more important. In addition, Sufi shows that lead arrangers are more likely to select participants that are familiar with the borrower when information asymmetry between the borrower and its lenders is more severe. Both of these actions help to align the incentives of the lead arranger with other participant banks. Drucker and Puri (2007) examine loan sales (an extreme form of syndication) and find that loans that are more likely to be sold contain more restrictive covenant packages. Taken together, these findings suggest (p.91) that contracting parties understand the incentive problem faced by lead arrangers when syndicating loans and they make adjustments to contracts to try and minimize its impact.

Credit Default Swaps

A credit default swap is an insurance contract that protects against corporate debt defaults. There are two parties to the contract, and the reference obligation is the particular credit that is insured (i.e., a corporate bond or corporate loan). The protection seller provides the protection buyer with the guarantee of purchasing the reference obligation at face value in the event of a default. In return, the protection buyer pays a periodic fee to the protection seller for the insurance (Longstaff, Mithal, and Neis 2005). The market for credit default swaps has exploded over the last five years, with the value of credit default insurance going from less than $2 trillion in 2001 to almost $50 trillion in 2007.

The existence of credit default swaps has the potential to severely reduce the incentives of financial intermediaries to screen and monitor loans. A financial intermediary that has bought credit insurance is immune to losses from defaults, and therefore has reduced incentives to engage in costly activities that reduce the probability of default of the borrower in question.

A recent example from the AIG collapse is instructive. AIG reported on March 15, 2009, that it paid $7 billion to Goldman Sachs, even though Goldman Sachs had previously reported that its exposure to AIG was not material (Hu 2009). The evidence suggests that these seemingly contradictory facts are due Goldman Sachs having a large amount of insurance through credit default swaps on AIG. As Hu notes, “Goldman Sachs was apparently an empty creditor of AIG…empty creditors have weaker incentives to cooperate with troubled corporations to avoid collapse, and if collapse occurs, can cause substantive and disclosure complexities in bankruptcy” (1).

Before the most recent crisis, the question was: Did market participants understand the potential incentive problem associated with credit default swaps? The evidence suggests that they did. For example, Longstaff, Mithal, and Neis (2005) show that only three out of the 68 corporate borrowers in their sample with active credit default swap markets are rated below investment grade. This suggests that market participants understand that the provision of credit risk insurance to financial intermediaries reduces the incentive to screen and monitor; as a result, credit default swaps are available primarily for the highest credit quality firms with little need for screening and monitoring. Ashcraft and Santos (2007) demonstrate that the onset of credit default swap trading for informationally-opaque and risky firms leads to a reduction in the ability of lead arrangers on syndicated loans to resolve (p.92) information frictions by holding a higher fraction of the loan. This suggests that market participants understand the reduced incentives after credit default swaps become available.

As a caveat, most of the research on credit default swaps focus on insurance contracts written for corporate bonds. There has also been an explosion in credit default swaps in which the reference obligation is a structured finance obligation such as a mortgage-backed security. Additional research is needed to see whether credit default swaps reduce incentives to screen and monitor in markets outside of corporate bonds.

Securitization

Securitization is a process used by financial intermediaries to convert less liquid financial assets (such as a loan) into more liquid securities backed by the pool of assets.1 The securitization process starts with an originator of a group of loans. The originator sells the loans to an arranger. The arranger has several duties, including due diligence on the originator, setting up the bankruptcy remote trust that will ultimately purchase the loans, and structuring the loans into tranched securities. In addition, the rating agencies provide a credit rating for the securities created through the securitization process. Finally, another key player is the servicer, which is the party in charge of collection and remittance of loan payments from the original borrowers, contacting delinquent borrowers, and supervising foreclosures and property dispositions (Ashcraft and Schuerman (2008)).

There are several steps in the securitization process that potentially reduce the incentives of financial intermediaries to properly screen and monitor borrowers. I focus on two issues here. First, the originating bank sells the loan to an arranger; as a result, the originator holds none of the credit risk and therefore has reduced incentives to properly screen borrowers. Second, the servicer of the mortgage pool does not always hold the credit risk of the loans, and therefore has reduced incentives to engage in ex post actions that will increase the value of the securities. These actions include quick foreclosure, renegotiation, or working out problematic loans.

Two contractual mechanisms are used to reduce these incentive problems in securitization. First, Gan and Mayer (2007) examine securitization of commercial mortgage-backed securities and show that the servicer is often required to hold the first-loss position in the securitization pool. They show evidence that holding the first-loss position provides additional incentives for the servicer to engage in value-increasing (p.93) activities. However, they also show that the first-loss position is often not retained by the servicer of the securitization pool. Second, Keys, Mukherjee, Seru, and Vig (2010) examine the securitization of residential mortgages, finding that most originating banks are required to take back onto their balance sheet any mortgage sold to the arranger if the mortgage defaults within three months of the sale. They show that default rates are much lower in the first three months after the sale of the mortgage to the arranger.

Although there is some evidence that contractual arrangements in the securitization process preserve lender incentives to some degree, there is also evidence that securitization is partially responsible for a sharp rise in mortgage default rates (Keys, Mukherjee, Seru, and Vig 2010; Demyanyk and Van Hemert 2008; Mian and Sufi (2009). These patterns have led Ashcraft and Schuermann (2008) to argue that “investors should demand that either the arranger or originator—or even both—retain the first-loss equity tranche of every securitization, and disclose all hedges of this position.” The relationship between securitization of subprime mortgages and the recent spike in mortgage default rates is the subject of the next section.

Securitization and the Subprime Mortgage Crisis

Foundational theories of financial intermediation suggest that the process of shedding credit risk reduces the incentives of debt originators to screen and monitor loans. Given that a sharp increase in securitization of mortgages precedes the recent mortgage default crisis, these theories are instrumental in understanding how and why the crisis occurs. In this section, I examine the sharp increase in the securitization of mortgages from 2001 to 2005 with a particular focus on the types of mortgages securitized, the effects on credit availability to subprime borrowers, and the impact on mortgage default rates from 2005 to 2007. In the final subsection, I explore the reasons securitization failed to preserve lender incentives to properly screen and monitor borrowers.

Data

The data set employed in the following analysis is a zip code–year level panel covering 17,009 zip codes in the United States from 1996 through the second quarter of 2007.2 There are two main sources for these data. The first source includes outstanding consumer credit amounts and defaults from Equifax Predictive Services. Equifax is a consumer credit rating agency that collects, organizes, and manages credit information for US consumers. The Equifax data have credit information for almost 170 million individuals. This data set allows me to construct aggregate mortgage debt (p.94) composition and defaults of every US zip code at an annual frequency from 1996 through the second quarter of 2007.

The second main source of data is mortgage origination information collected under the Home Mortgage Disclosure Act (HMDA). In order to supervise and enforce fair lending practices across that United States, Congress mandates that all loan applications related to home purchase, refinancing, and home improvement be reported to the federal government. The loan application information is publicly available through HMDA from 1996 through 2006. For every loan application, the public data records its status (denied/approved/originated); purpose (home purchase/refinancing/home improvement); loan amount; and applicant characteristics, including race, sex, income, and home ownership status. It also reports lender information, including the lender's reasons for applicant denial and type of lender, and whether the loan originator sold the loan to the secondary market within a year. Because our unit of analysis is a zip code, I aggregate the application-level HMDA data to zip codes.

In addition to these two main data sources, the final data set includes zip code–level demographic and income information from three additional data sources. Zip code level demographic attributes such as population, race, poverty, mobility, unemployment, and education come from the Decennial 2000 Census. The data set also includes annual measures of business opportunities available in a given zip code through the Business Statistics published by the US Census Bureau. These statistics provide data on wages, employment, and number of establishments at the zip code level. Finally, the final data set includes zip code–level average “adjusted gross income” as reported by the IRS. The IRS currently provides these data for 1998, 2001, 2002, 2004, and 2005.

Aggregate Trends

Figure 4.1 plots the growth rates in mortgage and non-mortgage consumer debt indexed to 1996. From 1996 to 2002, the relative growth rates of mortgage and non-mortgage consumer debt are similar. Beginning in 2002, there is a sharp increase in the growth rate of mortgage versus non-mortgage debt; this trend accelerates through 2006.

Figures 4.2A and 4.2B demonstrate the source of the increase from 2002 to 2006. Figure 4.2A plots the fraction of originated mortgages for home purchase that are not sold by the originator within the year of origination or are sold to a government-sponsored enterprise (GSE) such as Freddie Mac or Fannie Mae. These “traditional” channels of origination experience significant declines as a fraction of total originations. The fraction not sold by the originator declines from 0.50 to 0.40 from 2000 to 2005, and the fraction sold to a GSE decreases from its peak of almost 0.40 to less than 0.20 from 2003 to 2005. (p.95)

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.1 Consumer Mortgage and Non-Mortgage Debt, Indexed to 1996

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.2 Panel A Fraction of Originated Mortgages Not Sold and Sold to GSEs

Figure 4.2B explains why there is a relative reduction in traditional originations as a fraction of total originations. It demonstrates a sharp increase in mortgages sold to affiliates and mortgages sold for the purpose of securitization. The latter group is defined as mortgages that are sold for the purpose of private securitizations and mortgages that are sold to nonbank financial institutions such as mortgage banks or financing companies. (p.96)

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.2 Panel B Fraction of Originated Mortgages Sold to Non-GSEs

Although one cannot be sure that mortgages sold to nonbank financial institutions are sold with the purpose of securitization, Ashcraft and Schuermann (2008) show that the 10 largest issuers of mortgage-backed securities from securitization pools all fit into this category. The sharp rise in mortgages sold for the purpose of securitization is quite dramatic: It increases from less than 0.03 to almost 0.15 from 2003 to 2006.

Where Did Securitization Occur?

Figure 4.2B shows a dramatic rise in the fraction of mortgages sold by the originator for the purpose of securitization within the year of origination. In Table 4.1, I examine the characteristics of zip codes in which the growth in securitization from 2001 to 2005 is most dramatic. More specifically, for the analysis in Table 4.1, I create quartiles of zip codes based on the change in the fraction of mortgages originated and sold for the purpose of securitization. High securitization zip codes are those in the top quartile based on the change in the fraction of mortgages sold from 2001 to 2005 for the purpose of securitization. Low securitization zip codes are those in the lowest quartile based on the change in the fraction of mortgages sold from 2001 to 2005 for the purpose of securitization.

In Table 4.1, I compare the characteristics of low-versus-high securitization zip codes as of 2000, or the year before the securitization trend begins. The difference for high securitization zip codes is calculated after controlling for county-fixed effects. This methodology helps to compare low and high securitization zip codes (p.97)

Table 4.1

Where Did Securitization of Mortgages Occur?

This table presents a comparison of zip code characteristics as of 2000 for high and low securitization zip codes. The difference for high securitization zip codes is the coefficient estimate on a high securitization indicator variable in a county-fixed effects regression. The difference should be interpreted as the within-county difference for high-versus-low securitization zip codes for the characteristic in question. For example, high securitization zip codes have a fraction of residents with a FICO score under 619 that is 0.086 higher than low securitization zip codes in the same county.

Mean for low securitization zip codes

Difference for high securitization zip codes

Within-county standard deviation

Credit quality measures

Fraction of residents with FICO 〈 619

0.231

0.086**

0.076

Fraction of residents with FICO 〈 659

0.319

0.104**

0.088

Fraction of mortgages backed by FHA

0.051

0.067**

0.054

Fraction of mortgage applications denied

0.359

0.074**

0.073

Income measures

Median household income

38,214

–10,387**

11,290

Per capita income

19,128

−8,595**

6,556

Fraction in poverty

0.129

0.027**

0.057

Fraction unemployed

0.057

0.010**

0.028

Fraction with less than high school education

0.215

0.056**

0.076

(**) Coefficient estimate statistically distinct from 1 percent level.

within the same county. For example, the first row shows that low securitization zip codes have a fraction of residents with FICO scores below 619 of 0.231. High securitization zip codes in the same county have a fraction that is 0.086 larger than low securitization zip codes.

The top half of Table 4.1 demonstrates that residents in high securitization zip codes have worse credit quality before the securitization wave on all measures. High securitization zip codes have a higher fraction of subprime borrowers, a high fraction of mortgages that are backed by the Federal Housing Administration, and a (p.98) higher fraction of mortgage applications that are denied. The magnitudes are very large. When compared to the within-county standard deviation of the characteristic in question, the credit quality in high securitization zip codes is more than one standard deviation worse for all measures. These results demonstrate that the increase in securitization from 2001 to 2005 is concentrated in areas with much lower credit quality as of 2000.

The lower half of Table 4.1 examines measures of income. High securitization zip codes have lower household income and per capita income. They also have a higher fraction of households in poverty, unemployed, and with less than a high school education. As with measures of credit quality, the magnitudes are quite large. For example, high securitization zip codes have per capita income that is more than one full standard deviation lower than low securitization zip codes.

What Was the Effect of Securitization on Mortgage Growth and Default Rates?

Figure 4.3 demonstrates the effect of securitization on mortgage growth rates. It plots the growth rates in originated mortgages for home purchase from 1996 to 2005 for low and high securitization zip codes. From 1996 until 1999, low securitization zip codes experience a larger growth rate in originated mortgages. Beginning in 1999 and accelerating after 2001, high securitization zip codes experience a sharp increase in originated mortgage amounts relative to low securitization zip codes. In fact, from 1999 until 2005, high securitization zip codes experience a growth rate

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.3 Growth of Originated Mortgages for High and Low Securitization Zip Codes

(p.99) of mortgage amounts originated of 90 percent whereas low securitization zip codes experience a growth rate of only 60 percent.

Table 4.2 demonstrates a similar pattern in growth rates. From 2001 to 2005, the growth in mortgages originated for high securitization zip codes is 7.7 percentage points larger than for low securitization zip codes in the same county. The expansion in credit to high securitization zip codes is also evident from mortgage application denial rates. The fraction of mortgage applications denied increased by 0.030 in low securitization zip codes, but increased by only 0.015 in high securitization zip codes.

One potential reason for the relative increase in mortgage originations in high securitization zip codes could be a relative improvement in income prospects or

Table 4.2

Differential Growth Patterns in High-Versus-Low Securitization Zip Codes

This table presents a comparison of zip code growth patterns between 2001 and 2007 for high and low securitization zip codes. The difference for high securitization zip codes is the coefficient estimate on a high securitization indicator variable in a county-fixed effects regression. The difference should be interpreted as the within-county difference in the growth pattern for high-versus-low securitization zip codes. For example, high securitization zip codes have a growth rate of income that is 4.8 percentage points lower than low securitization zip codes in the same county.

Mean for low securitization zip codes

Difference for high securitization zip codes

Within-county standard deviation

Change in lending

Growth in mortgages originated, 2001 to 2005

0.619

0.077**

0.391

Change in denial rates, 2001 to 2005

0.030

−0.015**

0.046

Change in income profile

Income growth, 2001 to 2005

0.132

−0.048**

0.076

Employment growth, 2001 to 2004

0.029

−0.009

0.208

Change in debt to income ratio, 2001 to 2005

0.069

0.027**

0.079

Change in defaults

Change in default rates, 2005 to 2007

0.004

0.006*

0.040

(*, **) Coefficient estimate statistically distinct from 5 percent and 1 percent level, respectively.

(p.100) credit quality in these zip codes. In other words, despite the fact that these zip codes are worse on all credit quality and income measures as of 2000 (as shown in Table 4.2), this argument maintains that there are gains in these zip codes from 2001 to 2005 that justify increased lending. The evidence in Table 4.2 strongly disputes this argument. If anything, the evidence suggests that relative income growth for high-versus-low securitization zip codes is negative. For example, income growth was 5 percentage points lower in high-versus-low securitization zip codes in the same county, which is almost two-thirds of a standard deviation. In addition, the mortgage debt to total zip code income ratio increased more in high securitization zip codes relative to low securitization zip codes from 2001 to 2005. The findings in Table 4.2 demonstrate that high securitization zip codes did not experience improvements in credit quality or income during the relative mortgage expansion that would justify a relative increase in mortgage origination amounts.

Figure 4.4 examines the time series pattern of mortgage debt to income ratios and demonstrates a similar point. From 1998 to 2001, the change in the mortgage debt to income ratios of high and low securitization zip codes is similar. However, beginning in 2001, the mortgage debt to income ratio of high securitization zip codes accelerates at a more rapid pace. By 2005, the difference in the mortgage debt to income ratio for high and low securitization zip codes is 0.15, compared to only 0.05 in 2001.

Table 4.2 and Figure 4.4 show a relative securitization-driven expansion of mortgage credit to zip codes experiencing relatively negative credit quality trends. A key question is whether investors during this expansion were being compensated for the

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.4 Mortgage Debt to Income Ratios for High and Low Securitization Zip Codes

(p.101) increased credit risk of these mortgages. Unfortunately, I cannot directly answer this question using my data set given that the HMDA data do not include information on mortgage interest rates before 2004. However, Chomsisengphet and Pennington-Cross (2006) and Demyanyk and Van Hemert (2011) both provide evidence that the subprime–prime interest spread declined to historical lows from 2001 to 2005.3 Given that securitization was more widespread in subprime areas, these aggregate trends suggest that investors received lower interest spreads on lower credit quality mortgages than at any other period in the last 15 years.

Table 4.2, Figure 4.3, and Figure 4.4 demonstrate a securitization-driven expansion in mortgage availability from 2001 to 2005. Areas that experience a growth in mortgage availability due to securitization experience relative decreases in income growth and debt-to-income ratios when compared to other zip codes in the same county. The evidence also suggests that interest rate spreads on lower credit quality mortgages declined to historical lows during this time period.

Figure 4.5 examines subsequent default rates from 2005 to 2007. As Figure 4.5 clearly demonstrates, high securitization zip codes experience a sharp relative rise in default rates from 2005 to 2007. The change in the default rate over these two years for high securitization zip codes is almost 2 percentage points, whereas the increase

                      Lender Incentives, Credit Risk, and Securitization: Evidence from the Subprime Mortgage Crisis

Figure 4.5 Mortgage Default Rates for High and Low Securitization Zip Codes

(p.102) is only one-half a percentage point for low securitization zip codes. To put this into perspective, total mortgage debt outstanding for high securitization zip codes as of 2005 is $3.4 trillion, and so an increase of 2 percentage points in the default rate represents $67 billion in additional amounts in default.

Is the Evidence Uniquely Consistent with a Reduction in Lender Incentives?

Taken together, the evidence presented in the above subsection implies that securitization of mortgages expanded rapidly from 2001 to 2005 in zip codes with low credit quality and low income as of 2000. Despite the fact that these areas experienced a relative decline in income growth when compared to low securitization zip codes from 2001 to 2005, they experienced a rapid growth in the availability of credit and the mortgage debt to income ratios during this time period. Concurrently, interest spreads on risky mortgages declined to historical lows. Finally, default rates increased disproportionately in these areas from 2005 to 2007.

Are these results uniquely consistent with the view that securitization and its resulting negative impact on screening and monitoring incentives caused the mortgage default crisis? Although it is difficult to argue that securitization and its resulting negative effect on lender incentives is the only reason for the spike in mortgage default rates, the evidence above is certainly consistent with this hypothesis. Research by Keys, Mukherjee, Seru, and Vig (2010) also provide strong evidence that lax screening in securitization pools is responsible for a large fraction of recent mortgage defaults. They exploit the fact that borrowers with a credit score just above and below 620 have a differential ability to have their mortgages securitized. They show that mortgages to consumers just above a FICO score of 620 are much more likely to be securitized than mortgages just below, and they show that these securitized mortgages have much higher default realizations. This result is concentrated in low documentation loans, where the incentives of originators are particularly important.

An alternative explanation for the increase in default rates on securitized mortgages is that a fundamental economic shock occurred in 2005 that differentially affected poorer credit quality borrowers. This view maintains that the securitization process was fundamentally sound, but subprime borrowers experienced an unforeseen “bad luck” shock. This argument is supported by the view that securitization is designed to make credit available to lower credit quality borrowers. As a result, we should not be surprised that an aggregate economic shock leads to higher defaults on securitized mortgages.

The counterargument to this alternative explanation is quite simple: What is the bad economic shock? Default rates begin to increase sharply from 2005 to 2006 (p.103) despite the fact that aggregate personal income and employment growth remains strong. Indeed, one of the biggest anomalies of the current housing crisis is that it is not preceded by any large macroeconomic shock such as a recession. In fact, Mian and Sufi (2009) argue that housing price expansion and deceleration is driven by the mortgage expansion, not vice versa.

Why Did Securitization Fail to Preserve Lender Incentives?

The evidence above suggests that securitization failed to preserve the incentives of originators and servicers to properly screen and monitor low credit quality mortgages. As discussed above, financial markets with information frictions can function even with the existence of incentive problems for financial intermediaries. Rather than ignoring the incentive problems, contractual arrangements typically attempt to preserve a lender's incentives. Why did contractual arrangements fail to preserve incentives in the securitization setting?

There are many potential culprits to explain why securitization failed to preserve lender incentives. I choose to focus on three that I believe to be the most important. First, mortgage lending is secured lending, and the value of the underlying home protects the ultimate buyers of mortgage backed securities from default risk. Indeed, housing prices are often the most important input into mortgage default predictive models. From 2001 to 2005, investors likely did not fully appreciate that the mortgage supply expansion itself was in part driving house price appreciation in subprime areas. The lack of understanding of the feedback mechanism of lending on prices may have skewed investor perceptions of the true underlying mortgage default probabilities. Mian and Sufi (2009) demonstrate that house price appreciation was relatively stronger in low credit quality areas from 2001 to 2005. They also show evidence that stronger house price appreciation was driven by an expansion in supply rather than fundamental improvements in credit quality or income. If investors buying the mortgage-backed securities from securitization pool arrangers had a skewed sense of default probabilities given supply-induced house price appreciation, then they likely had a false sense of security that led them to be less mindful of the incentives of originators and servicers.

A related potential cause lies within credit rating agencies. It is likely that credit rating agencies also did not understand that house price appreciation in subprime areas was being driven by credit supply expansion instead of fundamentals. As a result, they provided unreasonably optimistic ratings on subprime mortgage-backed securities from securitization pools (see Greenlaw, Hatzius, Kashyap, and Shin 2008 for evidence on the dramatic increases in default probabilities for AAA-rated (p.104) mortgage-backed securities since 2007). There is evidence that institutional investors rely heavily on rating agencies to mitigate incentive conflicts between themselves and the originators of debt instruments (Sufi 2009). As a result of the high ratings on the securities, to preserve lender incentives, investors effectively relied on rating agencies rather than contractual arrangements.

A third potential reason for reduced lender incentives in securitization involves flaws in the contractual arrangements designed to preserve originator and servicer incentives. As mentioned above, one common contractual arrangement is to force the servicer or originator of the mortgages in a securitization pool to hold the most junior tranche at origination. This arrangement forces the originator or servicer to experience the first losses on mortgage defaults, and should therefore improve incentives to screen and monitor borrowers. A large body of research argues that such a contractual scheme has the potential to reduce agency and information frictions in securitization (DeMarzo and Duffie 1999; DeMarzo 2006; Ashcraft and Schuermann 2008; Gan and Mayer 2007).

However, there are two critical problems with the way this incentive scheme was implemented in mortgage securitization during the subprime lending expansion. First, the first-loss tranche of a securitization pool is often very small and is also often not actually held by the special servicer. Only half of the servicers in a sample used by Gan and Mayer (2007) of commercial mortgage-backed securities own the first-loss tranche of a securitization pool.

Second, even when the special servicer or originator holds the first-loss tranche, it may be able to sell or hedge the credit risk. For example, a January 14, 2008 article in the Wall Street Journal (Ng and Mollenkamp 2008) shows how a hedge fund called Magnetar bought up the riskiest tranches of securitization pools and subsequently hedged the risk. As the article reports, “Many hedge funds realized early on ‘that loans and securities that went into CDOs were extremely toxic, and the designed the structures to exploit that’” (p. C1). Magnetar bought the riskiest tranche from securitization arrangers who had strong incentives to sell given that “selling the riskiest pieces was ‘critical to getting the deals done’” (p. C1). At the same time, Magnetar was aggressively hedging the riskiest tranche by betting against the less risky tranches of the same securitization pools. In this manner, Magnetar actually profited when the defaults began to rise despite its holding the riskiest tranche of the mortgage pools.

This latter example shows the importance of contractually forcing originators and special servicers to hold at least part of the risk of the securitization pool and to disclose any hedging activity. It is not sufficient to allow them to sell the first-loss tranche to other investors, because other investors may have hedged the position in a way that leads them to buy the tranche at high prices even if the servicer has poor (p.105) incentives to monitor. There is precedent for such contractual restrictions on selling debt securities. For example, in the syndicated loan market, it is common to see contractual restrictions that only allow lenders to sell a piece of the loan with the approval of the borrower or the lead arranger (Pyles and Mullineaux (2008)).

Conclusion

I explore the relationship between foundational theories of financial intermediation based on lender incentives and the current subprime mortgage crisis. Incentive-based theories of financial intermediation argue that lenders must have risk in the credit they originate if they are to preserve incentives to screen and monitor borrowers. Although contractual arrangements in a variety of financial markets typically preserve incentives, the evidence presented here suggests a breakdown in lender incentives with the securitization of subprime mortgages. Securitization-driven mortgage expansion is concentrated in poor credit–quality areas with declining income opportunities over the period of credit expansion. Zip codes experiencing a large growth in securitized mortgages also experience a subsequent spike in default rates. The rise in housing prices, overly optimistic projections by credit rating agencies, and poorly structured contracts likely contributed to the failure of securitization to preserve lender incentives.

References

Bibliography references:

Ashcraft, Adam and Joao Santos. 2007. Has the Credit Default Swap Market Lowered the Cost of Corporate Debt?, Federal Reserve Bank of New York Staff Reports, Report no. 290, July.

Ashcraft, Adam and Til Schuermann. 2008. Understanding the Securitization of Subprime Mortgage Credit, Federal Reserve Bank of New York Staff Reports, Report no. 318, March.

Chomsisengphet, Souphala and Anthony Pennington-Cross. 2006. “The Evolution of the Subprime Mortgage Market,” Federal Reserve Bank of St. Louis Review 88: 31–56.

DeMarzo, Peter. 2006. “The Pooling and Tranching of Securities: A Model of Informed Intermediation,” Review of Financial Studies 18: 1–35.

DeMarzo, Peter and Darrell Duffie. 1999. “A Liquidity-Based Model of Security Design,” Econometrica 67: 65–99.

Demyanyk, Yuliya, and Otto Van Hemert. 2011. “Understanding the Subprime Mortgage Crisis,” Review of Financial Studies 24: 1848–80.

Dennis, Steven A. and Donald J. Mullineaux. 2000. “Syndicated Loans,” Journal of Financial Intermediation 9: 404–26.

Diamond, Douglas. 1984. “Financial Intermediation and Delegated Monitoring,” Review of Economic Studies 51: 393–414. (p.106)

Drucker, Steven and Manju Puri. 2007. On Loan Sales, Loan Contracting, and Lending Relationships, Working Paper, Duke University.

Gan, Yingjin Hila and Christopher Mayer. 2007. Agency Conflicts, Asset Substitution, and Securitization, Working paper, Columbia University.

Gorton, Gary B. and George G. Pennachi. 1995. “Banks and Loan Sales: Marketing Nonmarketable Assets,” Journal of Monetary Economics 35: 389–411.

Greenlaw, David, Jan Hatzius, Anil Kashyap, and Hyun Song Shin. 2008. Leverage Losses: Losses from the Mortgage Market Meltdown, Working paper, University of Chicago.

Holmstrom, Bengt and Jean Tirole. 1997. “Financial Intermediation, Loanable Funds, and the Real Sector,” Quarterly Journal of Economics 112: 663–91.

Hu, Henry T. C. (2009). “Empty Creditors and the Crisis,” Wall Street Journal, April 10.

Keys, Benjamin, Tanmoy Mukherjee, Amit Seru, and Vikrant Vig. 2010. “Did Securitization Lead to Lax Screening: Evidence from Subprime Loans,” Quarterly Journal of Economics: 125, 307–62.

Lee, Sang Whi and Donald J. Mullineaux. 2004. “Monitoring, Financial Distress, and the Structure of Commercial Lending Syndicates,” Financial Management 33: 107–30.

Longstaff, Francis, Sanjay Mithal, and Eric Neis. 2005. “Corporate Yield Spreads: Default Risk or Liquidity? New Evidence from the Credit Default Swap Market,” Journal of Finance 60: 2213–53.

Loutskina, Elena and Philip Strahan. 2009. “Securitization and the Declining Impact of Bank Finance on Loan Supply: Evidence from Mortgage Acceptance Rates,” Journal of Finance 64: 861–89.

Mian, Atif, and Amir Sufi. 2009. “The Consequences of Mortgage Credit Expansion: Evidence from the US Mortgage Default Crisis,” Quarterly Journal of Economics 124:4, 1449–96.

Ng, Serena and Carrick Mollenkamp. 2008. “A Fund behind Astronomical Losses,” Wall Street Journal, January 14: C1.

Parlour, Christine and Guillame Plantin. 2008. “Loan Sales and Relationship Banking,” Journal of Finance 63: 1291–314.

Pyles, Mark and Donald Mullineaux. 2008. “Constraints on Loan Sales and the Price of Liquidity,” Journal of Financial Services Research 33: 21–36.

Sufi, Amir. 2009. “The Real Effects of Debt Certification: Evidence from the Introduction of Bank Loan Ratings,” Review of Financial Studies 22: 1659–91.

——. 2007. “Information Asymmetry and Financing Arrangements: Evidence from Syndicated Loans,” Journal of Finance 62: 629–68.

Notes:

(1) For more information on the securitization process, see the excellent staff report by Ashcraft and Schuermann (2008) of the Federal Reserve Bank of New York. Much of the discussion in this subsec-tion comes from their report.

(2) For more information on this data set, see Section I of Mian and Sufi (2009).

(3) In particular, see Figure 1 in Chomsisengphet and Pennington-Cross (2006) and Figure 8 in Demayanyk and Van Hemert (2008). Demayanyk and Van Hemert demonstrate that investors are paid more in the cross section for lower credit quality mortgages, but the unexplained time series evidence suggests that spreads decreased to historical lows from 2001 to 2005.