Asset-backed securities

MISSING IMAGE

Material Information

Title:
Asset-backed securities the determinants of yield spreads
Physical Description:
v, 175 leaves : ill. ; 29 cm.
Language:
English
Creator:
Borgman, Richard H
Publication Date:

Subjects

Subjects / Keywords:
Finance, Insurance, and Real Estate thesis Ph. D
Dissertations, Academic -- Finance, Insurance, and Real Estate -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1994.
Bibliography:
Includes bibliographical references (leaves 168-174)
Statement of Responsibility:
by Richard H. Borgman.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 021604964
oclc - 33392490
System ID:
AA00013573:00001


This item is only available as the following downloads:


Full Text















ASSET-BACKED SECURITIES:
THE DETERMINANTS OF YIELD SPREADS







BY

RICHARD H. BORGMAN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA















TABLE OF CONTENTS



ABSTRACT ................................... ... ........... iv

CHAPTERS

1 INTRODUCTION ................................... 1

2 INSTITUTIONAL FEATURES OF SECURITIZATION ........ 8

The Market for Asset-Backed Securities ................... 8
The Framework of an Asset-Backed Security ................ 11
Implications for Pricing ............................... 26

3 PREVIOUS LITERATURE ............................ 32

Bond Pricing Studies ............................... .. 32
Mortgage-Backed Security Pricing Studies .................. 35
Asset-Backed Security Literature ......................... 45

4 ASSET-BACKED SECURITY RISK FACTORS AND
HYPOTHESES FOR PRICING ......................... 52

Default Risk (Part 1): Credit Rating ................ ..... 53
Option Characteristics of an ABS ........................ 57
Default Risk (Part 2): The Sufficiency of Credit Rating ....... 67
Marketability ...................................... 72
Collateral ................... ............... ........ 76
Summary: Hypotheses Concerning the Determinants of ABS Yield
Spreads ......................... .............. 77

5 METHODOLOGY AND DATA ......................... 80

D ata ................ .... ................. ...... 81
Sample Description: Entire ABS Sample ................... 85
Regression Sample ................................. 88








The M odel ........................................ 89

6 REGRESSION RESULTS ............................ 101

The Basic Absolute Spread Model ....................... 101
Interactive Variables for B Tranches and Multiple A Tranches ... 117
Reputation and Originator Rating ....................... 119
"Are Banks Different?" .......... ................... 122
Credit Enhancement ................................. 126

7 CONCLUSION ................................... 146

Implications of the Analysis ........................... 147
Extensions of the Pricing Analysis ...................... 149

APPENDICES

A ABS GLOSSARY .................... ............ 151

B REGRESSION TERMS AND ABBREVIATIONS ........... 155

C RELATIVE SPREAD REGRESSIONS ................... 162

REFERENCES ................................... .......... 168

BIOGRAPHICAL SKETCH ..................................... 175






















11iii














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


ASSET-BACKED SECURITIES:
THE DETERMINANTS OF YIELD SPREADS

By

Richard H. Borgman

December 1994

Chairman: Mark J. Flannery
Major Department: Finance, Insurance, and Real Estate



Asset-backed securitization has enjoyed rapid growth in its rather short existence

as a public market These nonmortgage, nongovernment-guaranteed asset-backed issues

have created a new source of fixed income securities for investors and a new source of

liquidity for banks and other lending firms. This study examines the determinants of

asset-backed securities' (ABS) equilibrium yield spread over Treasury, using a cross-

section of primary market issue prices.

The process of converting illiquid assets to traded securities involves a complex

set of institutional arrangements and structures that create an intricate set of risks for the

investor. These institutional arrangements, structures, and risks are analyzed as a

precursor for specifying a model of the determinants of pricing. This model extends








models previously used in pricing studies of corporate bonds and mortgage-backed

securities.

There is no standard data source for these securities, so there have been no

previous studies of how these securities are priced. A major contribution of the study is

the assembly and analysis of a substantial dataset that describes the pricing and

characteristics of over 700 ABS at issue. Ordinary least square regressions are utilized

in the pricing analysis, which includes issues from 1985 through 1992. This analysis

finds that ABS pricing (absolute and relative yield spreads) is rational and prices reflect

premiums for default risk, interest rate and reinvestment risk, and marketability. As one

would expect in a rapidly evolving market, institutional forms have changed frequently,

and new ones have been added, over the sample period. To some degree, the market has

required premiums for the unfamiliar or new and discounted for experience. ABS in

general do not exhibit negative convexity and are not subject to excessive prepayment

risk. Although the complicated structures utilized to separate the risk of the collateral

from the risk of the originator are effective, as indicated by the preponderance of AAA

rated issues, investors nonetheless require information in addition to credit rating

concerning an issue's pool and/or servicing when pricing the issue.














CHAPTER 1
INTRODUCTION

The significant growth of asset securitization by banks and other lenders via

nonmortgage, nongovernment-guaranteed asset-backed security issues has created a new

source of fixed income securities for investors and a new source of liquidity for banks

and other lending firms. While it has been suggested that asset-backed securities are

technically superior to traditional on-balance sheet means of financing because of

potentially lower costs and more efficient risk allocation, there has been no systematic

evaluation of how investors price these securities. This study examines the determinants

of asset-backed securities' equilibrium yield spread over Treasury, using a cross-section

of primary market issue prices.

The asset-backed security (ABS) market, in existence just since 1985, has rapidly

grown to over $60 billion of new issues a year. The process of converting illiquid assets

to traded securities involves a complex set of institutional arrangements and structures

(described in Chapter 2) that creates an intricate set of risks for the investor. There is no

standard data source for these securities, so there have been no studies of how these

securities are priced. A major contribution of the study is the assembly and analysis of

a substantial dataset that describes the pricing and characteristics of over 700 ABSs at

issue. This study offers the opportunity to analyze a new market and characterize the

market maturation process. It investigates the sensitivity of these asset-backed securities








2

to prepayment risk--do they exhibit sensitivity like mortgage-backed securities or exhibit

positive convexity like noncallable corporate bonds. It also analyzes the sufficiency of

credit rating to characterize credit quality in a heterogeneous market. Investors may

utilize other information in addition to credit rating, such as originator characteristics and

credit enhancement amount and type. If investors do use originator rating, the market

may not believe the claims of a "bankruptcy remote" legal structure.' Finally, the study

assesses whether standard proxies for marketability, default risk, and interest rate risk are

priced in this new market.

Securitization refers to the transformation of private loans negotiated between a

borrower and a lender into publicly traded (or at least saleable) securities. It separates

the originator from the ultimate investor. It is reflected by two streams: loan sales

(confined to banking) and the more narrowly defined (but more broadly practiced)

securitization, credit securitization or the issuing of asset-backed securities.2 Credit

securitization differs from single loan sales because it involves the pooling of assets--such

as mortgages, automobile loans, or credit card receivables--and the creation and sale of




'In the securitization process, a "bankruptcy remote" special purpose vehicle or entity
(SPV/SPE) is established to separate the risks of the pool of collateral from the originator. The
SPV is a specially chartered corporation that purchases the loans (collateral) from the originator
and is the actual issuer of the securities. Thus there is no liability created on the balance sheet
of the originator. For banks, the "bankruptcy remote" terminology is not exactly correct, since
banks cannot go bankrupt; rather, they can be declared insolvent. If this separation of the
originator from the issuer were not complete, it would imply an off-balance sheet liability for a
bank.

2Some authors would not include loan sales under the rubric of securitization. Greenbaum and
Thakor (1987) sharply distinguish between the two: "[S]ecuritization involves qualitative asset
transformation .... Thus, securitization enhances liquidity, reduces credit risk and restructures
cash flows. Loan sales merely separate funding from origination." (p. 380).










new securities collateralized by the pool of assets. Mortgage-backed securities usually

have an explicit or implicit guarantee from a U.S. government agency and are not the

concern of this study. Asset-backed securities with other types of collateral are addressed

here. These securities are usually rated triple A, a rating attained by the addition of some

sort of credit enhancement.3 Securitization transfers risk from the originator and often

restructures risk bearing to some degree.

The rapid growth of securitization over a relatively short time indicates that the

process has become an important tool for banks and other firms and an important subject

of study.4 The first nonmortgage backed, nongovemment-guaranteed, publicly issued

asset-backed security was introduced in March of 1985 by Sperry Corporation, a $192.5







'Credit enhancement refers to any additional protections for investors against default risk that
are included in the structure of the issue. These include guarantees of principal and interest
payments by third party insurers, senior/subordinate structures, and cash collateral accounts which
are a stated percentage of principal maintained as a separate account as a first loss protection for
investors.

4Asset securitization is used by many types of firms to reduce leverage, to diversify funding,
and, especially for lower rated originators, to raise less expensive funds. Asset securitization is
used by banks in fund raising, in asset-liability management, in meeting customer demand, and
in avoiding regulatory taxes (reserve requirements and deposit insurance premiums on funding
deposits, and regulatory capital requirements). Borrowers benefit through an increased supply of
loans, potentially lower costs, and increased credit on terms lenders might not provide to such a
degree otherwise, such as fixed rate mortgages. Investors have a greater supply and variety of
investment opportunities suited to specific needs. The banking system arguably has been made
more stable through an increase in liquidity for previously nonmarketable assets, through increased
opportunities for diversification, and because of an increased ability to compete for borrowers.
The overall economic system benefits through more optimal risk bearing. In fact, Roll (1987)
points to financial innovations such as collateralized mortgage obligations as helping to "complete
the market" and thus creating value (through lower required yields on the underlying mortgages).
Jameson, Dewan, and Sirmans (1992) attempt to measure the welfare benefits of CMOs and find
them to be substantial.








4
million issue backed by computer lease receivables.5 Yet in a short time the market for

these securities has grown dramatically. New issues for 1993 were just over $60 billion.

New issues for 1994 are expected to total between $62 billion and $70 billion, rising from

$1.23 billion in 1985.6

The growth in securitization is likely to continue, spurred by some recent

regulatory and legislative events. First, in November, 1992, the SEC passed Rule 3a-7

exemption to the Investment Company Act of 1940. Under this exemption, a broad

variety of assets not previously eligible can now be securitized.7 According to Asset

Sales Report, the adoption was "politically geared toward fostering the securitization of

small business loans" and thus to alleviate the credit crunch (ASR, November 30, 1992,

p. 1). However, the effects are more far reaching than just small business loans, allowing

for the securitization of student loans and unsecured consumer loans, for example.

Second, the Clinton administration has expressed support (with modifications) for the

Financial Asset Securitization Investment Trust (FASIT), proposed to Congress in May,


5The first private issue occurred slightly earlier in 1985 by Comdisco, a $35 million issue
backed by computer leases.

6Asset Sales Report, January 1, 1994, p. 1. Hereafter Asset Sales Report will be referred to
as ASR and references will be made parenthetically.

'Prior to this exemption, ABSs were issued under exemptions provided by Rule 3c-5 to the
1940 Act. Under Rule 3c-5(a), assets that primarily represented the purchase price of
merchandise, services, or insurance qualified--auto and credit card ABSs used this. It did not
apply (except for small amounts) to cash advances on credit cards. Under Rule 3c-5(c), assets that
represented an interest in mortgages or real estate qualified--home equity loans used this. Issuers
could also apply for special exemptions under Rule 6-c, but these were largely limited to "partial
pool" agency issues backed by mortgages. The new Rule 3a-7 expands the definition of assets:
eligible assets now cover "financial assets, either fixed or revolving, that by their terms convert
into cash within a finite time period, plus any rights or other assets designed to assure servicing
or timely distribution of proceeds to the security holders." By this definition, the only thing
clearly excluded is stock.








5
1993 (H.R. 2065) (ASR, June 28, 1993). FASIT would extend the tax benefits of the

Real Estate Mortgage Investment Conduit (REMIC) to assets other than real estate.8

Third, there has been legislative effort to encourage a secondary market for business loans

via the securitization technology. The sponsor, Representative Paul Kanjorski, said that

the "ultimate goal is to allow the private sector, through the new secondary market, to

take business, commercial, and community development debt and equity investments,

place them in securitized pools, and create investment products which are attractive to

pension funds and insurance companies."9 Whatever the political or economic merits of

these legislative proposals, it is clear that securitization is being pushed forward as if it

were a panacea for the credit crunch.

Given the significant growth of the ABS market and its future potential, a rigorous

evaluation of the factors that determine ABS pricing is important. This analysis finds that

ABS pricing (absolute and relative yield spreads) is rational and prices reflect premiums

for default risk, interest rate and reinvestment risk, and marketability. Spreads are found

to have widened over time due to market recomposition. But spreads for homogeneous




S"Under the proposal, the FASIT, a legal entity, could issue multiple classes of securities and
substitute assets. The latter action is essential for the securitization of revolving and short-term
assets like credit cards and trade receivables, home equity lines of credit and small business loans"
(ASR, June 21, 1993, p. 1).

'ASR, July 6, 1993, p. 1. This bill does not propose the creation of a new government
sponsored enterprise. Rather, it proposes a chartering process to allow a private entity to create
a market. The securities issued under the charter would be exempt from certain regulations, much
like securities issued by Fannie Mae and Freddie Mac. (According to ASR [April 26, 1993, p.
1] "This implies that small business loan securities would gain the benefits of the Secondary
Mortgage Market Enhancement Act of 1984, including favorable risk weighting.") However, it
also requires chartered entities to agree to goals to "promote lending to businesses in low to
moderate-income areas or enhance 'employment opportunities'."








6
securities show no trend. As one would expect in a rapidly evolving market, institutional

forms have changed frequently, and new ones have been added, over the sample period.

To some degree, the market has required premiums for the unfamiliar or new and

discounted for experience. ABSs in general do not exhibit negative convexity and are not

subject to excessive prepayment risk. Although the complicated structures utilized to

separate the risk of the collateral from the risk of the originator are effective, as indicated

by the preponderance of AAA rated issues, investors nonetheless require information in

addition to credit rating concerning an issue's pool and/or servicing when pricing the

issue.

Chapter 2 reviews the institutional features of the asset-backed market. First the

market for the securities is described, including its pattern of growth, range of assets

securitized, and buyers and sellers. Then the asset-backed security itself is examined in

some detail. This section provides background and impetus for the empirical study to

follow. Chapter 3 reviews the relevant pricing literature, drawn primarily from corporate

bond and mortgage-backed security (MBS) pricing studies. It concludes with a discussion

drawn from the ABS literature of additional ABS pricing considerations. Chapter 4

motivates the regression pricing model by examining the sources of priced risk and lists

the hypotheses under investigation. The methodology and data are presented in Chapter

5, while Chapter 6 analyzes the results of ordinary least square (OLS) regressions.

Chapter 7 concludes.

Appendix A contains a glossary of ABS terminology, and Appendix B defines the

variables and abbreviations used in the regressions. Absolute spreads are used in the








7
primary regressions in Chapter 6; Appendix C contains parallel regressions for the

relative spread model.














CHAPTER 2
INSTITUTIONAL FEATURES OF SECURITIZATION

The Market for Asset-Backed Securities

The growth of credit securitization has been truly astounding (see Figure 1). New

public issues of securities backed by assets other than mortgages were over $60 billion

in 1993 and may be as high as $70 billion in 1994 (ASR, January 10, 1994, p. 1).'

There is room to grow; by the end of 1990, for example, 40% of 1-4 family residential

mortgage debt was securitized, only 10% of consumer debt (Dreyer, 1991).

The variety of ABS collateral associated with 1992 issues is shown in Table 1.

All categories, with the exception of home equity loans and credit cards, increased over

1991.2 Although the market has been and continues to be dominated by just a few

collateral types, there has been an interesting variety of collateral involved in

securitization in the private and/or public markets. Among these are tenant leases (ASR,

March 22, 1993), payments due on travellers checks (ASR, October 5, 1992), cash flows



'"The asset-backed securities market, overall, has been the fastest-growing segment of the
fixed-income market." (Mitchell and Adler, 1991, p. C19). The ABS market, however, is still just
a fraction of the older mortgage-backed market. Outstandings of residential mortgage-backed,
pass-through securities went over $1 trillion by year-end 1990 (Dreyer, 1991). This popularity,
of course, has been due in large measure to the credit enhancements provided by government
agencies responsible for creating a secondary market for residential mortgages.

2Home equity backed issues dropped a whopping 40% from 1991, primarily due to the
sluggish economy and, especially, the increase in mortgage refinancing in 1992. Homeowners
often consolidate second mortgages and home equity loans into a new first lien mortgage.

8








9
from service contracts (ASR, November 23, 1992), third world debt (ASR, March 20,

1989), and U.S. military sales debt (by the Kingdom of Morocco, ASR, October 30,

1989).

The largest collateral category, the auto loan sector, offers a helpful

characterization of the market. While the market is broadening over time--a larger

number of smaller issuers--it is still dominated in dollar terms by a small number of high

volume issuers. In 1992, the top five issuers of auto ABSs originated 53% of the auto-

backed dollar volume (ASR, April 26, 1993), with General Motors the dominant issuer

(replacing Chrysler from 1991).3

Examination of the types of collateral points out a major characterization of credit

securitization. Some loans are more amenable to credit securitization than others. Loans

that are pooled must have similar characteristics so that the pool can be relatively easy

to evaluate by investors, rating agencies, and insurers. Residential fixed rate mortgages

used as collateral in mortgage-backed issues are relatively homogeneous and thus make

excellent pooled assets. Auto loans and credit card receivables are also relatively

homogeneous and thus form the bulk of nonmortgage asset-backed issues. Commercial

loans tend to be much less homogeneous and often must be evaluated individually. They

tend to be fairly large and complex and have terms that vary across borrowers. Thus they

are usually not pooled and are sold whole or in part by loan participation. Of course,

as the ABS technology develops and market conditions vary, this characterization is




'General Motors suffered a rating downgrade, which was its impetus for expanding in the ABS
market.








10
changing. For example, adjustable rate loans were generally considered less favorable

candidates than fixed rate loans for securitization because they do not have a fixed

payment stream, making them difficult to value and structure as an ABS. In addition, the

market preferred fixed rate issues until recently. But variable rate ABS issues have

become quite common since the fall of 1992, due to advances in the ability to structure

these issues and due to market demand because of cyclically low interest rates. Likewise,

some commercial loans have been securitized.

The largest sellers (originators) of ABSs are presented in Table 2. In dollar terms,

Citicorp is the largest originator, with over $30 billion of securities sold in 36 issues.

Chrysler, however, had 39 issues through the end of 1992, the most number of issues,

although Sears was not far behind with 37 issues. The top ten originators include five

banks and three automobile manufacturers, echoing the collateral types in Table 1 and

indicating the dominance of these two industries in the ABS market Banks, in fact, are

the largest originators with over $83 billion issued (see Table 2). But the captive finance

companies are not far behind with $80 billion sold. The captive finance companies

include primarily the consumer finance arms of the automobile manufacturers, as well as

providers of retail credit cards (e.g., department stores) and mobile homes (e.g.,

Fleetwood). The largest buyers of ABSs are pension funds, insurance companies, mutual

funds, thrifts, and commercial banks. Foreign buyers have become increasingly active,

especially with the increase in variable rate ABSs in late 1992, which are usually pegged








11
to LIBOR (London interbank offered rate). Recently, some ABS issues have tried to

attract money market funds by including a short-term money market tranche.4

The Framework of an Asset-Backed Security

The basic process of creating a nonmortgage asset-backed security involves six

steps: (1) pooling, (2) credit enhancement, (3) establishment of a special purpose vehicle

to own the assets, (4) repackaging cash flows, (5) rating the issue, and (6) selling the

securities to investors. There can be a large number of participants, with separate entities

possibly originating, issuing, structuring, distributing, servicing, insuring, monitoring, and

rating. The process is illustrated in Figure 2.

A large number of homogeneous assets (in terms of credit quality, maturity, and

interest rate risk) are pooled together by the originating firm in order to diversify credit

risk and to reach the minimum size required to justify a public securities offering. A

special purpose vehicle (SPV) is established to own the underlying securities. This trust

or corporation is set up to separate the risks of the pool of assets from all other risks of

the originator or fund raiser; the special purpose vehicle's business is usually restricted





4The first such ABS was a John Deere issue in September 1992, which included a money
market tranche with an expected average life of 0.42 years. It was structured as a pay-through
rather than a pass-through in order to create the short-term tranche (these terms are defined in the
next section). This issue has been successful, but other issues' money market trenches face
problems. The problem is that most money market trenches are fixed rate with expected average
lives of 4 to 8 months and with legal final maturities of 1 year. Because the exact maturity is not
sure, mutual funds must book these investments as 1 year holdings--i.e., to the final maturity date.
Because the SEC requires money funds to maintain an average maturity of 90 days, 1 year
holdings may increase the average too high. John Hollyer, manager of Vanguard's Prime
Portfolio money fund, rarelyey [makes] a one-year investment." The Deere tranche avoided this
problem by resetting the rate (against LIBOR) every three months. Money market funds are
permitted to book such resets as 90 day holdings (ASR, March 29, 1993, p. 3).








12
to the purchase of the assets and the issuance of securities backed by those assets.5

Because the assets in the pool are totally separated from the credit risk of the originator,

even should the originator default, investors have uninterrupted access to the underlying

assets. Thus, this is an improvement even over secured debt, where bankruptcy would

probably delay repayment. Removing the assets from the balance sheet via a true sale

to the SPV also has important regulatory advantages, especially for a bank, which thereby

reduces its required capital as well as the necessity of funding these loans with deposits

subject to reserve requirements. A nonbank also benefits by sale, because it lowers its

GAAP (Generally Accepted Accounting Principles) leverage which can affect debt

covenant compliance and/or pricing.6

Rights to receive the repackaged cash flows are sold to investors. An investment

banking firm typically underwrites the new securities. A rating agency will rate the issue.

A service, most often the originator, is responsible for collecting interest and principal

payments on the assets in the underlying pool and transmitting these funds to the

investors. The service is obligated to "manage and maintain control of the assets and

[the issue's] payment stream" (Watson and Joynt, 1989, p. 234). This is a key role.

These can be extremely complicated securities, and transaction management can be a

difficult task; it is important that it is done well and for the life of the issue. The

service must be of sufficient credit standing because among its functions is to collect and



5See Rosenthal and Ocampo (1988, pp. 48-63) for details on the tax considerations,
advantages, and limitations of various kinds of special purpose vehicles such as grantor trusts,
owner trusts, and debt

'Details on GAAP are provided in the next section on credit enhancement.








13
hold payments till passed on to the investor and to make advances on delinquent

payments. Also, the service, through collection procedures and policies, maintains pool

quality and can influence payment behavior.

Monitoring is performed by a trustee as well as accounting firms (periodic audits)

and the rating agencies. The trustee of an ABS, as with any security issue, represents the

interests of the investors. The trustee authenticates the issue's legality at the time of

issue, watches over the financial conditions and behavior of the issuer, and makes sure

all contractual obligations are carried out (Standard & Poor's, 1988, p. 19). The rating

agency continues to monitor the issue over its life as it reaffirms or changes the tranche

ratings.

Two apects of the securitization process demand special attention--credit

enhancement and the cash flow structure.

Credit Enhancement

Credit enhancement is added to the issue in order to raise the rating of the

resulting securities to investment grade. To be marketable and competitive with federal

agency-backed mortgage issues (i.e., government guaranteed), private issues need a

default risk comparable to that of Agencies'. The problem, of course, is that the

underlying securities to whose cash flows the ABS purchasers have claims are too

difficult or expensive for many market participants to evaluate for themselves. Hence,

successful securitization depends on the issuer devising a mechanism to separate the risk

of the ABS cash flows from that of the underlying securities. The credit enhancement

can occur in a variety of ways including a senior-subordinated structure, a recourse








14
provision, overcollateralization, letters of credit, surety bonds, spread accounts, or cash

collateral accounts. Many of these forms of enhancement can be provided by either a

third party or the originator.7

One way to divide the credit enhancement methods is between those primarily

provided by a third party and those that create a self-supporting structure. Third party

insurance was the primary form of enhancement throughout the 1980s; still very popular,

it includes surety bonds or guarantees by (primarily) insurance companies and letters of

credit issued by (primarily) large commercial banks. When the credit risk is shifted from

the pool quality to the balance sheet of the third party enhancer, risk analysis by the

ratings agencies focuses on that third party enhancer. For example, when insured by a

letter of credit, the Duff and Phelps credit rating on the tranche is determined by the

providing commercial bank's senior debt rating (see Gold and Schlueter, 1993, p. 156).8

Both letters of credit and surety bonds protect some stated percentage of principal

and interest payments, often up to 100%. The letters of credit must be irrevocable (not

standby) to avoid discretionary action by the providing bank at the time of need. Surety

bonds are "guarantees placed on the assets (i.e., student loan guarantees) or on the

individual notes or certificates (i.e., financial guarantees) to provide for payment of

principal and interest on the defaulted assets or securities" (Gold and Schlueter, 1993, p.



7Credit enhancement is a very important cost factor in credit securitization. Nine banks that
had issued card-backed securities in 1991 did not have any issues in 1992, partly because the cost
of credit enhancement had become too high (Kleege, 1993).

8S&P echoes this reliance on the third party enhancer's credit rating. However, Griep points
out that the failure of a few banks involved in the ABS business and the weakness of several
others have led to an increased examination of the underlying projects (Griep, 1993, p. 144).








15
156). That is, in the event of a borrower default, the insurer will immediately pay out

the remaining principal (a prepayment) and any interest due. A 100% "surety wrap" is

a guaranty of principal and interest on both A (the senior) and B (the subordinate) pieces.

The primary providers are financial guaranty firms, property-casualty insurers, and

banks.9 The originator may also be the enhancer if its corporate rating is high enough.

The use of self-supporting credit enhancement structures increased throughout the

late 1980s and into the 1990s for two important reasons. First, the credit quality of the

banking and insurance industries declined, accenting the event risk that the insurer could

have its rating lowered. An ABS can be rated no higher than that of the third party

insurer. Second, the banking industry experienced "the imposition of more conservative

bank regulatory treatment and risk-adjusted capital guidelines have increased the capital

charges for recourse" (Griep, 1993, p. 145). Banks withdrew from the ABS LOC (letter

of credit) business. Self-supportive structures include senior/subordination,

overcollateralization, spread account, and cash collateral structures. These enhancement

methods eliminate reliance on third parties, but they also reduce third party review of

collateral.

A senior-subordinate structure can have two or more trenches, with at least one

subordinate. In a typical deal, principal on the junior notes (the "B" class or tranche) is

not paid until the senior portion (the "A" class or tranche) is retired. If the amount of

subordination is 5% of the issue, the subordinate tranche absorbs the first 5% of any



9The financial guaranty firms are also known as monoline insurers and include MBIA
(Municipal Bond Investors Assurance Corporation), CapMac (Capital Markets Assurance
Corporation), and FGIC (Financial Guaranty Insurance Company).








16
losses on the underlying collateral, and thus provides the senior portion with credit

enhancement. The junior portion can be sold or retained by the issuer. If sold, the B

class requires a higher yield to compensate for this increased risk. In addition, the market

for these subordinate trenches is less liquid than the market for the senior trenches, also

contributing to a higher required yield. (Griep [1993] points out that these B securities

are increasingly being bundled and securitized themselves, increasing their liquidity.) A

potential problem with the senior/subordinate structure is a cash flow availability problem.

That is, with third party support, as soon as a loss is realized, a claim can be made to the

enhancement provider and funds are immediately forthcoming. However, a subordinate

class can only give up cash flows currently due to it, and these funds may not be able to

cover every loss in a timely manner.

With overcollateralization, the value of the underlying assets exceeds the face

value of the securities. The amount of overcollateralization thus provides a cushion for

the security holders and amounts to an equity layer in the SPV. This form of

enhancement is not usually a primary form of enhancement. It is usually used when

enhancement is "difficult to obtain but assets are plentiful" (Millard, 1993, p. 130).

A cash collateral account is basically a cash loan to the issuing trust (SPV) of an

ABS by the originator or a third party. The cash, often between 5% and 10% of the

issue, is invested in high quality securities and commercial paper. This highly liquid

account is then available to cover any borrower defaults. The cost for the SPV is the

difference between the rate charged on the loan and the rate earned on the cash collateral

investments. Cash collateral accounts have become a more prevalent form of credit








17

enhancement in recent years for two reasons--they are effective in eliminating the "event

risk" that the third party insurer may be downgraded, and they have been a way for banks

with less than triple-A ratings to get back into the credit enhancement business, since the

cash is provided up front and is therefore not dependent on the provider's rating.'0

Banks who previously acted as third party insurers by issuing letters of credit now

provide loans for cash collateral accounts. Cash collateral accounts are said to be more

expensive to provide than LOCs because they are funded assets, but it is felt that cost for

the SPV is made up in a reduced yield because of the event risk protection."

The type of credit enhancement utilized depends to some degree on the originator.

The goal of securitization in most cases is to remove the assets from the balance sheet

of the originator, i.e., to effect a true sale. Banks face stricter requirements than do

nonbanks in qualifying a transfer of assets as a sale. Under RAP (Regulatory Accounting

Principles), which banks must satisfy, sales with recourse are not "true" sales and the

securitized assets generally cannot be removed from the balance sheet (for regulatory

purposes)." However, GAAP (Generally Accepted Accounting Principles), which



'eThere is another way for less than triple-A banks to remain in the enhancement business and
still provide letters of credit, the rarely used "hybrid" structure. With this method a double-A
bank, say, provides a letter of credit with the stipulation that if the bank is further downgraded,
the letter of credit will be converted into a cash collateral account

"For example. Bank of New York, in a credit card deal in March of 1991, "saved 10 to 15
basis points in yield by using the cash backing, [said] market sources" (ASR, April 8, 1991, p.
8). The first issue to use a cash collateral account as a replacement for a letter of credit was
MBNA, also in March of 1991.

'However, a major exception to this rule applies to transfers of pools of residential mortgages
(not considered here) with a recourse feature. Such transfers can be reported as sales as long as
the originating bank does not retain any "significant risk of loss." The amount is vague, but
general consensus is that some amount less than 10% of the principal is not significant.








18
nonbanks must satisfy, does allow sales with recourse." This means that banks are

much more likely to employ third party guarantors, via a surety bond, a standby letter of

credit, or a third party cash collateral account. Nonbanks, on the other hand, are more

likely to employ the senior/subordinate structure and retain the junior position in the pool.

This junior position allows the originator to absorb the "usual" level of default risk

associated with the particular type of asset. These first loss obligations under the recourse

provision or the senior/subordinated structure make the sale one with recourse--a sale

under GAAP, not under RAP. Thus there could be no equity savings for banks using a

senior/subordinate structure where the subordinate position is retained. Instead, the

transfer would be treated as borrowing.

There is one exception, for banks, that allows them to offer additional credit

support themselves for nonmortgage collateral and still make a sale under RAP. This is

the "spread account" structure (also referred to as a reserve account). The spread account

is simply a type of escrow account. The way it works is simple: asset-backed certificates

carry promised payments below those on the underlying assets. The proceeds from this

spread, less a servicing fee, are collected in the spread account up to a stated level. After

that, the spread goes to the originator. The funds in this spread account provide credit

support for the asset-backed securities. After the securities are completely paid off,


"Asset transfers with recourse are treated more liberally under GAAP, where sales or
financing treatment are specified by Financial Accounting Standards Board Statement No. 77.
Three criteria establish a transfer of receivables with recourse as a sale: (1) the transferor
surrenders control of the future economic benefits relating to the receivables; (2) the transferor can
reasonably estimate its obligation under the recourse provisions; and (3) the transferee cannot
retum the receivables to the transferor except pursuant to the recourse provisions. Regulatory
Accounting Procedures (RAP) require (for there to be a sale) no risk of loss from any cause and
no obligation to any party for the payment of principal or interest for any cause.








19
anything left in the spread account will revert to the originator. Only at that time will the

originator take the funds in the spread account as income. Because it is not income until

that time, any use of the account to cover payment defaults will not result in a loss to the

bank. Consequently, regulators have determined that asset sales with spread accounts do

qualify as sales under RAP. Spread accounts are especially common with credit card

receivables.

It is quite common for issues to use a combination of credit enhancement

techniques to achieve the desired rating. A spread account may exist in tandem with a

third party guarantee, for example, thus lowering the cost of the third party insurance.

Cash Flow Structure

There are four major cash flow structures employed by ABSs: pass-through,

controlled amortization, bullet, and bond.14'1 These structures or similar predecessors

were developed in the mortgage-backed security market and are more familiar to many






"A passive grantor trust is the legal form of the entity (SPV) that will typically issue pass-
through securities. Controlled amortization and bullet structures (pay-throughs) and asset-backed
bonds are issued by corporations and owner trusts. For further discussion of these legal forms see
Pavel (1986, 1989) and Frankel (1991). Note that a passive grantor trust does not allow
transformation of cash flows, i.e., the cash flows are passed on to the ABS security holder when
and as received. Bonds and pay-throughs do allow transformation (e.g., monthly payments to
quarterly or semiannual payments).

"sWhile these are the major structures, commercial paper and preferred stock have also been
issued backed by pools of assets. Usually the commercial paper issuer or SPE is a conduit
established for the sole purpose of purchasing assets and issuing commercial paper. Preferred
stock ABSs are issued when the issuer has no tax liability because of tax-loss carryforwards, a
foreign tax credit, or an investment tax credit. Neither of these structures will be included in the
tests; commercial paper is not included in the sample, and preferred stock is different because of
the corporate dividend exclusion rule.








20
in that context.'" Credit securitization effectively began in 1970 with the Ginnie Mae

pass-through, developed by the Government National Mortgage Association (GNMA).

This is a mortgage-backed security collaterized by single-family Federal Housing

Administration (FHA) and Veterans Administration (VA) mortgage loans. A pass-through

represents direct ownership in a portfolio of loans that are similar in term to maturity,

interest rate, and quality." Certificates of ownership of the portfolio, which is placed

in trust, are sold to investors, who essentially have an equity position in the pool. The

loan originator services the loan by collecting interest and principal and passing them on,

minus a servicing fee, to the investors. There is often a second middleman, such as

GNMA, which receives the principal and interest from the originator and passes it on.

Because ownership lies with the investor (buyer), pass-throughs are removed from the

balance sheet of the issuer. GNMA is a direct agency of the federal government; the

government guarantees timely payment of principal and interest Investors therefore face

virtually no default risk, and an active secondary market provides a high degree of

liquidity for these securities.

Other pass-through mortgage-backed securities followed. The Federal Home Loan

Mortgage Corporation (Freddie Mac) developed the participation certificate (PC) in 1971

and the Federal National Mortgage Association (FNMA, i.e., Fannie Mae) developed the

mortgage-backed security (MBS) in 1981. Both are backed by portfolios of uninsured


'"The discussion here of the mortgage-backed market and its innovations is necessarily limited
in scope. For a more complete description of the securities and the historical perspective, see
Fabozzi and Modigliani, 1992, Chapters 2 and 11.

'7It is not permissible to substitute assets in this structure and usually there can only be a
single class, except for an additional single subordinate class retained by the issuer.








21

and privately insured mortgages. As indirect agencies of the federal government, there

is an indirect government guarantee of interest payments and full repayment of principal.

These mortgages tend to be paid off faster than Ginnie Maes. FNMAs have been

successful in part because of their swap program, where a mortgage lender can swap

whole mortgage loans for MBSs. The Mortgage-backed bond (MBB) is much less

common than the other types because the mortgages used as collateral remain on the

books of the issuer. The MBBs are reported as liabilities. The cash flows from the

mortgage collateral are not dedicated to the payment of interest and principal on MBBs;

rather, MBBs have a stated maturity and interest is usually paid semiannually. Credit

enhancement is provided by overcollateralization. The advantage to the issuer is that

through MBBs the loans are funded with long-term liabilities (usually between five and

twelve years).

The pay-through bond, like an MBB, remains on the issuer's books. However, the

cash flows are dedicated to servicing the bonds as with a pass-through. The collaterized

mortgage obligation (CMO), first issued by Freddie Mac in 1983, is a familiar example

(the volume of CMO issuance had become $59.9 billion by 1987). Typically, each CMO

issue is divided into a number of different maturity classes; four is common but there

have been as many as ten. The first (shortest maturity) class receives the first

installments of principal payments and any prepayments until class 1 bonds are paid off.

Then these payments go to class 2, and so on. These can be characterized as sequential-

pay classes. In this way the terms of the maturities are more certain and prepayment risk

is mitigated. Almost half of all CMOs are collaterized with federal agency pass-throughs








22
(a process referred to as resecuritization). The primary advantages of CMOs are the

creation of shorter maturities and the prepayment protection.

Despite these innovations, many potential investors did not participate in the MBS

market because of what they still perceived as significant prepayment risk. Thus,

structures were developed to create even more certainty in cash flows. In 1987, M.D.C.

Mortgage Funding Corporation issued the first Planned Amortization Class (PAC) bonds.

A PAC bond provides a fixed monthly payment as long as the prepayment rates on the

underlying collateral fall within a predetermined range. This cash flow stability is

achieved by giving principal payments of the PAC bonds higher priority than other CMO

classes, the companion or support classes which thus absorb a disproportionate share of

the overall prepayment risk. Rather than a sequential-pay characterization, as with the

earlier CMOs, these bonds are characterized as simultaneous-pay. Faster- or slower-than-

expected prepayment risk is absorbed by the companion trenches as long as prepayment

rates remain between the prepayment "collar" (generally between 80% and 300% PSA,

the Public Securities Association standard prepayment benchmark [Fabozzi and

Modigliani, 1992, p. 254]). A Targeted Amortization Class (TAC) is similar to a PAC,

but this class only offers protection from faster than expected prepayments. Their

expected average life can lengthen if prepayments are slower than expected.

Nonmortgage asset-backed securities utilize many of the same structures. The

pass-through is the most common structure and has been backed by auto loans, credit card








23
receivables, boat loans, and RV loans.'8 Despite their freedom from credit risk (due to

various types of credit enhancement), these securities still expose investors to interest rate

risk and prepayment risk. The latter is the risk that the investor who paid a premium for

the ABS will receive a lower than expected return because of faster than expected

repayment (generally occurring with a decline in interest rates) or that the investor who

bought the ABS at a discount will receive a lower than expected return because of slower

than expected prepayments.

Controlled amortization, bullets, and bonds offer progressively more certainty as

to the timing of cash flows and offer a reduction in prepayment risk. This is a result of

a transformation of cash flows not allowed with the pass-through structure. With

transformation of cash flows, not only can the payments be changed from monthly to, say,

quarterly, different classes of the liabilities can have different expected maturities and

payment schedules, such as interest only and principal only streams, or sequential classes

where prepayments of principal are first applied to one class, then the next, and so on.

Some of these securities may even have a stated maturity.

This restructuring of cash flows is advantageous for many institutional investors,

who prefer the more certain maturity and payments that match the frequency of their own

obligations. For example, thrifts prefer the shorter trenches, which match up better with

the short maturities of their liabilities. Pension funds have longer horizons. It also has

advantages for the issuer; by tailoring cash flows to meet investor demands, it may be



"Automobile loans were first pooled and sold in 1985 and are known as CARS (Certificates
of Automobile Receivables). Securitized credit card receivables are known as CARDS
(Certificates of Amortizing Revolving Debts).








24

possible to achieve a lower blended yield than with a single secured debt issue.

Rosenthal and Ocampo (1988, p. 55) claim that properlyry managed, several trenches of

securities of different maturities may be designed that produce a lower weighted average

interest rate than the interest rate that investors would receive for a single maturity

instrument." Presumably this translates to a net cost savings for the issuer.

Controlled amortization developed from the simultaneous-pay trenches of the PAC

structure. Controlled amortization refers to a cash flow structure with an initial revolving

period of uniform interest-only payments, followed by an amortization period of preset

length where interest and principal are paid to investors until the principal is paid off."

A typical structure might have a 36 or 48 month revolving period followed by a 12 month

amortization period (although there have been amortization periods as long as 24 months).

This differs from the PAC structure because instead of companion or support classes, the

usual structure uses some kind of reinvestment account or "principal funding account,"

whereby principal payments received before the amortization period are reinvested so as

to make likely sufficient cash flow to make all principal payments during the amortization

period. Similarly, if payments are slower than expected, the provider of the principal

funding account advances funds to make necessary principal payments. This is an

affordable option with nonmortgage asset-backs because prepayment risk is not as serious

a potential problem with the shorter maturity, less refinanceable collateral. Investors are

not totally free of prepayment concerns, however. Excessive prepayments could trigger


'There is a new variation on the controlled amortization structure, referred to a managed
amortization. In this structure, there is a managed amortization period, where principal is partly
paid out to investors and partly reinvested, followed by a rapid amortization period or payout
period. See ASR, December 14, 1992.








25

an "early amortization event," where the security immediately begins to amortize in the

interest-only period as a protection of principal. Such event "triggers" are described in

the issue's prospectus.

A bullet maturity structure is a refinement of the controlled amortization structure.

It implies a stated date where principal is paid off all at once, like a bond." Principal

payments received by the SPV before this date are invested in high quality short-term

investments until the expected maturity date. A soft-bullet is a variation on the bullet

structure; principal is paid off over a short period near the expected maturity date, thus

giving the issuer some flexibility.21

The advantage to the issuer of a controlled amortization structure is the ability to

pay out principal gradually. In a steep yield curve environment, a controlled amortization

structure is less appealing to an investor than a bullet, which has a single principal

payout, because investors face reinvestment risk, i.e., the possibility of reinvesting at a






"But note the difference from a bond--it need not be paid off at that time. To delay would
not imply default. "If the PFA [principal funding account] is not fully funded, a maturity
guarantor or the PFA provider will guarantee the investors' principal bullet repayment up to a
specific percentage of the initial offering" (Raab, 1990, p. 12).

""On the stated maturity date, the PFA may not be fully funded ... the current PFA balance
would be distributed to investors with the remainder of the outstanding balance amortizing until
the certificates are retired" (Raab, 1990, p. 11). One way that these principal repayment dates are
made more certain is through a "minimum principal repayment agreement," first used by GMAC
in 1987 (ASR, August 3, 1987, p. 1). Under this scenario, the security is issued under an assumed
prepayment rate. An up-front fee is paid to a guarantor (Morgan Bank in the GMAC issue) so
that (1) if the loans pay off faster than expected, the principal payments will be deposited with
the guarantor, who pays an agreed retum that is high enough to meet future principal and interest
payments; and (2) if the prepayment rate is slower than anticipated, the guarantor will advance
needed funds by acquiring issuer debt obligations.








26
lower rate. In a flat rate environment, controlled amortization is usually preferred, because

of the possibility of investing at a higher rate if the curve changes.

Even subordinate (B) trenches can have their cash flows made more certain. One

method is some form of "tail protection," i.e., a method used to shorten the payout period

on the security. This prevents an overly drawn out life for the subordinate tranche. An

example of "tail protection" is the "accelerated defeasance technique," which uses excess

cash flows from the deal to pay down the B securities after Class A has been paid. Or,

a stated maturity guaranty bond might be employed to "swap out the tail," that is, redeem

the remaining B securities at some predetermined date. Another method used is the

auction call option. In the case of a 1991 Household Finance home equity deal, for

example, after five years a call option takes effect if the original issue outstanding fall

below 25%. The trustee for the issue holds accepts bids for the option. If successful, the

B security holders are paid their principal. If the auction is not successful, the class B

holders receive a boost in yield to compensate them for the extra holding period (see

ASR, September 2, 1991).

Implications for Pricing

A few facts emerge from this discussion that could greatly affect the pricing of

an ABS. The collateral underlying these securities is broad-ranging, although there are

a few major categories. The originators vary as well--banks may be the most common,

but there are other financial companies as well as nonfinancial firms heavily involved.

These securities are typically credit enhanced to ameliorate default risk, but there are a

myriad of ways to accomplish this. The payment characteristics of the security can vary--








27
by different cash flow structures and by frequency. These are factors that will be added

to the traditional bond and mortgage-backed pricing models from previous studies. The

following chapter reviews this literature.








































1985 1986 1987 1988 1989 1990 1991 1992


FIGURE 1: ASSET-BACKED SECURITY NEW ISSUE VOLUME ($ BILLION)
(Annual Public Domestic Issues)


















Pass-through, asset-backed securities: structure and cash flows




Obligors Forwards
principal -id "Passes through" principal and
interest Tsee interest payments
--- payments Tstee -------- --------
Remit --*
principal and I
interest | Initial cash Initial cash Initial cash
payments proceeds proceeds purchase
from from of
Origminator/ securities securities securities
Sponsor/ Trust Underwriter Investors
Service Transfers Issues Distributes
loans on securities securities
P hareceivables Provides credit
Purchases
credit enhancement for the
enhancement asset pool, for example,
by a letter of credit

Credit nCash flows -
enhancer Scure _______


Source: Boeuctureo and Edwards, 1989

Source: Boemio and Edwards, 1989


FIGURE 2: THE ASSET SECURITIZATION PROCESS











TABLE 1: 1992 NONMORTGAGE ABS ISSUANCE, BY COLLATERAL


COLLATERAL $ (MILLIONS) %


Auto Loans 816.205 31.7%
Credit cards $15,677 30.6%
Home equity Loans $5,786 11.3%
Auto dealer floorplan Loans $3,500 6.8%
Manufactured Housing $2,645 5.2%
Leases computer, Railcar) $1,241 2.4%
Farm Equipment Payments $1,050 2.1%
Student Loans $373 0.7%
Exim Bank $352 0.7%
Smll Business Loans $350 0.7%
Home improvement $220 0.4%
RV, Boat Loans $198 0.4%
Other $3.603 7.0%


Total $51200


Source ASR, Fenruary 3.1993










TABLE 2: TOTAL ISSUE DOLLARS
BY LARGEST ORIGINATOR AND INSTITUTIONAL FORM


TEN LARGEST ORIGINATORS NUMBER TOTAL ISSUE MEAN
OF ISSUES DOLLARS MILL) SIZE

Citicorp 36 $30,133.6 $837.0
General Motors 27 $26,343.5 $975.7
Chrysler 39 $22,750.5 $583.4
Sears 37 $17,599.3 $475.7
Household Finance 21 $12,193.6 $580.7
Ford 8 $8,646.9 $1,080.9
Security Pacfic 24 $7,653.9 $318.9
Firlt Chicago 10 86,900.0 $690.0
Maryland National Corp 10 $5,024.7 $502.5
Chas Manhattan 8 $4,652.0 $581.5

INSTITUTIONAL FORM

Bank 190 $83,781.8 $441.0
Captive Finance Co. 149 $80,731.2 $541.8
Other Financial Institution 80 $28,609.7 $357.6
Savings and Loan 55 $9,127.4 $166.0
Non-Financial 29 $6,468.7 $223.1














CHAPTER 3
PREVIOUS LITERATURE

There has been no pricing study of ABSs. This study's analysis is anchored in

prior studies of the determinant yield spreads on corporate bonds and mortgage-backed

securities. This chapter contains a review of that literature and concludes with a

discussion of other possible ABS risk factors described in the ABS descriptive literature.

Bond Pricing Studies

The seminal bond pricing paper was by Lawrence Fisher (1959), whose basic

model is still employed in most studies today. Fisher's model states that a bond's risk

premium depends on a number of risk factors under the rubrics of risk of default and

marketability. The risk of default is captured by three factors: a measure of volatility

(Fisher uses the coefficient of variation of the issuing firm's net income), a leverage

measure (market value of equity to book value of debt), and the length of time the firm

has been operating without a default. Marketability is proxied by the market value of all

bonds outstanding by the firm. Fisher uses these four variables, all in log form, and

specifies a linear relationship. (He does examine some alternative measures.) The risk

premium depends positively and significantly on default risk, negatively and significantly

on marketability. The model has developed somewhat over time, especially in the

inclusion of variables for special characteristics such as call provisions, sinking funds, and








33
so on. But it is still common to assume a linear relationship between yield (risk premia)

and a number of risk measures or indicators.

The current study will use primary market issue prices. A number of bond pricing

papers have used primary market issue prices. Among them are Allen, Lamy, and

Thompson (1987), which examines alternative call provisions (call protection versus

refunding protection); Fung and Rudd (1986), which looks at the seasoning effect (new

issues versus existing issues) and the cost of underwriting; Sorensen (1979), which

examines the method of underwriting and bidder competition; Kidwell, Marr, and

Thompson (1984), which examines the effect of shelf registration on pricing; and

Billingsley, Lamy, and Thompson (1986) which looks at the pricing of convertible bonds.

All find that spread or yield measures (the dependent variable) depend positively

and significantly on default risk as measured by rating dummies and positively and

significantly on interest rate volatility. There are mixed results for issue size (a proxy for

marketability), call provisions, and maturity. The size of the issue enters negatively and

significantly except for Billingsley, Lamy, and Thompson (1986) and Kidwell, Marr, and

Thompson (1984), where the coefficient is negative but insignificant, and Fung and Rudd

(1986), where the coefficient is positive and insignificant. A call provision is generally

positive and significant, but shows no significance in Billingsley, Lamy, and Thompson

(1986) and, again, has the opposite sign in Fung and Rudd (1986). Maturity of the bond

generally has a positive, significant coefficient, except in Sorensen (1979) who uses an

"expected maturity" coefficient akin to the expected average life of an ABS. That is, he








34
averages time to maturity and time to call and this measure enters the regression equation

with a negative, significant sign.

The dependent variable in these models ("yield") can reasonably be measured in

several ways, and this choice has been shown to affect the stability of the regression

coefficients. The return on the bond, the dependent variable, can be expressed as the

absolute yield, the difference between the bond's yield and some index (the absolute

spread), or as a relative measure of yield which involves deflating the difference from the

index by the index (the relative spread). The concern that results in the latter

specification is that the absolute amount of the yield spread may somehow depend on the

level of the index.'

Lamy and Thompson (1988) present evidence that the relative yield spread model

is a superior specification. The authors refer to theoretical work by Bierman and Hass

(1975) and Yawitz (1977) that produce expressions that predict that risk premia are

positively related to interest rate levels.2 Lamy and Thompson empirically test this




'Lamy and Thompson (1988) review the use of these measures in the previous studies.
Studies that use the absolute yield include Marr and Thompson (1984) and Sorensen (1979).
Studies that use the absolute spread (and the index employed) include Kidwell, Marr, and
Thompson (1985; comparable maturity Treasury issues on the day of sale) and Fung and Rudd
(1986; a one day lagged Treasury Securities Index). Finally, studies using the relative spread
include Benson and Rogowski (1978), Cook and Hendershott (1978), and Billingsley, Lamy, and
Thompson (1986).

2Notice that this contrasts with the theoretical work of Merton (1974). Utilizing an option
pricing framework, he found the term premium or spread of a corporate bond to be a "decreasing
function of the riskless rate of interest." In a regression of absolute spread of corporate bonds on
(among other variables) an interest rate proxy, Fung and Rudd (1986) find a negative and
significant relationship. Rothberg, Nothaft, and Gabriel (1989), examining mortgage-backed
securities, find a similar result, which they find "anomalous" (see their note 7). Sorensen (1979),
again with corporate bonds, finds a positive and significant relationship.








35
concept. The absolute spread model demonstrates structural instability as the level of

interest rates change--especially the coefficients on default risk premia. The relative

spread specification is more stable.

Contrasting with this argument is the argument of Cook and Hendershott (1978).

In their example, assuming a constant risk factor, both absolute spreads and relative

spreads (the ratio of rates here) move with interest rates. The absolute spread rises as

interest rate level rises, the ratio of rates falls. But since the rise in spreads is

"approximately linearly related to the rise in yields.... an appropriate procedure when

estimating risk premium regressions is to use the spread as the dependent variable and the

level of rates as an independent variable to capture the effect on the spread of a constant

level of risk as yields rise" (p. 1180). Both specifications will be tried in this study.

Mortgage-Backed Security Pricing Studies

A mortgage-backed security such as a GNMA differs from a bond because it is

an amortizing security and because it is callable via prepayment or borrower default.

Thus, mortgage-backed security pricing models can be viewed as an expansion of bond

pricing models. That is, the basic linear model can still be used, but with different

independent variables which reflect the special nature of the securities. Credit risk,

because of the government guarantee, can be virtually ignored. Prepayment risk is a

major factor in these models and modelling or proxying for it is a major concern of

mortgage-backed studies.

Lacey and Chambers (1985) demonstrate empirically the existence of an option

premium in mortgage backed security returns. While GNMA security returns are








36
dominated by a fixed payment component, a separate option component coexists. Their

principal components methodology identifies the existence of the option premium by

examining four types of securities; it is found that returns from GNMA futures contracts

share orthogonal return characteristics with returns of futures contracts on fixed price

securities and futures contracts of option securities.

Given its existence, the pertinent question is how to proxy for prepayment risk in

the specification. Because the right to prepay can be viewed as a call option owned by

the borrower--the borrower has the right to call (buy) the mortgage at par--researchers

have looked to option theory to suggest proxies for prepayment risk. Hendershott (1986)

has reviewed the evidence and suggests that prepayment can be reasonably explained by

the observed term structure of interest rates and the volatility of spot rates.3 The intuition

is straightforward. A steeper (upward) slope of the term structure implies that lenders

(and investors) expect interest rates to rise in the future. The prepayment option on a

fixed rate mortgage is not as valuable in a rising interest rate environment. Thus, we

should expect an inverse relationship between slope of the yield curve and spread. This




3"[K]nowing the yield curve and the approximate variance of spot rates is sufficient to price
alternative mortgage features" (p. 505). It might seem preferable to use option pricing
methodology to price the call. Theoretical mortgage-backed valuation models do employ option
pricing theory to price the prepayment and default options (see, for example, Kau, Keenan, Muller,
and Epperson (1990) or Schwartz and Torous (1992)). The problem is that most popular models
are enormously complex, do not have closed form solutions, and require extensive numerical
simulations. In truth, according to Hendershott and Van Order (1987), "few studies attempt to
obtain realistic price estimates and even fewer compare estimates with market prices." Also,
Milonas (1987) has contrasted option pricing methodology with linear regression models. He
notes that option pricing techniques overestimate the value of the call because such techniques
assume an optimal call policy not practiced by these borrowers. Schwartz and Torous (1989)
allow for non-optimal prepayment in their valuation model and show that it fits the data better
than assuming an optimal call policy.








37
relationship follows from option pricing theory, which indicates that the value of a call

increases with the market value of the underlying security or asset. The market value of

a mortgage is equivalent to the present value of future payments, which is inverse to

interest rates. Thus the negative relation between interest rates and the value of the call.

Also, option theory tells us that volatility increases the value of a call. Intuitively, a more

volatile interest rate environment increases the probability of low rates in the future, thus

increasing the probability of prepayment, the value of the call, and the risk premium in

the spread.

The theoretical models from the mortgage pricing literature generally support

Hendershott and have implications for this study. Among the more important are Dunn

and McConnell (1981), Brennan and Schwartz (1985), and Hall (1985). Dunn and

McConnell (1981) attempt to document the effects of the amortization and call

(prepayment) features on the pricing of mortgage backed securities in comparison to non-

callable, non-amortizing bonds. They utilize an interest contingent pricing model with

a single state variable based on models developed by Brennan and Schwartz (1977) and

Cox, Ingersoll, and Ross (1978) in order to model GNMA securities. Specifically, they

model the effects of the call feature (optimal prepayments), amortization structure, and

suboptimal prepayments (caused by default, refinancing for equity, or sale without the

buyer assuming the mortgage) on the pricing, returns, and risk of GNMA securities.

GNMA's are, of course, default free. GNMAs differ from ABSs in several areas; for

example, all loans in a pool are required by GNMA to have the same coupon interest rate

and term to maturity and each is insured by the FHA or VA. However, Dunn and








38
McConnell's model and simulations are directly relevant to this study because of the light

they shed on the differences between amortizing and non-amortizing and between callable

and non-callable securities. Non-optimal prepayments are not truly an issue because

Dunn and McConnell assume that all suboptimal prepayments (including default) are

uncorrelated with all relevant market factors, are therefore unsystematic, and can be

costlessly diversified away.

Dunn and McConnell's model suggests that as the slope of the term structure goes

from negative to positive,4 the price (yield) of an amortizing, non-callable security

decreases (increases) relative to a non-amortizing, non-callable security. This occurs

because the later cash flows of a non-amortizing security are discounted more at higher

interest rates, likely in a flatter yield curve environment (the instantaneous risk free rate

is used as the discount rate in their model). This suggests, contrary to the negative

relationship between spread and yield curve slope suggested above, that the spread of an

amortizing, non-callable security from a non-amortizing, non-callable security such as a

Treasury bond should be a positive function of yield curve slope. Thus, if the call option

is not a dominant concern in pricing of an ABS, we should expect to find a positive

relationship between the slope and the ABS pass-through spread from Treasury. The

prepayment call option implies a negative relationship between spread and slope of the



4Although Dunn and McConnel employ a single state variable model, they run simulations
under different term structure environments. They can do this because the absence of arbitrage
in their model implies that the expected excess retum per unit of risk is the same for all interest-
dependent securities and thus their risk adjustment term (qr) is the same for all maturities,
including infinity. Thus they can assume various long-term interest rates R(-) and solve for q,
a risk-adjustment parameter that is in their model. However, as Brennan and Schwartz indicate,
this method assumes the variance of R(-) is zero.








39
term structure; the amortization feature implies a positive relationship between spread

and slope. For ABSs, we have to see empirically which dominates.

Also, their model predicts that amortizing securities are slightly less sensitive to

interest rate fluctuations than non-amortizing securities, again because the cash flows

come sooner for an amortizing security. Furthermore, the value of callable securities are

less sensitive to interest rate fluctuations than non-callable securities with the same

maturity. This interesting phenomenon occurs because the values of the call option and

a noncallable security both decrease as interest rates increase. The value of a callable

security is simply the value of an equivalent non-callable security less the call. This

difference is moderated because the two differencing values move together with interest

rates. Thus, according to the Dunn and McConnell model and simulations, the pass-

through structure, which is fully amortizing and exposes the investor to more prepayment

risk, would be expected to possess a less positive relation with interest rate volatility than

the more protected structures of controlled amortization, bullet, or bond.

Lacey and Chambers (1985), discussed above, point out that Dunn and

McConnell's one state variable model (instantaneous riskless rate of interest) is too

restrictive; Lacey and Chambers' results indicate that the return generating process of the

mortgage backed security responds to more than one source of uncertainty. Although the

principal components methodology they employ does not identify such a state variable,

the author's suggest interest rate variability as a likely candidate.

Brennan and Schwartz (1985) echo Lacey and Chambers by showing with

simulations that the two state variable model values the option better than the single state








40
variable model (i.e., Dunn and McConnell). Brennan and Schwartz use the instantaneous

riskless rate and the return on an infinite maturity consol (the authors are specifically

implying that yield curve slope is important). In comparison to their two state variable

model, the single state variable model underestimates the value of the call option.

Brennan and Schwartz's more accurate valuing of the call option properly simulates what

occurs as the slope of the term structure increases--the GNMA gains in value (reduces

promised yield) relative to the Treasury. This supports the negative relation between

slope and spread that we expect when the call feature dominates the amortization feature.

Further, Brennan and Schwartz's simulations support Dunn and McConnell's notion that

amortizing, callable securities are less sensitive to interest rate fluctuations than a

Treasury bond (non-amortizing, non-callable)--but only if the interest rate being measured

is the long-term interest rate. If it is the short-term rate that is measured, "quite the

opposite relation holds" (Brennan and Schwartz, 1985, p. 224). This is because the

GNMA behaves more like a short-term security. This is important, because ABSs would

be expected to behave even more like short-term securities. Thus, according the Brennan

and Schwartz's model and simulations and contrary to Dunn and McConnell's, the pass-

through structure would be expected to possess a more positive relation with interest rate

volatility than the more protected structures of controlled amortization, bullet, or bond.

Hall (1985) uses a two-state option pricing (binary) model to value the prepayment

option on a mortgage. He finds in simulations that the current level of interest rates has

little impact, but that the value of the option is negatively affected by the market's view








41
of the "direction of drift" of interest rates and especially and positively by the volatility

of expected future interest rates.

Empirical studies have supported the notion that interest rate volatility and yield

curve slope do proxy for the value of the prepayment call. Rothberg, Nothaft, and

Gabriel (1989) have utilized these proxies for prepayment risk in their study of the

relative yield spreads on mortgage pass-throughs. They find that both interest rate

volatility and the slope of the term structure significantly affect the magnitude of the

spread and are of the proper sign. In fact, they are the primary determinants (their

measure of marketability is not significant), thus lending some support to Hendershott's

conclusion that these two proxies effectively capture the prepayment risk.

Milonas (1987), like Rothberg, Nothaft, and Gabriel, utilizes a linear pricing model

to assess the determinants of the GNMA-Treasury spread. He utilizes both absolute

spreads and relative spreads with similar results. His pricing factors include proxies for

the prepayment option (current interest rate level, secondary market rates on FHA

mortgages, slope of yield curve (10 year Treasury less 3 month Treasury), and interest

rate volatility) and supply and demand variables (applications for GNMA's, number of

GNMA pools issued in a month, and new housing starts). He also finds the term

structure to be negative and significant However, the volatility measure is only used in

a regression not shown, and is insignificant Interestingly, the level of interest rates has

a negative coefficient and is "the major determinant of the yield spread," presumably for

both the absolute spread model (shown) and the relative spread model (not shown). The








42
number of GNMA pools issued in a month is not significant and is positive, not the

expected sign.

Arak (1986) finds that the factors that significantly affect GNMA-Treasury

absolute spreads are the slope of the yield curve (negatively), interest rate volatility

(positively), and the amount of adjustable rate versus fixed rate mortgages issues, a supply

proxy (negatively, i.e., as the amount of variable rate increases relative to fixed rate, the

yields on fixed rate GNMA decline). Arak hypothesizes that prepayment risk may be

linked to interest rate levels (the sign would vary depending if it was a premium or a

discount security), but the coefficient on this variable is insignificant. It should be noted

that the borrower, in addition to the option to prepay, also has the (put) option to default.

In a guaranteed mortgage-backed setting, when the mortgagor defaults on the mortgage,

the issuer is obligated to prepay to the mortgage-backed investor. This is a timing risk

that to the investor looks identical to regular prepayment risk. Optimally this option

should also be priced. Unfortunately, as Hendershott (1986) points out, valuing this

option in reality is very difficult because of unobserved values (the house price), unknown

rents (the value of future "dividends" in option pricing parlance), and an unavailable

series representing the volatility of individual house prices.

Epperson, Kau, Keenan, and Muller (1985) attempt to price the default risk in

mortgages themselves. The prime determinants are the volatility of the house (collateral)

price and the volatility of the spot interest rate. Kau, Keenan, Muller, and Epperson

(1987) include the value of the default option in their pricing model for commercial

mortgages and mortgage-backed securities. (Other independent variables include








43
uncertain future property values, a continuous term structure of interest rates, the value

of the scheduled payments, and the value of the prepayment option.)

Schwartz and Torous (1992) include the possibility of default on the underlying

loans in a two state variable, proportional hazards model of mortgage pass-through

securities. Default is important to the value of the mortgage backed security even when

principal is insured ("default free") because it affects the timing of cash flows.

Prepayment and default by the borrower result in an identical outcome to the insured

mortgage backed security holder--prepayment of principal. However, Schwartz and

Torous emphasize that such decisions by the borrower are taken under different economic

circumstances. Specifically, default dominates prepayment when both of the following

conditions hold: the value of the underlying collateral is less than the value of the

mortgage and the value of the underlying collateral is less than the principal outstanding.

The value of the mortgage is the present value of all future scheduled payments, which

is different than the principal remaining. If these conditions do not hold, prepayment

dominates default. The mortgage value is dependent on time to maturity, the value of the

collateral, and interest rates (here the instantaneous riskless rate of interest); the mortgage

value is inverse to interest rates, ceterus paribus. Thus default would be expected to

increase as the value of the collateral decreases (declining neighborhood conditions, "bad"

economic conditions, etc.) or as interest rates decline. In simulations, Schwartz and

Torous show that as interest rates increase, the value of the mortgage, insurance (the put

option), and the mortgage-backed security all decrease, all else equal. However, at high

interest rates the value of the mortgage backed security may not decline if the value of








44
the collateral is low enough. This is because security holders desire default and the

resulting prepayments because of reinvestment opportunities. Nonmortgage asset-backed

securities, where collateral value is much less likely to hold its value (consider mobile

homes or automobiles), may be likely to exhibit such a relation to interest rates. Interest

rate level and a measure of economic conditions, then, may be important to yield spreads.

In bad economic conditions, as the probability of default increases, investors would

require a higher yield to compensate for earlier than expected (or modelled by the issuer)

cash flows; but at high interest rates, such defaults may be desired, lowering required

yields.

Empirically, Vandell and Thibodeau (1985), using a logit model, more specifically

try to identify the factors which increase the probability of mortgage default. Their data

consists of a cross-sectional time series of 348 conventional fixed rate loan histories from

the Dallas area covering 1972-83. Explanatory variables include specific loan variables

(such as loan to value, payment to income), financial variables (such as corporate bond

returns, prime rate), borrower characteristics (such as marital status, employment history),

and housing market and general economic conditions (rental costs, neighborhood rating).

In addition to the specific loan variables, the only other significant factor was their proxy

for general economic conditions--specifically a "neighborhood" rating (a 1/0 indicator

variable). Thus, there is some evidence that economic conditions, although in this study

fairly localized, contribute to the probability of default

Pricing default risk or more specifically proxying for it is difficult at an aggregate,

non-individual level. Fortunately, as Hendershott points out, empirical work has shown








45
that the call dominates the put, indicating that ignoring the put option is less serious than

ignoring the call would be. However, the above studies indicate or at least suggest the

possibility that general economic conditions may proxy somewhat for default risk.

In summation, the mortgage-backed pricing literature provides indications of how

prepayment (call), default prepayment (put), and amortization should affect the ABS

spread from Treasury. Effective proxies for prepayment risk have been shown to be the

slope of the term structure and the volatility of interest rates. If prepayment is an

important concern, the spread should have a negative relation with the slope of the term

structure and a positive relation with interest rate volatility. However, the amortization

feature implies a positive relation between spread and the slope. For MBS, the call

(prepayment) has been found to dominate. The empiric results here will illuminate the

importance of prepayment risk with ABSs. Default caused prepayment risk may be

proxied by a measure of economic conditions; such a proxy should have a negative

relation with spread. There are some arguments that both prepayment risk (Arak, 1986)

and default caused prepayment risk (Schwartz and Torous, 1992) may depend on interest

rate levels, but there has been little empirical support for either argument.

Asset-Backed Security Literature

While the bond pricing literature and the mortgage-backed literature provide a

useful base for developing a pricing model for ABSs, nonmortgage asset-backs are more

complex than either bonds or mortgage-backs and require additional pricing

considerations. While there have been no pricing studies of ABSs, there is an extensive

literature on ABSs which illuminate these additional considerations.








46
Bryan (1988) describes ABSs in revolutionary terms; he sees ABSs as a key to

transforming the entire banking industry:

[Securitization] is better on all counts than the traditional lending system. It is
growing very rapidly precisely because it is the superior technology--one that, in
fact, is rendering traditional banking obsolete.... We estimate that it will take
10 to 15 years for structured securitized credit to displace completely the classic
banking system (Bryan, p. 65).

Further, "[s]ecuritized credit combines elements of traditional lending with elements of

traditional securities, but it also involves processes and structures not conceived of in

either traditional system" (Bryan, p. 70).

Specifically, he identifies four financial innovations that have been used to convert

loans into ABSs. The first, the creation of special purpose vehicles (the entity established

to purchase the pool from the originator and actually issue the securities), is important

because the SPV's "purpose is to isolate the risks inherent in the loans placed in it from

all the other risks of the funds raiser [i.e., the originator]" (Bryan, p. 72). This innovation

is significant, because if true than the financial condition of the originator should not

matter to investors in the securities. This separation or "bankruptcy remoteness" is an oft-

cited advantage of ABSs. However, as will be further explored in the next chapter,

investors may be concerned about the financial condition of the originator if this condition

affects the pool quality and the credit rating does not capture all pool quality effects, or

if the originator is also the service of the pool.

The second innovation is the pooling of borrowers, which diversifies credit risk.

This pooling, of course, is also an attribute of MBS. However, the third innovation,

credit structuring and enhancement, is more unique to ABSs because there is usually no








47

government guarantee of any sort. The guarantor assesses, underwrites, and guarantees

the credit risk, usually with the intent of raising the security to investment grade. This

allows investors, such as pension funds or individuals, to invest in these securities even

though they may have neither the skill nor desire to assess the credit risk themselves.

This would imply that if the resulting rating is a sufficient indicator of credit risk,

investors would not care about the amount or specific form of the enhancement.

However, investors may care about the enhancement if the rating does not capture all of

the potential risks to the investor resulting from underlying pool quality or type of

enhancement.

The final innovation is the repackaging of cash flows. This is most familiar in

MBS with the CMO, a sequential-pay series of trenches, or the senior-subordinate

structure. This allows the packager (investment banker who structures the issue) "to tailor

the cash flows of the different trenches to particular investor preferences ... [and] also

allows the packager to create trenches with different prepayment risk characteristics"

(Bryan, p. 73). However, MBS pricing studies have not attempted to model these

different tranches--they concentrate on the more homogeneous GNMA pass-throughs.

The ABS sample in this study (described in Chapter 5) allows a separation of trenches

into pass-through, controlled amortization, bullet, and bond. We would expect the more

structured and certain cash-flow structures to exhibit different responses to the prepayment

proxies.

Thus, Bryan's discussion of securitization innovations has illuminated three

additional areas of concern in developing a pricing model for ABSs. The effective








48
separation of originator and issuer, the sufficiency of credit rating, and the ability of the

various cash flow structures to combat prepayment risk.

Boemio and Edwards (1989), both members of the Federal Reserve Board's

Division of Banking Supervision and Regulation, delineate the risks that investors in

ABSs face. Investors face credit risk, that obligors may default on principal and interest

payments. More importantly in the current context, investors face "the risk that various

parties in the securitization process, for example, the service or trustee, will be unable

to fulfill their contractural obligations" (Boemio and Edwards, p. 663). To the extent that

a single organization performs several roles in the securitization process, investors face

a concentration risk of overexposure to a single organization. Boemio and Edwards also

indicate the possibility of "moral recourse," where originators, especially banks, may face

pressure to repurchase "securities backed by loans or leases they have originated that have

deteriorated and become nonperforming" (Boemio and Edwards, p. 664). Thus,

implications of the authors' analysis are that credit rating will be important to ABS

investors, but also that the originator may matter either because of its role as service

(operational and concentration risk) or because of some implicit moral recourse.

Goldberg and Rogers (1988) offer three areas of concern for investors when

evaluating an ABS: the characteristics of the collateral, the structural features of the

security, and the type and amount of credit enhancement provided. Collateral may matter

because of geographic diversity of the pool, whether the loans are secured by assets (auto

loans versus credit card receivables, for example), and because of different prepayment

risk profiles. While "virtually all ABS collateral have some degree of prepayment








49

uncertainty" (Goldberg and Rogers, 1988, p. 22), the gains from prepayment on ABSs are

limited by the short maturity and relatively small size of most ABS loans. However,

some ABS collateral, such as credit card receivables, have inherent payment uncertainty

because of a lack of fixed payment schedules.

The security structure is important to investors because of the independence of the

seller from the originator, the restructuring of cash flows, and the possible elimination of

prepayment risk. The creditworthiness of the originator should not affect the rating of

a properly structured ABS. However, the authors feel that investors "should distinguish

between external credit enhancement and internal credit enhancement" (Goldberg and

Rogers, p. 23). This is because with external or third party credit enhancement, the rating

of the ABS is subject to the "same risk of downgrade as the provider of the credit

enhancement." ABSs with internal credit enhancement, on the other hand, "will be

downgraded only if the quality of the collateral deteriorates significantly in relation to the

remaining credit support." Thus, the Goldberg and Rogers discussion suggests that at

least some types of collateral may be priced differently from others, that cash flow

structure is important, especially as regards prepayment risk, and that investors may

distinguish among types of credit enhancement

Similarly, Keighley (1993) lists four major areas of risk in an ABS issue: credit

risks, structuring risks including "the effectiveness of legal transfer of title of the assets"

(Keighley, p. 99), operational risks such as the continuing effective functioning of the

service, and financial risks "deriving from unexpected cash flows, such as pre-payments,

delinquencies" (Keighley, p. 99). Thus, Keighley echoes the above studies, and suggests








50

that investors may price default risk, prepayment risk, and default prepayment risk. In

addition, pricing of the originator's condition may reflect concern with operational risk

(if the originator is the service) and/or a concern over the effectiveness of the legal

separation of originator and issuer.

A final aspect that may be expected to effect ABS pricing is the newness and

maturation of the market. Barmat (1990), for example, feels the pricing of ABSs "has

evolved since the earliest issues. Initially, due in part to their novelty and similarity to

certain types of mortgage-backed debt, the yield of asset-backed securities in terms of

spread to Treasury bond yields was relatively high .... these spreads have trended down

since mid-1987 ... reflecting, among other things, increasing investor acceptance of this

type of security" (Barmat, 1990, p. 20). Some investors agree that wide spreads reflect

some premium for inexperience. Money manager Lawrence Harris of Alliance Capital

says: "Maybe that's why spreads are so attractive. They're paying us to take the risk of

inexperience" (in Sweig, 1989, p. 537). Thus, anecdotal evidence would imply a

narrowing of spreads over time. However, Barmat also points out that many new issuers

have entered the market and there has been a "significant broadening of the asset-backed

market according to asset type" (Barmat, 1990, p. 20). This broadening process might

imply a widening of spreads over time because these new originators and collateral types

are probably less well known to investors. Lawrence Harris also said, "There's a sort of

knee jerk reaction on the part of a lot of people ... these things are new and I'm not sure

I understand them ... It just takes time to overcome that" (in Sweig, 1989, p. 546). It

may be in fact that investors didn't completely understand them, that the "true" risks of








51

ABSs may have been greater (or lesser) than the inexperienced market first estimated.

Thus, detangling the time trend may prove of interest in understanding the development

of this market.














CHAPTER 4
ASSET-BACKED SECURITY RISK FACTORS AND HYPOTHESES FOR PRICING

As was made clear in the earlier discussion, ABSs are a rather heterogeneous

collection of securities, varying by collateral, cash flow structure, expected average life,

issue size, and type and amount of credit enhancement. This heterogeneity adds a level

of complexity to a pricing model. Like mortgage-backed securities, ABSs differ from

Treasury securities because of the additional call/prepayment option, because of payment

frequency, and because of their amortizing features. Like corporate bonds, ABSs differ

from Treasuries because they are not guaranteed by the government and carry varying

credit/default ratings. However, the basic risk factors in ABS pricing are familiar from

the previous literature discussion; the model's complexity derives from accounting for

the varying cash flow structures, subordination, and collateral.

ABSs can be viewed as amortizing bonds with embedded options; thus it is useful

to analyze the pricing factors of ABSs by first examining the option components of the

securities. These options, held by the insurer and the borrowers, are valuable and expose

investors to default risk and option induced interest rate or reinvestment risk. Then,

because ABSs involve the repackaging of illiquid assets into liquid assets, the features of

the issues and of the market that might influence marketability or liquidity are analyzed.

SPREAD, measured as either the absolute spread (the difference between the

expected yield on an ABS tranche and a Treasury of like maturity (i.e., equal to the








53
expected life of the tranche)) or the relative spread (absolute spread divided by the

Treasury of like maturity), would be expected to increase with an increase in default risk

(DEFRISK) or option induced interest rate or reinvestment risk (OPTRISK), and decrease

with an increase in marketability (MKT). In addition, the absolute spread (+), and

possibly the relative spread (-), may be a function of interest rate level (I) (Cook and

Hendershott, 1978).
+ +
SPREADi =f(DEFRISKb OPTRISKb MKTi, I. (1)

A discussion of these factors (and proxies included in the model) follows. In

addition, the complexity of these securities and the heterogeneity of the market suggest

other factors or characteristics may matter to pricing. Therefore, this chapter contains

discussions of the sufficiency of credit rating, what other features may be important, and

possible effects of differences in collateral. Credit rating may be an important but not

sufficient statistic for credit risk and investors will seek additional information. Similarly,

investors may seek additional information regarding non-credit effects such as undesired

cash flow timing changes or interruptions.

Default Risk (Part 1): Credit Rating

The exposure of ABS holders to default risk is reduced by the forming of a

diversified portfolio of assets to collaterize the security and the addition of private

(nongovernment) insurance in the form of some sort of credit enhancement. Thus, the

default risk of the tranche depends on the credit of the insurer for the amount of the

guarantee and the amount of enhancement relative to the true quality of the pool (i.e., the

percentage of principal guaranteed relative to the level of expected losses). ABSs are








54

sufficiently enhanced for the credit rating received. But even for a Triple A security,

default risk remains. ABS certificate holders are exposed to the risk that the credit

enhancer could default. Moody's and Standard & Poor's credit ratings of the trenches are

the proxies employed for measuring this default risk. While most ABSs are credit

enhanced, the ratings do vary and would be expected to be a major determinant of

pricing. A highly rated tranche (AAA) has a small probability of default; lower ratings

decrease the value of the security, and increase required yield.

There is evidence that financial risk premia will vary over time. (See, for

example, Cook and Hendershott (1978), Engle, Lilien, and Robins (1987), Ferson and

Harvey (1991), and Flannery, Hameed, and Harjes (1992) and the literature reviews

therein). This is a source of concern in this study because data are aggregated across a

seven year period. To help alleviate the problem, a control variable will be employed to

proxy for the changing price of risk. This proxy is the difference between Moody's daily

seasoned bond yields for BAA corporate bonds minus the yield for AAA corporate bonds.

This spread is indicative of the fluctuating price of risk (see Figure 3).

Thus, the amount of default risk (DEFRISK) is a function of the credit rating of

the tranche (RATE -) and the price of that risk is proxied by the corporate bond spread

(CSPR):
--+

DEFRISK, = f(RATE, CSPR). (2)

One is tempted to take the issue credit rating as a definitive, all-inclusive measure

of risk. However, with an issue as complicated as an ABS, other factors could matter as








55
well. Robert E. Pruyne, managing director of Scudder Stevens and Clark, says about

ABSs:

Some investors may rely very heavily on ratings, but we don't look at the
service ratings that closely. We're not tied to them. We rely on our own
homework and assign our own ratings .... we tend to be a little harsher
than the rating agencies. (in Zweig, 1989, pp. 541-42).

The rating agencies themselves explicitly do not evaluate some aspects of the

repayment flow. When rating a structured issue, S&P "ignores the obligor's

creditworthiness. Instead, a structured financing rating directly addresses the ability of

a specified asset or pool of assets to service payment obligations to the investor"

(Standard & Poor's, 1988, p. 19). They examine credit risk (including an evaluation of

portfolio quality that includes a review of the originator's credit and underwriting

practices, the service, and the trustee), cash flow (the ability to fulfill the "promise to

pay"), and legal issues (asset ownership and risks associated with the bankruptcy of the

seller/servicer). Credit enhancement is assessed in light of these risks.

But payment obligations for a pass-through, say, do not include the timing of the

cash flows, because the timing is not an obligation. Stress tests on the issue are run to

ensure that in worst case scenarios, principal will be protected. Further, while S&P does

consider that there should be "redemption provisions" in case of certain negative events

(Standard & Poor's, 1988, p. 23), it does not assess the probability of these provisions

being enacted. Redemption provisions include payout events and are viewed by S&P as

"safety valves" that minimize credit concerns. They end a nonamortization period prior

to schedule if the quality of the portfolio deteriorates substantially. Examples of payout

events include a substantial decline in yields, a significant increase in losses, a change in








56
borrower payment or borrowing habits that could adversely affect portfolio performance,

and issuer, trustee, or service default. "[T]he rating S&P assigns to a credit card

transaction generally does not address the likelihood of any of these early amortization

events and their resulting cash flow implications. the rating only addresses the

likelihood that investors will receive full return of principal by the 'final' maturity

date" (Griep, 1993, p. 138). Moody's ratings are similar: "Moody's further points out

that its ratings do not reflect the probability of occurrence of pay-out triggers" (Buerger

and Isely, 1989, p. 527, from ASR 8-31-87).

Thus, the agency ratings concentrate on the likelihood of principal repayment and

not on the timing of those payments. Whether these features of ABS cash flows not

considered in the credit rating matter to pricing depends on how investors feel about the

timing of their cash flows. The MBS literature identifies two features of timing, interest

related prepayments and borrower default related prepayments. These prepayments affect

MBS value because the cash flows are correlated with overall market conditions and thus

the effects cannot be diversified away. With insured ABSs, principal payments also occur

in connection with "early amortization events" mentioned in the quotations above. If

these cash flows are correlated with market conditions, market expectations about them

will be impounded in the initial ABS price (spread).

We now turn to an examination of these timing events using a simple option

perspective. Using the option framework hopefully leads to a better understanding of the

risks actually faced by an ABS investor by (1) separating the risks faced by the investor

and by the insurer, and (2) distinguishing the options owned by the borrower from that








57
owned by the insurer. Then we return to the discussion of the sufficiency of credit rating

and what other pieces of information may be utilized by investors to supplement the

agency rating.

Option Characteristics of an ABS

Like a corporate bond, an ABS can be valued as the difference between a riskless

bond (present value of future certain cash flows) and the value of embedded options.

That is:

B = PV(F) V(embedded options) (3)

In an ABS, these options belong to both the insurer and to the borrowers:

B, = PV(F) V,(embedded insurer option) V2(embedded borrower options) (4)

where B, is the value of an ABS.

In terms of loss of principal, the insurer faces the actual default risk from the

borrowers (up to the amount of credit enhancement); the investor faces the possibility

that the insurer may default on its promise to pay to the investors the defaulted principal

(exercise a put option). Thus, the embedded insurer put option represents to investors the

possibility of default effects after considering pooling and credit enhancement.

The embedded borrower options are the right to prepay without penalty and the

ability to default. The right to prepay is an American call option owned by the borrower

to purchase the loan at the face value of remaining principal (exercise price). The ability

to default is equivalent to an American put option to sell the collateral to the holder of

the loan (insurer of the ABS) for the face value of remaining principal. Because the

asset-backed securities are credit enhanced, the resulting cash flows from these two








58
options are identical to investors (once a loan is classified as in default, the guarantor

pays out the remaining principal (assuming the enhancement level is sufficient for the

level of defaults), which to the investor appears as a prepayment of principal). However,

as Schwartz and Torous (1992) point out, such decisions are taken under different

economic conditions. Thus, even though the result may appear the same to the investor,

regular prepayment risk and default prepayment risk must be differentiated in the pricing

model.

Only the borrower put option has relevance for the insurer. If the borrower

defaults, the insurer pays the borrower's unpaid principal to the security holder. When

the borrower exercises the call option to prepay, the payments are simply passed through

to the investor in a pass-through structure, or, in a controlled amortization, bullet, or bond

structure, placed in the guaranteed reinvestment account to earn interest hopefully

sufficient to make the future scheduled principal payments.

But both borrower options have relevance to the investor. The call is of concern

because prepayments are likely to increase when interest rates decline and the borrower

has cheaper refinancing opportunities or, likely with real assets such as automobiles,

trucks, manufactured homes, boats, and the like, the borrower pays off the loan when

he/she purchases a new similar asset at attractive financing rates. The investor in a pass-

through thus faces the interest rate risk of having to reinvest at lower rates. Similarly,

prepayments are likely to decline when interest rates rise and the investor would desire

prepayments with the resulting opportunity to reinvest at attractive rates. But even

investors in a non-pass-through structure have reason to be concerned. Excessive








59
prepayments faster or slower than expected can affect the timing of the cash flows in the

principal payout period of a controlled amortization structure and even extend the single

principal payment of a bullet structure into an amortization period. Furthermore,

excessive prepayments can trigger an early amortization event for either of these

structures, resulting in a payout of principal to investors much earlier than expected.

Since these prepayments are likely to be linked to interest rate levels, investors cannot

diversify away this risk.

Similarly, borrower defaults that result in insurer paid prepayments to investors

will matter to investors if they are linked to market-wide conditions and cannot be

diversified away and if these early payments of principal are undesired by the investors.

Finnerty (1993, p. 37) suggests that "general economic factors" and "general state of the

economy" influence the level of these defaults--a "bad" economy increases defaults. If

borrower default probability is linked to the overall economy, it cannot be diversified

away. If unexpected cash flows in a "bad" economy cannot easily be reinvested because

of a lack of investment opportunities or if the cash flows are received when interest rates

are low, then the borrower default put should require a premium by ABS investors.

Valuing the Borrower Options

We know from the comparative statics of the Black and Scholes option pricing

model that the value of a call is an increasing function of the value of the assets, the

volatility of the asset value, the risk-free rate of interest, and time to maturity. It is a

decreasing function of the strike price. And, the value of a put is an increasing function

of asset volatility and the strike price, and a decreasing function of the value of the assets








60
and the risk free rate of interest. Because this is an American put option, the value of the

option is a positive function of time to maturity (see Merton, 1990, p. 279). Of course,

we do not know the value of the underlying assets or their volatility. Thus pricing the

options using Black-Scholes pricing formulas is not possible. However, the previous

work in the MBS literature offers insight.

Borrower call option. As evidenced by the considerable work done in the

mortgage pricing literature, capturing the prepayment call option is of great importance

in mortgage backed pricing models. Whether prepayment is as important to ABSs is an

empirical question of this study. Regular prepayment risk (hereafter simply prepayment)

is the risk that cash flows may be received earlier or later than expected, i.e., than the

prepayment history of the pool or the prepayment model or algorithm used in the yield

calculations by the underwriter. When prepayments run higher than expected, security

holders face the risk of reinvesting elsewhere the early cash flows received, possibly at

a lower rate. The ex ante premium for prepayment arises because they tend to occur

precisely when rates are low. Thus, the security holder faces interest rate risk. When

prepayments run slower than expected, the security holder faces the problem of

liquidating outstanding securities at a lower value. That is, expected payments have not

been realized; if the investor has cash requirements (for example a pension fund), then

the securities must be sold. Thus prepayment is a valuable option held by the borrower

and has costs for the security holder.

Acceptable proxies for pricing this call option have been identified in the

mortgage-backed literature: the volatility of interest rates and the slope of the yield








61
curve. The value of the option increases with interest rate volatility and decreases with

an increasing yield curve slope as the probability of prepayment in the future decreases

if interest rates rise. For MBS, the link between market interest rates and mortgage rates

is fairly strong. Thus, the slope and volatility of Treasury rates effectively captures the

movement of mortgage rates. For ABSs, the connection is more tenuous, but still viable.

The movement of Treasury rates still proxies for the volatility of asset value and provides

a yardstick for future refinancing opportunities.

For an ABS, expected average life, akin to the maturity of a bond, could affect the

security's initial price and expected yield.' MBS pricing studies utilize similar maturity

securities (e.g., Milonas [1987] uses 30 year GNMA's). Thus, expected life is the same

for all securities and is not in the models used. But, the ABS sample contains trenches

of different maturities. Since a call option increases in value as time to expiration

increases, expected life should enter positively.

Thus the value of the prepayment call option (CALL) is a function of interest rate

volatility (VOL +), the future direction of interest rates (SLOPE -), and the expected life

(LIFE +):






'The average life is "the average time to receipt of principal payments (projected scheduled
principal and projected principal prepayments), weighted by the amount of principal expected
divided by the total principal to be repaid" (Fabozzi and Modigliani, 1992, p. 273). The average
life in years is:

average life= t (principal expected at time t)
12 ,. totalprincipal
where n is the number of months remaining.











CALLi =f(VOL,, SLOPE, LIFE). (5)

Borrower put option. The default prepayment option is much more difficult to

capture than the call. As with pricing the call, pricing the put directly with option

valuation formulas is impossible due to the lack of information on the underlying assets.

We can proxy for asset volatility with interest rate volatility. And, as with the call, the

value of the option should increase with expected life. However, most other parameters

are missing.

The mortgage pricing literature does provide limited evidence that economic

conditions provide some indication of borrower default risk. But these economic factors

tend to be localized, such as the neighborhood condition of Vandell and Thibodeau (1985,

see MBS literature review in Chapter 3). Asset-backed pools tend to be geographically

broad, making the collateral value less dependent on location. Thus a broader economic

measure is needed, akin to Finnerty's (1993) "general state of the economy." Note that

general economic conditions might influence both regular prepayment (call) and the

prepayments resulting from borrower default. But the effects should be opposite--a "bad"

economy lowering prepayment incentive and increasing defaults. Thus a negative relation

between the state of the economy and spreads may imply a premium for default

prepayment risk if the unexpected cash flows cannot be easily reinvested because of a

lack of investment opportunities in a poor economy.

The proxy employed here will be the previous six month percentage change in the

index of industrial production, although numerous national indexes and time periods








63
perform similarly.2 The problem with such a measure, of course, is that it presumably

captures much more than a default effect, but also especially supply and demand effects.

But if those effects are sufficiently captured by the other variables (issue dollars, interest

rate volatility, interest rate level, and so on) then this economic condition proxy may be

useful.

The value of the borrowers' option to default (BPUT) depends on economic

conditions (ECON -) and expected life (LIFE +):

-- +
BPUT =f(ECON, LIFE.). (6)

Effects of Cash Flow Structure

Amortization level. Adding complexity to the functional relationships of the

borrowers' call and put is that not all ABSs have pass-through structures. Controlled

amortization, bullets, and bonds have progressively more certain cash flow streams.

These structures should be less affected by prepayments, but they are not immune to the

effects, as discussed earlier. The principal payment stream over the controlled

amortization period could be affected by prepayments. The principal payout date for any

bullet--soft or hard--could be affected as well, resulting in an amortization period



2Other proxies tried were the composite index of leading indicators, the University of
Michigan Index of Consumer Expectations, the ratio of consumer installment debt to personal
income, the unemployment rate, percentage capacity utilization, and the change in the NYSE
composite price index. All of these were available in Kolb and Wilson, 1993. While most of
these proxies performed similarly, the percentage change in the Index of Industrial Production
entered with more significance and performed consistently. Logically it might seem that a
measure more inclined toward measuring consumer confidence might be an effective proxy,
because so many ABSs are collateralized by consumer loans. But it was felt that it is not
consumer confidence or expectations that matter so much as the aggregate consumer economic
situation, which might be effectively captured by industrial production.








64
following the predicted payout date (see Chapter 2, fn. 20). In order to account for these

perhaps differing prepayment effects on the various structures, interactive variables are

created by using structure dummies for controlled amortization, bullet, and bond

combined with the prepayment proxies.

However, we would not necessarily expect prepayment risk to be as important to

the pricing of an ABS as to an MBS. This is because the expected lives are generally

much shorter and because the incentives to refinance are not nearly as significant.

Prepayment patterns for auto loans and credit card receivables "are generally more stable

than residential mortgage prepayment patterns" (Finnerty, 1993, p. 36). In fact,

prepayment risk may not be an important factor at all, although this may depend on the

collateral involved. (Home equity loans, for example, are often refinanced when the

home is refinanced.) Dunn and McConnell's (1981) simulations show that as the slope

of the yield curve increases, the yield of an amortizing, non-callable security increases

relative to a non-amortizing, non-callable security. If an ABS pass-through is considered

an essentially non-callable security (i.e., the prepayment option is not a significant pricing

factor), then we would expect to find a positive relation between the slope of the yield

curve and the spread between an ABS pass-through and a Treasury security. This is

simply a way of saying that ABSs might exhibit positive convexity, like corporate debt,

rather than negative convexity like mortgage-backed securities. By extension, there

should be a predictable relation between the fully amortizing ABS pass-through and the

progressively less amortizing ABS structures of controlled amortization, bullet, and bond.

That is, the slope coefficients for these structures should be progressively less positive.








65
This is really a reinvestment risk problem due to the predicted cash flows, as

opposed to the unpredicted prepayments. In the trade press, bullet structures are said to

"roll down the yield curve" (ASR, November 16, 1992, p. 4) faster than a controlled

amortization structure, for example, and pass-throughs "roll down" slower (ASR,

November 23, 1992, p. 1; March 22, 1993, p. 1). The idea is simply that in a steep

positive yield curve environment, investors prefer receiving the principal farther out in

time, where it produces interest at the higher rate for a longer period. There is less

reinvestment/interest rate risk and a greater overall return for a bullet than for a similar

controlled amortization issue where principal is received earlier and must be reinvested

at a potentially lower rate. In a flat yield curve environment, there is little reward for

going long term and investors prefer receiving the principal earlier and facing the prospect

of reinvesting at potentially higher rates. Pass-through is preferred to controlled

amortization; controlled amortization is preferred to bullet. Thus, because we expect

opposite signs on the slope of the yield curve variable, the relative importance of

prepayment risk versus reinvestment risk for an ABS or, in other words, the convexity

of the security, is a testable hypothesis.

The proxy for default risk will be dummied by structure similarly to the call

proxies. In addition to pass-through, controlled amortization and bullets could pay out

principal early if an early amortization event is triggered. One early amortization event

typically occurs if defaults increase sufficiently to reduce portfolio yield below a base rate

set in the prospectus. Since neither Moody's nor Standard and Poor's considers the








66
possibility of these early amortization events when assigning their issue ratings, this type

of "default" risk should indeed be priced separately from ratings.

Subordinate and sequential-pay trenches. In effect subordinate (B) trenches have

negative leverage because B security holders own some small percentage of the pool yet

face 100% of the loss. The securities might be expected to be more sensitive than senior

trenches to the prepayment call option and the default prepayment put option proxies. 'A'

trenches in a multiple A tranche issue also face a possible sequential payment situation,

and some of them may be more sensitive than others to both types of prepayment. If the

splitting up of cash flows into multiple trenches is a more efficient distribution of risk,

then A trenches from a multiple A tranche issue may have a yield discount. Thus, both

B trenches and multiple A trenches will be represented by indicator variables in the

regression, and VOL, SLOPE, and ECON will be dummied to provide interactive

variables.

Issue Callabilitv

There is one additional option sometimes present in an ABS issue. The issue may

be subject to a call. Callability should matter, as long as the call is not a simple clean-up

call which occurs only at the end of the life of an issue when principal outstanding is

quite low. Investors should require a premium for callability.

To sum up, option induced risk (OPTRISK) is expected to be a function of interest

rate volatility (VOL +), the slope of the yield curve (SLOPE -), overall economic

conditions (ECON -), the expected life of the security (LIFE +), and the callability of the

tranche (CAL +):










+ + +
OPTRISK, =f(VOLi, SLOPE,, ECON, LIFE,, CAL). (7)

VOL, SLOPE, and ECON will be dummied by structure indicator variables to provide

interactive variables for different amortization structures and for differences in trenches.

Default Risk (Part 2): The Sufficiency of Credit Rating

Earlier, it was suggested that credit rating of the tranche proxied for the default

risk faced by the investor. But, it is not clear that tranche rating is sufficient to assess

risk in a complex, heterogeneous market such as the ABS market There are, potentially,

risks not captured by credit rating that may result in default, or more likely, in an early

amortization event or some other timing change. "Here, a transaction structure may

unwind (i.e., an early amortization event or event of default) due to even the most obscure

credit related problem" (Gold and Schlueter, 1993, p. 153). Due to these complex

structures, investors may utilize other information in addition to ratings that may be used

in the assessment of default risk specifically or, more likely, in the assessment of the

probability of undesired cash flow timing changes. Investors are likely to seek additional

information that indicates the current and future pool quality and/or that indicates the

probability of some sort of structural failure or early amortization event with an issue.

A low quality pool or a deteriorating pool increases the probability of future tranche

default, prepayments resulting from borrower default, and early amortization events.

Structural failure increases the probability of issue default, as well as increases the

possibility of an interruption of cash flows. The rating of the originator, the institutional








68
form of the originator, and the type and amount of credit enhancement may all be

considered.

The "bankruptcy remote" structure of an ABS would imply that the originator's

credit rating should not be important to the pricing of an ABS if the legal separation

between originator and actual issuer is believed to be effective. After all, unlike a bond

issue where the rating generally cannot be higher than the originating corporation's credit

rating, the "bankruptcy remote" structure of asset-backed issues allows this to commonly

occur. In fact it is an important reason a number of large originators such as Chrysler

or Citicorp have used the ABS market to fund assets. But despite the claims of

"bankruptcy remoteness," there may be legal risks associated with these asset transfers:

"experienced investors know that some of these provisions (e.g., nonpetition agreements)

may not be enforceable, that no entity can be made truly 'bankruptcyproof,' and that

consolidation cannot be judged by some 'safe harbor' checklist alone" (Buerger and Isely,

1989, p. 525). Due to ambiguities in the language of U.S. Bankruptcy Code and an

absence of definitive case law relevant to legal transfer, there is "the potential of a

bankruptcy court to recharacterize the legal transfer which in most transactions had been

intended as a sale of assets, to a pledge of assets, or in some cases, void the transfer

altogether" (Griep, 1993, p. 136). Thus, there is the risk that "bankruptcy remoteness"

may not be effective in all cases.

But even if the issue is "bankruptcyproof," the condition of the originator might

still be important for three reasons. First, bankruptcy of an originator might trigger an

early amortization event. Second, the originator has value to the pool after the sale of the








69
assets. The originator's financial condition is important to receivable collateral value,

borrower behavior, credit quality, and possibly the quality of servicing. Consider

warranties and product servicing. These can be important factors in resale value and

recovery value of collateral. (Buerger and Isely, 1989, offer the example of the depressed

resale value for Chrysler products during that firm's financial difficulties.) So a

bankruptcy by the originator could have repercussions on the value of a pool of assets.

Consider also credit card portfolios. Because of the nature of these receivables, the pool

consists of revolving credits and depends on new receivables being added to the pool.

If an originator is in trouble, it may be difficult to add receivables of the same quality or

any receivables at all. The problem is even worse for retail credit cards (issued by a

department store, for example) because if a bank fails, the credit card business would

most likely be divested and card usage continue. But with a retail card, when the

business fails most likely the credit card business would also shut down, resulting in no

new receivables.

Third, the originator is very often the service, and in "a single originator

transaction, it is likely that the originator may manage the SPV," i.e., "managing its debt,

interest rate swaps and foreign exchange payments, effecting cash movements, and

managing its accounts" (Keighley, 1993, p. 105). The servicing role is especially

important. A bankruptcy by the service could result in an interruption in the transaction

management duties and possible cash flow effects for investors. Worse, it could result

in a "permanent diminution in overall credit quality" (Buerger and Isely, p. 517.) These

would be effects beyond the historical experience used to estimate delinquencies, defaults








70

and recoveries. This is because the service, through collection procedures and policies,

maintains pool quality and can influence payment behavior. Efficient and effective

collection procedures are very important. Short of bankruptcy, it is still possible for

performance to deteriorate over time. Griep (1993) points out that troubled banks and

thrifts experience an increase in delinquencies and defaults. Service credit quality is

important for an additional reason: the service makes cash advances on delinquent

receivables and it collects and holds payments. This cash function could be affected by

service difficulties.

There is evidence that the originator's institutional form--bank, savings and loan,

captive finance company, and so on--might also matter to pricing. Previous studies have

found some evidence that banks, for example, are different. They may possess special

information or ability that other market participants do not, may structure loans

differently, or attract (or choose) different consumer clientele.3 Investors, then, may price

issues differently depending on the type of originator.

Finally, how an issue attains its rating may be important. Credit enhancement can

be provided via subordinate trenches, outside guarantees, inside guarantees, and many

other ways. Can the investor distinguish among these methods and does it matter to



3See Lummer and McConnell (1989) on the positive market response to bank loan renewals.
Billet, Flannery, and Garfinkel (1993), in an examination of commercial bank loans, find that
while loans from all originators (banks and non-banks) elicit a positive market response, the
quality (credit rating) of the lender, rather than the type, positively influences the market's
response to a loan announcement. Thomas Zimmerman of Prudential Securities researched home
equity loan prepayment speed. He found that the key to prepayment speed was where the
borrower is financing. Bank borrowers prepay faster, presumably because they are higher quality,
more sophisticated borrowers than those who borrow from finance companies (ASR, May 3,
1993).








71

pricing? Does the particular form and amount of the credit enhancement tell the market

anything in addition to what the credit rating indicates?

There is reason to believe the type of credit enhancement should matter. It has

been suggested that certain forms of enhancement are less sensitive to event risk, such

as a rating downgrade of the insurer. For example, unlike banks, the most common

provider of a letter of credit, no monoline insurer has ever been downgraded (see

Mortimer, 1993, p. 179). Also unlike banks, "monoline" insurers are not subject to

potential losses from other risk-carrying lines of business. Thus there may be some

pricing distinction between letters of credit and outside insurance. Also, a cash collateral

account has frequently been lauded as being without event risk altogether, because once

the cash has been provided it cannot be downgraded. Finally, a retained subordinate class

may offer advantages in either direction. Compared to third party enhancement, the

retention of a subordinate piece of the pool might be expected to decrease moral hazard.

On the other hand, Bhattacharya (1989) points out that a subordinate piece is not a perfect

substitute for other credit enhancement devices:

In the event of default or foreclosure, other credit enhancement alternatives
provide immediate coverage up to the limit protection. However, the ability of
the subordinated cash flows to meet shortfalls in the senior class cash flow is
limited by the balance of any fund created specifically for the purposes of such
contingencies and any current cash flow due to the subordinated holders. (p. 479)

The amount of credit enhancement may be important as well. It may be important

to the market if the credit rating does not assess the complete risk of a low quality pool.

The rating process on an ABS differs from the process with more traditional securities

because it "starts with the rating desired" and the rating agency tells the issuer what is








72
needed to achieve that (Watson and Joynt, 1989, p. 213). Thus the amount of

enhancement may be a direct indicator of pool quality. The quality of the pool, likely to

move with the economy and thus be undiversifiable, may be an indicator of the

probability of early payout events, and the amount of credit enhancement may be a signal

of this quality.

If originator rating, originator institutional form, or credit enhancement are

significantly priced, then the sufficiency of credit rating for assessing risk in an ABS

might be questioned. Such findings would indicate that credit rating of an ABS issue

does not assess all the implications of a low quality pool or weak structure, and therefore

the market seeks additional information. The market's assessment of credit risk

(DEFRISK), then, may be more than a function of issue rating (RATE -); it may also be

a function of originator rating (ORIGRT -), originator type (ORIGTP +/-), and credit

enhancement (CE +):

+
DEFRISK. -f(RATE,, ORIGRT,, ORIGTP CE). (8)

Marketability

A prime motivation for the restructuring and repackaging involved in an ABS

issue is to transform illiquid assets into liquid or marketable assets. Marketability may

be affected by the size of a particular issue or the size of the ABS market, by the

experience of investors with ABS structures, collateral, or originators, and by particular

issue characteristics such as frequency of payments.

The issue size is a common proxy for the marketability of any particular issue in

corporate bond studies (e.g., Billingsley, Lamy, and Thompson (1986), Kidwell, Marr, and








73
Thompson (1984), Fung and Rudd (1986), Allen, Lamy, and Thompson (1987), and

Sorensen (1979)). As the issue size increases, it is likely that the breadth of the market

for the securities will expand--there will be more investors holding the securities in their

portfolios and thus more regular trading. The expectation, then, is that larger issues are

more marketable and require lower yields. There has been a commonly found size effect

in bond pricing studies--size usually enters negatively and significantly (e.g., Allen, Lamy,

and Thompson (1987) and Sorensen (1979)). Alternative proxies for issue marketability

employed in this study are logged dollar value of an issue and the dollar value of the

issue deflated by the dollar value of corporate bond issues in the month of issue

(corporate bond issues were used because of the obvious time trend in ABS issues over

the period).

As a young, maturing market, the ABS market offers an opportunity to explore

the relative importance of market deepening versus market broadening over time, and to

determine if inexperience in these securities resulted in an initial overestimation or

underestimation of risks that was corrected over time. On the one hand, from its

beginnings as a public market in 1985, the ABS market might be expected to undergo a

deepening process as the instruments in the market get better understood by more

investors and the increase in the number of issues increased liquidity.4 This market

deepening would imply a narrowing of spreads over time. Spreads would also narrow if

the market's initial estimation of risk was overestimated due to inexperience.




4Money manager Lawrence Harris says: "It's been the same way with every new kind of
security that we've seen. It takes a while for liquidity to develop" (in Zweig, 1989, p. 535).








74

Similarly, two forces might widen spreads over time--a "market broadening"

process and a maturation process that reflects a correction for an initial underestimate of

risks due to inexperience. Over a relatively short period, the ABS market has absorbed

a tremendous increase in supply, has seen a large number of new, perhaps lesser known

originators enter the market, has broadened the types of collateral underlying the issues,

and has seen introduced a variety of instruments. Because of the rapidity of this growth,

the tendency toward the new and unusual, and possibly average collateral quality decline,

this broadening trend would be characterized by a widening of spreads over time. If ABS

risks were initially underestimated, spreads would also widen over time.

Two alternative proxies for a maturation or time trend that would capture the

dominant trend are a simple time count--this study uses a count by quarter with one being

the first quarter of 1985--and the logged total dollar value of issues in the entire ABS

public market prior to the day of a particular security's issue. Reasons for a significant

time trend include a changing composition of issuers, and/or collateral types, and/or

security designs, as well as changing risk assessments of stable issuer, collateral, or

security designs. Whether any trend holds for individual major originators and

homogeneous tranche types as well as the entire market should help determine whether

the effect is due to market deepening or broadening, whether it is due to an initial market

over- or under-estimation of ABS risk, or whether it is simply due to changes in market

composition.

A closely related marketability issue is that of "market saturation" or "hotness,"

that is, the dollar value of ABS securities currently available in the market. Because








75
secondary market information is not available, a proxy for market hotness is the logged

dollar value of ABSs issued in the previous three months from the day of a particular

issue. This quarter issue figure tends to follow an increasing time trend much like the

quarter count or logged total issue dollars. Most previous MBS studies expect a negative

relation with spread for such a supply variable (see Milonas [1987] and Rothberg,

Nothaft, and Gabriel [1989]), but find a positive, insignificant one.

The importance to pricing of investor "experience" with aspects of ABSs can be

illuminated more clearly, perhaps, by proxying for experience or inexperience more

specifically. First issues by a particular originator or of a particular collateral might

require a higher yield.5 Dummy variables for these issues will be included. The private

market, which tends to lead the public market, may mute the negative pricing effect of

this type of inexperience in the public market.6 Similarly, a very familiar, established

issuer who has established a reputation might be rewarded by issuing securities with

lower required yields. Proxies will include the logged dollar value of all previous issues

by an originator and, alternatively, a dummy variable for five or more previous issues.

Using either approach, experience may be a priced marketability factor.




5Ken Degen, Vice President of Structured Finance for MBIA, a major insurer, said "... we're
very cautious in taking on new asset types. To do a new asset type takes a long time.... A lot
of research and development go into new asset types-on all sides" (ASR, April 12, 1993, p. 6).

6Information on private transactions is more difficult to obtain. But Acheson and Halstead
(1988) note that a number of public issuers continue to use the private market, including Ford and
Chrysler. Also, Pavel indicates in the collateral category of lease receivables, for example, that
by June of 1988, there had been almost $1 billion issued publicly. Private issues at that time
totalled $500 million, a substantial percentage. ASR reports that private placement ABS issues
were $13.06 billion in 1993, a 12% increase over 1992. For 1993, that figure would place private
issues at 21.7% of the public market (ASR, February 28, 1994, p. 1).








76
Finally, specific tranche characteristics may increase or decrease tranche

marketability. For example, most ABSs have payments 12 times a year. However, some

trenches pay out only two or four times a year. If there is a demand for these less

frequently issued payment frequencies, perhaps because of a reduction in reinvestment

transaction costs or a better matching of required cash flows for some investors (a

clientele effect), then these securities may have a reduction in required premium. On the

other hand, if less frequent cash flows are not in demand, we may see the opposite effect.

Marketability (MKT) is a function of market maturation over time (TIME +/-),

market saturation or hotness (HOT +-), issue marketability (SIZE -), experience (EXPER

+/-), and payment frequency (FREQ +/-):

+
MKT =flTIME,, HOTS, SIZE,, EXPERT FREQ). (9)

Collateral

Collateral may reflect a great number of differences among issues. Unlike

mortgage assets, ABS pools are "characterized by an absence of standardized underwriting

and servicing. Consequently, portfolio quality and performance will vary widely"7

(Standard & Poor's, 1988, p. 71), perhaps across collateral type. Further, underwriting

standards could be affected by competition, growth strategies, and marketing methods

within an industry. Some industries/collateral-types may face different risks. For

example, competition in the credit card industry could affect pool cash flows if

competition forces a change in rates or fees. Or government legislation might change



7Mastercard and Visa, for example, unlike FNMA, FHLMC, or GNMA, do not require
adherence to specific underwriting standards.








77
pool behavior, for example the change in the tax treatment of interest expense that

occurred a few years ago. Goldberg and Rogers (1988) made the further point that some

distinctions may be made among collateral types because some securities are

collateralized by real assets (e.g., autos) while some are not (e.g., credit card receivables).

Thus, collateral type may be significantly priced because of different

characteristics or because they face risks omitted from the model and not captured in the

credit rating of the issue.

Summary: Hypotheses Concerning the Determinants of ABS Yield Spreads

Substituting (7) through (9) into (1) and including collateral and interest rate level,

we obtain the following model for ABS spreads.



SPREAD, =f(l, TIME,, HOT,, SIZE,, EXPER,, RATE., COLLAT,,
+ + + +
ORIGRT,, ORIGTP,, CE,, VOL,, SLOPE,, ECON,, LIFE,, FREQ, CAL). (10)

The conflicting signs on SLOPE and TIME should be clarified. For SLOPE, a negative

sign indicates that prepayment concern dominates, a positive sign indicates that the

amortization effect dominates. Zero would indicate that the slope is an ineffective proxy

or that the two effects net out. A negative sign on TIME would indicate that the market

deepening effect dominates and/or ABS risks were initially overestimated; a positive

coefficient would indicate that the broadening effect dominates and/or ABS risks were

initially underestimated.

The discussion of the ABS risk factors lead to several testable hypotheses which

are evaluated in this study. Except where noted, the null hypothesis is "no effect."








78
(1) Prepayment risk is a significant pricing factor with ABSs, and this risk diminishes

as cash flow structure becomes more certain. Thus, ABSs exhibit negative convexity and

the sign of the SLOPE coefficient is expected to be negative. The alternate hypothesis

is that ABSs exhibit positive convexity--prepayment risk is not a dominant concern while

concern over reinvestment of expected cash flows dominates. The slope coefficient sign

is positive and becomes less positive as expected principal payments are pushed out

further in the life of the tranche.

(2) Default prepayment risk is an important pricing factor and moves countercyclically

with the economy.

(3A) In a new and developing market, market "deepening" dominates and results in a

narrowing of spreads over time and/or ABS risks were initially overestimated.

(3B) Market "broadening" dominates and results in a widening of spreads over time

and/or ABS risks were initially underestimated.

(4) The market requires a premium for a lack of familiarity or experience such as a first

time originator or a new collateral type and discounts for an established reputation.

(5) Investors require information on pool quality in addition to tranche credit rating. In

particular,

(5A) Originator rating affects spread,

(5B) Originator institutional form affects spread,

(5C) Type and amount of credit enhancement affects spread.











BAAAAA -- RELATIVE SPREAD BAA-AAA
1.7 0.2



1.5 0.18


0.16
1.3


1.1 0.14


0.12


0.0.1


0.7
0.08


0.5 -+--- -1- --d -l --iN --I- 0.06







FIGURE 3: CORPORATE BOND YIELDS
SPREADS AND RELATIVE SPREADS














CHAPTER 5
METHODOLOGY AND DATA

The challenge of this study is to construct a model and develop a procedure to

price the relevant risk factors that acknowledges the heterogeneous nature of ABSs. The

model must allow for possible pricing differences among the different cash flow

structures, the level of subordination, the multiple tranche structures, and possibly the

major collateral groups. The plan of the study is to present an initial model that

highlights the relevant risk factors discussed earlier and, in particular, studies the

interaction between cash flow structure and proxies for prepayment and default

prepayment risk. Collateral type, subordinate B trenches, multiple trenches, payment

frequency, and first by an issuer or first by a collateral type are factored into the model

through the use of dummy variables. A time trend is included because this is a new,

developing market OLS regressions are used.

The initial regression establishes the ability of the model to explain a significant

portion of the variability in the spread from Treasury. It also assesses the importance of

prepayment and prepayment default risk to ABS investors and examines the ability of

different cash flow structures to ameliorate the effects of prepayment, prepayment default,

and reinvestment risk. Then, similar interactive effects are studied with regard to

subordinate trenches and multiple (sequential) trenches.








81

Further specific hypotheses are then addressed concerning the role of reputation

and the originator. The value of experience or reputation is expanded upon (from the

"firsts" of the initial regression) by proxying for previous experience in the market. The

role of the originator is first examined by adding originator rate to the model. It is

further studied by dividing the sample into three major collateral types. Dummies for

institutional form are added to the specification. The isolating of collateral not only

allows conclusions about why institutional form may be important, it also allows us to

assess the degree to which different collateral types face similar risks.

Finally, the role of credit enhancement is studied. Whether how an ABS attains

a certain rating matters to investors is analyzed by adding variables representing amount

and type of credit enhancement. The market may value enhancement differently by who

provides it or by the specific form of the insurance.

The analysis is conducted using both absolute spreads and relative spreads,

illuminating the effectiveness of each in capturing the above effects, and perhaps, as well,

offering insight into the stability of these different specifications (see discussion in

literature review).

Data

This study provides the first extensive examination of a large, fairly complete

sample of asset-backed securities. Data on the offer pricing and characteristics of ABSs

(seller (originator), issuer, class (tranche), principal, expected average life, collateral,

rating, yield, and lead manager) were collected from Asset Sales Report, a weekly








82
industry publication devoted to loan sales and asset-backed securities.' This information

was supplemented by detailed data on each issue's call, payment frequency, and payment

structure. These data were obtained from two sources: the Warga/Lehman Brothers

Fixed Income Data Base2 and Moody's Bond Record. The latter source also provided

tranche cusips (with which to access the Warga dataset), as well as issue ratings and

limited information on credit enhancement (usually just indicating the name of an insurer)

and payment frequency.3 Additional information, where required, is sought from the

"New Securities Issues" section of the Wall Street Journal and Dow Jones News Retrieval

Searches, including unpublished press releases from the various rating firms, insurers, or

originators.

Ratings for originators were based on the senior unsecured debt rating from

Moody's Bond Record at the time of issue. If unavailable in Moody's, Standard &

Poor's was consulted. In just a few cases, where bond ratings were unavailable, Moody's

thrift or bank long-term deposit rating was used. Originator ratings are for the parent

corporation, not the subsidiary (in consultation with Dunn and Bradstreet's annual Who



'Information from ASR was collected in two ways. Lexis/Nexis carries ASR from 1991 on.
Unfortunately, not all of what are called "New Issue Scorecards" are on Lexis (probably because
before mid 1992 they were considered tables or charts). Also, the editorial offices of ASR
allowed the author to examine their past issues. The author would like to thank Jeanne Burke,
editor, for her assistance. ASR is published by the American Banker.

2Developed by Arthur Warga of the University of Wisconsin--Milwaukee and based on the
Lehman Brothers Bond Indices, this new dataset contains structural information and monthly
pricing on fixed income securities. Even though asset-backs are not a primary component, this
served as an important source of data for this study.

3Until December, 1990, asset-backs were listed (sporadically) with regular bond issues. After
this date, ABSs are listed in a section titled "Structured Finance Issues." This section contains
no issue yields or secondary market pricing information.








83
Own's Whom), but in all but a few cases there is no conflict between the ratings of the

parent and the subsidiary (e.g., General Motors, not GMAC; Citicorp, not Citibank,

N.A.).

Credit enhancement information was gathered primarily from ASR, supplemented

by Dow Jones News Retrieval searches of the Wall Street Journal and the various press

releases included on this service. This information included type (letter of credit, surety

bond, cash collateral, etc.) and amount of enhancement (dollar and/or percentage), the

tranche(s) to which it applies, and the provider of the coverage. Moody's Bond Record

does not provide dollar or percentage of issue amounts of enhancement, but it does often

indicate the type of credit enhancement and the name of the insurer. Thus, this is

valuable supplemental information. In general, the credit enhancement information is

more complete for 1991 and 1992 (when ASR is carried on Lexis). The information that

is contained in the dataset created here is generally accurate and complete, except that

details on overcollateralization are scarce. Overcollateralization is commonplace, although

usually modest in amount. But details on the overcollateralization are generally available

only in the prospectus, which we do not have for most of these issues. Thus generally

we do not know the amount of overcollateralization. In two cases only, we know that

overcollateralization was the only form of credit enhancement utilized and we know the

amount. The overcollateralization amounts for these two issues only are included in the

credit enhancement sample.

Treasury rates are from constant maturity Treasury daily series as reported in

Table 1.35 of the Federal Reserve Bulletin. Spreads are calculated using a Treasury








84
security of comparable maturity, creating a term-structure adjusted risk premium. The

spreads utilized here will not exactly match the spreads quoted in the press for an issue.

There are two reasons for this. (1) Most spreads are quoted as the difference between the

ABS yield and the closest available Treasury. By contrast, the spreads in this study use

an linear interpolated Treasury rate for the maturity equal to the tranche's average

expected life. (2) Yields on ABSs are quoted as CBE (corporate bond equivalent) yields,

which are based on semi-annual interest and a 360 day year. Accordingly, constant

maturity Treasury rates, based on a 365 day year, are adjusted to CBE by multiplying by

360/365, thus yielding a slightly larger spread. Treasuries shorter than one year are

quoted as discount yields. These treasuries are first converted to BEY (bond equivalent

yields) based on 365 days and then adjusted by the 360/365 factor.

Yield curve slopes and interest rate volatility are calculated using rates reflecting

a 365 day year. The slope of the term structure is calculated as the difference between

the yield on the 5 year constant maturity Treasury and the six month treasury bill.

Interest rate volatility is calculated as the standard deviation of the yield of the 5 year

constant maturity Treasury for a period 30 business days before issue date.

The Index of Industrial Production is produced by the Bureau of Economic

Analysis of the U.S. Department of Commerce. It is published in The Survey of Current

Business and was collected through mid-1992 on computer disk in Realdata (Kolb and

Wilson, 1993).

The following section contains a discussion of the sample of ABSs created for this

study.










Sample Description: Entire ABS Sample

Tables 3 through 6 present descriptive statistics on the entire ABS sample

collected. As previously indicated, asset-backed securities are a rather heterogeneous lot--

which differentiates them strongly from the mortgage-backed market, for example. The

full sample (although not the regression sample) is a virtually complete representation of

the entire domestic, publicly issued ABS market from its inception in 1985 through the

end of 1992, including preferred stock issues, tax exempt issues, and interest-only

trenches. The sample contains 503 issues with 725 trenches totalling almost $209

billion.4 The impressive growth over time of the market is reflected in Table 3. From

7 issues (($1.2 billion) in 1985, the market grew steadily to 111 issues ($54 billion) in

1992.

The majority of the trenches are fixed rate (608), although variable rate branches

(117) have increased in recent years (see Table 4). Few of the trenches are callable (31)5

and most make interest and/or principal payments monthly. There are 97 subordinate (B

class) trenches and 606 non-subordinate (A class) trenches.6





4This figure is as close as possible to representing the actual public market for these issues.
For example, Gorton and Pennacchi (1992) have issue totals complete through 1988 provided by
Goldman, Sachs & Co. My figures compare exactly with their figures for 1985 and 1986. My
sample contains six more issues for 1987 and nine more issues for 1988.

'The call information, as with the payment structure and frequency information, is from the
Warga/Lehman Brothers dataset.

6Also categorized are interest-only trenches and preferred stock, which will be dropped from
the regressions, and mezzanine. A mezzanine tranche is senior to a B tranche, but subordinate
to an A. There are only 5 such identifiable trenches. Mezzanines are included (with B) as credit
enhancement to senior trenches.








86
There are five identifiable payment structures. The majority are simple pass-

through (310), but there are also a significant number of controlled amortization (113) and

bullet structures (88). There are 12 identifiable soft-bullet trenches; these are combined

with the bullets for the regressions. There are only 7 regular bonds. As expected, most

trenches (including many B classes) are rated Aaa, but tranche ratings range as low as B2.

Although much of the publicity for ABSs has centered on the variety of collateral

being securitized, in truth this is still a market dominated by just a few types of

underlying assets. While collateral (Table 5) is wide ranging, three types clearly

dominate--credit card receivables, auto and truck loans, and home equity lines or loans.

Together, these three collateral categories comprise over 80% of the market. Notice that

"credit card retail," such as department store cards, have been distinguished from "credit

card receivables" or general purpose credit cards, such as Visa, Mastercard, or Discover.

In 1992, credit cards comprised 30.1%, auto loans 31.7%, and home equity 11.3% of the

market.

The largest originators over this time period by dollars of securities issued were

Citicorp, with 36 issues totalling $30.1 billion, Chrysler with 39 issues totalling $22.8

billion, and Sears with 37 issues totalling $17.6 billion (see Table 2, Chapter 2). Note

that all three of these firms have seen some financial turmoil over the sample time period.

Chrysler, especially, has clearly been a major player in the ABS market in order to issue

securities at a higher credit rating than was available to them in the commercial paper or








87
bond markets.7 Banks (more precisely bank parent companies) have been the most active

single type of originator, with 190 issues. Captive finance companies, which include

GMAC, Ford Credit, and the like, are the second most active type with 149 issues.

The largest single issue was by GMAC for $4 billion in 1986, an issue containing

3 trenches. The largest single tranche was for $2.2 billion, a single tranche issue by Ford

Credit in 1989. The average issue (tranche), however, is much smaller at $415 million

($297 million). The average expected average life is 3.4 years, with a range from 0.2

years to 20.5 years. The average expected yield at offer is 8.09%, reflecting a spread

over similar maturity Treasury of 1.16%. A trenches, as would be expected, are larger

than B trenches ($337 million versus $60.3), have shorter lives (3.1 versus 5.2 years),

have lower yields (8.14% versus 8.3%), and have tighter spreads (1.1% versus 1.6%).

The slope of the yield curve, as measured by the difference between the 5 year

and 6 month Treasury rates, and interest rate volatility, as measured by the standard

deviation of the 5 year Treasury rate over the 30 trading days before an ABS issue, have

fluctuated quite widely over the sample period (Table 6, Panel D, and Figure 4). The

slope, in fact, ranged from a high of almost 3% in April of 1992 to barely negative slopes

in 1989 and 1990. Figure 5 illustrates the movement of the 5 year Treasury rate with the

slope of the term structure. They moved more or less together through early-1988

(correlation = 0.81) and they tended to move in opposite directions after this point


7See ASR, May 11, 1992, p. 1. "Though it is among the strongest finance companies, CFC
[Chrysler Financial Corp.] has been hampered by ratings downgrades attributable to the weakness
of its parent, effectively precluding CFC from funding itself in the unsecured commercial paper
and debt markets, said Stephen G. Moyer, high yield analyst at Kemper." At the end of 1989,
Chrysler had $10.1 billion of commercial paper outstanding, which had plummeted to $339
million by year end 1991.








88
(correlation = -0.75). Finally, Figure 6 traces the index of industrial production over the

sample period. The economy showed steady growth into 1989, where it began levelling

off. There were some dramatic fluctuations through the end of 1992, including a steep

drop in the second half of 1990 and into 1991.

Regression Sample

The sample used for the initial regression is limited by the requirement for cash

flow structure information and the exclusion of preferred stock trenches (15), variable rate

trenches (117), interest only trenches (2), and tax exempt issues (5). These were excluded

because it is likely that the pricing behavior of these securities would exhibit different

characteristics than the remaining sample.8 For most of the affected trenches, in fact, we

lack one or more of the required parameters--expected life or yield. Table 7 presents

summary statistics on the regression sample. The data are not significantly different and

indicate no systematic exclusion patterns from that presented on all ABS issues in Tables

3 and 6. The regression sample appears representative. Once branches with missing data

are excluded, the regression sample contains 452 trenches, down from the 725 total

trenches issued in this period.







'A preferred stock dividend rate is lowered by the 70% dividend exclusion enjoyed by
corporate purchasers of preferred stock. Tax exempt issues also have an obvious tax advantage.
Variable rate yields are expressed as a spread from some index, generally LIBOR, and they are
generally reset every three or six months. Thus, any link between the expected life of the tranche
and the expected yield will differ from fixed rate issues. They are effectively a series of short-
term securities (with a single transaction cost) which roll-over at the reset intervals. For more on
the valuation of floating-rate instruments and their differences from fixed rate instruments, see
Ramaswamy and Sundaresan (1986).










The Model

The specification of the initial model is


SPREAD, =a, + a, + ctVOL, + aSLOPE, + aSIZE, + axLIFE, + aTIME, (1)
+ aIP, + asRATE, + aCOLLAT + a,,FIRST + a,CALL, + a,,FREQ,
+ aoBDUM, + a,,MULTA, + asCSPR, + a INTERACTIVE,

where

SPREAD = Absolute spread calculated as expected yield on issue i minus the
yield on a U.S. Treasury issue with a comparable maturity at the
time of issue. When no matching maturity is available, linear
interpolation is used. For the relative spread specification, the
spread is divided by the interpolated treasury yield.

I = Interest rate level, five year constant maturity Treasury bond on the
day of issue. The coefficient is expected to be negative.

VOL = Interest volatility, the standard deviation of the annualized yield of
the five year constant maturity Treasury for a period 30 days
before issue date.9 The coefficient is expected to be positive.

SLOPE = the slope of the term structure calculated as the difference between
the yield on the 5 year constant maturity Treasury and the six
month Treasury bill.o1 The coefficient would be expected to be
negative if prepayment risk is a significant factor.


9The 30 day period is somewhat arbitrary. Rothberg, Nothaft, and Gabriel (1989) used this
period and found it significant in explaining the spread on mortgage pass-throughs. It assumes
market expectations of interest rate volatility are based on recent experience. The measure
employed here was chosen because it is simple, direct, and produced adjusted R Squares at least
as good as any altemative measure. Five other measures were substituted with very little effect
on the model. They were the absolute mean variation in the 5 year Treasury for periods of 10
and 30 days previous to issue; those absolute mean variations deflated by the five year Treasury
rate on the issue day (as employed by Chatfield and Moyer [1986]); and the standard deviation
using log differences for 30 days previous (as employed by Rothberg, Nothaft, and Gabriel
11989]), again using the 5 year Treasury.

1"In mortgage studies, the 10 year treasury is most often used. However, the rule of thumb
there is average life of 12 years for mortgages. ABSs have shorter expected average lives, so 5
years was selected as the far end of the pertinent yield curve. Slopes measured as the difference
between the 10 year and 6 month Treasuries and the difference between the 5 year and 3 month
Treasuries were also utilized with similar results.










LIFE = Expected average life, logged, measured in years." If risks such
as prepayment and default increase with time to maturity, the
coefficient will be positive.

SIZE = The dollar size of the issue, logged, a proxy for liquidity--larger
issues are expected to be more liquid. It has been demonstrated
generally in bond pricing studies that large issues tend to carry a
lower yield.'2

TIME = A numeraire to indicate time trend for market deepening or market
broadening, measured by a quarter count with one being the first
quarter of 1985."

IP = The state of the economy, represented by the previous six month
percentage change in the index of industrial production. The
coefficient is expected to be negative.

RATE = Dummy (1,0) variables for the issue's credit rating, a proxy for
default risk. The higher of either Moody's or S&P is used, if there
is a conflict.14 Aaa is the omitted rating (captured in the
intercept). Dummy rating groups are Aal/Aa2, Aa3/A1, A2, A3
and lower. The rating dummies should enter positively.

COLLAT = Dummy (1,0) variables for the type of underlying asset of the
issue. The sample is divided into 8 groups, with credit card
receivables the omitted group (i.e., captured in the intercept). The
other categories are credit card retail, auto/truck loans, home equity
loans, manufactured housing/mobile home loans, miscellaneous
commercial, and miscellaneous retail.


"There are two maturity dates for every class--expected maturity date and final maturity date.
The final maturity date is the actual final maturity of the underlying collateral. This ignores
prepayments which are expected and fairly predictable. The expected maturity date incorporates
expected prepayment and is the date, on average, investors can be expected to be repaid. This
date will be used in the analysis.

12An alternate proxy employed is the dollar value of the issue deflated by the dollar value of
corporate bond issues in the month of issue. There is no change in the results.

"'The logged dollar volume of all issues in the market up to, but not including, the day of
issue was also used, with very similar results.

"Note that Ederington, Yawitz, and Roberts (1987, p. 225) cannot reject the hypothesis that
"Moody's and S&P's ratings [are] interchangeable and equally reliable indicators of an issue's
creditworthiness."










FIRST = Two dummy (1,0) variables to indicate the first issue for an
originator or the first use of a type of collateral. It is expected that
the coefficient on FIRST will be positive, as the market will be
unfamiliar with something about the issue.

CALL = Dummy (1,0) variable with 1 indicating some type of call
provision. We expect a positive coefficient.

FREQ = Dummy (1,0) variables signifying interest and/or principal payment
frequency. 2, 4, or 12 times are possible in this reduced sample,
with 12 the omitted frequency. Signs may go either way;
reinvestment risk may be offset by a sort of preferred habitat effect
where buyers (mostly institutional) invest in securities with
payment frequencies to suit their needs.

BDUM = Dummy (1,0) variable to indicate that the tranche is a subordinate
B tranche. The sign is expected to be positive.

MULTA = Dummy (1,0) variable to indicate that the tranche is an A tranche
in an issue with multiple A trenches. Because of the possible
sequential pay nature of these trenches, they may exhibit different
characteristics than a single A tranche.

CSPR = The corporate bond spread between Baa and AAA, a control
variable for the time varying price of risk. In the absolute model
the absolute spread is used. In the relative spread model it is a
relative spread, the spread divided by the AAA rate. This ratio is
multiplied by 100 to put it in basis point form like the other
independent variables. Figure 3 showed the two move virtually
identically.

In addition, interactive variables are formed from cash flow structure dummy variables

multiplied by VOL, SLOPE, and IP, in order to investigate the interactions between

structure and prepayment and default risk. The indicator variables are for the structures

of controlled amortization, bullet, and bond. Pass-through is the omitted group. Later,

similar interactive variables will be formed using BDUM and MULTA, interacted with

VOL, SLOPE, and IP.








92

Variations on this initial pricing model, with five subsequent specifications ((2)-

(6)), constitute progressively smaller samples available. Variables added to the regression

are:

ORIGRT = Dummy (1,0) variables for the senior bond rating of the originator.
Originators were divided into five groups: Aaa to Aa3 (omitted),
Al to A3, Baal to Baa3, Bal to Ba3, and Bl to Caal.

REP = Reputation of the originator, proxied by the log of the total dollar
value of public ABS issues previous to the current issue.1
Reputation should enter negatively.

ORIGTP = Originator type. Dummy (1,0) variables for the type (bank, thrift,
non-bank financial, captive finance company, other). Banks are the
omitted group. Originator type may matter to investors if there is
some differences in origination standards, clientele, or servicing
and monitoring abilities.

INSURE = A vector of 13 credit enhancement variables, as logged percentages
of total issue, capturing the various types of enhancement available.
These factors should only be significant if investors require
additional information on pool quality beyond that contained in the
credit rating.

All regressions are repeated using the relative spread instead of the absolute spread. The

relative model is identical to the absolute model, with the exception that the interest rate

level is dropped for fear of spurious correlation with the dependent variable, and the

relative CSPR is used. Results will be discussed where relevant along with the relative

spread specification results. The results for the absolute spread regressions are contained

in Tables 8-16, and those for the relative spread specification in Appendix C.






"SAltematively, a dummy variable is utilized: 1 if the originator had 5 or more previous issue,
0 otherwise.














SLOPE

-.--- VOLATILITY


2.5 I
2,s !i


FIGURE 4: YIELD CURVE SLOPE AND TREASURY VOLATILITY 1985-92


0.6


0.5


0.4


0.3 -


0.2 5


0.1


S0











5 YEAR TREASURY SLOPE
12 3


11 i2.5



9 ifii i IN

I 1.5 S






i8 0.5


0


4 I1 1' --+ 1-l- -I -i 0.5






FIGURE 5:5 YEAR TREASURY AND SLOPE OF YIELD CURVE
(SLOPE = 5 YR TREASURY 6 MONTH TREASURY)











INDUSTRIAL PRODUCTION INDEX 6 MONTH % CHANGE IN INDEX
115




110




105




100




95




90
100 -- / ~^ l ij V


FIGURE 6: INDEX OF INDUSTRIAL PRODUCTION (1987 = 100)




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ELKGYA17D_RRL7XG INGEST_TIME 2013-04-08T22:57:40Z PACKAGE AA00013573_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES