1
Faculty of Economics and Business
Spreads to benchmark and credit qualities of Asset Backed Securities
The impact of 2007 subprime mortgage crisis
Author: Mina Lee
Student number: 10919120
Supervisor: Simas Kucinskas
Finish date: 26 June 2018
BSc Economics and Finance
2
Statement of originality
This document is written by Student Mina Lee who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
This paper assesses the role of subprime crisis of 2007-2010 in primary market spread at issue for European and the United States asset-backed securities from 2006 to 2016. More specifically, this study empirically investigates how bond and issuer characteristics differently affect issuance spreads during and after the subprime mortgage crisis that triggered financial turmoil. Furthermore, the paper discusses key aspects of ABS issues that affect issuance spreads and contribute to predicting the certain credit quality. The study’s main findings are that ABS issues in times of crisis have significantly lower spread compared to post-crisis ABS issues. Also, issues during the crisis are observed to be less likely to get highly rated while post-crisis issues are more likely to have higher credit rating.
Keywords: Asset-backed securities, Securitization, Structured finance, Special Purpose Vehicles, Credit ratings, Subprime mortgage crisis
4
Table of Contents
Chapter 1: Introduction ... 5
Chapter 2: Theoretical Background ... 6
Chapter 3: Methodology and Hypothesis ... 10
Chapter 4: Data ... 16
Chapter 5: Empirical results ... 20
Chapter 6: Conclusion ... 26
5
Chapter 1: Introduction
The subprime mortgage crisis from 2007 to 2010 has witnessed the need of review of the operation of asset-backed securities (ABS) market and securitization conduit in structured finance (Brennan, Hein, & Poon, 2009). ABS is one of structured finance products whose value and cash flow are
collateralized and derived from the pool of assets (Solomon, 2012). While the ABS market
encountered a severe breakdown, it points out a number of flaws and vulnerability to improper use of it. The global financial crisis of 2007-2010 has prompted regulators to implement more scrutiny and stronger measures to manage the risks and govern the vehicle. With lessons learnt from the collapse of ABS market as per the crisis, the implementation of new regulatory framework and measures helped to improve the reliability of credit ratings, develop the credit enhancement and restore the credibility of risk measurement (Standard & Poor's, 2008).
The financial turmoil demonstrates numerous vulnerabilities related to all parties associated with conduit securitization. However, much attention has been given to improper management of securitization process and failures of the credit rating agencies (CRAs). The key objectives of this study is to examine how the 2007 subprime mortgage crisis on ABS pricing and other common factors. Also, it aims to investigate the contribution of bond characteristics and crisis event on the credit quality of issues.
In addition to this motivation, another fundamental question which raised the research question was about how ABS market performed differently across different periods. In other words, the study is also aiming for examining the impact of crisis event on the performance of securities which will be
measured by the spread. Furthermore, the recovery of ABS market after the crisis has been proven that ABS can be highly beneficial for banks and investors if it is conducted by good management and correct guidelines (Pwc, 2011). Therefore, it is useful to look at characteristics of bond and issuer which are associated with spread.
Among several commentators who researched on the aspects of the crisis, Turnbull (2008) discusses about the early development of financial crisis and explains to what extent it has influenced the global financial system. Vink and Fabbozzi (2012) point out that the subprime mortgage crisis and collapse of ABS market have contributed to cause of financial crisis. Based on their model which empirically investigates the factors investors heavily rely on when pricing the yield, I further develop the study to examine the impact of the crisis and role of each bond characteristics in ABS pricing. In order to determine key factors which directly influence the pricing and help the recovery of ABS market, I regress bond and issuer characteristics on ABS pricing factors which can represent the performance of ABS issues. This way allows me to measure the effect of crisis on ABS spread and discover the key determinants which provide significant influence on it. Also, by running ordered probit model, I determine main characteristics of ABS issue which predict the certain credit quality (low, medium and high scale). Therefore, the model helps to interpret a number of factors that directly influence the quality of issues.
6
Chapter 2: Theoretical Background
2.1 The concept of asset-backed securities
Historically, banks played a role as financial intermediaries by collecting funds from depositors and providing debt financing to other borrowers. These funded loans were held in banks accounts until maturity. Gradually, banks began to diverge from this conventional funding and sought for new model of financing (Sebastian, 1990). Finally, securitization first introduced as a new financial instrument in 1970s in the fixed income market and it was observed to be highly beneficial and attractive to banks and investors over time (Solomon, 2012). This new source of funding, securitization, achieved steady growth since it allowed risk diversification of the underlying loans and to access funds more cheaply (Solomon, 2012) .
With a new banking model, banks began pooling the assets together selling underlying assets on their balance sheets to specifically established entities called special purpose vehicles (SPVs) (Solomon, 2012). A SPV is a separate legal entity created by originator or major investment banks (Efing, 2016). Once the underlying assets are sold to SPV and become the collateral for securities, investors evaluate the credit rating only on the collateralized assets but not on the originator (Sebastian, 1990). This aspect is critical to lenders since repayment and coupons from the SPV depend only on the underlying assets regardless of the originator’s financial abilities (Pwc, 2011).
Vink and Thibeault (2008) address that the securitization mainly consists of three different classes: asset-backed securities (ABS), mortgage-backed securities (MBS) and collateralized debt obligation (CDO). ABS which will be discussed in this thesis is structured finance instrument baked by
underlying assets or non-tradable consumer products such as mortgage loans, home equity loans, auto leases, credit card receivables and others (Choudhry & Fabozzi, 2004; Vink & Thibeault, 2008). As ABSs are created through securitization process, it is also necessary to look into features and uses of SPV.
2.2 Securitization process
In a basic form, the securitization process involves two steps. For the first step, a company which is known as originator identifies the assets it wants to remove from the balance sheet and pool together. Then, the originator sells the pooled asset to SPV which is an issuer playing a vital role in the efficient conduit and operation of securitization (Adams, 2005). It is formed to facilitate the securitization of the assets and protects the securitized assets from being administered as part of company’s bankruptcy estate (Taylor & Shane, 2013; Sebastian, 1990).
In the second step, this vehicle issues bonds by financing the acquisition of pooled assets as backing securities. Due to their high credit quality and rating which are evaluated independently from those of originator, SPV can borrow funds much more cheaply and uses borrowed money to fund the payment obligation to the originator for the assets (Agarwal et al., 2010). It groups the purchased underlying
7 assets or loans into tranches and sells them to a broad range of investors. By disaggregating the risk exposures and transferring to capital market investors who are willing to take on those risks, SPV provides investors the access to investment opportunities. Investors receive the floating or fixed rate payment derived and collateralized by a pool of underlying assets (Pwc, 2011). In this way, the legal entity SPV allows the sponsoring firm to generate a new source of revenue and meet the credit risk preferences of investors by tranching the pool of underlying loans (Almazan & Martin-Oliver, 2015).
The purpose of this approach is that the SPV can provide diversified and cheaper source of funding based on transfer of risk exposures from SPV to investors (Pwc, 2011). While it repackages the liability backed by the pool of assets into tranches with various risk profiles of investors, it can also create the advantage of cheaper borrowing on to the originator. Another feature of SPV that enables securitization to achieve a steady growth is the process of repayment while ensuring that it is formed for bankruptcy purposes as a separate legal entity from the originator (Ayotte & Gaon, 2011). Due to the fact that payment obligations and coupons from the SPV depend only on cash flow or pooled assets pledged as collateral to securities, acquisition of assets claims on the SPV but not on the originator. As a result, if the borrower fails to meet a financial obligation as a result of default, the originator is not obligated to repay the investors for the losses (Vink & Thibeault, 2008). Thus, as long as the SPV provides the revenue generation with sufficient cash flow, whether the originator becomes insolvent or not does not affect the repayments for the investors (Jobst, 2008).
2.3 Securitization history
Securitization is a financing process of issuing bonds or debt securities which derive the repayment of principal and coupons and collateralize by pool of illiquid assets (Cerrato, 2010). Over time,
securitization has steadily broadened the size and scope of structured finance products and grown its markets. The securitization generates ABS as one of the class of securities, first introduced in the U.S. mortgages market. Over the last three decades, ABS has evolved into a financial instrument which pools diverse types of illiquid assets and converts them to a tradable security (Vink & Thibeault, 2008). As a result of that, securitization became popular technique that provided alternative and cheaper source of funding to investment banks and corporations. Furthermore, financial deregulation of securitization market and easy credit access started to boom and increased the issuance volumes of debt financed securities (Pwc, 2011). International Monetary Funds (IMF) reports that investment banks and other financial institutions issued $1 trillion worth of securitizations which was
approximately five times that of Europe by 2000s (Miguel et al., 2013). Although debt financed consumption continued to soar, the U.S. subprime mortgage crisis of 2007-2010 which triggered financial crisis interrupted the constant growth of the securitization market (Coval et al., 2012). While causing financial turmoil, ABS market collapsed and lost a large number of investors from 2007 to 2010. The crisis prompted a series of major breakdowns in the global securitization market and caused the financial system to become fragile. Although the securitization market deteriorated drastically for several years, it demonstrated that the market has recovered since 2010 and the
8 number of investors also increased. Consequently, securitization issuance totaled $2.2 trillion in 2016 driving the increase the amount of $0.2 trillion from 2015 (Chris & Joseph, 2016).
2.4 Structural characteristics of securitization
The SPV improves the credit quality of a bond or securities with the use of two different types of credit enhancement mechanisms. This technique can be provided either internally or externally. The internal credit enhancement is provided by the sponsors and achieved through subordination which is the most common mechanism. Subordination is structured by the SPV through dividing the liabilities into tranches (Jönsson, Schoutens, & Eurandom, 2009). By tranching securities into different classes, it helps to optimize the risk profile of the bonds and manage diverse range of investors. The process of pooling and tranching leads to senior and subordination structure where senior tranches, rated AAA, hold the priority of collecting the payment over more junior or subordinated tranches which have lower credit ratings. The SPV generates cash flow backed by the pool of assets and allocates them with different priorities to classes in order or seniority (Klee & Butler, 2002). However, when the losses incur, those are distributed from the subordinated tranches with lower credit quality to the most senior tranche. Due to this fact, the subordinated junior tranches perform the function of providing protective layers to senior tranches which hold the first priority on cash flow. For instance, in case of default of underlying assets or financial distress, the subordinated tranche takes the allocation of losses while the senior tranche can still receive the payment of interest and payment due to priority rights and stay unaffected (Jönsson, Schoutens, & Eurandom, 2009; Pwc, 2011).
The external credit enhancement involves taking guarantees on ABS provided by a third party who is not a key player in securitization process. For example, an issuer can use credit enhancement technique to raise the credit quality of its securities by taking a guarantee which agrees on principal repayment and timely payment of interest (Klee & Butler, 2002).
In this way, credit enhancement improves the credit quality of the senior tranches and increase the value of bonds (Kotecha, Ryan, & Weinberger, 2010). In fact, most of the ABS receives the highest possible credit rating, AAA+ or AAA, which will be later discussed importantly in this study. As the senior/subordination structure of the securities increases to better credit rating, it lowers the cost of funding and allows borrowers to access the funds at a rate that would only be available to people who have higher credit quality in capital market (Pwc, 2011). Thus, the process of pooling and tranching provides investment opportunities and leads to more efficient use of capital (Frank J & Franco, Capital Markets: Institutions and Instruments, 2005).
Furthermore, the transfer of the credit risk lowers the capital requirement of the originator for the market risk (Revisions to the Basel securitisation framework, 2012). While the securitization
minimizes the risk weights and simplifies the management of asset and liability, it allows the originator to refinance at a lower rate because assets default risk is transferred into the capital market via SPV (Clinton, 2004). The assets default risk is independent of the originator’s credit rating or financial
9 ability, thus it does not solely depends on the originator. The structural characteristics of securitization often lead to higher yields of securities than treasury bonds of comparable maturity and date even though the securities have similar risk profiles or credit qualities (Sebastian, 1990).
2.5 The financial crisis and the role of securitization
During last three decades since 1970s, debt-financed purchases continued to increase substantially with easy credit conditions and financial deregulation in securitization market. Such financial condition and steady growth of securitization increased the number of investors (Pwc,2011).
Between 1997 and 2006 until before the financial crisis erupted, housing prices in the U.S. real estate market started to rise and marked 124% increase in property prices in that period (Pwc, 2011). With a real estate and credit bubbles, the number of financial securities which derived their value and payment from housing prices markedly increased. As of March 2007, the value of subprime lending in the U.S. housing market was $1.3 trillion in March 2007 (Schuermann, 2008).
However, since mid-2007, the real estate market crashed and property prices started to drop. As subprime borrowers did not meet their payment obligation for their mortgages, the SPVs could not provide payments of interests and coupons to investors. The massive number of investors who already made investments on mortgages with their pension funds had to undergo losses and lost their trust in securities. Also, the investment banks and sponsoring firms announced to write off bad debts which accrued more than $670 billion in SPVs (Pwc, 2011). As one of the key event related to this SPV failure, Bear Stearns collapsed after a failed bail out and subsequently sold to JP Morgan Chase in March 2008 during the subprime mortgage crisis. The global investment bank, Bear Stearns, was heavily involved in securitized debt market and having taken on large amounts of asset-backed securities which were mainly about mortgages (Saleuddin, 2015). This study defines the crisis period as lasting from March 2008 which is the date of fall of Bear Stearns. Accordingly, the credit availability declined and gave a detrimental impact on solvency of corporations or institutions. This consequence collapsed the securitization market and severely impacted on global financial markets.
The 2007 subprime mortgage crisis led experts in banking industry to identify a number of problems of all related parties associated with defects of securitization process. Among numerous issues indicated, experts have been paying attention to flaws of securitization and credit rating failures as two main issues. It was claimed that the transfer of risk to capital market was pseudo beneficial since the investors and originators are still exposed to systematic risk exposures (Pagano & Volpin, 2010). There is also a potential risk that whole market underestimates or wrongly measures the risk that determines the minimum capital requirement a bank must maintain to handle unexpected loss. Moreover, credit rating failures caused banks to pool the risky mortgages or other underlying assets into liquid securities and sold them to investors through SPVs (Brunnermeier, 2009). Several papers focus on the flaws of rating process. For instance, Pagano and Volpin (2010) criticizes the rating
10 inflation occurred between 2007 and 2008 and states that ratings only provides limited information about the risk characteristics of underlying assets.
Following the crisis, investment banks in Europe persistently encounter adverse regulatory changes. Basel III and Solvency II were implemented in 2010 to bring stricter supervisory regulations and strong measures on capital requirement. The stricter regulations on capital adequacy has restricted the risk-taking capacity of investment banks, also opportunities for risk allocation have constrained (Pwc, 2011).
Chapter 3: Methodology and Hypothesis
3.1 Methodology
This thesis performs two types of regression models in order to investigate the impact of the subprime mortgage crisis on ABS market. In this way, the estimation results from two models help to
understand how the ABS market experienced a major breakdown as a result of financial crisis and recovered.
3.1.1 OLS regression model
This subsection introduces the regression models applied in the study. The study aims to analyze impact of bond characteristics and crisis on the primary market spread, maturity, and principal from 2006 till the end of 2016. I run three OLS regression analyses with three different pricing factors as a dependent variable: Spread, Maturity and ln(principal).
In order to discuss the pricing factors that derive spreads in ABS primary market and dominant impact of each bond characteristics on the spread, the regression model is established as below:
Spreadi,t = αi + β1ln(principal)i,t + β2ln(price)i,t + β3Couponi,t + β4Maturityi,t + β5Europei,t + β6Crisisi,t +
β7Postcrisisi,t + β8(AA)i,t + β9(A)i,t + β10(BBB)i,t + β11(BB)i,t + β12(B)i,t + β13Crisis_AA + β14Crisis_A +
β15Crisis_BBB + β16Crisis_BB + β17Crisis_B + β18Crisis_ln(principal) + β19Crisis_Maturity+
β20Crisis_Europe + 𝜀 i,t
11 The histogram below shows that the distribution of the dependent variable, Spread, is close to a normal distribution which is one of the crucial conditions for linear regression analysis.
Figure 1 Histogram of primary market spread
The model above shows the variables which are included for the analyses of this study. In order to derive the Spread which is used as dependent variable, I used fixed rate securities which have information about coupon rate to calculate the yield of securities. For floating rate securities without a known index, the rate tied to Treasury bill rate which is a representative index benchmark. After the calculation of yield of securities with retrieved data, I used the U.S. treasury yield with comparable issue date and maturity as a proxy for the benchmark yield and subtracted it from yield of securities to create a Spread which describes primary market spread.
The Spread represents the risk premium on securities at issuance. Based on the information about the risk characteristics of pooled assets at the time of issuance, the risk premium measures the compensation for the risks investors undertake. This research defines the spread as the expected yield on a security at a time of issuance above the yield to maturity of a corresponding U.S. government yield for the risk-free rates. With regards to the procedure of obtaining a suitable benchmark yield which is U.S. government yield for the benchmark rates in this study, Vink and Thibeault (2008) suggest the following. First, the yields for the benchmark rates should be provided in the same currency. Those should also offer a comparable issue date and maturity. With the
corresponding U.S. government yield as a corresponding benchmark, I can calculate the difference between two yields to create the spread.
The study also aims to empirically investigate the characteristics of securities that affect the Maturity and ln(principal) of ABS issues with the help of the following models:
0 200 400 600
Fr
eq
ue
nc
y
-10 0 10 20 30 spread12
Maturityi,t = αi + β1ln(principal)i,t + β2ln(price)i,t + β3Couponi,t + β4Spreadi,t + β5Europei,t + β6Crisisi,t +
β7Postcrisisi,t + β8(AA)i,t + β9(A)i,t + β10(BBB)i,t + β11(BB)i,t + β12(B)i,t + β13Crisis*(AA) + β14Crisis*(A) +
β15Crisis*(BBB) + β16Crisis*(BB) + β17Crisis*(B) + β18Crisis*ln(principal) + β19Crisis*Spread+
β20Crisis*Europe + 𝜀 i,t
ln(principal)i,t = αi + β1Maturityi,t + β2ln(price)i,t + β3Couponi,t + β4Spreadi,t + β5Europei,t + β6Crisisi,t +
β7Postcrisisi,t + β8(AA)i,t + β9(A)i,t + β10(BBB)i,t + β11(BB)i,t + β12(B)i,t + β13Crisis*(AA) + β14Crisis*(A) +
β15Crisis*(BBB) + β16Crisis*(BB) + β17Crisis*(B) + β18Crisis*Maturity + β19Crisis*Spread+
β20Crisis*Europe + 𝜀 i,t
The variable ln(principal) is determined as the natural logarithm of principal which is the issuance volume at the time of issuance. The natural logarithm is also used to transform price to variable
ln(price) to measure a change in Spread associated with a 1% change in price. The credit rating of
each tranche is collected at the time of issuance for purpose of comparison. In order to examine the effect of credit rating on the spread, the model uses a dummy variable for all rating classes. In data sample, I leave the highest credit rating (AAA) out of the equation as a reference category.
Additionally, the interaction terms are added in the model to test whether the relationship between characteristics of issuance and the spread is different during the crisis than pre or post-crisis period. Adding interaction terms expands understanding of the relationships among the variables by assessing whether the relationship between bond characteristics and the spread is more dramatic during crisis than during other times. For instance, the coefficients of interaction terms test whether the sensitivity of spreads to certain credit rating, country of issuance or principal has been stable over time. Thus, multiplications of the dummy variable Crisis with composite credit rating, country of issuance, principal and maturity expand the understanding of the relationships among variables.
3.1.2 Ordered probit model
What data is used, what sample restrictions are made, what statistical methods are employed
Another regression model, the ordered probit model, is used to analyze the factors to credit quality of ABS issues.
The probit regression is a nonlinear regression designed for ordinal dependent variable. Whereas the previous subsection analyzes linear regression using OLS regression model, this method assumes non-continuous dependent variable. The regression models the predictors of credit quality of the securities and yields the predicted values between 0 and 1. However, these categories may not be equally spaced between the values. For instance, the study assumes the difference between
categories low and medium is much bigger than the difference between categories medium and high. In Statistics, variable described this way is known as ordinal variable or ordered outcome. In order to examine the thresholds which classify the space between those categories, the model uses the
13 standard normal cumulative probability distribution functions (c.d.f.) which denotes as 𝜙 (Stock & Watson, 2012).
The ordered probit model is defined by:
Pr(X) = 𝜙(β0 + β1Xi + 𝜀 i)
The ordered probit regression can be modelled as below:
Pr(X) = 𝜙(β1Spreadi,t + β2ln(principal)i,t + β3Couponi,t + β4Maturityi,t + β5Europei,t + β6Crisisi,t +
β7Postcrisisi,t + β8Spread_ln(principal))
Form the ordered probit model I set above, Y takes on the categorical and non-numerical values high, medium and low in a dataset. The central logic behind the ordered outcome is that there is latent continuous index underlying the three categories of credit qualities denoted as high, medium and low. A latent continuous variable, denoted as Y, is a linear combination of various predictors, x, and an error term that has a standard normal distribution. As previously mentioned, latent variable Y is unobserved at this stage and we only know when it cuts the thresholds.
A latent continuous variable model is given as:
Yi = β1Spreadi + β2ln(principal)i + β3Couponi + β4Maturityi + β5Countryi + β6Crisisi + β7c_Maturityi +
β8s_lnprincipali + 𝜀 i
Where Yi is the function of observed and non-observed variables, βi is the coefficient to be estimated,
Xi is observed explanatory variables measuring the contributions of bond characteristics i at time t,
and 𝜀 I is the idiosyncratic term following standard normal distribution.
Hence, the dependent variable considered in this thesis is the credit quality of ABS classified into three categorical levels: low credit quality (y=1); medium credit quality (y=2); and high credit quality (y=3). The probabilities for three categories are derived as:
Pr (Yi = 1) = 1 - 𝜙 (Xi𝛽 – u1)
Pr (Yi = 2) = 𝜙 (Xi𝛽 – u1) - 𝜙 (Xi𝛽 – u2)
Pr (Yi = 3) = 𝜙 (Xi𝛽 – u2)
Where u = {u1, u2, u3} are the threshold values for three credit qualities that define Yi.
In a linear regression model, βi is interpreted as the amount in the predicted value in Yi for each one
unit increase in Xi. However, due to the lack of interpretation of coefficients βi in ordered probit model,
I derive marginal effects which show the change in probability when the independent variable increases by one unit. The marginal effects in probabilities can be derived as follows:
14 𝜕Pr (Yi = 2)/ 𝜕X = 𝛽𝜙 (Xi𝛽 – u1) - 𝛽𝜙 (Xi𝛽 – u2)
𝜕Pr (Yi = 3)/ 𝜕Xi = 𝛽𝜙 (Xi𝛽 – u2)
Three categories indicate specific credit quality which is involved either in investment grade or in speculative or non-investment grade. The table below describes how the S&P’s classifies their credit quality ratings for securities.
Table 1 Standard &Poor’s Credit rating Scale
Investment grade
High
AAA Extremely strong capacity to meet financial commitments. Highest
rating
AA Very strong capacity to meet financial commitments
Mediu m
A
Strong capacity to meet financial commitments, but somewhat susceptible to adverse economic conditions and changes in circumstances
BBB Adequate capacity to meet financial commitments, but more subject to
adverse economic conditions
BBB- Considered lowest investment-grade by market participants
Speculative
grade Low
BB+ Considered highest speculative-grade by market participants
BB Less vulnerable in the near-term but faces major ongoing uncertainties
to adverse business, financial and economic conditions
B
More vulnerable to adverse business, financial and economic conditions but currently has the capacity to meet financial commitments
CCC Currently vulnerable and dependent on favorable business, financial
and economic conditions to meet financial commitments
CC Highly vulnerable; default has not yet occurred, but is expected to be a virtual certainty
C Currently highly vulnerable to non-payment, and ultimate recovery is
expected to be lower than that of higher rated obligations
D
Payment default on a financial commitment or breach of an imputed promise; also used when a bankruptcy petition has been filed or similar action taken
Ratings from ‘AA’ to ‘CCC’ may be modified by the addition of a plus (+) or minus (-) sign to show relative standing within the major rating categories.
Source: Standard & Poor's Guide to Credit Rating Essentials; p. 9
In this model, the estimation results in subsequent chapter of this thesis will investigate how changes in the predictors, x, translate into likelihood and probability of observing a particular ordinal response variable.
15 3.2 Hypothesis development
Before using statistical methods to discuss the characteristics driving the ABS and their expected impacts on the primary market spread, this section constructs hypotheses and explains the rationale behind those. Hypotheses explain why and how the explanatory variables might affect the spread which is the dependent variable in OLS regression and the qualities of securities in ordered probit models. These hypotheses are based on a ceteris paribus condition.
H1: ABS issues in times of crisis should have a lower spread than those during pre or post-crisis period.
From the first OLS regression model which uses the spread as a dependent variable, this study proposes that the subprime mortgage crisis negatively influenced the primary market spread. This hypothesis can be clearly supported based on the theoretical background which covers the
securitization history and the role of ABS in times of crisis. While the real estate market crashed and property prices greatly dropped as per mid-2007, risky mortgages were sold to investors. On a basis of this historical information, negative relationship between the crisis and spread can be clearly predicted.
H2: ABS with a higher spread should have a longer maturity than ABS with a lower spread.
Ayote and Goan (2011) suggest that ABS with a longer maturity should increase spreads than ABS with a shorter maturity due to higher risk. Since the spread represents the risk premium which is the compensation for the risks investors undertake, positive relationship between maturity and offered yield on securities constructs the following hypothesis.
H3: Principal amount of ABS issues is negatively related with spread
According to the study conducted by Ayote and Goan (2011), it also shows that ABS with a larger volume which is the principal should reduce spreads because a large issuance size leads to more liquidity in the markets.
H4: The crisis years have positive effect on the probability to rate ABS issues for higher credit qualities.
Pagano and Volpin (2010) identified the credit rating inflation as one of the main problems related to rating agencies who are involved in securitization process. It eventually caused wrong measure of risk and credit profile of borrowers. The paper also points out the fact that very highly rated securities have performed very poorly since mid-2007. From 2007 till end of 2008, it shows that the value of AAA rated MBS fell by 70% (Pagano & Volpin, 2010). However, following the crisis, new regulation was designed to mitigate the problems related with credit rating inflation and measure of risk characteristics.
16
Chapter 4: Data
4.1 Data samples
The dataset was retrieved primarily from Thomson One DataStream which provides information on ABS issued through 2006 to 2016. In this study, crisis period is defined as lasting from March 2008 to early 2009 on a basis of the date of fall of Bear Stearns. As shown in Table 2 which presents the total issuance per year, the dataset provides a uniform number of issuance over year. From the database, I obtained security’s initial coupon rate, rating, principal, offer price, country of issuance and other bond characteristics. Table 5 contains summary statistics for each characteristic of ABS issues.
Table 2 Total issuance over time
Year Total issuance per year
2006 170 2007 178 2008 184 2009 178 2010 150 2011 165 2012 184 2013 159 2014 173 2015 56 2016 40 Total 1,637
The data covers a various types of ABS with a heavy reliance on collateral assets based on issuer industry of financials. Table 3 below provides the sample split between different types of collateral assets. As it shows, the majority of ABS in the dataset is backed by credit card receivables or assets from financial industry. Meanwhile, other types of securities are backed by auto loans, mortgage loans, other real estate related loans and other smaller types of collateral assets. Furthermore, the underlying assets are located either in the United States or in Europe. Since the ABS issues are denominated in dollars, there is no need to consider the currency as a variable in the model. The proportion of floaters in the dataset is nearly half. However, the index does not provide estimates for coupon and maturity as the floaters tied to an interest rate benchmark such as Treasury bill rates or LIBOR.
17
Table 3 Data sample split between different type of collateral assets
Types of collateral Number of issuance backed by the assets
Alternative Financial Investments 3
Asset Management 15
Automobiles & Components 24
Banks 3
Brokerage 15
Credit Institutions 53
Discount and Department Store Retailing 2
Government Sponsored Enterprises 1
Insurance 14
Machinery 9
Other Financials 1,495 Other Real Estate 1
Transportation & Infrastructure 2
Total 1,637
As discussed in theoretical background, the subprime mortgage crisis and following financial market prompted the review of the role of all parties related to securitization and flaws of the process. Much attention has been paid to failures or CRAs which led to credit rating inflation until financial crisis erupted. Pagano and Volpin (2010) report that there has been a highest proportion for AAA rated securities up until crisis period even though highly rated debt products performed very poorly. After 2008, the graph shows there have been enormous and severe rating downgrades. Following the crisis, as tightened supervisory regulation and strict guidelines on credit rating are implemented, the average of rating falls over time.
18 In the dataset, the majority of issues were rated by S&P’s but ratings of Moody’s was frequently unavailable. Hence, I used only observations rated by S&P’s were used for this research and removed some of observations which were rated very differently by two agencies.
From another graph below, it presents that the average maturity has been decreased over time. In the dataset, the trend of the graph implicitly tells that the premium has been decreased as a maturity length got shortened.
Figure 3 Bar graph for average maturity of securities over time
4.2. Variables
Table 4 Description of variables
Variable Description
Spread Yield spread to benchmark
ln(principal) Log(Principal in US$ million)
ln(price) Log(Price in US$ millions)
Coupon Coupon rate in %
Maturity Years to maturity
Europe 1 for Europe and 0 for United States
Crisis 1 for issue date between 03/2008 and 03/2009 and 0 otherwise
Postcrisis 1 for issue date after 03/2009 and 0 for otherwise
AA 1 for AA+/AA/AA- and 0 otherwise
A 1 for A+/A/A- and 0 otherwise
BBB 1 for BBB+/BBB/BBB- and 0 otherwise
BB 1 for BB+/BB/BB- and 0 otherwise
B 1 for rating below BB- and 0 otherwise
Spread_ln(principal) spread*lnpricncipal
19
Table 5 Summary statistics of data for all sample
Variable Obs Mean Std. Dev. Min Max
Spread 1,637 1.968 3.421 4.514 27.781 ln(principal) 1,637 6.505 0.889 1.609 9.170 ln(price) 1,637 4.605 0.005 4.530 4.738 Coupon 1,637 2.420 2.072 0.002 10 Maturity 1,637 4.057 2.420 0.586 11.417 Europe 1,637 0.001 0.037 0 1 Crisis 1,637 0.002 0.044 0 1 Postcrisis 1,637 0.019 0.136 0 1 AA 1,637 0.007 0.084 0 1 A 1,637 0.002 0.048 0 1 BBB 1,637 0.004 0.059 0 1 BB 1,637 0.001 0.035 0 1 B 1,637 0 0.011 0 1 Crisis_AA 1,637 0 0.017 0 1 Crisis_A 1,637 0 0.012 0 1 Crisis_BBB 1,637 0 0.017 0 1 Crisis_BB 1,637 0 0 0 1 Crisis_B 1,637 0 0 0 1 Crisis_ln(principal) 1,637 0.468 1.704 0 8.279 Crisis_Maturity 1,637 0.335 1.196 0 4.661 Crisis_Europe 1,637 0.001 0.023 0 1 Crisis_Spread 1,637 0.132 1.093 -1.103 27.781
The average spread of ABS securities covered in a dataset is about 2 percentage points. While the principal levels average $ 6.5 million, the summary reports an average percentage point of 2.4 for coupon rates. It is remarkable that the mean maturity of the securities is relatively short which is 4 years since maturity has been decreased drastically over time and the majority of collateral are financial assets in a sample. I find that the variation in spread is relatively high whereas that of all control variables are is low.
20
Table 6 Data sample split across credit rating quality
Credit quality Frequency Percent Cumulative
High 1,196 73.06% 73.06% Medium 7 0.43% 7.49% Low 434 26.51% 100% 1,637 100%
The credit ratings given in the sample is a composite rating of S&P’s. About 73% (equivalent to 1,196), 7.5% (equivalent to 7) and 26.5% (equivalent to 434) of ABS securities were rated high, medium and low credit quality respectively. In other words, the greater part of index securities in data sample are rated AAA by Standard & Poor’s (S&P) at issuance with a fairly even distribution of lower ratings. However, speculative grade securities (those rated BB or below) comprises only a small minority of the sample.
Chapter 5: Empirical results
This thesis performs two types of regression models in order to investigate the impact of the subprime mortgage crisis on ABS market. In this way, the estimation results from two models help to
understand how the ABS market experienced a major breakdown as a result of financial crisis and recovered.
5.1 OLS regression analysis
This subsection examines the impact of crisis and determinants of spread, maturity and principal with the use of three regression models. I first run OLS regression for the total dataset with Spread as a dependent variable. After that, I run two other regressions separately with other dependent variables which are Maturity and ln(principal) to test the hypotheses and make the comparison of the results. They are analyzed in this way for two reasons. The first is to investigate whether the crisis event provides direct evidence to influence the spread of ABS issues by adding interaction terms in each regression analysis. The second reason is to investigate other key characteristics of securities that appear as significant measures for pricing factors which are mainly spreads, maturity, and principal.
20
Table 7 OLS regression analyses from 2006 to 2016
The table presents the regression output for three different regression analyses. Regressions (1), (5) and (8) are the regression result for a model that only includes dummy variables, so examines magnitude of credit ratings, country of issuance and issue date. Regressions (2),(6) and (9) repeat the regression but includes control variables. In regressions (3), (4), (7) and (10) interaction terms are added. Regressions (1) through (4) take dependent variable Spread, while (5) though (7) take Maturity and (8) through (10) take ln(principal) as a dependent variable. Superscripts *, ** and *** denote significance at 0.01, 0.05 and 0.1 respectively.
Dep.Variable: Spread Maturity ln(principal)
Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Constant -1.036*** 38.517 50.898 51.060 5.832*** 21.206 31.378 6.90*** 75.844*** 36.049* (0.168) (58.909) (59.069) (63.346) (0.121) (42.025) (42.342) (0.048) (19.015) (20.219) ln(principal) 0.835*** 0.859*** 0.872*** 0.377*** 0.427*** (0.074) (0.074) (0.079) (0.054) (0.058) ln(price) -8.888 -11.607 -11.660 -4.478 -6.774 -14.887 -6.253 (12.779) (12.815) (13.748) (9.117) (9.190) (4.129) (4.390) Coupon -0.116** -0.120** -0.120** 0.481*** 0.491*** -0.158*** -0.147*** (0.056) (0.056) (0.056) (0.038) (0.038) (0.018) (0.018) Maturity -0.656*** -0.655*** -0.656*** 0.078*** 0.075*** (0.031) (0.031) (0.031) (0.011) (0.011) Spread -0.334*** -0.350*** 0.088*** 0.086*** (0.016) (0.016) (0.008) (0.008) Europe -1.688*** 0.160 0.286 0.277 1.741*** 1.445*** 1.375*** -0.814*** -0.761*** -1.005*** (0.346) (0.298) (0.310) (0.358) (0.846) (0.210) (0.252) (0.10) (0.095) (0.111) Crisis 2.269*** 1.201*** 0.854*** 23.884 -1.393*** 0.161 3.400*** -0.033 -0.369*** 218.978*** (0.309) (0.271) (0.327) (189.134) (0.223) (0.194) (1.175) (0.089) (0.088) (56.936) Postcrisis 2.201*** 0.504** 0.516** 0.516** -2.168*** 0.429** 0.512*** -0.190*** -0.801*** -0.785*** (0.177) (0.246) (0.247) (0.248) (0.127) (0.175) (0.177) (0.051) (0.077) (0.077) AA 3.561*** 2.975*** 2.892*** 2.893*** -1.187*** 0.072 0.113 -0.219*** -0.425*** -0.366*** (0.174) (0.149) (0.152) (0.153) (0.125) (0.119) (0.121) (0.050) (0.053) (0.053) A 0.827*** 1.879*** 1.871*** 1.877*** 0.829*** 0.647*** 0.660*** -0.431*** -0.367*** -0.386*** (0.264) (0.229) (0.235) (0.235) (0.190) (0.166) (0.170) (0.076) (0.075) (0.076) BBB 1.483*** 2.504*** 2.491*** 2.501*** 0.395** 0.423*** 0.515*** -0.701*** -0.623*** -0.584*** (0.222) (0.202) (0.206) (0.207) (0.159) (0.151) (0.153) (0.064) (0.067) (0.067) BB 2.564*** 2.848*** 2.852*** 2.860*** -0.462* -0.188 -0.227 -0.499*** -0.460*** -0.455*** (0.357) (0.311) (0.312) (0.313) (0.257) (0.228) (0.228) (0.103) (0.103) (0.102)
21 B 3.931*** 3.834*** 3.825*** 3.833*** -1.342* -0.289 -0.208 -0.777** -0.834*** -0.828*** (1.087) (0.906) (0.905) (0.906) (0.782) (0.650) (0.647) (0.314) (0.295) (0.290) Crisis_AA 1.771*** 1.675** -0.749 -0.092*** (0.651) (0.705) (0.566) (0.234) Crisis_A 0.121 0.081 -0.436 0.338 (0.888) (0.958) (0.672) (0.306) Crisis_BBB 0.316 0.159 -1.511*** -0.523** (0.658) (0.723) (0.521) (0.212) Crisis_BB Crisis_B Crisis_ln(princip al) -0.118 -0.505*** (0.235) (0.167) Crisis_Maturity -4.826 -47.602*** (40.945) (12.362) Crisis_Spread 0.190*** 0.018 (0.052) (0.022) Crisis_Europe -0.006 0.429 0.459** (0.681) (0.479) (0.210) Observations 1,637 1,637 1,637 1,637 1,637 1,637 1,637 1,637 1,637 1,637 R-squared 0.305 0.523 0.525 0.525 0.515 0.516 0.522 0.257 0.273 0.283 F- statistics 89.23 148.42 119.56 99.48 79.72 143.76 98.04 33.27 46.78 35.38
22 The table reports F-statistics on whether the coefficients are jointly different from zero and R-squared at the bottom. R-squared values increased in regression (4), (7) and (10) as the model includes more variables which can have explanatory power of a regression.
From the estimation outputs above, the regression model for spread as a dependent variable can be translated as follows:
Spreadi,t = 51.06 + 0.872*ln(principal)i,t – 11.66*ln(price)i,t – 0.12*Couponi,t – 0.656*Maturityi,t +
0.277*Europei,t + 23.884*Crisisi,t + 0.516*Postcrisisi,t + 2.893*(AA)i,t + 1.877*(A)i,t + 2.501*(BBB)i,t +
2.86*(BB)i,t + 3.833*(B)i,t + 1.675*Crisis_AA + 0.081*Crisis_A + 0.159*Crisis_BBB –
0.118*Crisis_ln(principal) – 4.826*Crisis_Maturity – 0.006*Crisis_Europe
It presents that majority of variables in regression (4) are statistically significant at 1% level. The estimation results show that the variable ln(principal)i is positively related with the spreads. Controlling
for the natural logarithm of principal provides a coefficient of 0.872 at 1% significance level. In contrast, variable lnpricei has insignificant negative relationship which decreases spread 11.66 points
with additional 1% of price increase. Another remarkable result is that if the variable Europe is equal to one, the country of issuance is European country, it has insignificant positive relationship which increases spread 0.277 points.
As an another point, the output of regression shows that effects are different across the five credit rating groups at 1% significance level. For instance, spread increases significantly with a higher rating and lower rating. The increase of the spread is relatively large which is 2.89 and 3.83 in high and low rating classes respectively. However, the effects on spread are relatively less significant with medium credit rating classes such as A and BBB. In the meantime, the result demonstrates that effects are positively significant in all credit rating classes.
As for the analysis of first hypothesis, ABS issuance during crisis has a higher spread than issuance during other times. Thus, the finding contradicts the hypothesis which predicts that ABS issuance in times of crisis has a lower spread than those issued during other times. However, the result also shows the positive relationship between post-crisis issuance and spread at 5% significance level.
For further analysis, the interaction terms are added in the regression to assess whether the effects of each bond characteristics on spread are differ over time. Particularly, the coefficients of interaction terms provide the information about whether the relationship between crisis period and bond characteristics is stable or different during the crisis. In regression (4), some outcomes are almost identical to the results of regression (3). The significant outcome of the interaction effects is the coefficient of Crisis_AA which is 1.675. For post-crisis, AA rated issues have spread that are 2.893 higher than those of lowly rated issues after crisis. For AA rated issues during crisis, these differences become larger: 4.568 (2.893+1.675). Thus, the interaction term demonstrates this stronger effect of crisis for highly rated ABS issuance.
23 The second regression model is run with maturity as a dependent variable as below:
Maturityi,t = 31.378+ 0.427*ln(principal)i,t – 6.774*ln(price)i,t + 0.491*Couponi,t – 0.35*Spreadi,t +
1.375*Europei,t + 3.4*Crisisi,t + 0.512*Postcrisisi,t + 0.113*(AA)i,t + 0.66*(A)i,t + 0.515*(BBB)i,t – 0.227*
(BB)i,t – 0.208*(B)i,t – 0.749*Crisis*(AA) – 0.436*Crisis*(A) – 1.511*Crisis*(BBB) –
0.505*Crisis*ln(principal) + 0.19*Crisis*Spread+ 0.429*Crisis*Europe
The finding of regression (7) which shows that maturity would decrease with additional point of spread by 0.35 basis points contradicts the second hypothesis. Thus, the result does not explain the
expectation that a higher spread leads to a longer maturity. The dummy variable Crisis also emphasizes that maturity increases by significant amount, however its impact on length of maturity becomes greatly lower in post-crisis period.
The third model is translated as follows:
ln(principal)i,t = 36.049 – 6.253*ln(price)i,t – 0.147*Couponi,t + 0.075*Maturityi,t + 0.086*Spreadi,t –
1.005*Europei,t + 218.978*Crisisi,t – 0.785*Postcrisisi,t – 0.366*(AA)i,t – 0.386*(A)i,t – 0.584*(BBB)i,t –
0.455*(BB)i,t – 0.828*(B)i,t – 0.092*Crisis*(AA) + 0.338*Crisis*(A) – 0.523*Crisis*(BBB) –
47.602*Crisis*Maturity + 0.018*Crisis*Spread+ 0.459*Crisis*Europe
As for the third hypothesis, it shows a significant positive relationship between spread and principal amount which is the issuance size. As the evidence is consistent with the hypothesis, interpretation for this estimation would be that a larger volume of ABS can result in more liquidity in ABS market and produces a higher spread. It is also remarkable that the effect of crisis has less strong effect in
principal. From the interaction parameters, it indicates whether the effects of crisis on pricing vary over time. According to the interaction between crisis and every rating class, it shows that the negative effect of crisis is found to be stronger in lower ratings and weaker in higher ratings at 1% significance.
For all three regression models analyzed above, I find the highest R-squared which indicates a higher statistical explanation power. Also, the intercept is significant at 1% level which leads second and last regression to include more explanatory variables. After adding variables in the last regression for models for three different pricing factors as dependent variables, all characteristics of ABS issues are not significantly different from zero so that fall into rejection region of p-value.
In total, the first (H1) and second hypothesis (H2) cannot be proven true since the findings show the
negative relationship between variables. However, the failure of rejection to third hypothesis (H3)
means that a larger principal should have a higher spread since it generates more liquidity in ABS market.
24 5.2 Ordered probit regression analysis
Table 8 Ordered probit model estimation results
After running ordered probit regression with data of all samples for the whole period from 2006 to 2016, Table presents the output of Stata as shown below. The dependent variable has three categories, taking on the ordinal value “low” if the credit rating is equal or below BB, “medium” if the rating is between AAA and BBB and “high” otherwise. The output shows the regression performed on the total sample for. The z-statistics are based on the log likelihood with robust standard errors. Superscripts *, ** and *** denote significance at 0.1, 0.05 and 0.1 respectively.
Marginal effects
Variable Coefficient z-Stat. Low Medium High
Spread 0.368** 2.29 -0.114 0.001 0.112 (0.161) ln(principal) -0.241*** -1.71 0.074 -0.001 -0.073 (0.047) Coupon 0.472*** 13.09 -0.146 0.002 0.144 (0.036) Maturity 0.074*** 3.46 -0.023 0.000 0.022 (0.021) Europe 0.971*** 5.15 -0.359 0.002 0.357 (0.189) Crisis 0.073 0.39 -0.023 0.000 0.023 (0.186) Postcrisis 1.438*** 9.16 -0.340 0.005 0.335 (0.157) Spread_ln(principal) -0.092*** -4.25 0.012 0.000 -0.012 (0.022) cut 1 1.892 cut 2 1.910 Observations 1,637 Log Likelihood -746.689 LR test X^2(8) = 486.13 Pseudo R-squared 0.246
From the estimation results in the table, the study can interpret the coefficients which show whether the dependent variable Y increases or decreases with the regressors. The result also enables to examine what characteristics of ABS securities influence the credit quality rated by S&P’s. As shown, key covariates are spread, principal, coupon rate, maturity, country of issuance and issuance date.
From the results above, the model can be translated as below:
Pr(X)
= ϕ(0.368*Spreadi,t – 0.241*ln(principal)i,t + 0.472*Couponi,t + 0.074*Maturityi,t + 0.971*Europei,t25 To interpret the equation, it is crucial to note that a positive value of a coefficient which corresponds with an increase in its variable X (certain bond characteristic) would increase the probability of the higher credit quality level. The reverse is the case for negative value of a parameter; it would
decrease the probability of the highest credit rating class. Hence, the estimated outputs of coefficients which are different from zero imply that increases (decreases) in explanatory variables tend to
increase (decrease) the estimated propensity score, z, for the credit quality. Given that, the variables which show the positive sign are: spread, coupon, maturity, Europe, crisis and post-crisis. Thus, ABS issues which issued during crisis and post-crisis with higher spread, higher coupon rate, longer maturity and Europe as a country of issuance are more likely to be evaluated as highly rated issues. However, by comparing a parameter of postcrisis (β= 1.438) and that of crisis (β= 0.073), it concludes that post-crisis period has a larger impact on probability of higher credit quality of issues.
Three columns in the right hand side of the table present computation of marginal effects for the model. The additional unit of spread increases chance of rating the ABS securities highly by 11.2 percentage points and decrease the probability of lower rating by 11.4%. The chance of rating the securities as high credit quality would also increase by 14.4 percentage points with additional percentage of coupon rate. Thirdly, it shows the result of increase by 2.2% for the change for high credit grade with additional year to maturity. ABS securities issued in post-crisis period have 34 percentage points less likely to be rated as low credit quality (below BB) and 33.5 percentage points less likely to get high credit quality which consists of AAA and AA. Therefore, the result of analysis rejects the hypothesis by providing the evidence that crisis years have less positive effects for higher credit quality. Furthermore, ABS issued in European countries is 35.7 more likely to get rated as a high credit quality while it is 35.9 less likely to get low credit quality rated by CRAs.
The regression results estimate not only coefficients of repressors but also the thresholds which I denote as u1 and u2. Thresholds classify the line into a series of regions corresponding to the ordinal categories of credit qualities from low to high (27). Cut 1 indicates the threshold value which was previously unknown and now draws the clear line between high and medium credit quality rating. Similarly, cut 2 determines the real line that partitions the region for latent index between medium and low.
26
Chapter 6: Conclusion
This paper empirically investigates the ABS market during and after the subprime mortgage crisis. The research concentrates on ABS issues between 2006 and 2016 and examined 1,637 issues in a dataset all of which offered the complete information to run a full analysis. The analysis aims for describing the pricing and other characteristics of the issues and study the role of crisis which influenced the spread, maturity and pricing in ABS market over time.
The research examines the influence of the subprime mortgage crisis which triggered financial crisis on European and U.S. ABS market. Also, it clarifies the key characteristics of issues which predict the credit quality scale from low to high. The study adopts OLS regression model in order to test the three hypotheses with regard to changes in ABS pricing and impact of crisis. Through performing OLS regression models with total sample and with three different pricing factors as dependent variables, it has produced the results of testing three hypotheses and rationales behind them. The findings suggest that crisis has significant effect on pricing and quality of ABS issuance in securitization market. However, the regression results did not provide the evidence to support the first and second hypotheses. It demonstrates that ABS issues in times of crisis have a lower spread and higher spread does not necessarily lead to a longer maturity. On the other hand, the output for the third model was consistent with the hypothesis by proving that larger principal results in more liquidity in ABS market and produces a higher spread.
Furthermore, the order probit model was used to examine the influence of a number of bond characteristics on the quality of issues rated by S&P’s. The estimation results suggest that some of factors such as country of issuance, issue date and coupon rate appear to be the main contributors for the high credit quality. However, the results did not show the evidence to support the fourth hypothesis since the magnitude of effect of post-crisis was much larger than that of crisis.
This study concludes that there is a change in pricing and credit quality of ABS issues due to the impact of subprime mortgage crisis in 2007. Based on this analysis, I suggest that the study can be improved by considering the default probability of originator as another variable. Although theoretical background illustrates the ABS as a security which achieves bankruptcy remoteness, it is interesting to investigate to what extent investors can value the bankruptcy remoteness and default probability of the originator influences the ABS pricing. In addition to this, it is also necessary to reflect the different types of risk such as interest rate, systematic and default risk to improve the accuracy of the measure for risk exposures.
27
References
Adams, P. (2005). An Introduction to the European Securitization Market. The Journal of Finance, 33-39.
Agarwal, S., Barret, J., Cun, C., & De Nardi, M. (2010). The Asset-Backed Securities Markets, the Crisis, and TALF. Economic Perspectives, 34(4), 101-115.
Almazan, A., & Martin-Oliver, A. (2015). Securitzation and Banks' Capita Structure. Review of
Corporate Finance Studies, 4(2), 206-238.
Ayotte, K., & Gaon, S. (2011). Asset-Backed Securities: Costs and Benefits of “Bankruptcy Remoteness”. The Review of Financial Studies, 24(4), 1299-1335.
Brennan, M., Hein, J., & Poon, S.‐H. (2009). Tranching and Rating. European Financial Management, 15(5), pp.891-922.
Brunnermeier, M. (2009). Deciphering the Liquidity and Credit Crunch 2007-2008. Journal of
Economic Perspectives, 23(1), 77-100.
Cerrato, M. (2010). The rise and fall of the ABS market. University of Glasgo.
Chris, K., & Joseph, C. (2016). U.S. SECURITIZATION YEAR IN REVIEW . Sifma.
Clinton, N. (2004). Good news is no news? The impact of credit rating changes on the pricing of asset-backed securities. Board of Governors of the Federal Reserve System.
Coval, J., Jurek, J., & Stafford, E. (2012). The Economics of Structured Finance. Journal of Economic
Perspectives, 23(1), 3-25.
Efing, M. (2016). Arbitraging the Basel securitization framework evidence from German ABS investment. ESRB Working Paper Series, No 22.
Fabozzi, F., & Choudhry, M. (2004). The handbook of European structured financial products. Hoboken [etc.] : Wiley cop: Hoboken.
Frank J, F., & Dennis, V. (2012). Looking Beyond Credit Ratings: Factors Investors Consider In Pricing. European Financial Management, Vol. 18, No. 4, 2012, 515–542.
Frank J, F., & Franco, M. (2005). Capital Markets: Institutions and Instruments.
Jobst, A. (2008). What Is Securitization? . Finance & Development, 44-50.
Jönsson, B., Schoutens, W., & Eurandom. (2009). Asset backed securities : Risks, ratings and quantitative modelling. NARCIS (National Academic Research and Collaborations Information
System).
Klee, K. N., & Butler, B. C. (2002). Asset-Backed Securitization, Special Purpose Vehicles and Other Securitization Issues. Uniform Commercial Code Law Journal, 35(2), 23-68.
Kotecha, M., Ryan, S., & Weinberger, R. (2010). The Future of Structured Finance Ratings After the Financial Crisis. Journal of Structured Finance, 15(4), 67-74.
Miguel, S., Bradley, J., Peter, L., & Johannes, B. (2013). Securitization; Lessons Learned and the
28 Pagano, M., & Volpin, P. (2010). Credit ratings failures and policy options. Economic Policy, 5(62),
401-431.
Pwc. (2011). The next chapter Creating an Understanding of Special Purpose Vehicles. United Kingdom: PricewaterhouseCooper.
Revisions to the Basel securitisation framework. (2012, December). Retrieved from Basel Committee
on Banking Supervision: http://www.bis.org/publ/bcbs236.pdf.
Saleuddin, R. (2015). Regulating securitized products : A post crisis guide. Houndmills, Basingstoke, Hampshire ; New York, NY: Palgrave Macmillan.
Schuermann, T. (2008). Understanding the securitization of subprime mortgage credit. Federal
Reserve Bank of New York.
Sebastian, C. W. (1990). The Yield of Asset-Backed Securities Afterthe Financial Crisis: An Empirical Approach. Technische Universitat Damstadt, 2-53.
Solomon, D. (2012). The Rise of a Giant: Securitization and the Global Financial Crisis. American
Business Law Journal, 49(4), 859-890.
Standard & Poor's. (2008). The Basics Of Credit Enhancement. New York: Standard & Poor's.
Stock, J., & Watson, M. (2012). Introduction to Econometrics. Pearson.
Taylor D, N., & Shane M, S. (2013). The impact of securitization on the expansion of subprime credit.
Journal of Financial Economics, 454–476.
Turnbull, S. (2008). The Credit Crisis of 07. mimeo, University of Houston, Houston, Texas.
Vink, D., & Thibeault, A. (2008). ABS, MBS and CDO pricing comparisons : An empirical analysis. .
