Master Thesis
Reinier Zonneveld
The Effect of Regulations on the Risk and
Return of Asset-Backed Securities
Abstract
This paper investigates the effect of ABS regulation on return and systemic risk, using panel data for the period of 2000-2015. By using a quantitative approach, the magnitude of these effects can be estimated. The results show that there are indeed significant effects from regulation on return and systemic risk for most of the individual regulations in the sample. Credit rating controls and information disclosure both increase returns and decrease risk. Market interventions and capital requirements, on average, have an adverse effect— the former on risk and the latter on both risk and return. As a result, multiple policy implications appear. The relatively new approach in the context of the ABS market also provides a basis for future research in this area.
MSc Business Economics, Finance and Real Estate Finance Track Thesis supervisor: Dr. Giambona Date: October 2015
Table of Contents
Statement of Originality ... 4
1 Introduction ... 5
2 Literature Review ... 6
2.1 A Short Overview of ABS and the Financial Crisis ... 6
2.2 Risk and Return in the ABS Market ... 7
2.3 Common Sources of Risk of ABS and Their Effect on Returns ... 8
2.3.1 Credit ratings ... 8
2.3.2 Liquidity risk ... 9
2.3.3 Capital structure ... 9
2.3.4 Information asymmetry and intransparency ... 10
2.4 Regulatory landscape of ABS markets ... 10
2.4.1 Type I: Laws concerning credit ratings and their corresponding rating agencies 11 2.4.2 Type II: Government purchasing ABS in order to stimulate the market ... 11
2.4.3 Type III: Additional capital requirements ... 11
2.4.4 Type IV: Information disclosure ... 11
2.4.5 Type V: Initial specifications on the structure of ABS ... 11
2.5 Regulations in Practice: Findings Contradicting ABS Market Theory ... 12
2.5.1 Type I: Laws concerning credit ratings and their corresponding rating agencies 12 2.5.2 Type II: Government purchasing ABS in order to stimulate the market ... 12
2.5.3 Type III: Additional capital requirements ... 13
2.5.4 Type IV: Information disclosure ... 13
2.5.5 Type V: Initial specifications on the structure of ABS ... 13
2.6 Recapitulation ... 13
3 Methodology ... 14
3.1 Return and Individual Regulations ... 14
3.2 Volatility of the Returns and Individual Regulations ... 15
3.3 Return and Regulation Types ... 16
3.4 Volatility of the Returns and Individual Regulations ... 17
3.5 Control Variables ... 17 3.5.1 MBS and ABS ... 17 3.5.2 Macroeconomic variables ... 18 3.5.3 Liquidity measurements ... 19 3.5.4 ABS characteristics ... 19 3.5.5 Location effects ... 20
4 Data and Descriptive Statistics ... 20
4.1 Sample ... 20
4.2 Databases Used to Construct the Sample ... 21
4.2.1 Bloomberg ... 21
4.2.2 J.P. Morgan International ABS & CB Research ... 21
4.2.3 IMF & central banks ... 21
4.2.4 World Bank ... 22
4.2.5 CRSP / Compustat ... 22
4.3 Descriptive Statistics ... 22
4.3.1 Dependent and independent variables ... 22
4.3.2 Regulations on ABS ... 23
4.3.3 Sectors ... 27 4.3.4 Maturity type ... 27 4.3.5 Location ... 28 4.3.6 Cross-‐correlations ... 29 5. Results ... 29 5.1 Introduction ... 29 5.2 Individual Regulations ... 29
5.2.1 Return and individual regulations (regression 1) ... 29
5.2.2 Monthly volatility of returns and individual regulations (regression 2) ... 33
5.3 Grouped Regulations by Type ... 36
5.3.1 Return and regulations grouped by type (regression 3) ... 36
5.3.2 Monthly volatility of returns and regulations grouped by type (regression 4) ... 37
5.4 Causality ... 38
6 Robustness ... 42
6.1 The Effect of Omitting Non-‐core Variables ... 43
6.2 Using a Different Risk Premium Measure ... 45
6.3 Estimating return effects using control variables ... 45
7 Conclusions ... 51
8 Policy Implications, Limitations, and Future Research ... 53
References ... 54 Appendix ... 62 Canada ... 62 Canada 2009 ... 62 Canada 2010 ... 63 European Union ... 63 EU 2004 ... 63 EU 2007 ... 64 EU 2009 ... 64 EU 2011 ... 64 EU 2014 ... 64 Japan ... 64 Japan 2010 ... 64 Japan 2012 ... 65 Japan 2014 ... 65 Malaysia ... 65 Malaysia 2013 ... 65 Russia ... 65 Russia 2004 ... 65 Russia 2011 ... 66 South Africa ... 66 South Africa 2003 ... 66 South Africa 2013 ... 66 South Korea ... 67 South Korea 2003 ... 67 South Korea 2011 ... 67 United Kingdom ... 67 UK 2008 ... 67 UK 2010 ... 67 UK 2012 ... 67 United States ... 68 US 2006 ... 68 US 2007 ... 68 US 2009 ... 68 USA 2010 ... 68 USA 2011 ... 69
Statement of Originality
This document is written by Reinier Zonneveld who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
1 Introduction
Recently there have been multiple major developments in asset-backed Securities (ABS) and mortgage-backed securities (MBS) markets. The widespread use of MBS by financial institutions and the substantial growth in the size of the market, combined with developments in the United States housing market, led to the subprime mortgage crisis. Ultimately, this led to a global financial crisis. To illustrate, in 2014 alone there was over 225 billion dollars’ worth of ABS issued in the US alone (SIFMA, 2015). As a result, governments and financial institutions worldwide implemented regulations to prevent events like this from occurring in the future. Some examples are the Dodd-Franck Act in the US and the CMBS2.0 in both the EU and US (Demyanyk & Van Hemert, 2011; FSB, 2014).
Simultaneously with the growth of proposed regulations, the critique of and discussions about these laws increased as well. Purnanandam (2009) claimed that originators of the underlying assets of ABS should have screened their borrowers better before supplying a mortgage. Hanson et al. (2011) attribute the financial crisis to the shortcoming of the regulatory system, while Tropeano (2011) gives an overview of policy implications drawn from recent developments in the ABS market. However, not much research has quantified the effects of regulations on ABS, and the existing studies focus on only one regulation. Moreover, these papers find contradicting results in two aspects. First, they contradict theory set by other papers. Secondly, their quantitative results contradict with comparable papers (Amtenbrink & De Haan, 2009; Duygan-Bump et al., 2010; Stoebel and Taylor, 2009).
Combining these findings on market and governance developments and the small amount of quantitative research, the importance of a new study in this field is easily found. A major question is still not adequately answered: What are the effects of regulations on the risk and return of ABS? This question sums up the rationale and aim of the research. The paper tries to find an answer to that question by using large amounts of data from the ABS market, economies, and regulations and combines these in a model in order to estimate the effects of regulation on the risk and return of ABS. By creating categories of regulations, joint estimates for different governance approaches on risk and return can be made. Before-after analysis is applied on the data, which in itself does not estimate causality. However, through logical analysis, causality is proven, which makes the results much more interesting.
The finding in this research is that most regulations do have an effect on risk and returns. However, there are many differences in the magnitude of these effects. I also find that some regulations can even have an adverse effect. In general, the findings unveil the different effects of regulations — at both the individual and aggregated level — which makes it possible for interested parties to find targets for in-depth study and compare why certain regulations do (or do not) work. Furthermore, all findings are justified using economic theory; the results are not only quantified, but also explained. Another contribution to existing literature is shifting the perspective from a mostly theoretical discussion to a quantified discussion and providing estimates based on worldwide markets instead of using a limited timeframe and data sample. This makes it possible to view governance and ABS from a much wider viewpoint. Finally, the research will also set a basis for quantifying regulatory changes in relation to risk and return in ABS markets.
The remainder of this paper starts with an overview of the existing literature in chapter 2. It starts with a general, short introduction to ABS, followed by a framework for risk and return in the market. Sources of risk, like credit ratings and capital structure, are discussed, together with regulations and their theoretical effects on risk and return. The chapter concludes with summarizing the important debates in the literature regarding regulations and the implications for the hypothesis in the next sections. Chapter 3 covers the methodology and lays out the details of the model. Section 4 describes the data used and the sample set. Chapter 5 presents the findings, followed by robustness checks in chapter 6. Section 7 concludes the results of this research.
2 Literature Review
2.1 A Short Overview of ABS and the Financial Crisis1
ABSs are securities for which the coupon payments are consisting of capital generated by a pool of underlying assets. These appeared for the first time during the 1980s. The underlying assets are collateralized and mostly consist of illiquid assets, e.g. mortgage loans or credit card receivables. MBS is a special case of ABS, a subset for which the underlying is formed by mortgages only. The issuer selects the assets which are included in the ABS and defines a payout structure, based on multiple risk levels, through
1 Of course, the overview presented is incomplete and meant as a quick introduction for a reader without
any knowledge on the topic. For more information on ABS, I would suggest reading Bhattacharya and Fabozzi (1996) for example.
tranching. The securitization is often performed using a special purpose vehicle (SPV), which means the SPV constructs and sells the ABS to a trust. Finally, the trust issues the ABS on the market for the investors. The different tranches all have different claims on the cash flow generated by underlying— the highest tranche gets paid first, then the tranche below, and so forth, until the lowest tranche is reached. This system is called a waterfall structure and implies a higher risk level for lower tranches. Credit rating agencies (CRAs) assign a credit rating to the securities generated through this securitization process (Bhattarcharya & Fabozzi, 1996).
ABS, and more specifically MBS, played a major role in the financial crisis of 2007. The global financial crisis originated from the subprime crisis in the United States. Before the events that ultimately lead to the crisis, housing prices were rising. However, in 2006, the market showed signs of a reversion of that trend. By that time, a substantial amount of risky subprime mortgages were supplied by banks, without regulators questioning or controlling this behavior. During the same period, there was a fast growing market for MBS. These MBSs were repackaged into collateralized debt obligations (CDOs). The complexity of CDOs is high, since it makes it even harder to estimate the quality of the pool of underlying, compared to MBS. When the housing prices began to fall, the default rates of the collateral increased substantially. This led to sharp price decreases and an increase in uncertainty in the MBS and CDO markets. Since many major financial institutions held a large amount of ABS, the effect was so severe that it led to the Global Financial Crisis of 2007 (Sanders, 2008; Duchin et al., 2010; Bolton et al., 2012).
2.2 Risk and Return in the ABS Market
First, it is important to establish a framework from which the effects of risk and return in the ABS markets are clearly defined. Recall that in CAPM the risk of a security is divided into a diversifiable and a non-diversifiable portion. Diversifiable risk belonging to a specific security can be diversified away in a portfolio by investing in stocks with different correlations to the market. The remaining portion of the total risk, non-diversifiable risk or systemic risk, cannot be compensated for by changing the contents of a portfolio, which in turn means that investors require a compensation for bearing this sort of risk. To be precise, increasing systemic risk causes the required rate of return of an investor for a particular security to increase (Fama & French, 1992). Furthermore, systemic risk is an indicator of the risk of a market as a whole. If the non-diversifiable
risk is substantial, it could cause a major downturn in the market; the earlier mentioned subprime crisis is an example. In such an event, returns fall dramatically. Systemic risk also includes the uncertainty about the value of the underlying in the case of ABS. For example, if a large amount of the pool of mortgages underlying a MBS default, investors might lose on their return since the required capital to pay the coupon is non-existent. Thus, increasing non-diversifiable risk causes the required return on an investment to increase and the actual return to decrease in the ABS market (Ben-Horim & Levy, 1980).
Besides the negative effect of risk on return in the ABS market, there are also two ways in which systemic risk can be reinforced. Firstly, Borio and Zhu (2012) conclude that recent changes in the financial system increased the importance of the risk-taking channel. This means that monetary policy influences the risk-aversion of market participants, which influences financial conditions and, finally, the economic decisions and the market as a whole. In the current financial climate, investors therefore are more risk-averse and in turn command an even higher required rate of return on risky investments, ultimately lowering the actual return further. Secondly, information uncertainty is a source of systemic risk. The higher the information uncertainty, the higher the volatility. However, it also reinforces the effect associated with news disclosure: bad news will have a stronger effect. Thus, in uncertain, risky markets, bad news will cause returns to lower even more than the downward movement it already would have initiated in a market with less uncertainty. Finally, a measure for risk is the volatility of the returns, if one controls for idiosyncratic risk or, in other words, risk premium (Zhang, 2009).
2.3 Common Sources of Risk of ABS and Their Effect on Returns
In the previous section, an abstract approach to risk and return was discussed. However, to find logical relationships between regulations and their effects on risk in the ABS market, a more concrete definition of the sources of risk in ABS is required.
2.3.1 Credit ratings2
A substantial source of risk in ABS markets comes from credit ratings and credit rating agencies — CRAs. There is evidence on ratings being unreliable. The explanation is
2 CRAs, credit ratings, and their effects on risk and return in markets are very extensive topics. However,
fully discussing all channels through which these factors influence each other is beyond the scope of this research. I suggest reading, for example, Jeon and Lovo’s article “Credit Rating Industry: A Helicopter Tour of Stylized Facts and Recent Theories” (2013) as a starting point.
twofold: On one hand, the rating agencies have an incentive to assign a high rating to a customer. On the other hand, their strategy dictates to do so. Since the issuers of ABS pay the rating agency and not the investor, the agency is motivated to attach a rating high as possible to the security, since a content client is more eager to return when issuing ABS at a later point in time (Mathis et al., 2009). The strategy of the agencies, on the other hand, causes a delay in the adjustment of ratings. Ratings are said to reflect long-term investment, and by that reasoning, new information is not immediately implemented in the judgment. Furthermore, ratings are influenced by the state of the financial markets; during booms the ratings are higher on average than during recessions, which makes for a procyclical distortion in credit ratings. The influence of credit ratings on investors’ behavior is strong, It is an important channel in the financial economic chain (Bolton et al., 2012). The effect of uncertainty about the quality of the credit ratings of ABS leads to investors requiring a discount on ABS, which means a lower return. The information uncertainty about ratings is naturally reflected in a higher risk level (White, 2010).
2.3.2 Liquidity risk
Although ABS can be seen as a way to liquidize pools of mortgages, liquidity is still an urgent problem in ABS markets. ABSs are illiquid assets, for which trading speed can decline very fast in times of financial distress. During the subprime mortgage crisis, the demand for ABS was already lower. However, banks also tended to accumulate more liquid assets and they sold illiquid assets, including ABS. It follows that the liquidity problem reinforces itself during downturn markets, since supply grows and demand falls. Liquidity risk directly increases the risk of ABS; returns fall during illiquid times (Cornett et al., 2011).
2.3.3 Capital structure
Additionally, the capital structure of ABS can lead to more risk for the investor. For example, one of the contributing factors to the subprime crisis was an amount of leverage that was too high and risky debt, although one of the original purposes of ABS was to spread out risk from the originator to a large amount of smaller investors in the security. However, in the period preceding the subprime crisis, ABS was “misused” to evade capital restrictions, which resulted in highly risky investments (Acharya & Richardson, 2009). Without restrictions on the structure of capital in ABS, products could be designed with substantial risk for even higher tranches through a fragile composition of the underlying (Longstaff, 2010). Furthermore, repackaging the MBSs
into CDOs can result in very complicated products for which the composition of the original underlying assets is hard to determine for individual investors. Here moral hazard arises; the issuer or CDO’s manager should select high-quality assets. However, for short-term profit, it could be more interesting to select more risky assets. CDOs could be used to mask the actual capital structure of the individual MBS from which it consists (Duffie & Gârleanu, 2001). The manner in which capital is structured in ABS could directly increase risk because of incentives to increase short-term profits through the underlying and tranching, and repackaging could mask any risk caused by the structure proposed by the originators of the original ABS. Short-term returns could be higher when this is the case; however, evidence from the past suggests that market disruptions could eventually lead to negative returns.
2.3.4 Information asymmetry and intransparency
Closely related to the other source of risk is the problem of information asymmetry and intransparency of the ABS market (Riddiough, 1997). Information asymmetry in ABS is of special importance since these derivatives can be very complex. Channels through which this problem arises could be between the investor and the asset manager, since the latter has more information about the actual quality of the product than the former (Covitz et al., 2013). However, there is also uncertainty between the borrower and the originator and the originator and the arranger about the creditworthiness and mortgage fraud respectively. Finally, there are instances of introduction of ABS in countries where there were no regulations defining these types of securities, furthering the information asymmetry gap between potential investors and issuers (Ashcraft & Schuermann, 2008). Information uncertainty leads to higher risk, since it is harder to make an estimate of the risk level of a security. The required rate of return of an investor will be higher and the acquired returns therefore will be lower. Both risk and return are thus inversely related to information asymmetry and intransparency.
2.4 Regulatory landscape of ABS markets
This section will provide an overview of regulatory changes included in this research. As will be shown in the appendix and chapter 4.3.2, the sample set used consists out of the European Union (EU), United States (U.S.), Canada, Japan, Russia, South Korea, Malaysia, South Africa, and the United Kingdom. For these economic zones and countries, an analysis of the regulations is performed, which leads to five distinct types of regulations. The laws concern the following topics: credit ratings and agencies,
governments purchasing ABS, capital requirements, information disclosure, and initial specification of ABS. The types of regulatory decisions will be linked to aforementioned theory with regards to systemic risk and return in the ABS market. The appendix will summarize the contents and justify the classification of the individual regulations in the five different categories.
2.4.1 Type I: Laws concerning credit ratings and their corresponding rating agencies
Theory suggests that sub optimally performing rating agencies will lead to lower returns and higher risk. Logically, improving credit ratings by introducing laws should increase returns and lower risk. Examples are disclosure of information regarding the underlying pool of assets or mitigating adverse selection through increasing CRA independence.
2.4.2 Type II: Government purchasing ABS in order to stimulate the market
Although this is not a regulation, it is included. Governments create regulations to stimulate the market, and by influencing the market, these purchases could have a severe impact on ABS, especially considering the billions of dollars regularly involved in these market interventions. In the context of the aforementioned theory, a successful program of this kind should increase returns and reduce risk because of reduced liquidity risk.
2.4.3 Type III: Additional capital requirements
Theory suggests that capital structure can influence ABS markets’ risk and return. Numerous government regulations are introduced in order to influence capital structure problems. Some examples are leverage ratios and mandatory vertical tranche ownership by the originator. If developed correctly, capital requirements could lead to lower returns in the short-term. However, with a longer time horizon, the effects on returns should be positive since it could prohibit originators or CDO managers from excessive risk taking. Logically, risk could be lowered by type III regulations.
2.4.4 Type IV: Information disclosure
Regulations regarding information disclosure could be able to reduce information asymmetry and intransparency in ABS markets. Since these factors are inversely related to risk and return, both should increase as a result of successful type IV regulations . Some examples are reporting all cash flows of a security in a public database and extensive prospectus requirements.
2.4.5 Type V: Initial specifications on the structure of ABS
Type V regulations could be seen as a special variety of type IV, information disclosure. The difference is that this reform can only be introduced in countries or economic zones
where there are no legal definitions of ABS. It is the first step in the law making of ABS in the countries where such a regulation is seen. Effects of this regulation should be similar to type IV— lower risk, higher returns.
2.5 Regulations in Practice: Findings Contradicting ABS Market Theory
Theory seems clear about the effects of ABS market governance. However, the debate on whether regulations are really able to mitigate risk and return problems is not settled. There is enough critique in this area to doubt whether financial regulators are able to do so, and thus it is interesting to discuss some of the debates in the literature. Ultimately, this will influence the hypothesis tested in the model. The next section will therefore discuss some of the main critiques per regulation type.
2.5.1 Type I: Laws concerning credit ratings and their corresponding rating agencies
Common proposed advice in the context of CRA laws consist of renewing the supervisory regime and shifting the payment from the issuer to the investors. The first measure would ensure, for example, truthful and accurate information given by the CRAs. The latter could mitigate the misalignment of interest currently existing between the CRAs, issuers, and the investors (Jeon & Lovo, 2013; Amtenbrink & De Haan, 2009). Another suggestion is the mandatory use of a second best rating, which reduces the issuer to shop for the highest rating. Taking the ideas even further, the whole purpose of CRAs in the current regulatory landscape is doubtful, since it could be the case that the current system only exaggerates the issues regarding uncertainty about the quality of ABS (Bolton et al., 2012). Although many reforms regarding CRAs are proposed, especially in the EU and US, a common critique is that the regulations are not substantial enough to address the aforementioned issues. For example Amtenbrink and De Haan (2009) review the regulatory reform of 2009 in the EU and claim that it could not have a sizeable impact on the CRA problem. These findings contradict earlier mentioned theory and implicate that one cannot be certain on which direction the effects will go and if they are existent at all.
2.5.2 Type II: Government purchasing ABS in order to stimulate the market
The magnitude of literature on ABS market intervention is smaller than is the case for other regulations. However, for example, multiple studies have found conflicting evidence for the US ABS repurchasing program. Duygan-Bump et al. (2010) find a positive impact on liquidity and a reduced asset outflow from market mutual funds. On the other hand, Stroebel and Taylor (2009) found no significant effects of a similar
program for MBS by the Federal Reserve Bank. The effect of comparable measures taken by governments in different countries could not confidently be estimated just by theory.
2.5.3 Type III: Additional capital requirements
Calem and LaCour-Little (2004) find that the effect of capital requirements in, for example, Basel II are not homogeneous over the market and are dependent on individual characteristics of the targeted loans and geographic features. The result could be divergent effects cancelling out over the entire market. Merrill et al. (2012) suggest evidence for capital requirements provoking fire sales and therefore forcing prices to move downwards due to decreased liquidity in distressed MBS markets. Although capital requirements are often used, there is evidence the effect might not be as obvious as suggested by previous theory.
2.5.4 Type IV: Information disclosure
Glaeser and Kallal (1997) claim that market liquidity could move both ways as a result of information disclosure in ABS markets, which means returns and risk could be affected both ways through liquidity risk. Furthermore, they conclude that information disclosure results in asset bundling. In the context of MBS markets, this could, for example, mean repackaging of MBS in CDOs. However, since CDOs are known to be very complex financial products, it might increase information asymmetry between the investors and the regulators. Thus, an increase in information disclosure might have an unexpected adverse effect on ABS market transparency.
2.5.5 Type V: Initial specifications on the structure of ABS
There is no discussion regarding type V regulations. However, one could hypothesize an effect similar to that of type IV regulations. Ultimately, type V regulations establish a framework on which the other four regulations could be built upon. Since the effect of the other regulatory changes could be opposite to theory and thus unwanted from a financial market point of view, it could be the case that type V regulations have an adverse effect through this channel.
2.6 Recapitulation
Considering all the information in the previous chapter it can be concluded that
increased systemic risk can result in lower returns, through an increased required return from investors and by inducing market downturns. Common sources of risk in ABS
markets are CRAs, liquidity risk, capital structure, and information asymmetry and intransparency. Literature suggests clear theoretical effects of these types of risk on returns. The types of regulations imposed by governments all relate to some form of risk; CRA rules should effect CRAs, purchase programs are linked to liquidity risk, capital requirements affect the risk associated with capital structure, and information disclosure and defining of ABS could influence the information asymmetry and market
intransparency. However, there is substantial contradictory evidence from the literature to conclude that regulations could have adverse effects and not make the market behave as intended or as shown by theory. As a result, the hypothesis will be based on if
regulations do have an effect at all, as opposed to hypothesizing which direction the effects will have.
3 Methodology
This section will provide a detailed explanation of the methodology used to test the different hypotheses. All four regressions are basically of the before-after type, with the same control variables and regress panel data in time series format. However, the dependent variable and the independent variable of interest may differ. As a result of using before-after analysis, causality cannot be estimated from the model. However, this problem is addressed later in the paper.
3.1 Return and Individual Regulations
To test whether there is an effect of individual regulations on the returns of ABS, a difference-in-difference regression is used. The dependent variable is the logarithmic return on ABS. Using logarithmic form has the benefit of simplifying calculations for comparing returns. The return is based on mid prices, and not on the bid or ask prices for the securities. By using mid prices, one rules out the liquidity pricing component present in the bid-ask spread (Wang & Yau, 2000). The return is calculated as follows, for all observations in the dataset:
𝐿𝑜𝑔 𝑟𝑒𝑡𝑢𝑟𝑛 = 𝑙𝑜𝑔 !"#$!% [!!!]!"#$!% [!] (1a)
The independent variables of interest are the individual regulations. In the model, these regulations are a simple cross term between the country or economic zone and a certain point in time. If the area corresponds with a major regulation in place, the area dummy takes on the value 1 and is 0 otherwise. If the date corresponds with the
regulation, it will also take on the value 1 and 0 in a similar vein. The result is a variable which takes on the value 1 when a certain regulation is in place in a certain country and economic zone and 0 otherwise. The value changes to 0 again, when a new regulation replaces the old one or a new law appears that affects the same ABS characteristics of the old legislation. Using this dummy assignment process, one ends up with a vector containing dummies for all major regulations in the sample period. An overview of this vector is in section 4 and a detailed description is in the appendix.
𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛
𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 + 1
𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 + ⋯ , with r being the regulation at time t, in country c (1b)
Added to the model are the control variables and vectors of control variables, a dummy for MBS, macroeconomic variables, liquidity measurements, ABS characteristics per security, and location effects. These will be explained in more detail in section 3.5. When these controls are added in, the equation
becomes as follows: 𝐿𝑜𝑔 (𝑅𝑒𝑡𝑢𝑟𝑛) = 𝛽! !" ! 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 + 1 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 + ⋯ + 𝛽!𝑀𝐵𝑆!,!+ 𝛽!𝐺𝐷𝑃!,!+ 𝛽!%∆𝐺𝐷𝑃!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!𝑅𝑖𝑠𝑘 𝑃𝑟𝑒𝑚𝑖𝑢𝑚!,!+ 𝛽!𝐵𝑖𝑑 − 𝐴𝑠𝑘 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!!"#!!"# !"#$%&!"# !"#$% !,! !,! + 𝛽!"𝑅𝑎𝑡𝑖𝑛𝑔!,!+ 𝛽!!𝐴𝑚𝑜𝑢𝑛𝑡 𝐼𝑠𝑠𝑢𝑒𝑑!,!+ 𝛽!"𝐶𝑜𝑢𝑝𝑜𝑛!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝐿𝑒𝑛𝑔𝑡ℎ!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝑇𝑦𝑝𝑒!,!+ 𝛽!"𝑆𝑒𝑐𝑡𝑜𝑟!,! + 𝛽!"𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝐸𝑓𝑓𝑒𝑐𝑡𝑠!,!+ 𝜀!,! (1c)
time t, location c, security i
The hypothesis tested is as follows:
H0: Individual regulations do not have any effect on returns on ABS.
Ha: Individual regulations do have an effect on returns on ABS. 3.2 Volatility of the Returns and Individual Regulations
The volatility of the logarithmic returns will serve as a measurement of systemic risk in the ABS market. The reason volatility is a measure of systemic risk and not total risk consisting of idiosyncratic and systemic risk in this context is explained in the control variable section (3.5.2). Volatility is defined as the monthly volatility, per security, which
is common in multiple studies on bonds and securities, (e.g. Green & Ødegaard, 1997) and it is defined as follows:
𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 = 𝑣𝑜𝑙𝑎(𝑙𝑜𝑔 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 𝑝𝑒𝑟 𝑚𝑜𝑛𝑡ℎ (2a)
With inclusion of the regulation vector as presented in (1b), as well as the control variables, the total regression formula becomes the following:
𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 = 𝛽! !" ! 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 1 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 … 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑟 + 𝛽!𝑀𝐵𝑆!,!+ 𝛽!𝐺𝐷𝑃!,!+ 𝛽!%∆𝐺𝐷𝑃!,! + 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!𝑅𝑖𝑠𝑘 𝑃𝑟𝑒𝑚𝑖𝑢𝑚!,! + 𝛽!𝐵𝑖𝑑 − 𝐴𝑠𝑘 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽! 𝐵𝑖𝑑 − 𝐴𝑠𝑘 𝑆𝑝𝑟𝑒𝑎𝑑!,! 𝑀𝑖𝑑 𝑝𝑟𝑖𝑐𝑒!,! + 𝛽!"𝑅𝑎𝑡𝑖𝑛𝑔!,! + 𝛽!!𝐴𝑚𝑜𝑢𝑛𝑡 𝐼𝑠𝑠𝑢𝑒𝑑!,!+ 𝛽!"𝐶𝑜𝑢𝑝𝑜𝑛!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝐿𝑒𝑛𝑔𝑡ℎ!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝑇𝑦𝑝𝑒!,! + 𝛽!"𝑆𝑒𝑐𝑡𝑜𝑟!,!+ 𝛽!"𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝐸𝑓𝑓𝑒𝑐𝑡𝑠!,!+ 𝜀!,!
time t, location c, security (2b)
The hypothesis tested is as follows:
H0: Individual regulations do not have any effect on the systemic risk of ABS.
Ha: Individual regulations do have an effect on the systemic risk of ABS. 3.3 Return and Regulation Types
This equation is similar to (1); however, the regulation vector presented in (1b) is now replaced by a vector containing five categorical variables of the regulations across different countries. This makes it possible to compare the effect of a certain type of regulation on the returns of ABS. The categorical variables are constructed by identifying types of regulations and combining their dummies together. The equation for return and regulation types becomes the following:
𝐿𝑜𝑔 (𝑅𝑒𝑡𝑢𝑟𝑛) = 𝛽! !" ! 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 1 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 … 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 𝑞 + 𝛽!𝑀𝐵𝑆!,!+ 𝛽!𝐺𝐷𝑃!,!+ 𝛽!%∆𝐺𝐷𝑃!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!𝑅𝑖𝑠𝑘 𝑃𝑟𝑒𝑚𝑖𝑢𝑚!,!+ 𝛽!𝐵𝑖𝑑 − 𝐴𝑠𝑘 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!!"#!!"# !"#$%&!"# !"#$% !,! !,! + 𝛽!"𝑅𝑎𝑡𝑖𝑛𝑔!,!+ 𝛽!!𝐴𝑚𝑜𝑢𝑛𝑡 𝐼𝑠𝑠𝑢𝑒𝑑!,!+
𝛽!"𝐶𝑜𝑢𝑝𝑜𝑛!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝐿𝑒𝑛𝑔𝑡ℎ!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝑇𝑦𝑝𝑒!,!+ 𝛽!"𝑆𝑒𝑐𝑡𝑜𝑟!,! +
𝛽!"𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝐸𝑓𝑓𝑒𝑐𝑡𝑠!,!+ 𝜀!,! (3a)
The hypothesis tested is as follows:
H0: The five types of regulations do not have any effect on the returns on ABS.
Ha: The five types of regulations do have an effect on the returns on ABS.
3.4 Volatility of the Returns and Individual Regulations
The last equation is comparable to 2(b); however, like in (3a), the regulation dummies are replaced by regulation type variables.
𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 = 𝛽! !" ! 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 1 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 … 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑡𝑦𝑝𝑒 𝑞 + 𝛽!𝑀𝐵𝑆!,!+ 𝛽!𝐺𝐷𝑃!,!+ 𝛽!%∆𝐺𝐷𝑃!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒!,!+ 𝛽!𝑅𝑒𝑎𝑙 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!𝑅𝑖𝑠𝑘 𝑃𝑟𝑒𝑚𝑖𝑢𝑚!,!+ 𝛽!𝐵𝑖𝑑 − 𝐴𝑠𝑘 𝑆𝑝𝑟𝑒𝑎𝑑!,!+ 𝛽!!"#!!"# !"#$%&!,! !"# !"#$%!,! + 𝛽!"𝑅𝑎𝑡𝑖𝑛𝑔!,!+ 𝛽!!𝐴𝑚𝑜𝑢𝑛𝑡 𝐼𝑠𝑠𝑢𝑒𝑑!,!+ 𝛽!"𝐶𝑜𝑢𝑝𝑜𝑛!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝐿𝑒𝑛𝑔𝑡ℎ!,!+ 𝛽!"𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝑇𝑦𝑝𝑒!,!+ 𝛽!"𝑆𝑒𝑐𝑡𝑜𝑟!,!+ 𝛽!"𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝐸𝑓𝑓𝑒𝑐𝑡𝑠!,!+ 𝜀!,! (4a)
time t, location c, security i
The hypothesis tested is as follows:
H0: The five types of regulations do not have any effect on the systemic risk of
ABS.
Ha: The five types of regulations do have an effect on the systemic risk of ABS.
3.5 Control Variables
In order to test the hypothesis presented in 3.1-3.4, it is necessary to control for various aspects of the securities, market, and macroeconomic factors. All the control variables will be presented in this section.
3.5.1 MBS and ABS
MBS and other ABSs are fundamentally different in the sense that, although MBS is a type of ABS, the underlying pool of assets in MBS solely consist of mortgages. The difference between MBS and general ABS was an incentive for multiple countries to design laws particularly aimed at MBS, e.g., CMBS 2.0 in the EU and U.S. See the appendix for a detailed explanation of the differences between MBS and other ABSs. To
account for any structural differences, a dummy called MBS is included, which takes on the value 1 if it concerns a MBS and 0 otherwise.
3.5.2 Macroeconomic variables
The macroeconomic variables include GDP per year per country, the percent change in GDP, the real interest rate, the interest rate spread, and the risk premium.
There exists a chance that the underlying debt included in the securities, e.g., mortgages for MBS or credit card balance deficits for ABS, are prepaid (i.e. prepayment risk). The chance of this taking place depends on the state on the economy and the associated interest rates (Schwartz & Torous, 1989). The probability of the underlying being prepaid is included in the valuation of the securities. If the interest rate, for example, moves down, the price of the security should be adjusted accordingly because prepayment risk increases. Price movements caused by this mechanism could over- or underestimate the effects of regulations on volatility, through endogeneity. Thus, one needs to include interest rates and the state of the economy to isolate this effect so that a better estimation of the risk level can be acquired.
Interest rate variables are included in the form of real interest rates for a certain country at a certain point in time. In addition, the interest rate spread – the difference between the lending and borrowing interest rates – is added. The spread is known to be of influence to the prepayment decision making of individuals and thus should be accounted for (Quiqley, 1987).
As mentioned, the state of the economy influences prepayment decision making. To approximate these effects, adjusted GDP per country per year is included. Additionally, the percentage change from one year to another is represented by another variable: GDP percent change.
The last macroeconomic control variable is the risk premium, which originates from the CAPM and in this context should control for idiosyncratic risk, known as beta. By accounting for idiosyncratic risk, movements in volatility caused by regulations are based solely on systemic risk (Fama & French, 1992). Furthermore, risk premium and volatility are positively correlated, so ignoring the risk premium in the regression might cause overestimation of the volatility if the risk premium is high and vice versa (French, Schwert, & Stambaugh, 1987). Risk premium is built up using the MSCI World Index, which includes stocks, bonds and real estate. Then the risk-free rate of the country where the security originates is subtracted from the risk premium. A market portfolio consisting of not only stocks like the S&P500, but all these three types of assets should give
stronger results when some of the bonds are influenced by real estate– which is the case since MBS is included (Titman & Warga, 1986).
3.5.3 Liquidity measurements
It is known that changes in liquidity can cause movement in return and volatility. For example, in illiquid markets, investors require a higher return on their investment in order to compensate for the risk of not being able to quickly sell a security if demanded (Wang & Yau, 2000). A good measure of market liquidity for a specific security would be trading volume. However, given the fact that the market for ABS is private and OTC, this data is not available (Bloomberg, 2015).
Regardless of this, a control for liquidity should be implemented in order to obtain more precise estimators. The model uses mid prices for the calculation of the return and the volatility thereof. However, the bid and ask prices provide interesting details on the liquidity state of the market at that point in time for a certain security. Liquidity and bid-ask spread – defined as the difference between bid and ask prices – are found to be negatively correlated. For example, in illiquid times the bid-ask spread increases. Since other effects of the bid-ask spread, like higher expected rates of return, are accounted for by the various control variables, the bid-ask spread seems to be a proper measure of market liquidity (Wang & Yau, 2000; Amihud & Mendelson, 2015). Both the absolute bid-ask spread and the bid-ask spread, as a percentage of the mid price, are included.
3.5.4 ABS characteristics
The ABS in the sample varies based on certain important aspects: rating, issue size, coupon payment, maturity length, and maturity payment type. To correct for these differences, multiple control variables are added accordingly.
Arguably the most important characteristic of ABS is the corresponding rating. He, Qian, and Strahan (2011) point out that credit ratings influence the prices of ABS. Thus, these ratings have to be accounted for when modeling the risks of the securities. Ratings influence the beliefs of the investors, illustrated by the fact that the biggest part of the ABS trade depends on high-rated securities. To correct for rating differences, a dummy is created that takes on the value of -1 for the lowest ratings, 0 for intermediate rated products, and 1 for highest grade securities. The use of these categories makes it possible to include non-rated securities and securities for which the rating is not included in the multiple databases used to construct the dataset. These two types of securities are put in the 0 group, since I believe that on average, the ratings are normally distributed in
the sense that the average security with an unknown rating should fall in the 0 category. Detailed information about the composite rating consisting of Standard & Poor, Moody’s, and Fitch ratings and the rationale behind the classification of the three groups in the dummy rating can be found in the appendix.
Furthermore, a measure for size comparable to Fama and French (1992) is derived from the dollar amount of an issue, per security. Obviously this variable is fixed and takes on only one value per observation for each security. Additionally, in a similar vein to the aforementioned dollar volume, the coupon payment in dollars is added as a control variable as well. Coupon size in dollars and the dollar amount of the issue are included in multiple studies on risk and returns in bond markets, e.g. Lin, Wu,, and Wang (2010).
Additionally, time to maturity and the maturity type are added to the control variables. Time to maturity is defined as the time to maturity when the security was issued, thus maturity date minus issue date. A longer maturity means more exposure to interest rate changes, and the maturity payment type also changes this exposure. Again, these two control variables are added in multiple studies as well. Time to maturity is expressed in years, and maturity type is captured in a categorical dummy variable.
Finally, differences between the sectors in which the ABS operates are controlled for using a categorical dummy variable.
3.5.5 Location effects
The final control variable is a simple categorical dummy variable that captures country or economic zone specific fixed effects, and this variable is called location effects. In order to get a good understanding of the effect of the regulations or the set of regulations on the dependent variable it is necessary to exclude the country specific effects from the regulatory effects.
4 Data and Descriptive Statistics
4.1 SampleTo test the effect of the regulations or type of regulations on returns and volatility, one needs a large enough dataset, over as many regions as possible. Finding data on ABS is not simple. Many databases do not include ABS data, while others such as DataStream do include ABS but only the identifiers and no further information. Bloomberg provides a solid database about ABS. It lists around 18,000 ABSs for the period of 2000 to May 2015, for which more than 6,000 actually have pricing data. The sample period is this
range for multiple reasons. First, major regulations in the economic zones and countries in the sample only emerge in 2003. Secondly, there is enough room in the sample for a pre- and post-subprime crisis period. Directly after the crisis, most new rules were established, so it makes sense to include a long enough pre-crisis sample.
4.2 Databases Used to Construct the Sample
4.2.1 Bloomberg
A Bloomberg (2015) terminal is used to retrieve weekly data on bid-mid-ask price data, type of ABS, Bloomberg composite ratings, Standard and Poor’s, Moody’s, and Fitch ratings, issue dollar amount, coupon payment dollar amount, issue date, maturity date, maturity payment type, sector, country of origin, ISIN identifiers, ticker, currency, and issuer name. The output from the terminal was exported to Excel. To get the data in panel data format, a script was written that moves observations from the output in Excel in such a way that it is interpretable for Stata. The fixed characteristics are downloaded in a separate sheet. Using lookup formulas, the fixed characteristics are merged with the time series pricing data. Then the whole set is converted from the Excel to the Stata format. Afterwards, string variables are converted to numeric and N/A errors are removed in order to acquire a workable dataset.
4.2.2 J.P. Morgan International ABS & CB Research
The Director of Structuring at NIBC Bank, Wegener Sleeswijk, provided me with an extensive list of characteristics of ABS deals in my sample period, including, but not limited to the following: collateral type, collateral domicile, seller, benchmark, ratings, special comments, ISIN, and CUSIP. Since a sizeable amount of rating data was absent in the Bloomberg database, this list provided a basis to merge some of the data using ISINs and acquire a more complete sample of ratings (NIBC, 2015).
4.2.3 IMF & central banks
Data on regulations are extracted from several sources. IMF summarizes some major regulatory changes in their reports, which serves as a good starting point for finding details on them. In depth information on ABS regulations can be found on the websites of central banks, like ECB for Europe or Bank of Canada for Canada. In order to make the task of collecting data about ABS legislation more efficient, a script is constructed that scans reports on the words “asset-backed securities”, “mortgage-backed securities”, “regulations”, “laws”, “reform”, “act”, “ABS”, and “MBS”. This process makes looking
up regulations more convenient than solely reading and scanning manually. A data file per country containing the regulations with a short description is created. Then these regulations get divided into a certain category – one to five – and is tagged by year. The tagged categories together form the total pool of that type of regulation on a global scale. Corresponding dummies are generated in the Stata file. See the appendix for more details on regulations (IMF, 2015; ECB, 2015).
4.2.4 World Bank
The macroeconomic variables including adjusted GDP, real interest rates, and real interest spreads are acquired from the World Bank database (2015). The data is merged into the Stata file per country and per year, since it concerns yearly data.
4.2.5 CRSP / Compustat
The last database used is CRSP / Compustat, to download weekly pricing data on the MSCI world index, which consists of stock, bonds, and real estate worldwide and the S&P500 Index (WRDS, 2015).
4.3 Descriptive Statistics
In the next paragraphs an overview of the sample will be given using descriptive tables for the dependent and independent variables. Furthermore, an overview of the regulations, sectors, maturity types, and countries included in the sample is given, as well as a correlation matrix.
4.3.1 Dependent and independent variables
Table 1 shows that the amount of securities in the sample equals 6,108 and the sample period is 2000–2015. The logarithmic returns are slightly negative, being -0.39%. However, this is mainly because of big outliers during the subprime crisis, in which the return was as low as -726.66%. The percentiles show that most of the returns in the sample lie somewhere around zero. The monthly volatility of the log returns shows a similar pattern, with extreme observations during the crisis.
MBS is only a small part of the dataset, with only 183 securities with 15,309 observations, compared to 5,925 with 216,315 observations for other ABSs.
The macroeconomic variables show mostly results that should be expected from such data. Adjusted GDP in dollar per capita has a mean of around $ 27,486, a minimum of $ 12,563, and a maximum of $ 113,739. The change in adjusted GDP is denoted in fractions and has a mean of 0.06, which corresponds to 6%. The lowest observation is
minus 62% and the highest observation is a 50% increase from year to year. The real interest rate has a mean of 3.06%. The outliers are interesting, having observations of minus 70.43% and plus 93.94%. These kinds of extreme interest rates occurred during crisis years in Bulgaria, but these are excluded later on during the regression. The percentiles show that it concerns only outliers in this case. The real interest spread shows a comparable pattern with identical explanation, with a mean of 1.54%, a minimum of -1.11%, and a maximum of plus 216.38%. The risk premium on the MSCI is calculated as the difference between the MSCI market portfolio and the risk free rate of a corresponding country. The mean is -3.05%, which is caused by the return of this market portfolio not being very high during most observations.
For the liquidity measures, one can see a -$ 0.17 bid-ask spread mean and the values after winsorizing lie between approximately -$ 700 and $ 0. The bid-ask spread as a percentage of the mid price takes on values between zero and 11%; however, values very close to zero are common.
ABS specifications are as follows. By far, the most ratings fall in the zero category (See appendix for further explanation about the design of this dummy). The dollar amount issued varies greatly, which is illustrated by the high standard deviation. The smallest issue was $8.5 million, while the greatest issue amounted to $6.59 billion. Coupon payments have a mean of $4.75, with a standard deviation of $2.40. The lowest measured coupon equals to zero, while the highest coupon amount is $40. Maturity length averages at 5.76 years, with a standard deviation of 7.72 years. The longest maturity period is 47 years, will the shortest being zero years, i.e. less than one year.
4.3.2 Regulations on ABS
To analyze the key variable of interest, which is regulations, one needs to make an overview of the regulations for all countries and economic zones in the sample. Table 2 shows the result of the analysis of major regulations during the sample period. Laws enforced during this period all concern one of the following topics: credit ratings, governments purchasing ABS, capital requirements, information disclosure, and specification of ABS. However, not all of the regulations are aimed at the total spectrum of ABS; some only target MBS – as indicated in the table. For an extensive overview of the regulations included in this research, justification for the way in which they are assigned to the five categories, and the relevant literature from which the information is obtained, consult the appendix.
Table 1: Descriptive statistics of dependent and independent variables, excluding sector, maturity type, and countries. Type, observations, mean, standard deviation, minimum and maximum observations, as well as the 25%, median and 75% percentiles are reported.
Variable Type Obsservatio ns Mean Standard deviation Minimu m 25% Median 75% Maximum General
Security Identifier Bloomberg ID 231,624 2,417.52 1,633.62 1.00 6,108.00
Years Years 234,206 2,009.93 3.93 2,000.00 2,006.00 2,011.00 2,014.00 2,015.00 Dependent Log(Return) - weekly % 224,246 -0.39 15.13 -726.66 -0.31 0.05 0.16 560.11 Monthly Vola(Return) % 225,001 2.56 31.38 0.00 0.00 0.00 0.02 1,202.86 MBS MBS dummy (0, 1) [0 | 1] 231,624 0.07 0.25 0.00 0.00 0.00 0.00 1.00 Macroeconomic
Adjusted GDP $ per cap. 165,669 27,485.57 12,563.34 703.12 20,474.83 24,155.83 34,075.98 113,738.70
Change Adjusted GDP Fraction 165,639 0.06 0.09 -0.62 0.01 0.08 0.12 0.50
Real Interest Spread % 165,669 1.54 1.17 -1.11 1.38 1.69 1.75 216.38
Real Interest Rate % 165,669 3.06 1.96 -70.43 2.28 3.92 4.11 93.94
Risk Premium (MSCI) % 159,881 -3.05 2.18 -12.06 -4.46 -3.46 -1.92 34.40
Liquidity Bid-Ask Spread $ 223,169 -0.17 3.12 -669.95 -0.25 -0.01 0.00 0 Bid-Ask Spread/Mid price % 222,953 -0.04 29.46 0.00 0.00 0.00 0.00 11 ABS specification Rating Dummy [-1 | 0 | 1] 234,855 0.01 0.27 -1.00 0.00 0.00 0.00 1.00
Amount Issued $ 231,624 131 mill 332 mill 0.00 8.5 mill 11 mill 64 mill 6.59 bill
Coupon Payment $ 231,208 4.75 2.40 0.00 3.13 4.64 5.70 40.00
Note that there are significant differences between the economic areas in terms of which types of regulations they enforce and when the laws were passed. The EU and the U.S. regulatory climates for ABS changed the most as opposed to countries like South Korea, Malaysia, and South Africa. There seems to be a correlation between the size of the security market in a country and the amount and extensiveness of regulatory changes. For example, the markets in South Africa and Malaysia are of such small size that it is even given in an official statement by their central banks as a reason for delaying the legislation. On the other hand, South Korea did not acquire a great amount of new regulatory resolutions, although the market for ABS is substantial. A possible explanation could be that there were elaborate laws in place before the sample period, which averts the incentive to frequently adopt new legislation in the 2003–2015 period.
The programs specifically aimed at MBS in the EU and U.S. are part of the CMBS 2.0 framework and provide a broad spectrum of regulations concerning credit ratings, capital requirements, and information disclosure. Laws with respect to credit ratings in general focus on accessibility of rating data and reduction of adverse incentives of the rating agencies.
After the subprime crisis, governments of the EU, U.S., UK, and Japan decided to purchase large amounts of ABS in order to stimulate the market, which was obviously facing a downturn. Often the available ABS that was qualified for use in these programs was strictly defined by a set of rules, mostly concerning the credit ratings and origin of the securities.
Additional capital requirements are aimed at reducing systemic risk. Examples are the requirement for the originator to hold 5% of the most risky tranche and the leverage ratio specifications on the underlying asset pool.
Information disclosure is one of the most widespread laws in the sample. Examples are full disclosure of the underlying or continuous reporting of changes in the pool of assets, which was implemented in 2010 in Canada.
Initial specifications on the structure of ABS define what a certain security is and how it should be set up, and they set a framework for originators and regulators to identify and work with certain types of assets.
Table 2: Regulations on ABS
This table contains all the major regulations concerning ABS for all years and countries or economic zones included in the sample. See the appendix for an extensive overview including references and justification of the category indexation.
Country/
Zone EU USA Canada Japan Russia South Korea Malaysia South Africa United Kingdom*
Year 2003 V (MBS) V 2004 I IV 2005 2006 I 2007 I I 2008 I 2009 I + II I I + III 2010 II + III + IV IV II IV
2011 I + III + IV (MBS) I+II+IV(MBS) III + IV IV
2012 I + II II
2013 IV + V (Religious) IV
2014 II II
2015
Regulation indicator
I Laws concerning credit ratings and their corresponding rating agencies II Governments purchasing ABS in order to stimulate the market
III Additional capital requirements aimed at reducing risk (e.g. certain percentage of most risky tranches should be held by originator). IV Information disclosure (e.g. reporting ABS ratings in a public database)
V Initial specifications on the structure of ABS (e.g. basic tranching and payout scheme). MBS Indicates a regulation exclusively targeting MBS instead of all ABS
Religious The law is designed in such a way that it complies with religious beliefs about banking (in this case, Sharia)
4.3.3 Sectors
Table 3 shows the observations per sector. Communications, health care, and industrials are represented in such low amounts that they will most probably be insignificant in the model. However, while most ABS is part of the financials, there could be enough observations to measure differences between financials and the other sectors with thousands of observations such as consumer discretionary, consumer staples, energy, government, materials, technology, and utilities.
4.3.4 Maturity type
Table 4 Overview of maturity types in the sample.3
Maturity type Frequency Percent Cumulative
At maturity 158,108 68.26 68.26
Callable / Extendible 802 0.35 68.61
Callable / Puttable 2,270 0.98 69.59
Callable / Sinkable 1,139 0.49 70.08
Callable / Sinkable / Extendable 440 0.19 70.27
Callable 33,736 14.56 84.83 Convertible 22 0.01 84.84 Extendible 16,518 7.13 91.97 Normal 1 0 91.97 Puttable 182 0.08 92.05 Sinkable / Extendable 652 0.28 92.33 Sinkable / Puttable 203 0.09 92.42 Sinkable 17,551 7.58 100 Total 231,624 100
3 For a detailed explanation of these different types of structures, I would suggest reading Strumeyer (2012).
A precise description of the maturity types is beyond the scope of this research and is therefore not included.
Table 3: Overview of sector frequency in the sample.
Sector Frequency Percent Cumulative
Communications 30 0.01 0.01 Consumer Discretionary 5,803 2.51 2.52 Consumer Staples 1,040 0.45 2.97 Energy 1,616 0.7 3.67 Financials 212,511 91.83 95.5 Government 2,158 0.93 96.43 Health Care 30 0.01 96.44 Industrials 171 0.07 96.52 Materials 1,030 0.45 96.96 Technology 2,391 1.03 97.99 Utilities 4,643 2.01 100 Total 231,423 100
Table 4 shows the frequency of the different types of maturities. At maturity is the most widely used structure. Callable, sinkable, and extendible ABS are popular as well.
4.3.5 Location
Table 5 displays the number of observations per country or economic zone of the sample. Unfortunately, the information in the table is all the data available on Bloomberg. However, it seems to be a reasonable amount to make some interesting estimations using the model. Note the very high amount of South Korean ABS in the sample, which is probably caused by the substantial ABS market and the early implementation of information disclosure rules in the pre-2000 period. Most countries with less than approximately 500 observations cannot be included in the regression due to the small sample size. Fortunately, the EU adopts regulations concerning ABS for all member countries. Summed together, the EU countries still make up a fair share of the sample, being just short of 30,000 observations.
Table 5: Overview of countries in the sample
Country Frequency Percent Cumulative
Argentina 322 0.14 0.14 Australia 272 0.12 0.26 Britain 19,956 8.6 8.86 Bulgaria 66 0.03 8.89 Canada 10,466 4.51 13.4 Cayman Islands 5,331 2.3 15.7 Costa Rica 147 0.06 15.76 Czech 317 0.14 15.9 France 531 0.23 16.13 Germany 67 0.03 16.15 Guernsey 345 0.15 16.3 Ireland 78 0.03 16.34 Israel 1,029 0.44 16.78 Italy 191 0.08 16.86 Japan 18,245 7.86 24.73 Jersey 20 0.01 24.74 Kazakhstan 91 0.04 24.78 Luxembourg 3,242 1.4 26.17 Malaysia 303 0.13 26.3 Netherlands 110 0.05 26.35 Panama 1,308 0.56 26.91 Peru 66 0.03 26.94 Russia 1,719 0.74 27.68 South Africa 6,820 2.94 30.62 South Korea 156,666 67.53 98.16 Spain 2,173 0.94 99.1 Switzerland 384 0.17 99.26 Taiwan 40 0.02 99.28 Turkey 23 0.01 99.29 United States 1,652 0.71 100 Total 231,980 100
