Subprime Loan Performance

Transcription

1 Disclosure Regulation on Mortgage Securitization and Subprime Loan Performance Lantian Liang Harold H. Zhang Feng Zhao Xiaofei Zhao May 22, 2015 Abstract In 2006, the US Securities and Exchange Commission adopted Regulation AB (Reg AB) that mandates the disclosure of originators with 20% or more of the pool assets in mortgage backed securities. Using data on non-agency mortgage backed securities (MBS), we uncover a jump in the fraction of mortgage deals consisting of origination stakes just below the disclosure threshold (low stake) after Reg AB. These deals have significantly larger losses than those without low stake when issued after Reg AB. Further, our analysis on loan level data provides evidence that securitized loans show higher delinquency when their originators are associated with more low-stake participation after Reg AB. Our findings suggest that mortgage securitizers attempt to circumvent the disclosure requirement under Reg AB when riskier loans are securitized. Keywords: Regulation AB; Disclosure Threshold; Low Origination Stake; Securitized Loan Performance We thank Allen Berger, Mark Flannery, Andra Ghent, Kathleen Weiss Hanley, Igor Kozhanov, Marco Rossi, and participants at seminars at Fordham University, 2015 Fixed Income Conference, and the US Securities and Exchange Commission for helpful comments. The authors are from the Naveen Jindal School of Management, University of Texas at Dallas, 800 West Campbell Road, Richardson, Texas, 75080, lantian.liang@utdallas.edu, harold.zhang@utdallas.edu, feng.zhao@utdallas.edu, xiaofei.zhao@utdallas.edu

2 1. Introduction The regulation reform on disclosure in the non-agency mortgage backed securities (MBS) market remains a virtually unexplored area in the aftermath of the 2007 to 2008 subprime mortgage crisis. To meet the insatiable demand from global investors reaching for higher yields, the entire supply chain of mortgage securitization has increasingly expanded lending to riskier borrowers. Compelled by the rapid growth of the loan securitization market and the lack of explicit regulations directed towards the distinguishing features of this market, the US Securities and Exchange Commission introduced Regulation AB (Reg AB) in January Despite the creation of Reg AB, no study has examined its effects on the loan securitization market. Our investigation represents the first such attempt. In this study, we focus on the impact of disclosure regulation under Reg AB on the residential mortgage securitization market. Securitized residential mortgages accounted for a large fraction of the new issuance of securitized loans in the period leading up to the financial crisis. 1 The research attributes the financial crisis to the sharp increase in the defaults of mortgage loans. 2 Collecting detailed deal and loan level data on securitized residential mortgages, we conduct an in-depth investigation and provide a quantitative assessment of the impact that the disclosure mandate in Reg AB has had on the composition of mortgage deal originations and securitized loan performance. To identify the impact, we use the cross-sectional variation in mortgage deal originations composition just below the disclosure threshold and examine different performance reactions to the regulation. This is motivated by different disclosure requirements under Reg AB on loan originators based on the percentage of loans included in a mortgage deal from each originator. Specifically, Reg AB Item 1110 requires disclosure of information regarding the 1 According to former International Monetary Fund chief economist Simon Johnson, the total volume of private mortgage-backed securities (excluding those issued by Ginnie Mae, Fannie Mae and Freddie Mac) grew from $11 billion in 1984 to over $200 billion in 1994 to close to $3 trillion in See, Mian and Sufi (2009), Nadauld and Sherlund (2013), Keys, Mukherjee, Seru, and Vig (2010), Keys, Seru, and Vig (2012), Purnanandam (2011), among others. 1

3 size and composition of the originator s portfolio as well as information material to an analysis of the performance of the pool assets, such as the originator s credit-granting or underwriting criteria for the asset types being securitized, if the originator or group of affiliated originators originated or is expected to originate 20% or more of the pool assets. We collect information on the originators from mortgage deal prospectus supplements of privately securitized residential mortgages between 2003 and 2007 from Bloomberg terminals. For each mortgage deal, its prospectus offers information on the performance-related characteristics such as the FICO score, loan-to-value ratio, and the collateral s pool size. More important for our analysis, it provides information on the composition of the mortgage loans from different originators. To link individual loans to a particular originator in a mortgage deal, we use the First American Corelogic LoanPerformance database. The Corelogic data provide the name of the original lender for each loan. We collect identity and affiliation information for the original lender of each loan to determine if the original lender is one of the originators for the mortgage deal or is affiliated with one of these originators. Using this information, we assign individual loans to the originators listed in the prospectus supplements. Then we perform a loan level analysis on the implication of the use of the origination stakes just below the disclosure threshold on a loan s performance under Reg AB. For mortgage deal performance, we examine the cumulative net loss that is defined as the sum of all of the losses of the deal s principal suffered up to a specific date divided by the total original balance of all of the mortgages. For the individual loan s performance, we use the delinquency within 24 months of the loan s origination, a standard measure used in an analysis of a loan s performance. Our investigation demonstrates a significant impact of the regulatory disclosure mandate on the use of origination stakes just below the disclosure threshold in mortgage deals. This in turn has an important implication for the performance of the securitized mortgages. First, we find that the percentage of deals with an origination stake from an originator and its affiliates that is less than 20% of the collateral pool the threshold for mandatory 2

4 disclosure increased significantly after Reg AB (we refer to a low origination stake in a mortgage deal as a low stake or LS, deals with at least one low stake as LS deals, and deals without a low stake as non-ls deals, hereafter). In particular, the percentage of LS deals more than doubled after Reg AB relative to before Reg AB. Second, LS deals have larger cumulative net losses than the non-ls deals, and the performance difference becomes significant only after Reg AB. Further, by introducing variables that represent more low-stake participation after Reg AB we find that deals of originators with more lowstake participation after Reg AB have larger cumulative net losses. Our findings are robust to the deal s cumulative net losses measured at different dates after Reg AB. Our analysis on loan level data provides further support that securitized loans show worse performance when their originators have more low-stake participation after Reg AB. It is worth noting that originators low-stake participation does not necessarily have any implication for the mortgage s performance if it is used to simply lessen the burdens associated with meeting SEC disclosure requirements. Rather, it is more likely driven by the intention to avoid scrutiny and to potentially withhold some adverse information on riskier loans. The latter implies that more low-stake participation in mortgage deals indicates more risker mortgages being securitized in those deals after Reg AB. Our paper offers the first empirical investigation on the impact of regulations on mortgage securitization directed at the practices of financial institutions participating in this market. It contributes to two strands of research. First, our paper contributes to the literature on the economic consequences of financial reporting and disclosure regulation. 3 In particular, our study relates to the papers that explore the unintended consequences of regulation changes, such as the going dark activities after SOX. 4 Our findings shed light on the effect 3 Many recent papers on disclosure regulation mainly focus on the impact of regulation changes under Regulation Fair Disclosure (Reg FD) and Sarbanes-Oxley Act (SOX). See Leuz and Wysocki (2008) for a comprehensive review of the related studies; Granja (2013) examines the effect of disclosure regulation in the commercial banking industry, among others. There are also studies examining disclosure regulation on OTC bulletin board firms (Bushee and Leuz (2005)) and JOBS Act (Chaplinsky, Hanley, and Moon (2014)). 4 For example, Gao, Wu, and Zimmerman (2009) provide evidence on the unintended consequences of Sarbanes-Oxley Act exemptions for small companies (i.e., firms with a public float of less than $75 million). They find that the size-based exemptions provide incentives for firms to stay small by curbing growth to 3

5 of mandatory disclosure on financial institutions and its implication for the performance of securitized assets issued by these financial institutions. Our evidence adds to this literature in a new and important setting and supports the view that understanding the firms potential responses and avoidance strategies is crucial when evaluating the costs and benefits of disclosure regulation and also when designing the rules in the first place (Leuz and Wysocki (2008)). Second, it adds to the fast growing literature that explores the relation between mortgage securitization and subprime loan performances, and represents the first study on directly assessing the impact of Reg AB on the MBS market. 5 The rest of the paper is organized as follows. Section 2 describes the information disclosure in Reg AB. Section 3 describes data and provides summary statistics. In Section 4, we present and discuss our empirical findings at the deal level. In Section 5, we provide the findings on the loan level analysis. And, Section 6 concludes. 2. Regulation AB The Securities and Exchange Commission defines asset-backed securities (ABS) as securities that are backed by a discrete pool of self-liquidating financial assets. The ABS market has experienced rapid growth in the last two decades. Bank One Capital Markets estimates that the annual issuance of US public non-agency ABS grew from $46.8 billion in 1990 to $416 billion in Thomson Media estimates that the new issuance for 2003 was at $800 billion. 7 Prior to the introduction of Reg AB, there were few SEC initiatives directly related to ABS. In this section, we describe the most relevant item of Reg AB on originator avoid crossing the compliance threshold. Leuz (2007) and Leuz, Triantis, and Wang (2008) show that going dark is associated with SOX-related events. 5 For studies on various issues related to mortgage loans and mortgage-backed securities, see, e.g., Mian and Sufi (2009), Keys, Mukherjee, Seru, and Vig (2009), Keys, Mukherjee, Seru, and Vig (2010), Purnanandam (2011), Ben-David (2011), Keys, Seru, and Vig (2012), Demiroglu and James (2012), He, Qian, and Strahan (2012), Nadauld and Sherlund (2013), Piskorski, Seru, and Witkin (2014), Demyanyk and Loutskina (2014), Griffin and Maturana (2014), Stanton, Walden, and Wallace (2014), Garmaise (2015), Rajan, Seru, and Vig (2015), among others. 6 See Bank One Capital Markets, Inc., 2004 Structured Debt Yearbook. 7 See Asset Securitization Report (pub. by Thomson Media Inc). 4

6 information disclosure for mortgage securitization and discuss the implications of the use of low origination stakes for the securitized loans performance. According to the SEC, Asset-backed securitization is a financing technique in which financial assets are pooled and converted into instruments that can be offered and sold more freely in the capital markets. In a basic securitization structure, a financial institution known as sponsor constructs a pool of financial assets, such as mortgage loans, which can be self-originated or acquired directly or indirectly through an affiliate. Securities that are backed by a pool of financial assets are then sold to investors by investment banks known as underwriters. Payment on the ABS depends primarily on the cash flows generated by the assets in the underlying pool, and the other rights designed to assure timely payment such as guarantees called credit enhancements. Asset-backed securities and their issuers differ from corporate securities and operating companies in that there is generally no business or management to describe in offering these securities. Instead, information about the transaction structure and the quality of the asset pool and servicing is often what is most important to investors. 8 According to the SEC, prior to Reg AB, many of its existing disclosure and reporting requirements, which were designed primarily for corporate issuers, did not elicit the information that is relevant for most ABS transactions. Regulation AB, which became effective in January 2006, thus represents a comprehensive treatment of ABS under the Securities Act of 1933 and the Securities Exchange Act of It consolidates and codifies the SEC s positions and industry practice which the SEC has done through no-action letters and the filing review process over time. Regulation AB covers four primary regulatory areas: Securities Act registration, disclosure, communications during the offering process, and the ongoing reporting under the Securities Exchange Act. 9 The new rules on disclosure represent the most dramatic changes 8 See Securities and Exchange Commission Asset-Backed Securities Proposed rule Release NOS ; See Securities and Exchange Commission Regulation AB Final Rule

7 in the ABS markets. Prior to Reg AB, there was no disclosure items specifically tailored to ABS. While eliminating unnecessary boilerplate and de-emphasizing unnecessary legal recitations on terms, Reg AB requires that issuers disclose information material to an ABS transaction such as the background, experience, performance, and roles of various transaction parties. Specifically on loan originators, Reg AB Item 1110 requires as an initial level of disclosure the identification of any originator or group of affiliated originators that originates, or expects to originate, 10% or more of the pool assets. If the originator originates or expects to originate 20% or more of the pool assets, then the regulation further requires disclosure of information regarding the size and composition of the originator s origination portfolio as well as information material to an analysis of the performance of the pool assets, such as the originator s credit-granting or underwriting criteria for the asset types being securitized. Thus, loan originators of 20% or more of the collateral pool represent an important disclosure threshold that did not exist prior to Reg AB. Our empirical investigation on the implication of the disclosure mandate in Reg AB for the securitized loans performance focuses on the change in the percentage of loans originated by lenders surrounding the disclosure threshold and the associated cross-sectional variations in the loan performance of these originators. Prior to Reg AB, the SEC s position on an ABS issuance was done through no-action letters. These positions and industry practice are consolidated and codified under Reg AB. The disclosure mandate on originators under Reg AB thus subject MBS issuers to more scrutiny and higher litigation risk. Consequently, after Reg AB, riskier loans could be placed in a mortgage deal at a stake below the threshold to avoid mandatory information disclosure. Therefore, we expect the number of LS deals to increase after Reg AB. This gives us the following testable hypothesis. Hypothesis 1: All else being equal, the fraction of low stake deals is higher after Reg AB than before Reg AB. There are two main motives behind using low origination stakes. The first is to reduce 6

8 the burdens to comply with SEC disclosure requirements. In the final ruling of the Reg AB, the SEC states that the initial proposed breakpoint for disclosure was 10%. However, several commenters argued that the disclosure threshold should be higher to lessen disclosure burdens. After incorporating the comments from various market participants, SEC adopted 20% as the breakpoint of disclosure in the final rule of Reg AB. 10 However, low stakes should have no impact on the deal s or the loan s performance under this motive. The second possible motive is to avoid disclosure of information on riskier loans that constitute a low stake in a deal. Before Reg AB, loan stakes did not have to be under 20% of the deal to avoid information disclosure because no mandatory disclosure threshold existed. After Reg AB, if riskier loans are included in a deal, they have to be kept under the 20% threshold to avoid scrutiny by investors and regulators. In other words, if avoiding information disclosure is the primary motive, we expect the performance to be worse for deals with more lowstake participation by these originators after Reg AB. This expectation leads to our second hypothesis. Hypothesis 2: All else being equal, the increased use of low stakes is associated with worse performance in securitized mortgages after Reg AB than before Reg AB. 3. Data description and summary statistics Our data come primarily from two sources: Bloomberg and First American Corelogic Loan- Performance. We collect information on deal characteristics from Bloomberg terminals. Bloomberg provides information on the mortgage originator and underwriter extracted from the deal prospectus supplements filed with EDGAR. 11 Our sample consists of publicly issued non-agency mortgage deals that were issued between 2003 and 2007, the period immediately preceding the financial crisis. Each deal in our database has detailed information on its characteristics at issuance. In the meantime, our loan level data consist of information on 10 Please refer to the final ruling for more details: 11 We use publicly issued non-agency mortgage deals due to data availability. 7

9 securitized mortgages constructed by Corelogic LoanPerformance. Corelogic provides information on loan origination dates, the mortgage loan pools, the identities of the securitizers, the MBS where the loans are placed, and on the borrowers and loan characteristics. We also construct variables from various sources on regional housing and economic conditions. Bloomberg reports the identities of the originators and the percentage of dollar principal that each of them originates for the deal. Not every deal provides origination information, thus we focus on a sample of 2,248 deals for which the origination information is available for our investigation. From the detailed origination information, we identify deals that have origination stakes in 10-20% or below 20% of the pool assets from an originator and its affiliates. Considering the disclosure requirements of Reg AB, we use 10-20% as the main measure of a low origination stake and use below 20% as an alternative measure. 12 We also calculate the aggregate percentage of low stake loans for each deal. Our deal level performance measure is the cumulative net loss rate measured as the sum of all of the losses of principal suffered up to September 2014 divided by the total original balance of all of the mortgages. As a robustness check, we also use the cumulative net loss rate measured as the sum of all of the losses of principal suffered up to December We use the deal s characteristics as control variables that comprise deal original collateral balance, an indicator for high issuer reputation following Griffin, Lowery, and Saretto (2014), the number of tranches, an indicator for higher than mean share of loans that have limited or no documentation in the collateral, weighted average FICO score, weighted average loan-tovalue (LTV) ratio, percentage of adjustable rate mortgages in the deal, an indicator for the presence of negative amortization, percentage of purchase loans (as opposed to refinancing), percentage of loans for single family houses, percentage of loans for owner-occupied houses, percentage of loans for equity take out, percentage of loans for refinance, and percentage of second lien loans. To construct a sample for the loan level analysis, we first identify the link between 12 Under Reg AB, originators contributing less than 10% to the collateral pool do not have to reveal their identities. This explicitly precludes using below 10% as a separate threshold in the analysis. 8

10 each securitized loan and its originator in a deal with multiple originators. The Corelogic database provides the name of the original lender for each loan whether it is a direct lender or a mortgage broker. We collect identity and affiliation information for the original lender of each loan to determine if the original lender is one of the mortgage deal s originators or is affiliated with one of the deal originators. When such a link can be made, we assign individual loans to the originators listed in the prospectus supplements. When the original lenders cannot be linked to any of the originators as is often the case with the loans acquired by the originators, we set the originator s information for these loans as missing and exclude them in our loan level analysis. We merge the deal level originator variables with the loan level data by originator name and deal number. The definitions for all of the variables at both the deal and the loan levels are described in the appendix. We begin our investigation with the deal level analysis. Table 1 reports the summary statistics for the deal level variables. For our full sample, the average deal cumulative net loss is 13.1% with a standard deviation of 12.4%. The deals with 10-20% (less than 20%) stakes from an originator and its affiliates are 18% (23%) of the sample. For the full sample, low stake loans account for 4.8% (5.7% for less than 20% stakes) of a collateral pool with a standard deviation of 13% (14% for less than 20% stakes). The highest percentage of aggregate low stake loans is 100%. In other words, in the extreme case, a deal could consist entirely of low stake loans. For deals with 10-20% (less than 20%) low stakes, the percent of low stake loans are on average 25.8% (24.5%) of the pool assets. Table 1 about here Table 2 reports the correlation coefficients on the main variables of interest at the deal level. The cumulative net loss is significantly and positively correlated with the presence of low stakes and the aggregate percentage of low stake loans in these deals. The results are very similar for both measures of low stakes: the percentage of loans in a deal within 10-20% or the percentage of loans in a deal below 20%. Consistent with the findings in the literature, the deal s cumulative net loss is negatively correlated with the average FICO score, which 9

11 suggests that the high credit worthiness of the borrower is associated with lower defaults. However, the deal s cumulative net loss is positively correlated with the average loan-tovalue ratio, percentage of adjustable rate mortgages, the presence of negative amortization loans, percentage of purchase loans, and the percentage of loans with a second lien due to the higher default risk associated with these characteristics. Our correlation estimate also suggests that the deal s cumulative net loss is negatively correlated with the percentage of single family home loans. Table 2 about here 4. The change in the use of low stakes and its implication for deal performance We start our empirical analysis by examining the impact of Reg AB on the use of low origination stakes. We then focus on investigating the implication of the change in the use of low stakes for the performance of securitized mortgages at the deal level The change in the use of low stakes under Reg AB We refer to a low stake as a group of loans from an originator and its affiliates in the collateral pool backing an MBS deal that is below the threshold necessitating mandatory disclosure by SEC under Reg AB. To test our hypotheses on the impact of Reg AB, we define a low stake as a group of loans from an originator and its affiliates that accounts for 10-20% of a collateral pool. As a robustness check, we also use an alternative definition for a group of loans that accounts for less than 20% of the collateral pool. In Figure 1, we plot the number and the percentage of the deals with low stakes in our sample period. The top panels present the plots for deals with 10-20% stakes before and after Reg AB. Both the number and the percentage of deals with low stakes show similar 10

12 pattern surrounding Reg AB. Specifically, the number of deals with low stakes experience a sharp increase from 121 before Reg AB (before 2006) to 303 after Reg AB (after 2006). In percentage terms, the increase is more than double from around 11% before Reg AB to 27% after Reg AB. Moreover, the bottom panels show that the percentage of deals with low stakes were relatively stable before Reg AB and the sharp jump occurred right after Reg AB became effective and then remained high. Figure 1 about here The increase in the percentage of low stakes in a deal before and after Reg AB is statistically significant. We apply probit regressions to evaluate this change by controlling for other factors that might affect the use of low stakes in deals by using the following specification. LS Deal = f(β Post Reg AB + Deal and Macro controls + Fixed effects). where LS Deal is a dummy variable that represents the presence of low stakes in a deal. Because changes in the house prices and the macroeconomic environment might have an impact on the mortgage s performance, we include additional control variables in our analysis. We calculate the pre-deal run-up in house prices for the representative geographic area using the house price index for the corresponding state reported by the Federal Housing Finance Agency (FHFA). Specifically, we compute the weighted average change in the house prices that is associated with a deal during the four quarters preceding the quarter the deal is issued in. Table 3 reports the marginal effects from the probit regression estimates. In column (1), the dependent variable is a dummy variable equal to one if a deal has at least one 10-20% stake from an originator and its affiliates and equals zero otherwise. We find that there is a very significant increase in probability that a deal involves at least one low stake after Reg AB. In particular, the probability that a deal involves at least one 10-20% low stake increases by 18% after Reg AB. We find a similar result in column (2) when we use the alternative 11

13 dependent variable to capture the presence of at least one low stake that is less than 20% of the collateral pool. Table 3 about here To demonstrate that the change in the use of low origination stakes occurs around the 20% threshold, we examine the difference between the percentage of mortgage deals with stakes just below 20%, at [10,20)%, [15,20)%, and [18,20)% and the percentage of mortgage deals with stakes just above 20%, at [20,30]%, [20,25]%, and [20,22]%, respectively, before and after Reg AB. We next apply a difference-in-differences test to the differentials in the percentages of the deals with just below 20% stakes and the percentages of deals with just above 20% stakes. We observe a differential of 7.5% for [10,20)% versus [20,30]%, 5.8% for [15,20)% versus [20,25]%, and 2.9% for [18,20)% versus [20,22]%. Our test results show that the increase in the percentage of deals with just below 20% stakes relative to those just above 20% stakes is statistically significant at the 1% test level for all three comparison brackets. We also evaluate the increase in the use of low stakes quantitatively, controlling for the other factors that might affect the use of low stakes as well as the lead underwriter fixed effect. For each deal, we create a dummy variable to represent the presence of low stakes in a bracket just below 20% and another dummy variable representing the presence of stakes in a bracket just above 20%. The difference between these two dummy variables is denoted as diffa20b where [A,20) is the bracket just below 20% and [20,B) is the bracket just above 20%. The combinations of {A,B} in our analysis include {10,30}, {15,25}, and {18,22}. Table 4 reports the results for the OLS estimation (panel A) and the ordered probit (panel B) regression analysis. Controlling for the deal s characteristics, the underwriters reputation, and macroeconomic variables, our OLS estimation shows that the change in the fraction of deals with stakes just above the threshold of 20% to just below the threshold increased by 15% from [20,30]% to [10,20)%, 8% from [20,25]% to [15,20)%, and 4% from [20,22]% to [18,20)%, respectively, from before Reg AB to after Reg AB. Given that the average fractions of LS deals 12

14 in the [10,20)%, [15,20)%, and the [18,20)% brackets are 10.8%, 5.2%, and 1.9%, respectively, before Reg AB, our estimates show that the percentage of deals with stakes just below the threshold increased by 136% for [10,20)%, 140% for [15,20)%, and 190% for [18,20)%, respectively, after Reg AB. The ordered probit regression analysis produces qualitatively similar results. For instance, the probability of mortgage deals with stakes just above the threshold to just below the threshold is higher by 7% for [20,30]% to [10,20)%, 6% for [20,25]% to [15,20)%, and 3% for [22,22]% to [18,20)%, respectively, after Reg AB. Relative to the average fractions of LS deals before Reg AB, these estimates represent increases of 65%, 115%, and 158% respectively after Reg AB. Table 4 about here 4.2. The implication of the change in the use of low stakes on deal performance Now that we have documented a significant increase in the use of low stakes after Reg AB, we explore its implications for securitized mortgage performance. On one hand, we expect the use of low stakes to have no implication for performance if it is primarily driven by the motive of reducing the burdens of SEC disclosure requirement. On the other hand, the use of low stakes might be associated with a deal s worse performance if it is used by the originators to avoid disclosure of information on riskier loans. To discern these two implications, we regress the deal s cumulative net loss on variables that capture the presence of low stakes and their interactions with a post-reg AB dummy variable. The inclusion of the interaction term allows us to assess whether the use of low stakes has an incremental effect after Reg AB rather than before Reg AB. Specifically, we use the following specification for our regression analysis: Cumulative net loss = α + β 1 Post Reg AB + β 2 LS Deal 13

15 +β 3 Post Reg AB LS Deal +Deal and Macro controls + Fixed effects. where LS Deal represents the presence of low stakes in mortgage deals. In addition to the LS deal measure defined above, we also use a continuous variable that captures the aggregate percentage of low stake loans in a deal. We do so for low origination stakes at the 10-20% level of a collateral pool and for low origination stakes below 20% of a collateral pool, respectively. We include house price change which we compute as the weighted average change in the house price associated with a deal from the quarter that the deal is issued in to the third quarter of The results are reported in Table 5. Columns (1) to (4) present the findings for the 10-20% stakes. It is clear that prior to Reg AB, the disclosure threshold has no significant implication for a deal s performance. However, after Reg AB, the LS deals have significantly worse performance. Specifically, the estimate in column (2) indicates that LS deals have 2.38 percentage points higher deal cumulative net loss. This represents 18% of the average cumulative net loss in our full sample period (2.38/13.12). When using the aggregate percentage of LS loans as the measure of low origination stakes, our estimate shows that a one standard deviation increase in this aggregate percentage of LS loans is associated with a 1.03% higher cumulative net loss. This represents an 8% average cumulative net loss for our full sample (1.03/13.12). Our results are robust if we redefine a low stake as less than 20% (see columns (5) to (8)). Table 5 about here To provide a placebo test for the effect of low stakes on the mortgage deal s cumulative net loss, we conduct a regression analysis that includes both the presence of 10-20% stakes and 20-30% stakes, a bracket just above the threshold. Table 6 reports the results of our analysis. For the dummy variable that represents the presence of low stake loans and the continuous variable that represents the percentage of low stake loans in a collateral pool, we 14

16 find that the 20-30% stakes have no significant effect on the deal s cumulative net loss. On the other hand, the 10-20% stakes are associated with a significantly larger cumulative net loss. More important, the larger cumulative net loss is concentrated in the deals with the 10-20% stakes after Reg AB. This finding highlights the implication of 10-20% stakes, an amount just below the disclosure threshold, post-reg AB for deal performance. Table 6 about here 4.3. Deal performance and increased use of low stakes Next, we investigate the implication of the increased use of low stakes on the mortgage deal s performance. For this analysis, we introduce a dummy variable to represent the increased use of low stakes for each originator before and after Reg AB. Specifically, for each originator, we count the number of deals in which the originator participated and compute the change in the percentage of its LS deals before and after Reg AB. In particular, we identify specific originators with more low-stake participation after Reg AB and refer to these originators as increased low stake (ILS) originators. 13 For each deal, we define the dummy variable as equal to one if at least one of the originators is an ILS originator and refer to such deals as ILS deals, and equals zero otherwise. Notationaly, for originator k, we define L pre k and L post k as the percentage of low stake deals this originator and its affiliates originate before and after Reg AB, respectively, and LS k = (L post k L pre k ) as the change in the percentage of its low stake deals before and after Reg AB. For mortgage deal j, we define the dummy variable ILS as follows: ILS = I(max k O j LS k > 0), where I( ) is an indicator function, and the O j represents the set of originators for deal 13 In practice, the use of low stakes can be decided by originators and/or deal sponsors. The latter can choose how many loans from each originator to be included in a deal and may lower the fraction of loans from a specific originator if its loan quality is not as good. For brevity, we simply use ILS originators to refer to the increased use of low stakes in deals. 15

17 j. The ILS originators have LS greater than zero, and ILS deals have at least one ILS originator. Similarly, we define a dummy variable ILS <20% to represent the increased use of low stakes when low stake is defined as below 20% of the collateral pool. As an alternative, we identify the originators as ILSH originators if they have an above average increase in the use of low stakes, and deals with at least one ILSH originator are referred to as ILSH deals. We define the dummy variable ILSH <20% to represent a larger than average increase in the low stake usage when the low stake is defined as below 20% of the collateral pool. For our sample of 149 originators, the average increase in the use of low stakes is 2.0% for 10-20% loans and 6.4% for below 20% loans after Reg AB. 14 We use the following specification for our analysis on the implication for the mortgage deal s performance associated with the increased use of low stakes. Cumulative net loss = α + β ILS +Deal and Macro controls + Fixed effects. Table 7 reports the results of our analysis. Our estimate shows that deals with increased low stake usage (the ILS deals) are associated with a 1.68% higher cumulative net loss than the non-ils deals (column (1)). A similar result is found when we use the dummy variable ILS <20% (column (2)). When using the alternative measure ILSH, we find that ILSH deals are associated with a 1.94% higher cumulative net loss than other deals (column (3)). A similar effect is also found when we use the dummy variable ILSH <20% (column (4)). Table 7 about here To demonstrate that our findings are robust to the deal s cumulative net loss measured at different dates, we construct an alternative measure of cumulative net loss at December 2012 that is scaled by the original collateral balance. The results based on this cumulative net loss 14 The ILS originators participated in more deals after Reg AB than the non-ils originators. This is why there is a sharp increase in LS deals after Reg AB. 16

18 variable are reported in Table 8, Table 9, and Table 10. Consistent with our earlier findings, the implication of the increased use of low stakes for the mortgage deal s performance remains significant and qualitatively similar. This finding indicates that the adverse effect associated with the use of low stakes in mortgage deals is not specific to the cumulative net loss measured at different dates. It reflects the response of MBS securitization to changes in the regulation, and this change has important implications for mortgage deal performance. Table 8, Table 9, and Table 10 about here 4.4. The implication of the change in low stakes for deal yield spreads and credit enhancement One question is whether the higher cumulative net loss of mortgage deals with increased low stake usage is reflected in the deal s initial yield spreads and credit enhancement. This is relevant for how investors evaluate the implication of the disclosure mandate for credit risk protection and deal pricing. For credit enhancement, we focus on the subordination that is measured as the percentage of the face value of trust securities not rated AAA by Moody s or Standard & Poor s at the deal s close. For deal yields, we use the initial average yield spread for all of the securities issued by the trustee of the mortgage deals. This is the difference between the average yield of all of the securities issued by the trustee weighted by the face value of the securities and the yield on the 10-year Treasury bond. The former is calculated using the standards of the Bond Market Association and reported by Bloomberg. We use the same specifications for our yield spread and credit enhancement as for the cumulative net loss. Table 11 reports the results of our regression analysis. Panel A shows that LS deals do not offer significantly different yields from the non-ls deals. The Reg AB does not change this result, which is true for both 10-20% stakes (column (1)) and below 20% stakes (column (2)). Furthermore, the presence of low stakes has a significant impact on credit enhancement 17

19 measured by mortgage deal subordination. Under both measures of LS deals, the presence of low stake deals is associated with a higher average subordination before Reg AB and a lower subordination after Reg AB (columns (3) and (4)). This shows that LS deals actually offer less credit enhancement after Reg AB than before, in contrast to their worse performance after Reg AB. Combining the findings on deal yield spreads and subordination provides evidence that suggests investors might not have impounded the higher risk associated with the use of low stakes into the mortgage deal s yields and credit enhancement after Reg AB. Table 11 about here Panel B reports the results of the analysis on whether the increased use of low stakes after Reg AB versus before Reg AB is reflected in the deal s yields and credit enhancement. We find no evidence that the presence of ILS or ILSH is reflected in the yield spreads or the subordination. Overall, our results on Reg AB have two important implications. First, the disclosure threshold has increased the use of low stakes in mortgage deals after Reg AB. Second, the increase in the use of low stakes is associated with higher deal cumulative net loss. These findings suggest that the increased used of low stakes are likely driven by the need to avoid scrutiny and to potentially withhold some adverse information on riskier loans. 5. Loan performance and the change in the use of low stakes In this subsection, we investigate the implication of the change in the use of low stakes for the mortgage s performance at the loan level. We first directly test whether the change in low stake usage ( LS) by originators is associated with a different quality in the loans that they originated. 15 A consistent result from this analysis will reinforce our deal level findings 15 We use LS for each originator in the loan level regressions because we no longer need to aggregate LS at the deal level. 18

20 since loan level analysis enables us to directly compare loans from originators with different low stake usage before and after Reg AB. Furthermore, we investigate whether the effect of LS on the loan s quality is stronger for low origination stakes after Reg AB. It is possible that loans in different stake sizes have different quality. We control this characteristic by including stake sizes in our regression. Following the definition in Table 7, we compute the change in the percentage of LS deals before and after Reg AB, that is, LS = (L post L pre ) for the originator of the loan under consideration. Merging the deal level information on the originator with the loan level data and excluding missing observations, we have more than three and a half million loans in 1,603 deals. Following the standard practice in the literature, we use the securitized loan s delinquency, defined as 60 days or more past due within 24 months of the origination as the performance measure in the loan level analysis. To control for the macroeconomic environment in the loan level data, we compute the appreciation in house prices over the 24 months after origination by using the house price index for the borrower s metropolitan statistical area (MSA) reported by the Federal Housing Financing Agency (FHFA). We also compute the change in the state-level unemployment rate over the 24 months after origination using data reported by the Bureau of Economic Analysis and collect the median household income in 1999 for the borrower s zip code as reported by the US Census Bureau in Additionally, we include the credit spread and the 10-year Treasury yield as macro control variables. To control for the different qualities of loans securitized at different time periods, we include deal s issuing semester fixed effect. The fixed effect mitigates both the vintage effect and other macroeconomic changes in the sample period not captured by our macro control variables. In Table 12, we report the summary statistics for the loan level variables for the full sample and subsamples of loans in the 10-20% and 20-30% stakes respectively. We use loans in the 20-30% stakes for comparison due to their comparable stake size. We observe that the sample averages for these variables are close between the whole sample and subsamples, 19

21 and even closer between the two subsamples. Table 12 about here We take two steps in our loan level analysis. In the first step, we examine whether the loans from originators with more low-stake participation after Reg AB are more likely to be in delinquency. We expect that the loans from the originators with more low-stake participation after Reg AB are riskier than the loans by other originators. This expectation suggests that the loan delinquency will increase in LS. In the second step, we investigate whether the effect of LS on loan delinquency is particularly strong for the 10-20% stakes after Reg AB. We use loans at the 20-30% stake level as a control group because of their close proximity to the 10-20% stake level. In the regression specification, we interact LS with a dummy variable that indicates whether the stake size is 10-20% and expect the interaction term to be significantly positive if the 10-20% stake size is being used to avoid disclosure on riskier loans as opposed to the 20-30% stake size. We conduct this analysis separately for the before and after Reg AB subsamples of loans and expect the interaction term to be significantly positive after Reg AB and insignificant before Reg AB. Table 13 reports the marginal effects from the probit regression for the whole sample with Stake size (column (1)), the whole sample with both Stake size and LS (column (2)), the subsample of loans in the 10-20% stakes (column (3)), and the subsample of loans in the 20-30% stakes (column (4)). Our estimation shows that for the full sample the loans from larger stake sizes have lower delinquency. This finding makes it necessary to control for the stake size in our subsequent analysis. As expected, LS is positively associated with delinquency when controlling for the stake size and various loan level controls. The economic magnitude is significant in that the loans from originators with a 10% more lowstake participation after Reg AB have a 0.5% higher delinquency rate. For the subsample of loans in the 10-20% stakes, we observe the same effect for LS. On the other hand, we find no such effect for loans in the 20-30% stakes, demonstrating a striking contrast around the 20% threshold. 20

22 Table 13 about here Next we explicitly test whether the effect of the increase in the low stake usage is particularly strong after Reg AB. This test will provide evidence on the 10-20% stake size being used to avoid scrutiny after Reg AB. Table 14 presents the results of a probit regression on loans from the 10-30% stake group for the pre-reg AB subsample ( ) and the post-reg AB subsample ( ). Our estimation for the pre-reg AB subsample shows that the variable LS is positively associated with loan s delinquency, yet statistically insignificant for the pre-reg AB period. Also, there is no significant difference in the effect of LS between the 10-20% stake group and the 20-30% stake group, which is consistent with our expectation that there should not be a jump at the 20% threshold before Reg AB. In contrast, our estimation for the post-reg AB subsample shows that the effect of LS is much stronger for loans in the 10-20% stakes than those in the 20-30% stakes. However, LS is negatively associated with delinquency for loans in the 20-30% stakes. This is consistent with the explanation that originators use the 10-20% stake size to avoid disclosure on riskier loans. Table 14 about here 6. Conclusion How to design and implement effective regulation has received widespread attention following the 2007 to 2008 subprime mortgage crisis. Very little is known about the impact of regulation on the non-agency mortgage backed securities (MBS) market implemented during the height of the housing boom right before the crisis. Even less is known about the effects of these regulations on the market participants and its resulting economic impact. In this paper we fill this void. One of the most important aspects of Reg AB is the disclosure mandate on lenders who contribute loans to collateral pools backing MBS. Specifically those originators who con- 21

23 tribute more than 20% of the loans in a collateral pool are required to provide detailed information material to the investors analysis of the collateral assets. The purpose of this requirement is to encourage transparency and therefore accountability. Using data on mortgage deals constructed before and after Reg AB, we find that the disclosure requirement might have been circumvented in the cases where riskier loans were included in the mortgage pool. Our loan level analysis provides supporting evidence on our deal level findings. Overall, our study on how these regulations change the market participants behavior and the ensuing economic impact can shed light on future research and the policy-making directed at the asset-backed securities markets. Coincidentally, the recently adopted Regulation AB II has tightened the disclosure requirement on originators that originate less than 10% of the pool assets. 16 There are also other aspects of these regulations such as communication in the offering process and on-going reporting that could potentially change these markets; we leave those topics for future research. 16 Regulation AB II adopted on August 27, 2014, requires that if the cumulative amount of pool assets originated by parties other than the sponsor or its affiliates is more than 10% of the total pool assets, then any originator that originates less than 10% of the pool assets also must be identified in the prospectus. 22

24 References Ben-David, Itzhak, 2011, Financial constraints and inflated home prices during the real estate boom, American Economic Journal: Applied Economics pp Bushee, Brian J, and Christian Leuz, 2005, Economic consequences of sec disclosure regulation: Evidence from the otc bulletin board, Journal of Accounting and Economics 39, Chaplinsky, Susan, Kathleen Weiss Hanley, and S Katie Moon, 2014, The jobs act and the costs of going public, Working Paper. Demiroglu, Cem, and Christopher James, 2012, How important is having skin in the game? originator-sponsor affiliation and losses on mortgage-backed securities, Review of Financial Studies 25, Demyanyk, Yuliya S, and Elena Loutskina, 2014, Mortgage companies and regulatory arbitrage, Working Paper. Gao, Feng, Joanna Shuang Wu, and Jerold Zimmerman, 2009, Unintended consequences of granting small firms exemptions from securities regulation: Evidence from the Sarbanes- Oxley Act, Journal of Accounting Research 47, Garmaise, Mark J, 2015, Borrower misreporting and loan performance, The Journal of Finance 70, Granja, Joao, 2013, Disclosure regulation in the commercial banking industry: Lessons from the national banking era, Working Paper. Griffin, John, Richard Lowery, and Alessio Saretto, 2014, Complex securities and underwriter reputation: Do reputable underwriters produce better securities?, Review of Financial Studies 27,