NBER WORKING PAPER SERIES ASSET QUALITY MISREPRESENTATION BY FINANCIAL INTERMEDIARIES: EVIDENCE FROM RMBS MARKET Tomasz Piskorski Amit Seru James Witkin Working Paper 18843 http://www.nber.org/papers/w18843 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 February 2013 We are grateful to Equifax and BlackBox Logic for their data which were invaluable for the analysis in this paper. We thank Charlie Calomiris, John Cochrane, Gene Fama, Chris Mayer, Lasse Pedersen, Amir Sufi, and Luigi Zingales as well as seminar participants at Columbia Business School for valuable comments. We thank Ing-Haw Cheng and Andrew Ellul for graciously sharing their data on incentive compensation and the risk management index. All errors are our own. Piskorski thanks the Paul Milstein Center for Real Estate at Columbia Business School for financial support. Seru thanks the Initiative on Global Markets at Booth for financial support. Contact Authors: Tomasz Piskorski ([email protected]) and Amit Seru ([email protected]) The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peer- reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. © 2013 by Tomasz Piskorski, Amit Seru, and James Witkin. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. Asset Quality Misrepresentation by Financial Intermediaries: Evidence from RMBS Market Tomasz Piskorski, Amit Seru, and James Witkin NBER Working Paper No. 18843 February 2013 JEL No. G14,G21,G24,G28,K22 ABSTRACT We contend that buyers received false information about the true quality of assets in contractual disclosures by intermediaries during the sale of mortgages in the $2 trillion non-agency market. We construct two measures of misrepresentation of asset quality – misreported occupancy status of borrower and misreported second liens – by comparing the characteristics of mortgages disclosed to the investors at the time of sale with actual characteristics of these loans at that time that are available in a dataset matched by a credit bureau. About one out of every ten loans has one of these misrepresentations. These misrepresentations are not likely to be an artifact of matching error between datasets that contain actual characteristics and those that are reported to investors. At least part of this misrepresentation likely occurs within the boundaries of the financial industry (i.e., not by borrowers). The propensity of intermediaries to sell misrepresented loans increased as the housing market boomed. These misrepresentations(cid:3)are costly for investors, as ex post delinquencies of such loans are more than 60% higher when compared(cid:3)with otherwise similar loans. Lenders seem to be partly aware of this risk, charging a higher interest(cid:3)rate on misrepresented loans relative to otherwise similar loans, but the interest rate markup on misrepresented(cid:3)loans does not fully reflect their higher default risk. Using measures of pricing used in the literature,(cid:3)we find no evidence that these misrepresentations were priced in the securities at their issuance. A(cid:3)significant degree of misrepresentation exists across all reputable intermediaries involved in sale of(cid:3)mortgages. The propensity to misrepresent seems to be largely unrelated to measures of incentives(cid:3)for top management, to quality of risk management inside these firms or to regulatory environment(cid:3)in a region. Misrepresentations on just two relatively easy-to- quantify dimensions of asset quality could result in forced repurchases of mortgages by intermediaries up to $160 billion. Tomasz Piskorski James Witkin Columbia Business School Columbia Business School 3022 Broadway, Uris Hall 810 [email protected] New York, NY 10027 [email protected] Amit Seru Booth School of Business University of Chicago 5807 South Woodlawn Avenue Chicago, IL 60637 and NBER [email protected] Section I: Introduction Market rules and regulations that require disclosure of information and prohibit misleading statements on the financial products being manufactured by intermediaries play an important role in the functioning of capital markets (Akerlof, 1970). However, the nature of intermediation has changed dramatically over the past decade–with the introduction of more agents in the supply chain of credit (Loutskina and Strahan, 2009; Keys et al. forthcoming; Nadauld and Sherlund, 2013) --- potentially weakening the ability of existing market arrangements and regulatory oversight in ensuring truthful disclosure of asset quality. This concern has gained momentum in the aftermath of the crisis, with a precipitous decline in the value of supposedly safe securities as well as large investor losses (Acharya, Schnabl, and Suarez, forthcoming).1 This paper adds to this debate, and more generally to one on the design of market rules and future financial regulation, by quantifying the extent to which buyers may have received false information about the true quality of these assets by sellers of these securities and by exploring factors that may have moderated such behavior by financial intermediaries. Our study focuses on misrepresentation of asset-quality that consist of securities collateralized by residential mortgages that are originated without government guarantees (non-agency RMBS) -- representing a $2 trillion market in 2007.2 These misrepresentations are not instances of the usual asymmetric information problem in which the buyers know less than the seller. Rather, we contend that they are instances where, in the process of contractual disclosure by the sellers, buyers received false information on the characteristics of assets.3We identify misrepresentations by comparing the characteristics of mortgages in the pool that were disclosed to the investors at the time of sale with actual characteristics of these loans at the same time and show that such misrepresentations constitute a significant proportion of the loans. We investigate if lenders and investors were aware of such behavior and assess characteristics of intermediaries involved in the sale of assets, such as incentives for managers, that could have moderated this behavior. In doing so we demonstrate the limits of existing market and regulatory arrangements in preventing such behavior in the capital market. The RMBS securitization process involves aggregating mortgages into loan trusts, either through direct origination or indirect acquisition, and using their underlying cash flows to issue securities. The sale of these securities is organized by underwriters who, as part of this process, must collect and verify information regarding the quality of the underlying collateral backing these securities. 1 Critics of imposing more regulation argue that reputational concerns of large, well-established financial intermediaries would prevent such violations of investors’ rights. In contrast, proponents of increased regulation argue that intermediaries were able to exploit investors despite their reputation (and existing regulation). 2 See Agarwal and Ho (2007), and Keys et al. (2012), and the Federal Reserve Board March 2009 Statistical Release (http://www.federalreserve.gov/econresdata/releases/mortoutstand/mortoutstand20090331.htm) 3 Akerlof (1970) also considers settings where quality of goods can be potentially misrepresented by sellers. 1 The underwriters in this market are large, reputable financial intermediaries who are considered to be more sophisticated than the buyers in this market, which are typically institutional investors such as pension funds, mutual funds, and insurance companies. Thus, given the lighter regulatory oversight relative to public capital markets that are open to retail investors, this market serves as a good laboratory to study whether existing market mechanisms based on implicit and explicit contracting are sufficient to safeguard investors’ rights. In general, several factors make it difficult to perform a systematic study of the nature and determinants of asset misrepresentation. For instance, most studies of corporate fraud define fraud as events in which the rights of the shareholders of a firm are violated by its managers. However, such violations of investors’ rights involve firms with varied production technologies and in different time periods, making such comparisons difficult. Moreover, most of these studies focus only on the cases where fraudulent behavior was ex-post discovered. We focus on the RMBS market and study the disclosure activity related to sale of assets by intermediaries, which circumvents some of these challenges. Since residential mortgages are fairly standardized and have a simple structure, comparisons of disclosure activity are relatively easier. In addition, detailed histories of such activity by financial institutions involved in sale of RMBS are available, allowing us to study asset misrepresentation by these firms over time as well as in the cross- section. Finally, our methodology will identify asset misrepresentation regardless of whether or not it is ex-post discovered by the investors. We start our analysis by combining two sources of data, loan-level data on mortgages from BlackBox with data on consumer credit files from Equifax, to construct two measures of misrepresentation regarding the quality of mortgages backing the RMBS pools. The mortgage- level data include characteristics of loans that were disclosed to the investors at the time of asset sale. The consumer credit data, which were not available to investors at the asset sale date, contains the actual characteristics of loans at the same time. The matching between the two datasets is done by the credit bureau. Using the combined dataset we identify two, relatively easy-to-quantify, dimensions of asset quality misrepresentation by intermediaries during the sale of mortgages. The first misrepresentation concerns loans that are reported as being collateralized by owner-occupied properties when in fact these properties were owned by borrowers with a different primary residence (e.g., a property acquired as an investment or as a second home). The second form of misrepresentation concerns loans that are reported as having no other lien when in fact the properties backing the first (senior) mortgage were also financed with a simultaneously originated closed-end second (junior) mortgage. The advantage of looking at this second type of misrepresentation is that it clearly indicates that the distortion of information occurred within the boundaries of the financial industry, as some institutions (e.g., a lender financing a second lien loan) had to be aware of the presence of such higher liens. Our hope in both instances is to 2 directly identify asset misrepresentations relative to sellers’ contractual disclosure. This differentiates our study from the literature that infers the decisions of agents in the supply chain of credit (e.g., screening by lenders) based on outcome variables such as delinquencies (e.g., Keys et al. 2010). A high-quality match between the dataset that contains actual loan characteristics and that containing characteristics reported to investors is critical for constructing measures of misrepresentation. We conduct extensive analysis that suggests that the misrepresentations we identify are not likely to be an artifact of matching error between the datasets that were merged by the credit bureau. In particular, we follow a conservative approach and conduct our analysis only on loans that were matched across the datasets with highest confidence level by the credit bureau. We confirm the high quality of this merge when we conduct an independent analysis of various data fields and find that, for the vast majority of cases, these fields are the same across the two databases. More importantly, as we will demonstrate, the misrepresented loans we identify strongly predict ex post delinquencies, further confirming our assertion that these loans misreport meaningful information about asset quality. This is not the pattern we find for a placebo test where we use few incidences of likely data errors across datasets to construct a measure of pseudo misrepresentation.4 Finally, we cross-validate the merge quality using an internal database from a large subprime lender. We find that almost all loans we identified as having second liens that are not disclosed to investors do indeed have such liens reported in the bank's internal data. Using our measures we find a significant degree of misrepresentation of collateral quality across non-agency RMBS pools. More than 6% of mortgage loans reported for owner-occupied properties were given to borrowers with a different primary residence, while more than 7% of loans (13.6% of loans using a broader definition) stating that a junior lien is not present actually had such a second lien. Alternatively put, more than 27% of loans obtained by non-owner occupants misreported their true purpose and more than 15% of loans with closed-end second liens incorrectly reported no presence of such liens. The propensity of banks to sell loans that misrepresented asset quality increased as the housing market boomed. This pattern is consistent with the model of Povel, Singh, and Winton (2007). Overall, more than 9% of loans had one of these misrepresentations in our data. Note, however, that because we look only at two types of misrepresentations, this number likely constitutes a conservative, lower-bound estimate of the fraction of misrepresented loans. 4 The placebo test uses the notion that incorrectly merged records should not have such a strong relationship with the subsequent adverse performance of loans. In particular, we focused on the few records in our database for which the loan balance of the first mortgage does not exactly match across the two databases. We find that the subsequent performance of such loans in terms of their default risk is similar to that of loans with perfectly matching balances across the two databases (see Appendix A for more details). 3 We find that these misrepresentations have significant economic consequences. In particular, loans with misrepresented borrower occupancy status have about a 9.4% higher likelihood of default (90 days past due on payments during the first two years since origination), compared with loans with similar characteristics and where the property was truthfully reported as being the primary residence of the borrower. This implies a more than 60% higher default rate relative to the mean default rate of owner-occupants during our sample period. Similarly, loans with a misrepresented higher lien -- which we find are likely to be fully documented loans -- have about a 10.1% higher likelihood of default compared with loans with similar characteristics and no higher lien. This is again a large increase, about 70%, relative to the mean default rate of loans without higher liens. Thus, our results indicate the same pattern for the two measures, which confirms that the misrepresentations we identified capture economically meaningful information about asset quality. Because of their substantially worse performance, misrepresented loans account for more than 15% of mortgages that defaulted in our sample, a higher share than their proportion in the overall sample (about 10%). Next, we investigate if misrepresentations on the two dimensions, given that they are associated with higher mortgage defaults, appear to be priced in by lenders and investors. To do this we first assess whether lenders charged higher interest rates for loans with misrepresented quality. We find that mortgages with misrepresented owner occupancy status are charged interest rates that are higher when compared with loans with similar characteristics and where the property was truthfully reported as being the primary residence of the borrower. Similarly, interest rates on loans with misrepresented second liens are generally somewhat higher when compared with loans with similar characteristics and no second lien. Given the increased defaults of these misrepresented loans, this suggests that lenders were partly aware of the higher risk of these loans. Strikingly, however, we find that the interest rate markups on the misrepresented loans are much smaller relative to loans where the property was truthfully disclosed as not being primary residence of the borrower and as having a higher lien. This suggests that relative to prevailing interest pricing of that time, interest rates on misrepresented mortgages did not fully reflect their higher default risk. We also examine whether pools with a higher incidence of misrepresented assets were perceived to be of lower quality by investors. For that purpose, we investigate the relation between misreporting in a pool and the measures of pricing used in the literature, i.e., imputed average yield of the pool as well as the subordination level that protects its AAA-rated tranches.5 Using these measures, we find little evidence that investors were compensated for a greater risk of securities involving a higher share of misreported assets. These results suggest that RMBS 5 As is the case in the literature, we do not have access to data on actual prices paid by the investors at the time of pool issuance (see Keys et al. 2012). Instead, consistent with literature as well as industry practice, we use proxies of prices such as average balance-weighted coupon. These proxies are the same as those used by prior work to investigate whether risk of collateral was reflected in investor prices (e.g., Faltin-Traeger, Johnson, and Mayer 2010; Demiroglu and James, 2012; He, Qian, and Strahan, forthcoming). 4 investors had to bear a higher risk than they might have perceived based on the contractual disclosure. As a result, investors could argue that the ex ante value of the securities with misrepresented assets that were sold to them was less than the price paid, and truthful disclosure of the characteristics of the assets could have prevented some of their losses. Assuming that our estimates are broadly applicable to the entire stock of outstanding non-agency securitized loans just prior to the crisis, enforcement of contractual guarantees by investors in response to these misrepresentations could result in forced repurchases of mortgages with combined balance of up to $160 billion.6 We also investigate the variation in asset misrepresentation across underwriters. We demonstrate that there is substantial heterogeneity in the extent of these misrepresentations across underwriters. The propensity to misrepresent is largely unrelated to measures of incentives for top management and to quality of risk management inside these firms. While misrepresentations appear to be less prevalent among commercial banks, underwriters with more RMBS experience, and underwriters with more high-powered incentives given to its top management and better internal management, none of these associations is statistically significant. Importantly, a significant degree of misrepresentation exists across all reputable intermediaries in our sample. Finally, we also find no relation between share of misrepresented loans and leniency of regulatory environment in a region. Our earlier analysis suggests that lenders were partly aware of the risk of misrepresented loans since it was reflected in the interest rates charged on these loans. In the last part of the paper we examine where in the supply chain of credit--i.e., at the borrower, lender, or underwriter level-- these misrepresentations likely took place. This analysis requires significantly richer micro data than are available to us, but we are able to provide some evidence that suggests that part of the misrepresentation occurs at the level of the financial institution. In particular, we use an internal database of a large subprime lender in which we observe the data that were available to the lender as well as data that were disclosed to the investors. To the extent that practices in this subprime lender are representative, our findings suggest that misrepresentation concerning owner-occupancy status was made early in the origination process, possibly by the borrower or broker originating a loan on behalf of the lending institution. In contrast, the lender was aware of the presence of second liens, and hence their misreporting likely occurs later in the supply chain.7 This last result is also consistent with our earlier evidence that loans misrepresented on the dimension of higher liens are more likely to be fully documented. Our findings contribute to the debate on the recent crisis. We provide evidence showing that 6 Asset misrepresentations by intermediaries could induce other costs beyond those directly incurred by the investors. For instance, these misrepresentations could have also led to misallocation of scarce investor capital among the proposed real investment projects in the economy. 7 Our review of mortgage deeds records from the FirstAmerican database indicates that both first-lien and closely situated second-lien mortgages on a given property were commonly financed by the same lending institution. 5 investors bought assets that not only proved to be ex post risky but may also have been, at least in part, ex ante misrepresented by financial intermediaries. More broadly, our results suggest that the current market arrangements -- based on reputational concerns and explicit incentives -- may have been insufficient to prevent and eliminate misrepresentations of asset quality in a large capital market. These findings suggest that a critical inspection of the protection of investors in other capital markets, especially those with more passive investors like the high-grade investment debt market, may be warranted. Our findings resonate well with studies that suggest that the existing regulatory framework may have also been insufficient in preventing such behavior by various actors in the supply chain of credit (see Keys et al. 2009; Keys et al. forthcoming, and Agarwal et al. 2012). Our work most directly relates to the recent empirical literature on the housing and financial crisis.8 In this literature, our research on the extent and consequences of loan misrepresentation in the mortgage securitization market is closest to studies that attempt to infer decisions of borrowers and financial institutions (e.g., lenders) using data based on outcome variables such as delinquencies (see Keys, Mukherjee, Seru, and Vig (2010), Ben-David (2011)). It is also related to Jiang, Nelson, and Vytlacil (2012) that provides evidence of misrepresentation of income for low documentation borrowers, and Garmaise (2012) that provides evidence that suggests that borrowers overstated their assets and that loans granted to such borrowers had higher default rates. In contrast to these papers, we attempt to directly identify misrepresentation of asset quality rather than infer them from outcomes, and we do so across the main players in the market. Our analysis is also closely related to recent research by Haughwout et al. (2011) who employ credit bureau data to identify properties acquired by real estate investors. Like our paper, Haughwout et al. also identify some proportion of such buyers as misreporting their intentions to occupy a property. Their focus is on explaining the role such buyers played in the recent boom and bust in the housing market.9 In contrast, our focus is on identifying the consequences of such misrepresentations on additional defaults and on understanding whether these misrepresentations were priced by investors and lenders as well as which underwriter characteristics relate to such behavior. Our work is also related to a large theoretical and empirical literature in accounting and finance that investigates various aspects of corporate fraud (e.g., Richardson et al. 2002; Burns and Kedia 2006; Kedia and Philippon 2009; Povel, Singh, and Winton, 2007; Efendi et al. 2007; 8 See Mian and Sufi 2009, forthcoming; Mayer et al. 2009; Loutskina and Strahan, 2009 and 2010; Campbell et al., forthcoming; Keys, Mukherjee, Seru, and Vig 2010; Keys, Seru, and Vig, forthcoming; Rajan, Seru, and Vig, forthcoming; Piskorski, Seru, and Vig 2010; Melzer 2010; Berndt, Hollifield, and Sandås, 2010; Mian and Sufi 2011; Agarwal et al. 2011; Demyanyk and Van Hemert 2011; Demiroglu and James (2012), Nadauld and Sherlund, 2013; Purnanandam, 2012; Acharya, Schnabl, and Suarez, forthcoming and He, Qian, and Strahan, forthcoming. 9 See also related research by Chinco and Mayer (2012) who use deeds records to identify distant real estate investors and study their role in fueling the recent boom in house prices. 6 Dyck et al., forthcoming; Wang, Winton, and Yu 2010; Wang forthcoming). It is also connected to the work in law and economics that focuses on securities fraud litigation (e.g., Choi et al., 2009 Griffin et al. 2001). Finally, our analysis is also related to recent studies that explore the determinants of risk taking undertaken by financial intermediaries (e.g., Cheng et al., 2010 and Ellul and Yerramilli, forthcoming). We contribute to these areas by providing systematic evidence on the characteristics of asset misrepresentation -- in the cross-section of financial intermediaries and across time -- in a large debt market. Section II: Information Disclosure on Collateral Quality of RMBS The vast majority of mortgages originated in the United States are not held by the banks that originated them but are instead securitized and sold as securities to investors. In this paper we focus on residential mortgage-backed securities (RMBS) that are collateralized by mortgage loans that are originated without government guarantees (non-agency RMBS). This sector was a significant portion of the overall mortgage market and reached more than $2 trillion of securities outstanding (Keys, Piskorski, Seru, and Vig, forthcoming). In our analysis we are concerned about the misrepresentation of collateral backing RMBS. When a pool of mortgages is sold to investors, the underlying characteristics and quality of these mortgages are disclosed in the prospectus of the pool in the section related to “representations and warranties.” This disclosure usually contains information about risk-relevant variables, such as loan-to-value ratios of loans in the pool, their interest rates, borrowers’ credit scores, and the occupancy status and location of the properties backing the mortgages in the pool. Our analysis focuses on two misrepresentations that our data enable us to identify. The first misrepresentation concerns the occupancy status of the property backing the loan because, as we explain later, such status can impact how much risk is associated with the mortgage. In particular, we are interested in identifying loans reported as being collateralized by owner-occupied properties when in fact these properties were owned by borrowers with a different primary residence (e.g., a property acquired as an investment or as a second home). We now provide an example of a contractual disclosure that refers to asset quality on this dimension -- namely, the prospectus supplement for Series 2006-FF15 states: The prospectus supplement will disclose the aggregate principal balance of Mortgage Loans secured by Mortgaged Properties that are owner-occupied.10 The second case of asset-quality misreporting we consider concerns loans that are reported as having no associated higher liens when in fact the properties backing the first (senior) mortgage were also financed with a simultaneously originated second (junior) mortgage. Prospectus 10 The prospectus supplements will also usually describe the frequency of loans for second homes or investor properties. For example, the prospectus supplement for the Wells Fargo Mortgage Backed Securities Series 2007–8 states, “Approximately 0.02% (by aggregate unpaid principal balance as of the cut-off date) of the mortgage loans in the mortgage pool are expected to be secured by investor properties.” 7 supplements commonly make statements regarding the total value of all liens on the collateralized property. As we explain, omitting information on the junior mortgages on a property will also understate the true risk associated with the pool collateral. Again, it is useful to provide an example of such contractual disclosure. The prospectus supplement of WFMBS Series 2007-7 states: For purposes of Appendix A, the “Combined Loan-to-Value Ratio” or “CLTV” is the ratio, expressed as a percentage, of (i) the principal amount of the Mortgage Loan at origination plus (a) any junior mortgage encumbering the related Mortgaged Property originated by the Sponsor or of which the Sponsor has knowledge at the time of the origination of the Mortgage Loan or (b) the total available amount of any home equity line of credit originated by the Sponsor or of which the Sponsor has knowledge at the time of the origination of the Mortgage Loan, over (ii) the lesser of (a) the appraised value of the related Mortgaged Property at origination or (b) the sales price for such property.11 As we explain in detail in Section IV, in order to identify such a misrepresentation we use loan- level data concerning the characteristics of mortgages in the pool that were disclosed to the investors at the time when the pool was sold. This data contains the most detailed information that is contractually provided to investors by asset sellers and pertains to the quality of loans that back the mortgaged-backed securities. Then, using another proprietary matched dataset that contains information on the actual characteristics of these loans, we verify whether the information disclosed to the investors was accurate. Thus, it is important to keep in mind that our analysis is from the perspective of an investor who used detailed loan level information available from the trustee at the time of sale of the pool. We discuss how this assumption could impact our inferences in Section VIII. The information contractually disclosed to investors allows them to assess the risk of the security. Previous research has shown that mortgages with a higher loan amount relative to property value and loans backed by non-owner-occupied properties are often associated with greater default risk (see Mayer et al. 2009). Thus, both of these misrepresentations could imply that RMBS investors had to bear a higher risk than they might have perceived based on the contractual disclosure. As a result, investors could argue that the ex ante value of the securities with misrepresented collateral that was sold to them was less than the price paid, and full disclosure of the true characteristics of the collateral could have prevented their loss. Misrepresentation can also have consequences for the party securitizing the mortgages (i.e., the intermediary who does the underwriting of the pool) since such parties are often contractually 11 Similarly, the prospectus supplement for First Franklin’s Series 2006-FF15 reports that “original full Combined Loan-to-Value Ratio reflects the original Loan-to-Value Ratio, including any subordinate liens, whether or not such subordinate liens are owned by the Trust Fund." 8