«May 2012 Collateral Damage: Sizing and Assessing the Subprime CDO Crisis Larry Cordell Yilin Huang Meredith Williams1 ...»
WORKING PAPER NO. 11-30/R
SIZING AND ASSESSING THE SUBPRIME CDO CRISIS
Federal Reserve Bank of Philadelphia
Federal Reserve Bank of Philadelphia
Federal Reserve Bank of Philadelphia
Sizing and Assessing the Subprime CDO Crisis
1 Cordell is vice president, Huang is a senior specialist, and Williams is a systems analyst in the Risk Assessment, Data Analysis, and Research (RADAR) Group at the Federal Reserve Bank of Philadelphia. We wish to thank Jeremy Brizzi, Mike Hopkins, Paul Willen, Bill Lang, and the staffs at the Federal Reserve Board and the Federal Reserve Banks of Boston, New York, and Philadelphia for helpful comments. The views expressed here are those of the authors and do not necessarily reflect those of the Federal Reserve Bank of Philadelphia or the Federal Reserve System. This paper is available free of charge at www.philadelphiafed.org/research-and-data/publications/working- papers/.
1 Abstract This paper conducts an in-depth analysis of structured finance asset-backed securities collateralized debt obligations (SF ABS CDOs), the subset of CDOs that traded on the ABS CDO desks at the major investment banks and were a major contributor to the August 2007 financial panic. We identify these CDOs with data from Intex©. We estimate that 727 publicly traded SF ABS CDOs were issued between 1999 and 2007, totaling $641 billion. We describe how and why multi-sector CDOs became subprime CDOs and show why they were so susceptible to catastrophic loss. We then track the flows of subprime bonds into CDOs to document the enormous cross-referencing of subprime BBB bonds and credit default swaps (CDSs) into CDOs. We also show that lower rated tranches of CDOs were not sold and were largely recycled into CDO2s and other CDOs. We estimate that total write-downs on SF ABS CDOs will be $420 billion, 65% of the original issuance balance. We then analyze the determinants of expected losses on the deals and AAA bonds and examine the performance of dealers and rating agencies. Finally, we discuss the implications of our findings and the many areas for future work.
2 I. Introduction How much will total write-downs be on the universe of CDOs at the center of “the Panic of 2007”? 2 We set out to answer this question not only to resolve speculation about the amount of the write-downs but also to get an understanding of the exact size and composition of the market.3 Resolving this question and determining our bottom line figure of $420 billion of write-downs on $641 billion of issuance turned out to be a complicated undertaking, but for reasons much different than expected. The actual pricing of the CDO securities was among the more straightforward parts of our analysis. What proved to be much more difficult were several of the more basic parts of our research. First among them was defining the universe of publicly traded CDOs that traded on the major ABS CDO desks; these CDOs are considered a major factor in the panic that erupted in financial markets in August 2007 (Gorton (2008); Covitz, Liang, and Suarez (2009)). Developing a robust classification for the SF ABS CDO market and then identifying the 727 CDOs that comprise this market was complicated because we could not find any source that attempted to define this market in a systematic way. Once we identified them, it cleared up much of the confusion about the size, composition, and institutional features of the SF ABS CDO market as well as making clear how and why this market came to be dominated by subprime securities, increasingly of the synthetic type. Surprisingly, tallying life-to-date write-downs proved more difficult than the valuation exercise for still active securities, which is most important to do since write-downs already incurred make up 71% of our $420 billion estimate. Finally, standardizing data across the many different structures presented a number of challenges that, once resolved, gave us valuable information for our analysis.
Academic studies suffer from informational gaps when attempting to investigate this market because researchers lack the primary data necessary to undergo a thorough analysis of the SF ABS CDO market.
To do this analysis, one needs access to, and expert knowledge of, monthly data files and valuation software from Intex, which we will show is the source data for the universe of publicly issued privatelabel mortgage-backed securities (MBS)4 as well as publicly traded SF ABS CDOs.5 For investment banking research and trading, Intex provides the primary source data and valuation tools (see Goodman, et al. 2008).6 What does make our study unique is that investment banks have no interest in conducting a study of a market that has completely shut down and which has generated such extraordinary writedowns, many at the same banks.
This paper is organized as follows. In Section II we pose a series of questions whose answers provide insights into how $641 billion of SF ABS CDOs could generate $420 billion of write-downs. First, we describe our methodology for defining and sizing the SF ABS CDO market. We estimate that 727
2 This characterization of the financial crisis beginning in 2007 is by Gorton (2008).
3 One early figure from a credible source had losses of $500 billion on a trillion dollars of issuance. See CreditFlux Newsletter, January 8, 2008. Lewis (2010) ended his book without knowing what losses on CDOs were or what the size of the market was.
4 Private-label, or nonagency, MBS refers to those securities not issued by the three agencies, Fannie Mae, Freddie Mac, and the Government National Mortgage Association (GNMA).
5 This surprisingly little-known company is immortalized in a 2009 article in New York magazine by Osinski (2009), who explains how he wrote the program for the Intex DealMakerTM that became “the bomb that blew up Wall Street.” 6 Barnett-Hart (2009) was able to indirectly use Intex with “Lehman Live,” a database of some 735 CDOs compiled by Lehman Brothers. We confirmed that Intex is the source data, but LehmanLive includes 142 deals that we don’t include and misses 134 deals that we do.
3 publicly traded SF ABS CDOs were issued between 1999 and 2007 totaling $641 billion. All told, $201 billion of the underlying collateral of CDOs was composed of synthetic references, or credit default swaps (CDSs). Next we describe why these CDOs were so susceptible to catastrophic loss by examining subordination levels of different bonds and ex ante views about losses and house price appreciation in 2005 when CDO issuance exploded. We then describe how multi-sector CDOs evolved into subprime CDOs. Then we track through the flows of subprime securities into CDOs to show how $64 billion of BBB-rated subprime bonds became $140 billion of CDO collateral. We also document how most lowerrated tranches of CDOs were mostly recycled into other CDOs. In Section III, we describe how we extracted data from Intex and other sources and produce summary statistics. In Section IV, we describe our process for first compiling write-downs and then our approach for generating expected write-downs to arrive at our $420 billion figure. In Section V, we extend the work of Barnett-Hart (2009) to analyze the determinants of write-downs on the universe of SF ABS CDO bonds.7 In Section VI, we conclude by summarizing our findings, assessing the subprime CDO crisis, and discussing areas for future work.
II. The SF ABS CDO Market We begin our analysis in subsection A by describing the structural features of SF ABS CDOs and estimating the exact size and composition of the market. In subsection B we show why these CDOs were so susceptible to catastrophic loss. In subsection C, we describe why the market came to be dominated by subprime securities. In subsection D, we track the flows of subprime mortgage bonds into CDOs and CDO bonds into other CDOs and CDO2s to document the astonishing amount of cross-referencing that took place in these CDOs.
A. What Is the Exact Size and Composition of the SF ABS CDO Market?
Before defining the SF ABS CDO market, we briefly describe the private-label mortgage securitization process and how CDOs were constructed. Figure 1 is a stylized visualization of the transformation of mortgage loans to mortgage-backed securities to SF ABS CDOs and, finally, to CDO2s. While this chart is represented elsewhere,8 unique to this work is the inclusion of actual figures on the size of the various submarkets from the analysis we describe below, as well as some structural features of the securities. As shown, between 1998 and 2007, a total of $3.3 trillion of mortgage loans were placed into RMBS securities (i.e., prime or Alt-A securities) and $2.5 trillion into home equity (HE) securities (i.e., mostly subprime but also some junior lien and “scratch and dent” loans), for a total of $5.8 trillion of privatelabel MBS issuance.9 Mortgage loans are the assets (collateral) for the RMBS and HE securities;
liabilities are issued in a senior/subordinated structure. Exactly why prime and Alt-A securities are classified as RMBS while subprime securities are classified as ABS is described below.
7 We are especially indebted to Anna Katherine Barnett-Hart, who shared her data with us, allowing us to understand her sources and learn from them as we developed our own database.
8 This depiction was originally done by UBS in Goodman et al. (2008) and reprinted in Gorton (2008), but using only representative numbers for tranche sizes.
9 This figure matches almost exactly the figure of $5.66 trillion of private-label MBS issuance from 1998 to 2007 reported by Inside Mortgage Finance (2010), the unofficial keeper of U.S. ABS/MBS data. The IMF obtains its figures from independent sources. The minuscule difference is likely due to a small number of privately placed MBS.
4 CDOs are constructed using RMBS and HE securities as assets. CDO liabilities are also set up in a senior/sub structure. Generally, bonds with a credit rating of A or above were placed into so-called “high grade” CDOs; BBB-rated bonds were placed into “mezzanine” CDOs.10 Based on our classification described below, $342 billion of high grade and $299 billion of mezzanine CDOs were issued from 1998The final link in the chain is the CDO2s, whose underlying collateral is primarily CDO bonds.
Forty-eight SF ABS CDO2s were issued totaling $31 billion. Our classification for CDO2s was done by using the simple rule that CDOs made up at least 50% of total deal collateral.11 As described, Intex contains the universe of publicly traded private-label MBS; it also contains the universe of 12 publicly traded “144A” SF ABS CDOs issued through these markets. But Intex contains some $1.4 trillion of CDOs issued between 1998 and 2007, so our central challenge is to define the subset of CDOs that traded on the “ABS CDO desks” at the major investment banks and asset-management firms. This is important because these desks were where trading took place and where pricing and fair value information was generated and exchanged. In particular, information generated at the ABS CDO desks in 2007 played a critical role in launching the financial crisis, so we are most interested in identifying the universe of securities that traded there.
First, CDOs are classified as “structured finance” in Intex if the CDOs can be “actively managed.” SF CDOs generally have a reinvestment period, usually up to five years, when collateral managers are allowed to purchase new assets or sell credit risky assets from the CDO. Mortgage-backed CDOs are allowed to be actively managed because prepayment risk is high and CDOs can pay down quickly without replacement. In contrast, CMBS or CRE CDOs are mostly “static pools” because commercial mortgages have prepayment penalties or yield-maintenance clauses that effectively eliminate prepayment risk.13 The static pool feature of CMBS and the whole loan feature of CRE are reasons CMBS/CRE traded on the CMBS/CRE desks separately at the large investment banks. This is important because studies frequently mix CRE CDOs with SF CDOs.
Of course, since CDO structures comprise whatever dealers can sell, there are exceptions to this classification, most notably the 68 static ABS CDO pools that emerged with the growth of the synthetic market.14 We continue to define these deals as SF CDOs if they included subprime MBS, since these deals also traded on the ABS CDO desks.
A second distinction is made between corporate CDOs and ABS CDOs. It is also the case that there are separate desks where corporate CDOs, collateralized loan obligations (CLOs), and high-yield collateralized bond obligations (CBOs) were underwritten and traded. Therefore, by adding the “ABS” qualifier, we exclude from
10 This is only a stylized model because, in practice, it is the weighted average rating that determines the classification calculations we do below, so bonds with all different ratings can appear in each.
11 We needed to establish a cutoff because 628 of the 727 CDOs had at least some CDOs as collateral.
12 Rule 144A of the Securities Act of 1933 allows private companies to sell unregistered securities (the Rule 144 securities) to qualified institutional buyers (QIB) through a broker dealer. The rule also permits QIBs to trade these securities among themselves. To be a QIB, the institution must control a securities portfolio of $100 million or more.
Because of the unregistered status of these securities, disclosure is often not as complete as in public securities.
13 Fabozzi (2007) argues that, because of these features, CMBS trade more like corporate bonds.
14 For the synthetics, which make up most of the static deals, investors often opted not to allow replacement, since they were made up entirely of CDSs.