Goldman CDO: The Loan Data Was There for All to See
May 5, 2010
The Securities and Exchange Commission's (SEC) suit against Goldman Sachs centers on the allegation that the Wall Street firm failed to disclose to investors the role an opportunistic hedge fund, Paulson & Co., played in choosing the assets underlying a CDO, before betting against it.
But detailed and updated data about these assets — mortgage loans populating 90 MBS that made up the $2 billion CDO called ABACUS 2007-ACI — were there for all to analyze, providing they paid for it.
"What's strange about the idea that institutional investors would need to be protected from each other is that there was enormous amount of information available to institutions, even those with modest budgets," says Kenneth Posner, an analyst at Morgan Stanley, who until 2008 covered specialty finance companies including lenders such as subprime mortgage broker Countrywide Financial Corp.
If a SEC proposal to require more detailed information on ABS comes to pass, that loan-level data and more will actually be provided by the issuer on a regular basis - and be downloadable from the Web for electronic analysis. This would make claims against Wall Street firms that they provide inadequate disclosure about a deal's contents all but obsolete.
That data, about the types of loans and their status relative to the value of properties they financed, much less information about the borrowers' credit worthiness, was not disclosed by Goldman in the Abacus prospectus or otherwise. This was common practice at that time.
But Goldman did list the mortgage-backed securities transactions referenced by the synthetic CDO and provided their Cusip numbers, which identify the securities.
"Synthetic" CDOs derive their values by references to a set of securities whose values can rise and fall. The owners of the CDOs might have no ownership stake in any of the referenced securities.
Still, even if a prospectus for a product like Abacus did not include loan-level data and Goldman did not provide current data about the underlying assets, that did not mean the information was unavailable to investors.
Loan-level data about the mortgages pooled in those MBS was available to investors in Abacus and other such deals throughout the frothy heyday of CDOs, before the financial crisis struck full force in September 2008.
And that data would have provided investors with far more substantive information about the deal's health than whether a hedge fund selected some of its loans. At least one technology vendor provided loan-level data for virtually all securitized mortgages. But instead most investors relied almost solely on the now discredited ratings from the ratings agencies, perhaps in part because they were bond investors unaccustomed to the nitty-gritty loan analysis traditionally performed by commercial bankers.
Those investors could have analyzed a wide range of metrics about the financial status of the borrowers and their loans, as well as the performance of the securities comprising those assets. So an investor in Abacus could have analyzed the real estate underlying each of the loans in JPMAC 2006-FRE1 M8 Midprime, a mortgage-backed security referenced by Abacus, as well as the underlying loans of the remaining 99 mortgage-backed securities in the CDO. The JPMAC security was serviced by JPMorgan and rated 'BBB' by Standard & Poor's.
David Hurt, senior vice president of business development at Loan Performance, a unit of CoreLogic, says his firm has provided the relevant loan-level data and many of the tools to analyze it since before the housing bubble.
A CoreLogic database aimed at mortgage servicers accepts information about loans from the servicers without identifying the borrower or the specific property. That prevents customers from using the information opportunistically to solicit competitors' borrowers. But it does show what city and state the loans are based in, the types of loans, whether borrowers are paying off the loans consistently and related data.
CoreLogic then normalizes the servicer-provided data into a standard format, enabling its servicer customers to compare the health of their loan portfolios with the market nationwide or a specific region, or against some other parameter. The data service can also be licensed by investors or other parties.
The servicer database, however, does not provide information about specific mortgage loans in a mortgage-backed security. But investors can retrieve that information by tapping CoreLogic's securities database. There, an investor can analyze the individual loans in each security according to parameters such as the loan-to-value ratio of the mortgage, whether second liens such as home equity loans are also attached, the geographic region of the property, and the borrowers' credit scores. These details can be used and analyzed through digital tools to determine whether a concentration of mortgages is at risk in a given security and ultimately the CDO.
If an investor finds such a concentration, he or she can use the servicer database to see how that pool of loans compares with similar loans in the geographic region, or to another parameter.
For example, an investor in MBS can compare a current estimate of the value of a property, which is based on property sales in the vicinity, to the value of the property when the loan was originated. And since CoreLogic updates property information with county-recorder offices, the investor could also retrieve current mortgage balances on the properties from CoreLogic. That would enable the investor to determine how many mortgages are "under water" and exceeding the values of the properties, and how many are at risk of becoming delinquent.
The investor can also check to see whether a home equity loan or other subordinate financing was taken out on the property. And the investor can do this for every property in the security and throughout the CDO to determine with a high likelihood what portion of loans in the transaction will end up in default.
"You could do this for every mortgage in the pool and determine what the overall risk characteristics of the bond are," Hurt says.
A synthetic CDO investor can order those reports regularly for each of the referenced mortgage-backed securities or order them on a one-off basis, if he or she suspects a shift in the market.
CoreLogic declined to provide fee details for its services, noting they depend on factors such as the number of reports customers request and the difficulty in normalizing the data.
But Hurt notes that the firm had only two hedge fund or private equity customers four years ago. "Now we're north of 300," he said.
CoreLogic's services undoubtedly aren't cheap, given the volume and complexity of work involved, but neither are they out of reach. "This costs tens of thousands of dollars, not millions," says Posner, whose recently published Stalking the Black Swan discusses a hedge-fund analyst who suspected mortgage-market problems and used CoreLogic data to reach bearish conclusions.
"The people who made the right bets recognized the bigger picture, the causal drivers, and they were able to get enough information to test their theories," Posner says.
Hurt declined to say whether Paulson & Co. used the CoreLogic databases and the hedge fund declined to comment.
Without the loan-level mortgage data provided by a vendor such as CoreLogic, however, investors had faced a daunting task.
Upon request, servicers of MBS typically have provided a loan tape containing data about the mortgages in their pools. CDOs, however, sometimes referenced hundreds of mortgage-backed securities, requiring the investor to obtain loan tapes from each of the servicers.
Jon Thompson, vice president of structured finance at Advantus Capital Management, says that upon obtaining the loan tapes, investors would have then needed a program to translate their often differently formatted data fields, which frequently carried different labels for the same type of data.
"Issuers typically have done their own thing in terms of formatting the data fields. You would have needed a key for each issuer to translate what column headings meant," Thompson says.
Thompson adds that each investor would then require software enabling it to select different data fields and combine them, in order to analyze the range of data according to the investor's own risk assumptions.
Vendors such as CoreLogic, Lewtan, Intex and 1010data provide many of those tools.
"Our technology makes it very easy to roll down to the loan level and look at each piece of information loan by loan," says Greg Munves, vice president in charge of capital markets at 1010data, adding that the firm has more than 100 customers for which it hosts data and provides software to analyze it at the loan level.
However, 1010data doesn't generate the loan-level data. Lewtan began generating it in 2009, and Intex did not respond to inquiries about whether it provides loan-level data. Other firms that advertise providing data at the mortgage-loan level include Lender Processing Services, Blackbox Logic and MBS Data.
The SEC proposal to amend Regulation AB, affecting asset-backed securities, was issued April 7 and comments-likely to be plentiful-are due Aug. 2. The proposal requires ABS issuers pursuing public offerings to disclose significantly more loan-related data in offering prospectuses and update that data regularly.
More significant: Issuers in the private 144A market, where CDOs were offered, would have to provide the disclosures upon request from investors.
While this data has been largely available in the mortgage-backed securities market through third parties and investors' own efforts, now issuers will be responsible for providing it, along with the so-called "waterfall program" that models the securities cash flows and allows for forecasts based on investor market and risk assumptions.
Should synthetic CDOs eventually return in force, analyzing them may still be a difficult task requiring a vendor such as CoreLogic or a big staff.
"There's no mandate [in the proposal] for the CDO issuer to go out and reference each loan in the residential mortgage-backed securities" that make up the CDO, says Ned Myers, chief marketing officer at Lewtan.
Nevertheless, the issuers of the underlying mortgage-backed transactions, if they are registered deals, would be required to submit a computer file to the SEC that contains loan-level data and provide ongoing updates.
Those files would be downloadable by investors from the SEC's EDGAR database.
"They would have to show not just loan balance, rate, and term and who the servicer is, but also the borrowers' income range, credit scores geographic location-all the kinds of data that would help investors determine whether borrowers can pay back a loan. This for every single public securitization," says Edward Gainor, a Washington, D.C.-based partner of law firm Bingham McCutchen. Issuers in the 144A market would have to provide essentially the same data upon request.
That should improve access to the data, which the SEC would require to be presented in Extensible Markup Language, providing a standard for the first time that will format information for easy sharing over communication networks and computer analysis.
"As a user, standardization goes a long way in helping make sure all the required information is presented. I believe it would solve a lot of the issues, such as not requiring a key from each issuer about what every column heading means," Thompson says.
For more information on related topics, visit the following: