The Department of Justice’s 119 page complaint against Standard & Poor’s is a devastating depiction of S&P’s ratings of residential mortgage backed securities (RMBS) and collateralized debt obligations (CDOs). S&P is vigorously fighting the government’s action, claiming that the evidence presented in the complaint is incomplete and misleading. S&P has a strong legal track record, having won dismissal of 31 cases with another 10 voluntarily withdrawn (another 30 cases are pending.) However, this case is different. It is being brought under the Financial Institutions Reform, Recovery, and Enforcement Act (FIRREA), which combines the scope of criminal statutes with the lower burden of proof of civil statutes. Politically, the government needs scalps, and FIRREA seems like an effective tomahawk.
The complaint is extensive and well-researched. As an ex-S&P person, I found that it reads like investigative journalism. But I have the advantage of having known most of the executives mentioned in the complaint.
The government’s case rests on two basic arguments. The complaint alleges systematic delays in updating models and criteria which would have negatively impacted S&P’s market share. The second leg of the government’s case is that S&P rated and affirmed existing ratings on CDOs even when it knew the RMBS assets contained in the CDOs were impaired.
Alleged Manipulations to Maintain Market Share
S&P based its RMBS ratings on a model called LEVELS which predicted defaults based on data obtained prior to 1999. By 2002, S&P had purchased more extensive data, which included the default experience of riskier mortgages being added more often to the securities S&P was rating. In 2004, S&P got around to incorporating the new data into a proposed new version, LEVELS 6.0. The new version was planned for release at the end of 2004. However, the new version was never released because, according to the complaint, it would have negatively impacted S&P’s market share by requiring more stringent collateral in RMBS deals.
The complaint documents internal discussions in 2005 that “S&P was underpricing risk on RMBS deals by having loss coverage levels that were too low.” It cites an email dated March 2005 which stated “We have known for some time (based upon pool level data and LEVELS 6.0 testing)” that loss coverage levels for sub-prime “B and BB levels need to be raised” and that loss coverage levels for Alt-A “B, BB and BBB levels need to be raised (we have had a disproportionate number of downgrades).” Another email dated March 2005 said “Version 6.0 could’ve been released months ago and resources assigned elsewhere if we didn’t have to massage the sub-prime and Alt-A numbers to preserve market share.”
In June 2006, over two years after the new mortgage data was first modeled in LEVELS, a more rigorous version of the model (LEVELS 5.7) was released. Although not highlighted in the complaint, subsequent RMBS delinquencies were lower for deals rated in the second half of 2006, after the release of the new model.
In 2003, S&P planned to update the model it used to evaluate CDOs. At the time, S&P had the dominant market share in non-investment grade cash CDOs (collateralized pools of existing securities; fees up to $500,000 per transaction) but lagged in the investment grade synthetic CDO market (collateralized with default swaps; up to $750,000 per transaction). A key goal for the new model was to preserve S&P’s market share for cash CDOs while improving its share for synthetics. Two teams worked on the task and came up with two differing approaches. One approach resulted in higher ratings for the synthetics but hurt the ratings on the cash deals, while the other approach was no help increasing S&P’s share of synthetics. Both were rejected.
In frustration the executive overseeing the project selected the attractive parts of each approach and glued them together, ignoring the fact that the two approaches were incompatible. This approach was also rejected as lacking analytic justification. The version released in 2004 had the same “historically inaccurate” default matrix as the previous version, according to the complaint.
A new CDO model was planned to be released in July 2005. An contemporaneous internal memo described the existing CDO model as “not analytically sound, given that it contains PD [probability of default]/correlation assumptions that are inconsistent with historical data.” The model was given to issuers to test, and received negative feedback. From an internal report: “Importantly, Bear Stearns pointed out that the potential business opportunities we would miss by effectively having to walk away from such high yield structures [low grade synthetic CDOs] would NOT be compensated for by any increase in rating volume for highly rated collateral pools.”
The release was delayed until December 2005. According to the complaint, S&P created an alternative version of the new model which was more lenient and not based on historical research or analytical data. The alternative model was not mentioned publicly. The complaint alleges that during 2006 S&P used the more lenient model to rate synthetic and cash CDOs which were not ratable using its official model.
S&P will argue that the delays in releasing its models were caused by factors other than business concerns. An S&P spokesperson is telling the press that LEVELS 6.0 had other flaws, such as predicting adjustable-rate mortgages were less risky than fixed-rate loans. Doubtless there were other non-business issues with the S&P models, but the complaint cites considerable evidence that business considerations were paramount. Perhaps enough to establish a ‘preponderance of evidence.’ Also, the complaint cites internal S&P statements alleging that rating standards were inadequate or flawed, irrespective of the reasons for delayed improvements. If substantiated, they raise issues of negligence at the very least, and they set the stage for the second part of the government’s argument.
Alleged Failures to Incorporate RMBS Risk
The government’s second argument is that S&P rated CDOs which included RMBS that it knew were impaired or likely to be impaired.
In the fall of 2006, S&P noticed rising delinquencies in the mortgages underlying non-prime RMBS, especially those issued earlier that year (using the allegedly faulty version of LEVELS.) “Indeed, the performance of the mortgage loans underlying the 2006 subprime RMBS was so bad that analysts initially thought the data contained typographical errors.” The person in charge of monitoring existing RMBS ratings “regularly expressed frustration to her colleagues that, notwithstanding the dire performance of subprime RMBS, she was prevented by [senior managers] from downgrading the ratings of subprime RMBS because of concern that S&P’s ratings business would be affected if there were severe downgrades.”
Soon after, S&P apparently established an internal standard that any RMBS which experienced delinquencies in excess of 50% of its credit support (collateral) required a negative rating action. This rule seems to have been largely ignored, according to the complaint.
In February 2007, a decision was made to put 50 RMBS tranches on CreditWatch Negative. However, a review by senior executives decided to take action on only 18 tranches. At the same time, S&P identified 583 tranches in which delinquencies exceeded 75% of collateral, but, contrary to the standard established the previous fall, apparently took no rating actions.
Meanwhile, issuers had figured out that the mortgage market was tanking and were rushing to get the mortgages off their books so they would not have to be marked to market. The CDO market was still functioning, so there was a stampede of issuers trying to package their rapidly deteriorating mortgage assets into highly rated CDOs. According to the government’s complaint, S&P executives were well aware of this rush for the exits. In March 2007, the head of the cash CDO group informed his team that he expected the analysts to be very busy because “CDO deals need to be priced and closed to reduce issuers’ exposure to the underlying RMBS collateral,” according to the complaint.
Also in March 2007, S&P analysts did an analysis of sub-prime RMBS issued in 2006, finding that the ‘overwhelming majority’ of non-investment grade RMBS were at high or medium risk of default. An email comment from one of the S&P executives viewing the data: “Wow, these deals are in huge trouble.” One of the analysts who conducted the study did the now-infamous parody of the Talking Heads’ Burning Down the House. One of the verses: “Hold tight/CDO biz – has a bother/Hold tight/Leveraged CDOs they were after/Going – all the way down, with/Subprime mortgages.”
At a UBS conference in March 2007, the head of S&P’s cash CDO group stated that “Standard & Poor’s has an integrated surveillance process to ensure ratings in our rated RMBS bonds and CDO transactions reflect our most current credit view.” An S&P article also published in March 2007 claimed that “[S&P’s] RMBS Surveillance and CDO Surveillance processes function in an integrated fashion.” However, according to the government’s complaint “S&P continued to rate new CDOs without making adjustments to account for continuing deterioration of underlying non-prime RMBS and in disregard of internal analyses and reports demonstrating the extent of this continuing deterioration.”
According to the complaint, S&P issued and/or confirmed ratings on 61 CDOs backed at least in part by RMBS valued at $51 billion in March 2007. In April 2007, S&P issued or confirmed 47 CDOs valued at $24 billion. In May, 29 CDOs valued at $33 billion. In June, 32 CDOs valued at $30 billion. The wheels finally started coming off the bus in late June, but according to the complaint S&P still managed to rate 8 more CDOs in July valued at $4.5 billion.
Meanwhile the internal alarm bells were ringing about RMBS securities. Reports generated in April 2007 showed 590 sub-prime and 481 Alt-A RMBS deals with over 75% delinquencies relative to credit support, averaging 121% delinquencies versus credit support for sub-prime and 254% for Alt-A. These numbers were far in excess of the standards (50%) set in the fall of 2006, yet no actions appear to have been taken.
Also in April 2007, S&P conducted an analysis of RMBS issued in 2005 and 2006. The analysis projected average default rates of 47% for BBB tranches rated in 2005 and 56% for those rated in the first half of 2006 (before the release of the improved LEVELS model), and 35% for those rated in the second half of 2006. Meanwhile CDOs were being rated with BBB default rate assumptions of 3%.
By May 2007, the number of RMBS exceeding 75% delinquencies had risen to 631 sub-prime deals and 511 Alt-A deals. The average deliquencies: 126% and 202%. Apparently, no action.
At the same time, the head of the CDO group reported new records for CDO issuance to S&P senior management: “Because of the effect of the subprime RMBS situation, in March we experienced the highest monthly deal volume ever, doubling the total from the previous two months. The cash flow area closed an impressive 72 deals.” The report noted the acceleration of cash CDOs was in part “due to preferential ‘mark to market’ treatment a dealer can receive for CDO ‘priced’ liabilities versus owning warehousing risk in its ‘raw form’ (i.e. at the underlying subprime mortgage level.).” In other words, it was better to have AAA CDOs than rapidly deteriorating RMBS securities.
For the first four months of 2007, S&P received $74.82 million in revenues from rating CDOs. This was double the $36.23 million it received for the same period in 2006. The head of the CDO group expected May to generate $24.48 million in revenue, “the highest total for the month of May to date.”
By June 11 2007, S&P had run analysis of all 18,000 RMBS ratings it had issued and found that, on average, the tranches rated BBB and below had greater than 100% delinquencies versus credit support. According to the complaint, “S&P analysts and executives knew that an SD versus CS ratio in excess of 100% meant that the RMBS tranche at issue would in the near term almost certainly be subject to a negative Rating Action.” The complaint alleges that S&P managers withheld the June RMBS analysis from those rating CDOs and S&P continued to rate CDOs without making any allowances for the deterioration in RMBS delinquencies.
On or about June 28, 2007, the head of S&P’s structured finance group finally decided to authorize large scale downgrades of non-prime RMBS. The complaint alleges that S&P managers withheld this information from CDO analysts rating deals incorporating RMBS collateral. On July 10, 2007, S&P announced that ratings on 612 classes of RMBS were placed on CreditWatch Negative. It also announced that it had toughened its requirements for new sub-prime transactions, although, according to the complaint, these new standards were not implemented on some of the deals subsequently rated.
According to the complaint, analysts in the CDO group proposed the CDO ratings be discontinued until rating actions on underlying RMBS had settled down. Managers however rejected the proposal according to the government, insisting that issuers still needed to clear out their warehouse lines and that S&P should not “close the window.” The decision was made to “notch” the ratings on RMBS, considering the RMBS as lower rated than the existing ratings. However, the complaint references 16 CDOs valued at $13.3 billion issued subsequently which allegedly were rated with RMBS taken at face value.
The Government’s Case
According to a white paper by Schiff Hardin, FIRREA is a tool that the government has recently dusted off, and has begun to use on banks. The appeal of FIRREA is twofold:
FIRREA is a comparative walk in the park for government attorneys compared to the rigorous burden of proof required in criminal cases and the criminal discovery rules, which are focused largely on protecting a criminal defendant’s rights. First, FIRREA has the benefit of only requiring the civil burden of proof, meaning violations must be proven by a preponderance of the evidence rather than beyond a reasonable doubt. Providing government attorneys with a means to utilize this lower burden of proof is particularly dangerous when coupled with the fact that FIRREA allows civil actions involving the broadest of the criminal statutes. Mail and wire fraud have always been regarded as among the broadest and sweeping of criminal statutes, encompassing virtually any interstate activity where the mail, telephone, or computer is used to accomplish a fraud.
FIRREA also has the advantage of a longer statute of limitations, ten years.
S&P’s past success in averting lawsuits rests on its first amendment assertion that ratings are opinions. However, the first amendment gives no protection against fraud, which basis for the government’s complaint. Also, past suits have not had the benefit of the vast amount of confidential information included in the government’s case. S&P will argue that the information in the complaint is selective, out-of-context, incomplete and misleading. Nevertheless, the sheer volume of the government’s case is overwhelming. Since the criterion for a conviction under FIRREA is ‘preponderance of evidence’, the heft test is relevant.
There is another factor which would help the government’s case: witnesses. S&P fired most of the staff in its RMBS and CDO groups after the markets collapsed, and we suspect there are quite a few who would be willing to testify. The complaint contains a number of allegations which were clearly taken from interviews with former staff. Only a handful of S&P’s former executives and managers are explicitly named in the complaint. Knowing some of the individuals whose identities are hidden behind “Executive X” and “Senior Analyst Y”, I suspect they are eager to help the DOJ bring its case to trial. I don’t believe past lawsuits against S&P have featured testimony from ex-employees.
The government’s case is particularly dangerous because it focuses on the the weak underbelly of rating agencies, timeliness. Investors have always complained about the timeliness of ratings (with transparency being a close second.) Detractors have assumed that ratings agencies are incompetent because ratings actions often lag market reality. The crux of the government’s case is that S&P’s ratings on RMBS and CDOs lagged reality because it was in S&P’s economic interest to do so.
As a former employee of S&P, I read the government’s complaint with dismay. In my experience, the majority of S&P ratings personnel tried very hard to maintain their independence, conscious of their duties to investors as well as issuers. The memos and other documents included in the government’s complaint depict managers who were beholden to the issuers they worked with, focused primarily on helping the investment banks ease their inventory problems with fast deteriorating mortgage securities.
By their very nature, ratings on structured securities involve a high degree of interaction between the rating analysts and the issuers. The issuers (bankers) constantly iterate with analysts on the criteria. If I add more BBB to this tranche, what does that do to the rating on the transaction? How much do I need to add to raise the rating 2 notches? Analysts become part of the banking process and extensions of the financial engineering.
In a blog I wrote in 2008, I wondered about the checks and balances on the structured finance group within S&P: “The structured finance group had a lot of hubris when I was around, and I can only imagine how much autonomy they received as their size and profitability grew. I suspect it became a business within a business, isolating itself from the rest of the ratings group.” It appears from the government’s complaint that there was very little oversight from the rest of S&P on the criteria and standards developed in the structured group, unlike when I was at S&P. The business in structured had become so large and profitable that there seemed to be little accountability to other units within S&P.
Now most of those people are gone and S&P is fighting for its life. S&P will be a healthier and presumably more chastened place without them. Provided of course that it survives this lawsuit and its progeny.