Will the Securities and Exchange Commission (SEC) ever bridge the Grand Canyon-sized gap between issuers of asset-backed securities and their investors when it comes to waterfalls and the Python language?
The SECe_SSRqs proposed Regulation AB, for which comments were due Aug. 2, would require issuers to provide loan-level disclosures to better measure risk and require them to maintain significant "skin in the game" by retaining a 5% slice of offerings. While industry players voiced criticism about those provisions in their comment letters, the more technical waterfall and Python provisions prompted much shriller tones.
The American Securitization Forum (ASF), in a supplemental comment letter focusing on the waterfall, summarizes its issuer constituents by saying that the proposal's "daunting challenges, staggering costs and inestimable risks" will result in securitization no longer representing a rational and cost-effective source of funding. It added: "Indeed, if the commission proceeds with its proposal, issuers question whether a viable securitization market will exist at all."
Waterfall models display the flow of repayments between the numerous shareholder classes in an ABS security. They are typically provided by third-party vendors and include adjustable parameters to predict the outcome of the waterfall payments under different market scenarios - complex models issuers vehemently deny producing themselves today.
The SEC's proposal would require issuers to provide electronic versions of waterfalls for each deal - written in the Python open-source programming language - that would be downloadable from the Internet. And that's where the differences between issuers and investors widen into a chasm.
The ASF represents both issuers and investors and typically reaches consensus among its members on issues. Not this time.
"Despite special efforts to find common ground and to reach practical compromises that addressed these competing interests, consensus could not be reached on the commission's proposal," the ASF noted in its comment.
"Because we have such a fundamental disagreement, we don't present a firm industry view [in the comment letter]," said Tom Deutsch, executive director of the ASF, in an interview.
The crux of issuers' waterfall concerns is addressed by JPMorgan Chase in its comment letter when it says that the proposal would require issuers to create a "predictive" model. The model would not be limited to describing how projected cash flows are allocated to each tranche in the ABS, and it would "combine the functionality of a collateral engine and valuation model, which will use investor-assumed performance data, allocations and distributions in order to predict the cash flows for a transaction."
Bianca Russo, the bank's associate general counsel, writes that the model "would be extremely complex, time consuming and expensive for sponsors and issuers to build," since a distinct program would have to be built and maintained for each transaction.
More importantly, though, the predictive nature of the model proposed by the SEC could open issuers' up to liability for programming errors and errors by users. "Suffice it to say," wrote the American Bankers Association, "the waterfall program as proposed is unworkable, and the commission should not go forward with this aspect of the proposal."
On the other side of the fence, major institutional investors, such as MetLife and Vanguard, are strongly supportive of the waterfall provision and its predictive components.
"We fully agree to the commission's approach to waterfall modeling, and issuer responsibility for modeling, updating and liability for accuracy," wrote Robert Auwaerter and Bob Behal, principals at Vanguard. They added, "We believe this is one of the most important changes proposed by the commission."
Vanguard said it supports issuers providing a model that allows for input of "all relevant collateral assumptions such as prepayments, losses, delinquencies and recoveries," and it should accurately enable investors to reproduce the expected cash flows as well as independent scenarios.
"We're looking for accurate models from issuers that they've tied out to the waterfalls outlined in the prospectuses," said Behal in an interview, adding that that could help minimize potential inaccuracies that arise when third parties have to translate a prospectus' written waterfall into a computer program.
The Risk Management Association's (RMA) Securitization Risk Roundtable also strongly supports the waterfall proposal. It noted in its comment letter that in the lead-up to the recent financial crisis, investors relied heavily on the opinions of third-party vendors because there was a lack of readily available and affordable tools to make investment decisions. It added that the waterfall proposal would make it easier for investors to analyze ABS offerings at the time of their initial investment decisions, as well as monitor ongoing performance of those investments.
"Investors would spend less time chasing the English language statement of the waterfall within the deal documents and going through the steps to convert to computer code in a programming language, and more time analyzing and understanding the transactions," wrote Joseph Masri, chairman of the RMA's securitization roundtable.
With investors eager for the prospect of readily available data over the Internet and some issuers suggesting the proposal should be scrapped altogether, the SEC's final rule will likely disappoint one side. Nevertheless, participants in the securitization industry did leave some room for compromise.
Behal, for example, said the issuers "should be accountable for the accuracy of the model, but not specific outcomes." So if a cash flow priority cannot be modeled with accuracy, it should not be included in the waterfall model.
Nevertheless, Vanguard's comment letter said issuers should retain ongoing liability for accuracy and completeness of model assumptions. As a result, "We believe issuers are likely to see improved pricing and economics of ABS securities as the market develops confidence in the models used to predict expected cash flow," the Vanguard principals wrote.
Russo at JPMorgan instead recommended limiting the SEC's waterfall proposal to the cash flow model describing how projected cash flows move through the payment waterfall, and eliminating the predictive components.
"Investors would then have the ability to use the filed waterfall computer program with either their own proprietary modeling program or a commercially available modeling program in order to analyze the risk profile of a certain ABS transaction based on their own predictive model requirements," Russo said.
Deutsch, in his letter to the SEC on behalf of the ASF, noted investors' support of the SEC proposal. He said some investors build waterfall models internally, and nearly all use the more elaborate models built by vendors. And because issuers would have to build models for each transaction, their usability and functionality would probably be inferior, resulting in continued use of third-party models.
He added that investors view issuer models resulting in more precise pay rules as greater transparency in the offering process. "Investors believe that the waterfall computer program would also aid their ability to internally perform cash flow analysis and improve model quality among third-party providers," he wrote.
Despite some investors support for issuers being strictly liable for their models, they already use models from software companies that are held to a lower liability standard.
Kevin McCarthy, managing director at Intex Solutions, a major provider of waterfall models along with Trepp and Bloomberg, said no issuer or software developer can always guarantee the performance of a model. "This is why the market for models has developed with limited liability typical of software contracts rather than 'lOb-5' liability ordinarily associated with securities offerings," McCarthy wrote in his firm's comment letter.
JPMorgan also calls for a lower liability standard. It requests that the SEC deem its pareddown version of the waterfall program "not to be ... incorporated in the issuer or sponsor registration statement," and so "not subject to liability under the Exchange Act or Securities Act ..."
Behal declined to comment on JPMorgan's specific request. He said, however, that Vanguard supports a legal structure for the waterfall model that maintains an issuer's legal liability for its accuracy. "It may not be an absolute liability threshold, but we do support the concept of making issuers have liability for the models' accuracy going forward," he said.
The SEC's related proposal for issuers' waterfall programs to be written in the Python programming language concerns mainly issuers, since their staff would have to be fluent in it to write the programs and presumably to provide tech support to program users. But it also elicited comments from technology giants such as Microsoft.
Issuers' opposition to adopting the Python language is summed up by the ASF, which said the language is currently used by no one in the ABS industry and consequently has no industry coding standards. Without standards, "variations in the form and functionality of waterfall computer programs across issuers would be inevitable, and investors would have to invest significant amounts of time and effort just to operate the programs, much less to read and evaluate their output," said Deutsch, summarizing issuers' concerns. In contrast, he added, "Third-party vendors offer standardized controls and formatting, thereby facilitating comprehension and comparison of ABS across issuers."
Despite additional Python-related costs investors might have to bear, at least according to issuers and vendors such as Intex, few investors submitting comments even mentioned the issue. Perhaps that's because mostly very large institutional investors commented on the proposal. As Intex noted, however, Python could be much more challenging for smaller investors who might not have the technology expertise on staff.
The programming language has the strong support of the RMA, which said that requiring a single language for the waterfall programs would enhance the transparency of the ABS issuing process. "Python is a superb choice of language for [waterfall computer programs] because it is concise and expressive ... it allows developers to write programs that will run without further modification on any computing platform where the language interpreter has been installed," RMA's Masri wrote.
Microsoft's criticism doesn't focus on the Python language, which it has implemented in the Windows environment. Rather, it opposes requiring a single language. "In allowing for choice, the commission can attract the widest array of market competitors to address the problem it has identified," wrote Craig Shank, general manager for interoperability at Microsoft.
