Written by Bob Schneider Posted on September 25, 2010
In response to the highly-criticized lack of transparency in asset-backed securities (ABS) – often cited as a key cause for the recent financial crisis and subsequent recession --- the SEC issued a proposed rule in April. Totaling some 677 pages, the Rule covers a multitude of complex issues concerning the offering process, disclosure, and reporting.
The purview of this post is far more narrow, limiting itself to the relevant text in the SEC proposal -- and, especially, the Comments to it -- regarding the use of XBRL. I believe the research and links provided will be useful to those who want to explore the possible use of XBRL for ABS reporting.
With respect to XML/XBRL, the Rule contains these proposals:
For each loan or asset in the asset pool, we are proposing to require disclosure of specified data relating to the terms of the asset, obligor characteristics, and underwriting of the asset. Such data would be provided in a machine-readable, standardized format so that it is most useful to investors and the markets. Under our proposal, issuers would be required to provide the asset-level data or grouped account data at the time of securitization, when new assets are added to the pool underlying the securities, and on an ongoing basis.
We are proposing to require the filing of a computer program (the “waterfall computer program,” as defined in the proposed rule) of the contractual cash flow provisions of the securities in the form of downloadable source code in Python, a commonly used computer programming language that is open source and interpretive. The computer program would be tagged in XML and required to be filed with the Commission as an exhibit. Under our proposal, the filed source code for the computer program, when downloaded and run (by loading it into an open “Python” session on the investor’s computer), would be required to allow the user to programmatically input information from the asset data file that we are proposing to require as described above. (pp. 16-17)
In its proposal, the SEC describes its reasons for choosing XML over XBRL:
We are proposing to require asset-level information and grouped account data (with respect to credit cards) related to an offering and ongoing periodic reporting be filed on EDGAR in XML as an asset data file….For asset-backed issuers, we believe that XML is the appropriate format to provide standardized asset data disclosure… XBRL allows issuers to capture the rich complexity of financial information presented in accordance with U.S. GAAP. In contrast, the proposed asset data file will present relatively simpler characteristics of the underlying loan, obligor, underwriting criteria and collateral among other items that are well suited for XML. We are proposing XML, rather than XBRL, because there are many commercial products that can be used with XML including parsers that would allow investors to insert data into a relational database for analysis, data extensions available in XBRL are not applicable to this data set, the nature of the repetitive data lends itself to an XML format and the schema could be easily updated. (pp. 190-2).
I performed a search on the Comments for keyword XBRL. Of the more than 140 Comments the SEC has received on the Rule, 15 mention XBRL. Six of the 15 -- Association for Competitive Technology, DTCC (Depository Trust and Clearing Corporation), JPMorgan Chase, TickLab partners, ABA Business Law Section, and Ariel Blumencwejg (structured credit trading professional) – make only in-passing and generally insignificant references.
Of the remaining Comments, those of XBRL US, UBmatrix, PwC, and MERS (Mortgage Electronic Registration Systems) generally support the use of XBRL. Mortgage Bankers Association, Mortgage Industry Standards Maintenance Organization, and Real Analytics are, to varying degrees, negative about the use of the data standard for ABS purposes. Risk Management Association and Microsoft are mixed or neutral in their assessments.
Comments Generally in Favor
XBRL US provides a detailed analysis that is well worth reading in its entirety. Here is an (oversimplified) summary of its main points:
1. XBRL US supports the use of the waterfall model, adding that “…Any such automated model must be consistent with the prospectus…It is important that the results of the Waterfall model match the prospectus as closely as possible.” (pp. 5-6)
2. It believes, however, that the output of the model is not clearly defined in the Rule. That output “should be in XBRL and, at a minimum, XML…XBRL is preferable over XML as the semantics associated with a cash flow are already incorporated into the XBRL standard….” (pp. 1-2)
3. The output should form part of a bond remittance file that represents the information about the cash flows associated with each bond. XBRL US has created a draft taxonomy to assess the feasibility of using XBRL for this purpose. (p.3)
4. XBRL is preferred to XML for capturing details of the loan file; in its Comment, XBRL US counters the argument that the sheer size of mortgage loan files makes XBRL impractical (bottom p.3).
UBmatrix has also provided a thoughtful and compelling response which deserves to be read in full. Summarizing its argument, UBmatrix states that:
The SEC proposal offers some key insights into how to help investors make sense of ABS filings. However, ultimately the proposed approach is undermined by a simple implementation detail -- the intent to use a Python-based computer program as the issuer's vehicle of delivery.
After outlining objections to a Python-based computer program (pp. 3-5), the Comment presents XBRL as “…a viable alternative, both as a mechanism for submitting ABS data and as a platform for modeling cash flows for investor analysis.” It enumerates five characteristics of XBRL -- data accuracy, data comparability, data transparency, extensibility, and sufficiency (i.e., the ability to offer a single solution that adequately meets issuer and investor needs, without the need to build or acquire from third party vendors) – that make it the better choice.
As part of a longer response to various ABS issues, PwC addresses the use of XML for the asset data files and states:
Adopting XML without also adopting an additional framework – be it XBRL, OAGIS, UBL or some similar public standard – means that numerous data file design decisions will have to be made by each reporting entity for the ABS disclosures. This may increase, not decrease, the variety of reporting formats and practices to the market…Some of the ABS disclosures, particularly those related to credit ratings (e.g. credit ratings, financial information on underlying properties, etc.), already exist in publicly available XBRL Taxonomies or dictionaries of common disclosure definitions. It would enhance the consistency of market use to enable the reuse of these definitions by all entities reporting and using ABS information.(p. 5)
Regarding cash flow waterfalls, the PwC Comment says:
We believe that providing a standardized basis for reporting waterfalls is useful, but that any programming language, no matter what the reputation is for readability, may be too technical to be easily understood by all potential users. Python is not a standardized information reporting tool. As such, it is possible to create Python code that returns correct results but is coded in such a way that may be very challenging for users to understand how the provided code creates the end results. The use of existing standards for the data (such as the use of the XBRL Specification to define appropriate taxonomies and instances for such data) and the formulas and business rules (such as XBRL Formula, RuleML, FPML or similar standard formula and rules languages) would facilitate the creation and population of waterfalls that are more easily understood and widely used by third parties while still allowing the standardized publication of online tools that work with these models. (p. 5)
MERS (Mortgage Electronic Registration Systems) includes this response to one of the SEC’s questions from the proposed rule:
SEC Question We note that there are several different standards under which asset-level data is already required. Would our requirements impose undue burdens on ABS issuers?
MERS Response MERS supports the creation of a common reporting requirement, and urges the SEC to work with other government agencies to harmonize the reporting requirements and move towards a common set of definition and data reporting requirements, optimally employing MISMO XML standards and XBRL. (p.6)
Comments Generally Against
The Comment of the Mortgage Bankers Association includes this section:
SEC Question In what format do issuers currently provide asset data information to investors (as may be required, for example, under transaction agreements)? Do any market participants currently provide asset data in accordance with a technical specification or schema commonly used across a particular asset class? If so, would our data points cause divergence from current practice? Please tell us which specific proposed data points would be of concern and why. How can we address those concerns? Is another format preferable, such as XBRL?
MBA Response ...XBRL is designed for the financial accounting industry and would require significant modification to be adapted to the U.S. mortgage industry. (p.54)
Mortgage Industry Standards Maintenance Organization, which is a not-for-profit subsidiary of the MBA, responds to the same question:
Issuers currently provide investors with information using multiple types of formats. There is no single format used for all disclosures across all ABS industries. Given that the type of data requested under the Proposal is very detailed, MISMO believes that XML is the better solution [emphasis added]. Within the real estate finance industry, MISMO XML standards represent the broadest implemented base of open, nonproprietary data standards in the real estate finance industry. (p. 19)
The Comment of Real Analytics offers this analysis:
Use of XML and Mortgage Data Standards
A central proposition of the SEC’s proposed regulations is to require that data pertaining to mortgage-backed securities and their related collateral be reported in XML. This proposition is consistent with the SEC’s similar existing requirement for financial reporting using a version of XML known as XBRL. Industry-specific versions XML are widely used in U.S. industries and throughout the world’s financial markets. While XBRL was developed for financial accounting purposes, it is not well suited for mortgage data or real estate. (p.1)
Risk Management Association includes this text in its Comment:
XML is a suitable format for asset data files and we support the use of XML schemas, specified either with the XSD language or the more specialized XBRL, to describe the required input structures.…We recommend that the SEC avoid using XML altogether in the representation, storage, or transmission of Python source code…Instead, we recommend that the SEC require that the waterfall computer program be submitted in the form of a text file, also known as a plain text or ASCII file. (p. 9)
Microsoft does note in its Comment that it is a founding company and ongoing supporter of XBRL. It is unenthusiastic about the waterfall proposal. It suggests a “data-oriented approach” similar to the Open Government Directive issued by the OMB on December 8, 2009; it does not mention XBRL in its recommendation.
In closing, let me reiterate that the XML versus XBRL question is only one part of the much larger debate of what disclosures and processes should be implemented for ABS. As this blog post indicates, the waterfall proposal itself has been subject to much scrutiny and is deeply controversial. How these larger reporting issues are resolved will inevitably determine what role either XML or XBRL will have in ABS reporting.