At Data Transparency 2013, Issa Confronts SEC’s Stalled Progess on Open Data

Learn more about our members’ COVID-19 Projects. Click Here

Financial Regulation

At Data Transparency 2013, Issa Confronts SEC’s Stalled Progess on Open Data

Last Tuesday at Data Transparency 2013, Rep. Darrell Issa (R-CA), chairman of the House Committee on Oversight and Government Reform, announced a new investigation aimed at persuading the U.S. Securities and Exchange Commission to restart the stalled transformation of its corporate disclosure filings from documents into open data.

Issa’s criticisms are an opportunity for the SEC to clarify its commitment to the principles of President Obama’s Open Data Policy. It’s also a lesson in the challenges of converting any complex area of government reporting from static PDFs with page breaks into open data–standardized, structured, and accessible to all.

In 2009 the SEC began phasing in a requirement for public companies to submit their financial statements (and related footnotes) in the eXtensible Business Reporting Language (XBRL) structured data format, alongside the old-fashioned plain-text document versions of those financial statements. The SEC publishes the XBRL files on its EDGAR website. The SEC’s attempts at standardization and publication–the two fundamental steps of open data –should, by now, have brought benefits to the agency, investors, and public company filers.
  • Since structured data–with each element electronically identified–can be analyzed automatically, the XBRL requirement should allow the SEC to review corporate financial statements faster and use Big Data analytics to find errors and fraud.
  • Since structured data can be downloaded readily into databases and spreadsheets, the XBRL requirement should allow investors to make faster, better decisions.
  • Since structured data, as long as each element has an agreed definition, can be produced automatically by software, the XBRL requirement should allow filers’ internal accounting software to create their financial statements automatically and submit them more easily.
These three benefits exemplify the three promises of open data in government: better management, better public transparency, and automated compliance. Other federal agencies choosing to standardize and publish their reports–notably the Recovery Accountability and Transparency Board in 2009 and the FDIC in 2006–have achieved all three.

But so far the SEC has missed all three opportunities, to the detriment of its own mission, investors’ interests, and the compliance burden of the companies it regulates. On Tuesday Rep. Issa announced he had sent a letter to SEC chairman Mary Jo White requesting answers and documents in each of these three areas. Here’s Issa’s letter:

Rep. Issa is not alone in his criticisms. Congressional appropriators, the SEC’s own Investor Advisory Committee, the American Institute of CPAs, Columbia Business School, and our coalition of tech companies have all made the same points. But his letter provides the most detailed explanation to date of exactly why the SEC’s attempt at open data has so far missed the benefits that the Recovery Board, the FDIC, and others have realized.

Disagree with Issa? Ask the President.

Moreover, on open data, the Republican Chairman of the House Oversight Committee is aligned with the Democratic President of the United States. On May 9, President Obama released his Open Data Policy, which requires agencies to open all data they collect and generate. The SEC’s corporate disclosure regime, taken as a whole, flunks all seven parts of President Obama’s seven-part definition (page 5) of open data.

1. The SEC Spent $11 Million Double-Checking Financial Statements with Calculators.

First, the SEC has missed its opportunity to use structured data to find and correct errors (primarily the responsibility of its Division of Corporation Finance, or “CorpFin”) and accounting fraud (primary the Division of Enforcement’s job) in the financial statements. Why? Because the agency has not tried to incorporate the electronic structure of XBRL into CorpFin’s reviews of corporate disclosures. Nor has it (yet) deployed Big Data analytics to help Enforcement sift through financial statements for fraud.
I’m accustomed to writing our Coalition’s blog in the third person, but I must shift to the first person to explain a relevant backstory. In 2009 I was an attorney in the SEC’s office of interactive disclosure. I interviewed lawyers and accountants working for CorpFin to learn how they review incoming corporate disclosures, particularly the financial statements, to find errors and confirm that accounting rules and the SEC’s regulations are followed.

As might be expected, CorpFin’s reviewers need to verify that all the required items are present in the financial statements that the SEC receives. They must check the math of financial statements to see if any calculations are wrong. They also apply accounting rules. For instance, under U.S. Generally Accepted Accounting Principles (U.S. GAAP), no “miscellaneous” line item should be greater than ten percent of the larger category it’s in. If, say, “Miscellaneous Revenue” is more than ten percent of “Total Revenue,” CorpFin contacts the company to request that it revise the financial statement and further break down “Miscellaneous Revenue.”

I was shocked to find that CorpFin’s attorneys and accountants do all of this work manually.

In CorpFin, the attorneys and accountants print each financial statement as a document, pile the paper documents on their desks, fill them with underlines, and tap away on plastic Texas Instruments calculators. They eyeball whether or not each required disclosure item is present, make a handwritten note, and then turn to their desktop computer and check a box (or leave it unchecked) in a Microsoft Word document. They use a calculator to add up the line items in balance sheets and income statements. They check mechanical GAAP principles, like that ten-percent rule, the same way–with a calculator. As they confirm that the math is correct and accounting rules are followed, they note each confirmation by hand and in the Word document. The Word document later becomes a review memo that goes to supervisors, who summarize all the review memos for their higher-ups.

Our most sophisticated Enron detection device.

It’s easy to understand why this practice is wasteful. With financial statements now available as data, not documents, software could highlight missing disclosure items, the math mistakes, or any disregard of accounting rules automatically–in real time, as the financial statements are received by the SEC and CorpFin. The agency could even adjust its submission system to automatically reject financial statements with such shortcomings.

Of course, not everything that CorpFin’s reviewers do could be automated. They also carefully read each submission as a whole, and their experience helps them notice when some aspect doesn’t match its context. They compare narrative portions of each disclosure form, like the Management Discussion and Analysis, with numeric portions, like the financial statements. Humans, for now, are better than computers at figuring out whether descriptive text and numeric facts are consistent with one another.

But, five years ago, I estimated that about half of the reviewers’ time was spent verifying and checking aspects of the filings that structured data could, and should, automate.

CorpFin reports 448 full-time equivalents (FTE) in fiscal year 2012 and 471 in 2013 (page 80 of SEC fiscal 2014 budget justification). Judging from the division’s total salary and benefit costs (page 78) of nearly $91 million in 2012 and just over $103 million in 2013, compensating these staffers cost CorpFin about $205,000 each in 2012 and $219,000 each in 2013. Let’s assume, very conservatively, that about one-quarter of CorpFin staffers’ time is spent reviewing existing companies’ disclosures, and three-quarters is spent handling new IPOs, confidential treatment requests, and rulemaking. (Anecdotally, the real proportion of CorpFin time spent on existing company reviews is much higher, especially as new IPOs have cratered since the 1990s.) Half of that review time, or one-eighth CorpFin’s total FTE–56 FTE in 2012 and 58.9 FTE in 2013–could be automated using structured data.

The bottom line? The SEC spent $11.5 million last year and $12.9 million this year on manual pencil-and-calculator labor that could, and should, be instant and automatic.

Chairman Issa has asked Chairman White to explain why, “[r]ather than using the wealth of structured financial data it is already collecting to automate any level of review, the SEC continues to ask Congress for more money to pay for increased staff” (page 3, emphasis added). Backed by the House Oversight Committee’s power to request documents and information from federal agencies, Issa asks the SEC to brief his staff on how–and whether–the agency intends to “incorporate XBRL-formatted financial disclosures into its review process” (page 4, question 4).

(The picture at the Division of Enforcement is not as bleak as at CorpFin. This year the SEC announced that it will use structured data to spot anomalies in financial statements that might indicate fraud. But that project is in its early stages. No Enrons have yet been found. And it will not have an impact on the day-to-day reviews of corporate disclosures.)

2. The SEC Has Never Tried to Enforce the Quality of Financial Statement Data, so Investors Don’t Use it.

Second, structured data hasn’t made life any easier for investors. Last year, Columbia Business School asked investors whether they were using the XBRL versions of corporate financial statements. Fewer than ten percent were (page 61). The overwhelming reason? Data quality (page 28). Investors are not certain that the data in the XBRL files matches the numbers they see in the old-fashioned documents that are submitted at the same time.

(Remember, the SEC has not completely adopted structured data. Companies must file the old-fashioned form as a PDF or HTML document, and submit an XBRL file alongside it. The XBRL file contains a structured-data version of the financial statements and notes, but not the rest of what’s in the old-fashioned form.)

Investors’ concerns are understandable. XBRL US has tracked over 1.4 million errors in the structured data filings since the XBRL reporting requirement began.

It’s not surprising that a new reporting requirement–particularly in an unfamiliar format like XBRL–might have been hard, at first, for companies to figure out. What’s surprising is that the SEC has never asked for a single correction, not even for errors that would have earned an instant reproof if they appeared in the old-fashioned document version. For instance, companies have flipped positive and negative signs on accounting line items–a very simple error, easily spotted and easily corrected–over 400,000 times (table) since XBRL filing began in 2009. The SEC, in all its correspondence with the companies about other aspects of their filings, never asked for any corrections. Small wonder that investors distrust the data.

Investors are calling on the SEC to finally start enforcing the quality of the structured financial data it has been receiving for four years. In July, the agency’s official Investor Advisory Committee recommended “active monitoring” of structured data (page 3). The agency hasn’t yet responded to the recommendations, though it is required by the Dodd-Frank Act to do so.

Chairman Issa asked Chairman White to explain why the SEC “has not issued even one comment letter [from the Division of Corporation Finance] on any of the more than 1.4 million errors identified” (page 3). A bit of enforcement, even as a symbolic signal, would give investors more confidence that the SEC is serious about the quality of structured data submitted to it and encourage them to start using the data to make better, faster investment decisions.

3. Unless the SEC Clarifies its Plans for Open Data, Investors and Companies Can’t Plan for the Future.

Issa’s third criticism is perhaps the most important. The SEC has stopped moving toward open data in corporate disclosures. It has adopted the XBRL format for financial statements and notes, but it still requires the old-fashioned document version to be submitted too.

Of all the securities regulators around the world adopting XBRL (a partial list includes the UK, the European Union, China, India, Japan, South Africa, Singapore, and the United Arab Emirates), not one has adopted the SEC’s half measure. When a jurisdiction chooses to open up its corporate securities disclosures by applying the XBRL format, it switches all the way from documents to data. The only jurisdiction, anywhere, that requires filers to submit two versions of the same information is the SEC.

Moreover, the SEC has given no indication whether it intends to transform other aspects of corporate disclosure–executive compensation, proxy statements, prospectuses, corporate structure, stock ownership–from documents into open data. That means companies don’t know how their reporting requirements will change in the future. They don’t know whether to invest in software that could automate their SEC reporting, because they don’t know whether those reports will be structured data or still be documents and paper.

Investors are demanding open data. The Investor Advisory Committee, when it asked the SEC to fully enforce the quality of existing XBRL filings, also called on the agency to “integrate data tagging into all future rulemaking and rule revision efforts that involve the collection of data by the SEC” (page 3, emphasis added) and to prioritize certain forms for the transformation, including portions of the proxy statement, mutual fund voting disclosures, and corporate current reports on Form 8-K (page 5). The SEC has yet to respond.

Chairman Issa has asked the agency to produce “all documents relating to current use and planned future use” of structured data formats like XBRL (page 5, emphasis added). But he probably would be happy with just one document: the SEC’s plan for the future transformation from documents into open data–if such a plan exists exists.

Issa’s conclusion (page 4) is worth repeating. It applies not only to the SEC but to all attempts to adopt open data in regulatory reporting:

Structured data in financial regulatory reporting has the potential to create profound, positive changes: better enforcement through analytics, more efficient and more accurate reviews, improved market efficiency, cheaper capital costs, and the open data investors are demanding. These revolutionary improvements will only occur as the SEC integrates structured data into its existing review processes, enforces the quality of data submitted under the Interactive Data Rule, and articulates a vision for the transformation of its whole disclosure system from inaccessible documents into structured data.