Back to Blog

Inside The AQM, The SEC’s Analytical Tool For XBRL Data

Merrill Disclosure Solutions | December 17, 2013

Dimensions_Dec2013-XBRL Errors

Now that all publicly held companies in the United States must file their SEC financial disclosures in XBRL, the regulatory focus is shifting to the quality of XBRL-tagged data in corporate filings. Moving toward this goal, the SEC is expanding its internal use of XBRL and the quantitative analysis needed to parse XBRL instance documents for review. This push comes at a time when many voices in Washington are urging the SEC to adopt a more forward-thinking approach to the data it collects. The most notable voice so far has been that of Rep. Darrell Issa (R-CA), chair of the House Committee on Oversight and Government Reform, who expressed his views through a letter to the SEC in September 2013. (See Game Changer: Congress Challenges The SEC Over XBRL Quality on page 2 in this issue of Dimensions.)

Perhaps unbeknownst to some observers, the SEC's command of XBRL analytics is already becoming increasingly sophisticated - a fact that filers should remember when considering the quality of their XBRL in disclosures. Given the growing possibility that the SEC will start to issue comment letters and impose liability for filings that are inaccurate because of poor XBRL quality, companies should be familiar with the SEC's analytical methods for reviewing disclosures.

Dimensions_Dec2013-XBRL Errors-SEC Issues Comment Letters

The SEC's New Structured Analytic Tool
One of the SEC's most important new initiatives in the XBRL arena is the Accounting Quality Model (AQM), created while Craig Lewis has been Director and Chief Economist at the SEC's Division of Economic and Risk Analysis (DERA). The AQM is a quantitative analysis tool the SEC will use to review filers' financial disclosure with XBRL tags, flagging apparent anomalies for further consideration by a human examiner. Dr. Lewis discussed the details of the AQM, then under DERA development, in an April 2013 Dimensions interview and again during his keynote address to the 2013 XBRL US National Conference, held September 23-25, 2013, in Las Vegas.

Craig M. Lewis

A Tool To Assist The Review Process, Not Replace It
Dr. Lewis began his address by characterizing his DERA division as “the group that houses the economists.” About half of its 110 staffers have advanced degrees in economics, statistics, mathematics, or computer science. He told the conference that development of the new Accounting Quality Model began with the desire not just to build a risk-assessment tool but also to put “a platform in place that could deliver the results from a risk-assessment program broadly throughout the agency to the various individuals [who] would…be responsible for performing the actual risk assessment.” The platform as created will allow examiners throughout the SEC to perform queries and generate reports. Built to feed this infrastructure, the Accounting Quality Model is “a structured analytic model that takes filer information and identifies outliers,” as Dr. Lewis summarized it. He initially saw the AQM as a tool primarily to assist the Division of Corporation Finance in its  review process, but he thinks the tool also has applications for enforcement. “You can see that the Division of Enforcement has actually ramped up its investigations into accounting fraud, and this is one of the tools that is being used by the Division as they try to identify accounting fraud cases.”

However, he indicated, “the real intent of [DERA creating the AQM] was to assist the review process - in trying to improve the actual quality of the financial disclosures that are being made in firms' 10-Ks and 10-Qs.” The AQM was conceived as a way to help the SEC's review teams “score firms on a aggregate level on the basis of risk, but then actually have an ability to point to areas in a firm's financial statement, before [they] actually open the 10-K, that might [free up] a little more time and energy than taking it and starting from scratch.”

Numerical Factors That Feed The AQM
When analyzing XBRL-tagged financial statements, Dr. Lewis told the conference, the AQM attempts to model a company's “discretionary choices” in its total accruals. Two types of factors feed the model: risk-inducers, reflecting circumstances that might “motivate a firm to manage its income in a particular way”; and risk-indicators, which are elements that suggest a company “may have been actively engaged in some type of earnings management.” If risk-inducers appear in a filer's score, it does not necessarily suggest untoward accounting practices, he explained, but it may indicate to an SEC review team that the company's financials are worth examining in greater detail. The risk-indicators are partly informed by any past SEC accounting or enforcement actions against the company, but they may also stem from the company's recent behavior - for example, a sudden change of auditor. Again, the presence of risk-indicators in a filer's score does not mean the SEC thinks the company has erred or transgressed in its current financial statements. “It's part of a broader profile,” he clarified.

Scores derived from quantitative data in the AQM can be used to rank companies according to the degree of their potential accounting risk. “You could think about using that to actually schedule the review process in CorpFin,” Dr. Lewis observed. “The factors that light up, or are important, also point [you] in the direction you might want to take your review.”

Beyond The Numbers: The AQM's Textual Analysis 
The AQM will eventually reach beyond just numerical data in financial statements. DERA is developing a way for the system to examine the textual information that companies include in their 10-K and 10-Q filings. “The next step in our risk-assessment model,” Dr. Lewis said, “has been to try and use textual analysis - find out what firms are actually saying in their 10-Ks - and try to understand that as well.”

Dimensions_Dec2013-XBRL Errors-AQM Detects Vocabulary Choices

As used for textual analysis, the AQM tries to detect distinctive vocabulary choices in financial statements that are known, from past SEC enforcement actions, to be associated with accounting fraud. Research by DERA has shown that companies engaged in accounting fraud “tend to speak to benign elements in their financial statements much more than non-fraud firms [do], and they tend to under-discuss risks - key risks that are facing the industry.” The AQM converts these textual findings into scores like those generated from numerical data.

“The reason why I am particularly excited about the textual-analysis piece,” added Dr. Lewis, “is that as the factors in the quantitative model point to specific elements that are numeric in a firm's financial statement, the text analysis allows us to point to particular topics…that are either being overreported or underreported.” He cautioned that the textual analysis, like the numerical analysis, is not primarily a fraud-detection device. “It's not necessarily about finding fraudsters. It's about identifying areas where firms can improve the quality of their textual disclosure to investors.”

Just The Beginning For XBRL: "The Utility Will Only Go Up"
He closed by explaining that the AQM may eventually be used for more than helping SEC review teams. It may also be deployed to make XBRL-tagged corporate financials conveniently available to the public. To that end, DERA is considering ways to aggregate corporate XBRL data for use by the market. “One of the things that we're looking at and exploring is the idea of actually putting the base financial statement data out in an aggregated way in a standard data format.” In DERA's vision, analysts and investors would be able to download this data directly from the SEC's website. The first stage of this endeavor would focus on base financials (e.g., income statement, balance sheet, statement of cash flows), and the second would make data from detailed footnotes available.

Dr. Lewis is optimistic about the future not only of the SEC's use of interactive data but of XBRL itself. “I think that there are huge opportunities to do additional academic-style research around footnote disclosures, around MD&A sections, and it will only become more valuable, the utility will only go up, as time goes by and the [XBRL-tagged] database itself grows as more years are added to the database.”

Click here to access all Dimensions eNewsletters

I agree

This site uses cookies to offer you a better experience. For more information, view our privacy policy.