Will XBRL Filings Produce Reliable Data?
March 2, 2009
As of June 15, 500 of the largest U.S. companies will have to file their quarterly and annual reports in the extensible business reporting language (XBRL). The newly interactive filings are intended to supply investors with financial data more quickly and accurately, but there are concerns that high error rates and relaxed auditing requirements will lead to unreliable information.
The Securities and Exchange Commission in December approved rules that will, over the next three years, phase in mandatory filing by all public companies in the Web- and machine-friendly format. But Jeff Mahoney, general counsel for the Council of Institutional Investors in Washington, D.C., pointed to what the commission isn't mandating.
"The SEC could have, but chose not to, require that the annual financial information in the interactive data file be audited by the company's independent auditors, consistent with the data in the company's existing financial statements," said Mahoney. Nor is the agency requiring that the interactive reports be certified by a company's CEO and CFO to the same extent as its current filings, he added.
That makes it hard, asserted Mahoney, to have confidence "that the data has some level of accuracy."
In October 2000, former SEC chairman Arthur Levitt recommended consideration of XBRL, which began life in 1999 as an accounting standard, as a way to increase the amount of data that could be automatically extracted from companies' reports for comparison and analysis. A voluntary filing program started in 2005 but attracted no more than 100 companies in any given year.
In August, the SEC proposed its new filing rules, which Christopher Cox, chairman at the time and a long-time proponent of XBRL, heralded as bringing an "entirely new architecture for disclosure that is based on information, not on forms and transactions."
However, the interactive reports submitted in the voluntary program-participants include Microsoft Corp., Pfizer and United Technologies Corp.-have been plagued by errors as filers have tried to get used to tagging numbers, footnotes and comments, according to a study.
Sixty-eight percent of the XBRL-formatted forms submitted in 2007 were inconsistent with the companies' regular filings, up from 64 percent in 2006 and 59 percent in 2005, say accounting professors Won Gyun No of Iowa State University and J. Efrim Boritz of University of Waterloo in Ontario. But that could be attributed to the number of new filers each year: In 2005, 22 files were created by nine companies; two years later, 67 participants produced 188 reports.
"People should be wary about whether the XBRL-tagged items will be the same ones that currently appear in a table in a document," said William Murphy, CTO of Capital IQ and Compu-Stat, units of Standard & Poor's that provide financial data. "You have to dig deeper than just those numbers."
While complete data for 2008 is not yet available, the quality of the filings "seems to be improved," said No. But a fair number still show errors similar to those found in previous years. A Microsoft filing in October, for example, included several errors, according to No. Missing in one set of calculations was the company's number for short-term debt.
David Blaszkowsky, director of the SEC's office of interactive disclosure, noted that the voluntary XBRL program aimed to highlight potential problems so they could be fixed. "Treating the sandbox like the real world is not helpful to either the sandbox or the real world," said Blaszkowsky. "It runs the risk of raising issues that are not going to be real, while maybe providing ammunition to stop doing something that could actually be important."
