FORTUNE — It was supposed to be the great democratizer in financial information.
Without a pricey subscription to a data service like Bloomberg or Capital IQ, with eXtensible Business Reporting Language (XBRL), even the smallest investor would be able to access and analyze company financial results the instant they became public.
But the quixotic implementation of XBRL has created a costly and almost useless product whose many flaws are now serving as justification to slow the inevitable and important march toward low-cost, machine-readable financial results for all.
Currently, almost all publicly traded companies are required to file their annual 10-K and quarterly 10-Q financial reports in XBRL format, but a bill before Congress would eliminate the XBRL filing requirement for “small” companies, which would include more than half of all filers.
The justification for eliminating this is as flawed as the data many companies publish in XBRL. Executives argue that XBRL filings create an unnecessary expense because so few people are using the filings, despite their potential to vastly expand access to information, improve market efficiencies, and reduce the likelihood of another Enron-style fraud.
MORE: Warren Buffett: Coke exec compensation plan was excessive
So few people use these filings, in part, because public corporations produce embarrassingly poor XBRL documents. There are errors throughout XBRL filings. In its most recent filing, for example, Pinnacle Financial Partners reported dozens of numbers as 1,000 times larger than they should have been. Perhaps more disturbing: Companies may be intentionally abusing extensions or custom tags used to identify numbers.
But first, a bit about how XBRL works. Companies attach tags to each number and footnote in their filings. There are more than 9,000 tags approved by the Financial Accounting Standards Board to identify financial information. Since it is impossible to create tags to cover every item for every company without the system becoming unwieldy, companies are allowed to create custom tags, called extensions.
Extensions damage XBRL’s usability. If an investor wants to compare total revenue growth for companies across an industry, she would need only to search for the revenue tag and the industry code. But if one company in that industry uses an extension to describe revenue, that company will be excluded.
Managers who want to hide results can do so simply by creating extensions, and the SEC, which is expected to use XBRL to help it search for accounting fraud, does not oversee the use of extensions or punish companies for abusing them.
And one doesn’t need to go far down the Fortune 500 list to find signs of possible abuse. Exxon Mobil (XOM) uses an extension to describe its line item “Total revenue and other income” in its income statement, while Chevron (CVX) uses an approved tag to describe its revenue, also described as “Total revenue and other income.” Exxon did not respond to a request for comment.
This practice is common. Recent research has found that 40% of extensions in SEC filings are semantically identical to machine-readable tags that have been approved. So, why do companies use extensions? Are they simply too lazy to search the approved list, or is it more intentional?
In research I conducted with 28msec, a company that has developed a platform to extract information from XBRL filings, I found evidence that companies may be learning how to use extensions to effectively hide bad results. As companies become more proficient at using XBRL, they use more extensions, contradicting the SEC’s prediction that extension use would decline with increased familiarity with the system. In addition, increased extension use is associated with stock mis-pricing, as well as greater errors and less agreement in analysts’ forecasts on a given company’s earnings.
MORE: Meet Synack, a new security startup started by NSA veterans
Firms that use extensions are more likely to have performed poorly in the given earnings period and have a greater spread between reported income and actual cash flow, an indicator of possible accounting errors or manipulation and one of the factors that the SEC looks for when it uses machine readers to hunt for fraudulent activity.
XBRL is broken, but that does not diminish its potential importance to markets and regulators. By voting to eliminate reporting requirements for a majority of publicly traded firms, Congress is doing a disservice to investors and signaling to companies that make questionable accounting decisions that they have a sympathetic ear in Washington.
What is needed for XBRL now is not a step back to reflect, but a sprint forward to increase oversight and improve its usability.
Ethan Rouen is a Ph.D. candidate in accounting at Columbia Business School.