In his continuing dig into XBRL submissions to assess utility and functionality, Charles Hoffman is finding income statement data hard to figure out.

Hoffman is often credited with founding the concept of XBRL and inspiring the American Institute of Certified Public Accountants to invest in developing it. He has been studying more than 5,500 XBRL filings to the Securities and Exchange Commission to get a sense of the semantic model of financial reports that XBRL produces.

The assessment of the balance sheet and the cash flow were straightforward enough, he said. But the income statement is a lot more challenging. In fact, he calls it maddening.

“The income statement has a lot more moving parts,” he said. Not all income statement elements apply to all filers in the same way. Items like income from operations, discontinued operations, noncontrolling interests, extraordinary items, perhaps even income taxes make it impossible to build a one-size-fits-all income statement.

That makes it hard for filers to sort through the GAAP Taxonomy, he says, and find the concepts that are appropriate to their particular data. If companies had a series of examples to follow, it might give them a clearer roadmap to submitting data that would ultimately be more comparable, he says.

The income statement also doesn’t carry subtotals in the same way as the balance sheet or the cash flow statement, making it more difficult at time to cross reference or verify numbers, he said. All of that, combined with the nature of XBRL user software, makes it difficult for users of income statement data to find what they might be looking for, he said.

To illustrate, Hoffman says he has discovered that filers are generally using three different concepts in the Taxonomy to report net income or loss. He worries perhaps there is too much ambiguity as to which of the three most chosen concepts should be used under different reporting scenarios.

Of nearly 4,000 filings he has examined, nearly three-fourths report income (or loss) from continuing operations before taxes, but a wide variety of labels are used, and many filers even created extensions for this concept, giving it the exact same name as the concept provided by that name. Hoffman wonders if some filers carried it forward from an earlier XBRL submission when the taxonomy in place did not have a concept for this item.

He found revenue for 84 percent of the filings where he searched for it. He’s pretty sure the other 16 percent report revenue somewhere, but he’s just not sure where. “There are so many different ways to report revenue,” he said. Some companies provide a total for revenue, making it easier to find, but some do not, making it much more difficult to find, he said.

Hoffman said he’s hopeful his exercise will help practitioners focus more on how to make XBRL more consistent across companies and therefore more useful to users of financial information.

Charles Hoffman, often credited with founding XBRL, has been studying more than 5,500 XBRL filings to the Securities and Exchange Commission to get a sense of the semantic model of financial reports that is emerging through the XBRL process. He’s curious to see how well SEC filings adhere to that model.

He analyzed both 10-Ks and 10-Qs looking for some patterns of consistency in how companies are selecting tags to relate common financial reporting concepts. Hoffman found that 21 different generators were responsible for creating his sample of 5,525 filings, which logically suggests there might be more diversity in implementation than if the entire sample were generated by a single source or only a few sources.

The analysis focused on some fundamental elements that should be common among virtually all financial reports for public companies: assets, liabilities, equity, profit, loss, cash, revenue, basic entity information and the like. Despite the large number of sources, 3,170 of the filings, or 57 percent, had a combination of all the key characteristics where Hoffman was hoping to see some patterns.

It’s an impressive indicator of the quality of the XBRL submissions given how new the technology is, in Hoffman’s view. “There is definitely a leveragable semantic model peering out from that set of 5,525 SEC financial filings,” he wrote.

Hoffman found every sample reported the concept for the entity’s name, but only 21 percent reported the concept for their trading symbol. He surmises registrants must not find the trading symbol concept to be useful. While every sample reported the document period end date, Hoffman found on closer examination that about 20 filings reported it as the submission date or some other date rather than the balance sheet date.

While 97 percent of all filings reported a concept for assets and for liabilities and equity, he found five filings where they didn’t balance, either in the HTML or the XBRL filings. Hoffman found that 96 percent of filings reported one of four concepts for equity – stockholders equity, stockholders equity including the portion attributable to noncontrolling interest, partners’ capital, and partners’ capital including the portion attributable to noncontrolling interest – and he sensed some inconsistency in the use of the equity concepts.

Hoffman was disturbed to note that net income/loss is being reported under three different concepts. There are separate concepts in the taxonomy for profit/loss, net income/loss, and net income/loss available to common stockholders. “Basically, there is ambiguity as to which of the three of those concepts to use in different reporting scenarios,” he wrote.

Hoffman says his next move will be to drill another level lower into the cash flow statement, which he expects to be easy. Then he will take on the income statement, which he expects to be more difficult because there is more variety in how companies report income, so it will be more difficult to detect patterns. “A bottom-up approach may be superior to a top-down approach,” he wrote.