The prediction of future earnings using financial statement information: Are XBRL company filings up to the task

Financial statement data for large companies became available to the public in XBRL format starting in 2009 in the United States. Proponents of XBRL, along with the SEC, argue that XBRL filings offer several advantages over data provided by data aggregators, such as lower cost, faster availability, and broader coverage. The purpose of this study was to contribute to the common body of knowledge by investigating whether current XBRL company filings are useful in the prediction of future earnings and to attempt to interactively obtain the balances of 70 accounting concepts needed to create an earnings prediction model from a sample of XBRL filings. Current XBRL filings do not allow for interactive extraction of required accounting elements because too many accounting elements are missing from the XBRL filings. Accordingly, an additional data set was created by manually populating missing accounting concepts within the XBRL filings if sufficient component accounting concepts existed within the same XBRL filing (e.g., if current liabilities and long-term liabilities were tagged in the XBRL filing, total liabilities could be calculated). This process mimicked what could be performed by added functionality built directly into the XBRL taxonomy. This functionality would not create any excess time, effort, or cost for preparers or users. This fully populated XBRL data set allows the user to create earnings prediction models interactively, whereas the current XBRL data set does not. This indicates that current XBRL company filings are likely to be limited in their usefulness in other areas as well, while a more fully populated set of XBRL company filings that includes additional data has the potential to improve the usefulness of XBRL data.