#Fail: Social Media, Firm Distress, and Going Concern Opinions
Abstract
Audit firms and regulators have both commented extensively on the potential for new sources of data to transform the audit process. Focusing on auditors' going-concern opinions, we use deep learning to measure the "bearishness" of posts on social media and find it strongly predicts the likelihood of firm failure. This association is incremental to other market-based signals, such as a firm's default likelihood or short interest. Interestingly, this signal appears largely orthogonal to an auditor's going-concern opinion, implying that social media predicts future events that precipitate failure not fully considered by auditors. While we fail to observe a direct association between bearishness and going concern opinions, our evidence does suggest that going concern accuracy improves with bearishness. Finally, we consider potential channels for these results and find that bearishness foreshadows difficulties in raising capital, predicting the likelihood of future credit downgrades and equity issuances. Our evidence should be informative to regulators and audit firms, both of whom are currently evaluating the usefulness of "new" data to auditors.