MD&A Disclosure and the Firm’s Ability to Continue as a Going Concern

In January 2012, the Financial Accounting Standards Board decided by a narrow margin of 4-3 not to require management to perform an assessment of the entity’s ability to continue as a going concern. By May 2012, the FASB reconsidered this requirement and in June 2013 issued an exposure draft that mandates going concern disclosures as part of the financial report. Proponents of this requirement contend that more information is needed from management to inform investors and creditors of impending firm failure, particularly given the spate of recent bankruptcies that have occurred seemingly without warning from either the management or the firm’s auditors. Opponents contend, among other reasons, that managers already disclose sufficient information in their MD&A voluntarily. As such, their view is that an additional disclosure mandate would be an unnecessary imposition on management. In our paper, MD&A Disclosure and the Firm’s Ability to Continue as a Going Concern, which was recently made publicly available on SSRN, we directly inform this debate by assessing whether, to what extent, and when existing disclosures in a firm’s MD&A inform about a firm’s ability to continue as a going concern.

We use a matched pair sample and conditional logistic regression models to assess the ability of current year MD&A textual features to discriminate between (control) firms that continue as a going concern and (treatment) firms that subsequently file for bankruptcy. We examine two textual features of the MD&A: (1) the explicit mention by management of the possibility that the firm may not be able to continue as a going concern, and (2) the overall linguistic tone of the MD&A. Our estimation reveals that firms which explicitly mention the potential inability to continue as a going concern, as well as those using more (less) negative (positive) words in the MD&A, are more likely to cease as a going concern. In terms of economic significance, these three determinants alone yield an area under the ROC (receiver operating characteristic) curve (AUC) of 85%. As 50% represents chance levels and 100% represents perfect ability to discriminate, the evidence is consistent with existing MD&A disclosures providing substantial predictive ability to determine if a firm will continue as a going concern.

After controlling for other information from the financial statements such as financial ratios and the presence of an auditor’s going concern opinion, we find that positive and negative words continue to have predictive ability and the model’s overall diagnostic accuracy improves to 91%, an improvement of 6% over the MD&A only model. The incremental predictive ability of linguistic content is robust to the inclusion of market variables as well as alternative empirical specifications. That linguistic content in the MD&A is an incremental predictor of bankruptcy supports the FASB’s position that mandating disclosures of going concern assessment would help improve bankruptcy prediction.

Naturally, our results alone are insufficient to conclude whether mandatory disclosure is appropriate, because such a determination will require a careful cost benefit analysis. Our analysis, however, offers the following additional insights to inform this cost benefit discussion. Among our treatment sample firms that ultimately cease to continue as a going concern, in only 38% of the cases managers discuss the possibility of ceasing as a going concern. This seemingly low percentage is potentially consistent with the view that managers could do more to inform markets about a firm’s uncertainty related to continuing as a going concern. However, the current voluntary disclosure frequencies by management regarding the ability to continue as a going concern are slightly below what auditors currently provide and already outpace what we observe under the mandatory disclosure regime in Canada. In particular, auditors of our treatment firms issued a going concern opinion in only 43.5% of the cases. A recent study by the Staff on the Ontario Securities Commission using a sample of firms that ceased operations in Canada (OSC STAFF NOTICE 52-719 Going Concern Disclosure Review), where going concern disclosures are currently mandatory, only 27% had specific going concern disclosure in their MD&A prior to bankruptcy.

It is worth noting that while managers may not provide information regarding going concern when it is warranted, we also find that in 8% of the control sample (where firms continue to remain a going concern), managers discuss going concern problems. This suggests that even managers face inherent difficulties in predicting bankruptcy. When examining the time series predictive ability of the financial variables and MD&A disclosures we find intriguing results. Despite the inherent difficulty that always surrounds the prediction of uncertain future events we find that the relative ability of voluntary textual MD&A disclosures over financial variables to predict bankruptcy improves as the prediction horizon increases. That is, the predictive accuracy of a model including only voluntary textual MD&A disclosure (financial statement ratios) is 85% (90%) in the year prior to bankruptcy. Three years prior to bankruptcy, not surprisingly, the predictive ability of voluntary MD&A disclosure (financial statement ratios) decreases to 73% (66%). What is intriguing, however, is that the MD&A disclosures become more predictive relative to financial ratios. While financial ratios are incrementally better predictors of bankruptcy closer to the date of bankruptcy, management opinions and linguistic tone provide relatively more information content three years prior to bankruptcy. This suggests that textual MD&A disclosures are leading indicators of bankruptcy with an overall diagnostic accuracy of 73%. This finding continues to hold when we evaluate the incremental predictive ability of MD&A relative to other predictors.

Although our evidence does not directly speak to whether mandating going concern disclosures by management would improve the predictive power of MD&A disclosure, at a minimum our findings suggest that the voluntary MD&A disclosures in firms’ financial statements provide important insights over financial ratios in assessing a firm’s ability to continue as a going concern. In addition to providing important and timely insights into the current regulatory debate, this study more generally adds to the literature on the informativeness of the MD&A (Brown and Tucker, 2011; Feldman et al. 2010; Cole and Jones 2005), to the bankruptcy prediction literature (Altman 1968; Ohlson 1980; Zmijewski 1984; Shumway 2001; Beaver et al. 2005; Beaver et al. 2012), and to the literature studying the boundaries of the auditor with respect to going concern assessments (Carson et al. 2012). The study also adds to the growing literature on the importance of qualitative disclosure using automated linguistic techniques (Tetlock 2007; Tetlock et al. 2008; Li 2010) and in particular fills the void noted by Li (2011) that linguistic analysis may be useful for bankruptcy prediction.

The full paper is available for download here.