Do going concern disclosures in the management report and audit report signal bankruptcy risk? Evidence from privately held firms

This article provides evidence of the relations between disclosures in the management report of conditions that could challenge the continuance of operations, auditors' going concern reporting and the likelihood of bankruptcy for privately held firms in an environment characterized by low litigation risk. We conduct a matched case – control study using data for 125 firms filing for liquidation bankruptcy and 125 non-bankrupt firms. The results show that disclosures in the management report and auditors' going concern modifications are positively associated with the likelihood of bankruptcy and that their predictive ability is incremental to the control factors. Despite the lower litigation pressure facing private Swedish firms, the informational value of going concern disclosures is comparable with those in litigious environments.

This article provides evidence of the relations between disclosures in the management report of conditions that could challenge the continuance of operations, auditors' going concern reporting and the likelihood of bankruptcy for privately held firms in an environment characterized by low litigation risk.We conduct a matched casecontrol study using data for 125 firms filing for liquidation bankruptcy and 125 nonbankrupt firms.The results show that disclosures in the management report and auditors' going concern modifications are positively associated with the likelihood of bankruptcy and that their predictive ability is incremental to the control factors.
Despite the lower litigation pressure facing private Swedish firms, the informational value of going concern disclosures is comparable with those in litigious environments.

K E Y W O R D S
auditor opinion, bankruptcy prediction, going concern, management's disclosures

| INTRODUCTION
We investigate the association between textual disclosures regarding a firm's future development in the management report, 1 the audit report and a firm's ability to continue as a going concern.In particular, we analyse the extent to which managers' going concern uncertainties and auditors' going concern modifications are associated with the likelihood of bankruptcy using private companies in a low litigation environment.
The quality and reliability of the information provided in firms' financial statements are vitally important for accurate resource allocation decisions.Auditors are responsible for evaluating a firm's ability to continue as a going concern.International Standards on Auditing (ISA) No 570 requires auditors to evaluate the appropriateness of the firm's use of the going concern aspect of accounting, assess whether significant doubt exists about the firm's ability to continue as a going concern and evaluate the adequacy of the financial statement disclosures regarding going concern.The Swedish Annual Accounts Act requires managers to present the information essential for assessing the development of the company's operations, position and results and significant events that have occurred during the financial year in the management report.The management report is a compulsory part of the annual report, and these disclosures indicate whether there are events or conditions that could cast significant doubts on the firm's ability to continue as a going concern.Therefore, both aspects (i.e., management disclosures and auditor going concern opinion) are crucial to the appropriate reporting of going concern uncertainties.
However, very few studies have considered the predictive role of the above two aspects in the probability of bankruptcy.Uang, Citron, Sudarsanam, and Taffler (2006) focus on listed financially distressed companies in the United Kingdom and find that only an auditor's going concern opinion is significantly associated with the severity of following financial stress outcomes.They conclude that the information conveyed by management is mostly arbitrary and unhelpful to users.Mayew, Sethuraman, and Venkatachalam (2015) compare U.S.-listed companies facing bankruptcy in the subsequent year with companies not facing bankruptcy and report the importance of both the auditor's opinion and the management opinion for bankruptcy prediction.They also find that these two opinions are not substitutes and that management disclosures provide incremental information beyond the auditor opinion.
The above (Anglo-Saxon) studies provide evidence of going concern reporting for public companies from strictly regulated settings.
High litigation risk in Anglo-Saxon countries usually works as a means to deter inappropriate reporting.Our study is warranted because it investigates the informative value of the management report and the auditor report from the Swedish institutional environment, which is considerably less rigid than Anglo-Saxon countries.In particular, the litigation risk is relatively low in Sweden (i.e., there are only a few court cases against auditors).Although the Swedish setting differs from Anglo-American countries, it is similar to many European countries characterized by relatively low litigation risk (e.g., Hope & Langli, 2010;Kinney, 1994;Vanstraelen, 2003).Prior studies (e.g., Arce & Mora, 2002;Christensen, Hail, & Leuz, 2013;Francis & Wang, 2008) suggest that the quality of financial reporting is considerably affected by countries' institutional settings.The reason for this is that countries' legal environments affect managers' and auditors' incentives differently due to the varying severity of consequences for undesirable actions.
In addition to this, another fact that adds further relevance to our research is that it is focused on private companies.Given the important differences between public and private companies, it is not obvious without testing that results for public companies will generalize to private companies (e.g., Chaney, Jeter, & Shivakumar, 2004;Hope, Langli, & Thomas, 2012;van Tendeloo & Vanstraelen, 2008).Private companies constitute a significant part of economic activity in Sweden, like in other countries, and are vital for the European economy (Vanstraelen & Schelleman, 2017).This study focuses on two aspects of going concern reporting (i.e., the firm's disclosures and the auditor opinion) in financial reports issued by private firms.A private firm setting is particularly appropriate to examine going concern disclosures.There are several reasons why disclosure practices of managers and auditors in this setting might be different compared with a public firm setting.First, private companies differ from public companies in terms of the agency conflict they encounter (Langli & Svanström, 2014;Minnis & Shroff, 2017).Private firms are often managed by the majority owners; the separation of ownership and control is less of a problem in those firms (Minnis & Shroff, 2017).Minnis and Shroff (2017) argue that the absence of separation and control reduces information asymmetry costs among owners and, thereby, reduces the importance of disclosure to the disclosing firm.Second, unlike public companies, private companies are typically smaller and have a simpler hierarchy, which implies fewer control activities.The relaxed control suggests fewer constraints for dysfunctional management disclosure behaviour (Abdel-Khalik, 1993).Third, because private companies are not scrutinized by financial analysts or stock exchange regulators and are not scrutinized as much by investors as public companies, the probability of audit failure detection and the risk of litigation is much lower in private firms (van Tendeloo & Vanstraelen, 2008).Fourth, it also has been shown that the risk of auditor reputation loss in private firms is lower (Bell, Bedard, Johnstone, & Smith, 2002;Johnstone & Bedard, 2003).These special features of private firms suggest that both managers and auditors face fewer constraints in private firms than in public firms.Moreover, these companies operate in a less rigorous litigation environment, which may also affect managers' and auditors' incitements to disclose the going concern-related information.On the other hand, the lower threat of litigation suggests a lower pressure on managers and auditors and might reduce the number of going concern warnings issued only for fear to cover a litigation risk.It is, therefore, reasonable to examine management and audit reporting in a less litigious environment.
To bridge this gap in our knowledge, we investigate whether and how (i) disclosures of adverse events of conditions in management reports are associated with the subsequent bankruptcy and (ii) auditors' going concern opinions are associated with the likelihood of bankruptcy.
To address the above issues, we use a sample of bankrupt and non-bankrupt private firms.Following prior bankruptcy and going concern research (e.g., Altman & McGough, 1974;Bruynseels & Willekens, 2012;Charitou, Neophytou, & Charalambous, 2004), we adopt a matched pair design matching 125 bankrupt and 125 nonbankrupt firms.First, we find that managers seldom explicitly express their doubts about the firm's ability to continue its operations.This finding differs from prior studies of publicly listed firms in more strictly regulated environments.We further find that managers yet report significant conditions important for assessing the firm's development and other events of material importance to the firm.The disclosure of such conditions and events is associated with a higher probability of bankruptcy.Second, we report a positive and significant association between auditors going concern opinions and bankruptcy outcome.The firm financial condition reflected in the auditor's and the management's reports in Sweden (a less strictly regulated environment) is associated with the subsequent bankruptcy in a way similar to highly regulated (i.e., the United States) environments (Mayew, Sethuraman, & Venkatachalam, 2015).Thus, Swedish managers and auditors seem to have incentives to express opinions indicating going concern uncertainties in a private firm context.However, the proportion of management and auditor reports disclosing going concernrelated information is considerably lower in our study than in Mayew, Sethuraman, and Venkatachalam (2015).Finally, we observe a low proportion of erroneously issued going concern audit opinions.
We contribute to the going concern and bankruptcy literature in the following ways.First, we study two aspects of going concern reporting (i.e., management disclosures and auditor going concern opinion) in the context of privately held firms in the less strictly regulated institutional environment, whereas prior studies (Mayew, Sethuraman, & Venkatachalam, 2015;Uang, Citron, Sudarsanam, & Taffler, 2006) focus on publicly held firms in Anglo-Saxon countries characterized by strictly regulated environments.Our study, thus, provides some insight into managers' and auditors' disclosure practices for private firms that face different incentives than do public firms.Second, we extend the studies on the effects of issuing going concern opinions on the probability of bankruptcy in private firms (Gaeremynck & Willekens, 2003;Vanstraelen, 2003).In addition to auditor opinion, we analyse the effects of the information provided in the management report on the probability of bankruptcy.Furthermore, we extend these two studies by reporting newer evidence.
When these Belgian studies were conducted, a going concern standard had not yet been adopted 2 and 'its predecessor, a non-binding "circular letter" allowing for considerable discretion' was used (Carcello, Vanstraelen, & Willenborg, 2009, p. 1396).Our study covers a period when an auditing standard based on the ISA has been adopted.The adoption of a formal audit standard reduces the extent of discretion, which can influence auditor reporting accuracy (Carcello, Vanstraelen, & Willenborg, 2009).Furthermore, the studied period is characterized by the implementation of the European Audit Directive (2006/43/EC), including the requirement to disclose the engagement partner's identity, which might influence auditor reporting behaviour (Hardies, Vandenhaute, & Breesch, 2018).
The remainder of the paper is organized as follows.Section 2 describes the Swedish institutional setting.Section 3 discusses the prior literature and develops study hypotheses.Section 4 introduces the sample and explains the research design.Section 5 presents the results.Section 6 concludes the study.

| INSTITUTIONAL SETTING
Sweden adopted national auditing standards, Revisionsstandard (RS), based on the ISA in 2004.The standards were translations of the corresponding ISA standards with mostly minor additions or changes.

ISA standards as issued by the International Auditing and Assurance
Standard Board (IAASB) were adopted in 2011.However, the going concern standard ISA 570 was similar to the national standard RS 570 in all relevant aspects.
Most of the observations in the empirical part are from the period when the version ISA 570 issued in December 2009 was effective, and we therefore refer to paragraphs and terminology in that version. 3  ISA 570 § 19 specifies that the auditor may express an unmodified opinion and includes an emphasis of matter paragraph if any events or conditions cast doubt on the entity's ability to continue as a going concern and the disclosures provided by managers in the financial statements are adequate.A qualified or adverse opinion is required if the disclosures are not adequate or if the auditor concludes that the going concern assumption is incorrect (ISA 570 § § 20-21).More specifically, according to ISA 570 § 18, the auditor should evaluate whether the firm adequately describes the principal events or conditions that could cast doubt on the entity's ability to continue as a going concern and the management's plans to deal with those events and conditions.The auditor should also evaluate whether the financial statements disclose whether there is a material uncertainty that could cast significant doubt on the entity's ability to continue as a going concern.Finally, in certain circumstances, the auditor may believe it necessary to request management to extend its assessment.Furthermore, in situations involving multiple material uncertainties that are significant to the financial statements as a whole, the auditor may also consider in extremely rate cases it appropriate to issue a disclaimer of opinion instead of adding an emphasis of matter paragraph (ISA 570, § A 22). Thus, the auditor may issue an emphasis of matter, qualified, adverse or disclaimer of opinion related to going concern uncertainties.
The financial statement disclosures on events that could cast doubt on a firm's ability to continue as a going concern are typically provided in the management report in Sweden but may also be reported in notes to the financial statements.The Swedish Annual Accounts Act (Chapter 6, § 1) regulates that the management report should include information about the conditions that are important for the assessment of the development of the company's operations, position and results and also the events that have occurred during the financial year and are of material importance to the company.The act also includes, since 2005, a requirement to disclose significant risks and uncertainties.However, the regulation includes little guidance on how this disclosure requirement should be applied for the going concern reporting. 4This requirement applies only to large firms (i.e., those that meet more than one of the following criteria in two consecutive years: revenues > SEK 80 million or total assets > SEK 40 million or number of employees > 50). 5   The main laws regulating insolvencies are the Bankruptcy Code and the Reorganization Law.For firms filing under the Bankruptcy Code, the court appoints a trustee who organizes the sales of the assets.Proceeds from the sales are distributed to creditors according to the priority of their claims.A petition under the Bankruptcy Code means that the firm ceases to exist as a juridical entity, although all the assets may be sold to a party that continues the operations (see Thorburn, 2000).In the empirical analyses, we only include firms filing a petition under the Bankruptcy Code (i.e., a 'liquidation bankruptcy' petition).

| LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
A considerable amount of research has been conducted over the years on the association between auditors' going concern reporting and the likelihood of bankruptcy.The vast majority of studies are based on the data on public companies from Anglo-Saxon countries (mainly the United States) (e.g., Altman, 1982;Altman & McGough, 1974;Berglund, 2020;Carson et al., 2013;Citron & Taffler, 1992;Menon & Schwartz, 1987).These studies focus on different periods (between 6 months and 2 years) before bankruptcy (e.g., Altman & McGough, 1974;Citron & Taffler, 1992;Hopwood, McKeown, & Mutchler, 1989) and find that a frequency of issuing of going concern opinion to failing companies does not exceed 50%.In later studies (Berglund, 2020;Carson et al., 2013), however, the proportion of companies receiving a going concern audit opinion 1 year before they filed for bankruptcy is considerably higher.However, there is relatively limited prior evidence on private firms.Gaeremynck andWillekens' (2003) andVanstraelen's (2003) studies are based on private firms in a less rigorous regulation environment, Belgium, and focused on self-fulfilling propensity effects of auditors' going concern opinion.Gaeremynck and Willekens (2003) find that out of 52 bankrupt firms, 79% received a qualified opinion 1 year before failure.Vanstraelen (2003) studies larger private companies and reports that only 37% out of 392 bankrupt companies received a going concern opinion 1 year before bankruptcy.She also controls for bad news disclosed by the board in the annual report.She reports a positive impact of significant events after the closing of the fiscal year on bankruptcy probability.Our study differs from her study in that we explicitly analyse whether management's disclosure of factors that potentially could threaten the continuation of the business is related to the bankruptcy probability.
The above studies indicate that the rate of failing companies receiving going concern opinions is generally not high.The question is whether the relatively low rate results from auditor failure or other explanations for auditors not always qualifying their opinions.Kida (1980) points out that a reluctance to issue qualified audit opinions may be attributed to the unwillingness to lose the client.
Reporting errors can arguably negatively affect auditor reputation.For instance, Xu and Kalelkar (2020) find that Type I errors (if the auditor issues a going concern opinion to the client that does not subsequently fail) are negatively associated with the change in audit office market share and the increased number of clients dismissing an audit office during the following year.Accordingly, several studies suggest that Type II errors (the auditor issues a clean opinion to the client who subsequently files for bankruptcy) can seriously damage auditor reputation (DeFond, Raghunandan, & Subramanyam, 2002;Reynolds & Francis, 2000).A recent study (Berglund, 2020), however, has not found any evidence on audit committees' response to such errors, while negative investor responses appear to be only short-term.The results suggest that the reputational costs of Type II errors to audit firms may not be as extensive as it was previously assumed.
However, according to Mutchler (1984), narrative disclosures provided by management are important for auditors and can influence their going concern decisions.A reliable forecast predicting declining conditions or the lack of a reliable forecast in combination with contrary information could lead to a modified audit opinion, even if the financial statements do not look that bad.Contrarily, given financial statements indicating poor conditions and a reliable forecast predicting improvement, auditors will not be generally inclined to modify their opinions.A few studies (Smith & Taffler, 2000;Tennyson, Ingram, & Dugan, 1990) have explored the association between a management narrative disclosure alone and the probability of bankruptcy and find the importance of the information provided by management for explaining bankruptcy.
The main concerns for inappropriate disclosure seem to be the threat of litigation and reputational cost (e.g., Kasznik & Lev, 1995;Skinner, 1994;Skinner, 1997).Healy and Palepu (2001) suggest that the threat of litigation can affect managers' disclosure decisions in two ways.On the one hand, litigation concern can encourage firms to disclose more information.On the other hand, the risk of being sued by shareholders can negatively influence managers' incentives to disclose particularly forward-looking information.Other management concerns stem from career considerations related to management compensations, promotion opportunities and potential determination (Nagar, 1999;Nagar, Nanda, & Wysocki, 2003).
Prior research suggests that the quality of financial reporting can be influenced by a country's regulatory environment.For example, Geiger, Raghunandan, and Rama (2006) find that the adoption of the Private Securities Litigation Reform Act of 1995, which reduced the threat of litigation against auditors, significantly decreased the probability of issuing going concern modified reports.Fargher and Jiang (2008) report that the proposition of a draft of the Corporate Law Economic Reform Program (CLERP 9) in 2002 in Australia has resulted in an increase in the proportion of firms receiving going concern opinions from 8% to 12%.However, many of the companies receiving modified audit reports have not failed (Type I error).They conclude that the increased propensity to issue modified audit reports does not necessarily increase overall audit quality.
Research on going concern reporting taking into consideration both auditors' going concern paragraphs and disclosures in the management report is relatively limited.Uang, Citron, Sudarsanam, and Taffler (2006) focus on the U.K. regulatory environment and study a sample of 179 listed companies receiving auditors' going concern opinions from 1994 to 2000.Using content analysis methodology, they find that only the tone of auditors' going concern opinions is related to the severity of the subsequent outcomes.In contrast, the tone in directors' statements is not significantly associated with the level of severity of the subsequent outcome.They emphasize the arbitrariness and unhelpfulness of managements' information and conclude their reluctance to signal their going concern-related problems.Mayew, Sethuraman, and Venkatachalam (2015) investigate a sample of 460 listed U.S. firms filing for bankruptcy from 1995 to 2012 and find that both audit opinion and management opinion are significantly related to the probability of bankruptcy.Their analyses also suggest that management disclosure has an incremental informative power of bankruptcy.Therefore, both studies find a significant informative power of auditors' going concern opinions, although the evidence regarding the predictive ability of management opinions in predicting subsequent financial stress outcomes is not consistent.Hence, more research is needed to clarify the role of the two aspects of going concern opinion in predicting bankruptcy.

| Management going concern disclosures
What is similar in the above research is that the studies focus on listed firms and are performed in Anglo-Saxon litigation environments.
These studies, however, provide mixed evidence on the informative ability of management reports.In contrast, our study focuses on smaller privately held firms performing in a less litigious environment.
Lower risk of litigation may tempt managers to withhold their doubt regarding the company's ability to continue its operations.Another reason for management reluctance to disclose their doubt is that such negative information will contribute to the company's bankruptcy (i.e., self-fulfilling prophecy effect).Also, the context of private firms suggests fewer levels of control activities, which can provide more latitude for dysfunctional management behaviour (Abdel-Khalik, 1993).
In addition, less information asymmetry costs (because of more concentrated ownership) reduce the importance of disclosure to the disclosing firm (Minnis & Shroff, 2017).These factors may enable managers to conceal significant negative information about their firms, leading to a weaker association between the disclosure of significant negative information and subsequent bankruptcy.
On the other hand, managers have superior information about the company, and they are responsible for providing a true picture of the firm's position for creditors, suppliers and other stakeholders.The moral obligation to do so might outweigh other reasons.Similarly, it may be important for managers to maintain successful long-term relationships with their creditors, which requires building up a significant degree of confidence.For these reasons, managers could be interested in disclosing the relevant firm information.Moreover, the environment with a greater risk of litigation can negatively influence managers' incentives to disclose forward-looking information (Healy & Palepu, 2001).This suggests that managers might be more inclined to provide a true picture of the firm's position in a lower-litigious setting.
In such case, a stronger association between management reports and subsequent bankruptcy will be observed.This leads to our first hypothesis: H1.The disclosure of events or conditions that may cast doubt on the entity's possibility to continue operations in the management report is positively associated with subsequent bankruptcy.

| Auditor going concern disclosures
Auditors in a lower litigious setting may be less motivated to report a client's going concern uncertainties objectively.Lower litigation risk and absence of pressure from the public capital market may reduce the severity of consequences associated with improper audit reporting.Prior literature suggests that the risk of litigation is the important factor in the auditor's opinion decision.For instance, Vanstraelen (2002) finds that the auditor's decision to disclose going concern uncertainty is influenced by economic incentives such as loss of audit clients and audit fees in the business environment characterized by a low risk of litigation, such as Belgium.In contrast, in a highly litigious U.S. business environment, no evidence that auditors' incentives affect their going concern decisions is found (Louwers, 1998).Vanstraelen (2002) concludes that litigation may be an overriding factor in the economic trade-off made by the auditor.
The lower litigation risk implies that other auditors' incentives may not come into play.The above arguments lead us to predict that auditors are less inclined to issue going concern warnings in this environment.One would observe a weaker association between the auditor going concern disclosure and subsequent bankruptcy in such a case.Some 100 disciplinary proceedings are initiated each year.There are around 4000 auditors in Sweden during the sampling period, suggesting that around 2.5% of auditors are sanctioned each year. 6  Swedish auditors seem to have incentives to provide appropriate reporting of going concern-related issues.Besides the incentives to issue appropriate going concern reports, there is another factor that may increase auditor propensity to disclosure going concern uncertainties in the environment with less rigorous regulation.The low pressure on auditors (because of a lower litigation threat) may reduce the number of erroneously issued going concern warnings (that are only issued to cover litigation risk), leaving only warnings for the most distressed companies.This may strengthen the association between the auditor going concern disclosure and subsequent bankruptcy.Accordingly, our second hypothesis is as follows: H2.The going concern paragraph in the audit report is positively associated with subsequent bankruptcy.
The financial data analysed in this study is obtained from Serrano, a database including financial information on all Swedish firms.The sample is composed as follows.We first searched for large firms that had filed for liquidation bankruptcy in Sweden over the 2006-2015 period.The reason why we exclude smaller firms is that only large firms have to disclose information about future prospects in the management report.Furthermore, we require that the last available annual report should be prepared between 30 and 730 days before bankruptcy.The financial data in our study are, therefore, from the years 2004 to 2014.
There are 595,838 firms in Serrano with financial data for one or more years over the 2004-2014 period, and of these, 16,575 firms meet the criteria in the Accounting Act for being a large firm (i.e., at least two of the following criteria are met in two consecutive years: revenues > SEK 80 million or total assets > SEK 40 million or number of employees > 50).The auditor's assessment of a firm's ability to continue as a going concern normally covers 12 months (ISA 570 § 13), which means that we predict bankruptcy over a longer period than what is covered by the auditor.The drawback of this is that a failure to issue a going concern paragraph cannot necessarily be interpreted as a Type II audit error.However, our main purpose is to study the informational value of audit report and management report variables.
Many privately held firms overlook the requirement to publish their financial statements before bankruptcy, and we would end up with a smaller and less representative sample by studying only firms with financial reports 1 year before bankruptcy.There are 39,230 bankrupt firms in Serrano with financial data available for the period covered, and of those, 320 are large firms.The omission of one bankrupt financial institution leaves 319 large bankrupt firms.We next omitted publicly traded firms and required that auditor data and variables to calculate the probability of bankruptcy would be available, leaving 165 firms.
We searched for matched pairs from the same year and industry (measured at the two-digit level).We required that the non-bankrupt firms had not filed for bankruptcy or initiated a liquidation in any of the years covered by the Serrano database (i.e., until the end of 2017).Furthermore, we required that the difference in the probability of bankruptcy of the bankrupt firm and the matched pair would be less than 30%. 10These criteria resulted in an omission of 18 pairs, leaving 147 matched pairs.Finally, we excluded one pair due to the

| Empirical model
We use the following equation to test the hypotheses: where BRUPT is an indicator variable taking the value 1 for the bankrupt firms and 0 for the matched pair.The definitions of the variables included in the model are provided in Appendix A.

| Independent variables
To test Hypothesis 1, we consider the following measures to capture qualitative information in the management report: (i) GC_MGMT, The coding of GC_MGMT follows Mayew, Sethuraman, and Venkatachalam (2015).The variable is coded with 1 if the firm explicitly refers to the term going concern or discuss the firm's ongoing operations. 12However, firms frequently describe adverse events and circumstances without explicitly stating that those events may cast doubt on the firm's ability to continue as a going concern.Therefore, we use CONDITIONS_MGMT as a second measure.This variable takes the value 1 if the management report (or notes to the financial statements) includes the following types of adverse information: (i) The firm has (or may get) problems paying its debt (e.g., that a firm has defaulted on debt payments); (ii) the firm is undergoing (or about to undergo) a restructuring; (iii) losses have consumed a significant part of the equity, and the firm may have to liquidate; (iv) the firm has a large receivable, and its collectability is uncertain; or (v) the management report discloses other unfavourable events or conditions.These disclosures arguably suggest events or circumstances that cast significant doubt on the firm's ability to continue as a going concern.
The following motivates the inclusion of categories (ii) and (iii).
The outcome of a restructuring process is uncertain, and typically, an unsuccessful reorganization results in bankruptcy.Therefore, by including the information about a reorganization in the management report, a firm discloses an unfavourable condition that may affect its ability to continue operations.Similarly, a firm should, according to the Company Act, liquidate if losses have consumed more than half of the firm's share capital, also indicating that the firm's ability to continue to operate is threatened.Appendix B includes examples of statements expressing the circumstances resulting in the coding of CONDITIONS_MGMT with 1.

Third, we use INFO_MGMT(POS) and INFO_MGMT(NEG) to test
Hypothesis 1.To classify this variable, we use a principle similar to that of Skinner (1994) and Boo and Simnett (2002).Based on the information content of the management report, the variable INFO_MGMT(POS) takes the value 1 if the information is positive, and INFO_MGMT(NEG) takes the value 1 if the information is negative.The variables take the value 0 if the provided information is neutral.The researchers and a research assistant's coding of the management report variables have been done based on PDF files with annual reports.The variable GC_AUD is used to test Hypothesis 2. It assumes the value 1 if the audit report includes a going concern remark irrespective of type. 13

| Control variables
We include the following control variables.Research suggests that the probability of bankruptcy depends on the size of the firm (e.g., Ohlson, 1980).The size is measured by the log of sales (LNSALES).Further, prior research suggests that various board characteristics affect the probability of bankruptcy.To control for the association between board size and default probability (e.g., Laitinen, 1999;Fich & Slezak, 2008;Schultz, Tan & Walsh, 2017), we include LNBOARDSIZE.Studies document the various positive and negative effects of multiple board memberships (e.g., Kaczmarek, Kimino, & Pye, 2014).To study whether multiple board memberships are associated with bankruptcy risk, we include the log of the average number of assignments of the board members (LNASSIGNMENTS).Finally, we control for the possible negative effect of board members' resignations by including the dummy variable RESIGNED.
We also control for various auditor variables.Firstly, we include a BIG 4 dummy. 14Sundgren ( 2009) studied a sample with financially weak firms and found that liquidating bankruptcy was less common for firms with a Big 4 auditor.We also include AUD_SWITCH to control for the possible effects of audit partner changes on bankruptcy risk.Studies suggest that auditor resignations are positively associated with client risk factors (e.g., Ghosh & Tang, 2015), and we, therefore, expect a positive association between auditor switches and bankruptcy risk.
We use an algorithm that minimizes the difference between the probability of bankruptcy (PROBZ) for the bankrupt and non-bankrupt pair when we select the pairs.Although it is possible to identify a match with a similar probability in most cases, the difference is quite large for some pairs.Following the suggestion of Cram, Karan, and Stuart (2009), we include PROBZ to control for possible effects of imperfectly matched variables.
As firms in corporate groups may offer financial support to other firms in the group (e.g., Beaver, Cascino, Correia, & McNichols, 2019), we include the indicators PARENT, SUBS_DOM and SUBS_FOREIGN to control for their impact on bankruptcy risk.SUBS_DOM and SUBS_FOREIGN are dummy variables that take the value one for subsidiaries with a domestic and foreign parent, respectively.Firms that do not belong to a group are in the comparison category.Finally, the regressions include indicators for each matched pair (PAIR_FIXED_EFFECTS).  Vanstraelen, 2007).Hardies et al. study a sample with Belgian firms of a similar size as in our sample, and they find that 31.7% of the bankrupt firms in their sample received a going concern opinion.However, our sample size is relatively small and, therefore, is not directly comparable with the prior studies.

| Going concern uncertainties in the management report and the audit report
Considering the management report variables, Panel A shows that GC_MGMT takes the value 1 only for 2.40% (3/125) firms, showing that the management very seldom explicitly state that there are uncertainties about whether a firm can continue as a going concern.
The proportion is 5.88% (2/34) when we study firms with financial statements less than 1 year before bankruptcy.
In one of the three reports, management concludes that the operations will be discontinued.In the other ones, they conclude that a prerequisite for the firm's survival is that a reorganization plan is approved or that the firm's survival is uncertain until a reorganization plan is approved.Panel B reveals that no non-bankrupt firms disclose going concern uncertainties.
Mayew, Sethuraman, and Venkatachalam (2015) use a similar coding for a sample with publicly listed firms from the United States.
They find that management expresses uncertainty about being a going concern in 39.3% of the bankrupt firms' reports.The text in the American going concern standard is similar to ISA 570.An explanation for the difference in results is that the more litigious environment in the United States may bring management to disclose more information.If the evidence in the American study were used as a minimum benchmark for adequate reporting practices, there would be a great need for improvement in Sweden.
The auditor can issue an emphasis of matter, qualified, disclaimer or adverse going concern opinion.As discussed in more detail in Section 2, the appropriate kind depends on the information disclosed on the going concern issue in the management report.Untabulated results show that out of the 31 bankrupt firms' audit reports with a going concern paragraph, 29 include an emphasis of matter opinion.
There are no observations in the sample when the auditor has issued a qualified or adverse opinion related to the going concern uncertainties.Thus, the auditors have not for a single firm in the sample concluded that disclosures in the Management Report are inadequate enough to issue a qualified opinion.
Furthermore, the auditor issues an emphasis of matter opinion for 26 of 29 firms although the firm has not explicitly expressed uncertainties about the firm being a going concern in the management report, that is, when GC_MGMT = 0 (not reported in tables).This suggests that auditors do not require an explicit statement about going concern uncertainties in the management report to issue an 'emphasis of matter' opinion.The likely reason is that the regulation is ambiguous.Although ISA 570 requires explicit disclosures to warrant an emphasis of matter opinion, the accounting rules are less precise, leaving leeway for managers.Furthermore, there is no authoritative guidance in Sweden on applying that paragraph when the firm's ability to continue as a going concern is uncertain.Because there is no precise guidance in accounting regulation, managers may find it pointless to disclose going concern uncertainties.Disclosure may even be harmful as there is a risk of it becoming a self-fulfilling prophecy (Gaeremynek & Willekens, 2003;Vanstraelen, 2003).
CONDITIONS_MGMT focuses on the disclosure of events and conditions that may adversely affect the firm's ability to continue operations.However, different from GC_MGMT, we do not require an explicit statement that the events and conditions may cast significant doubt on the entity's ability to continue operations to code The sample includes firms that filed for bankruptcy within 2 years from the balance sheet date.When we limit the sample to firms that filed for bankruptcy within 1 year, the percentage of firms with CONDITIONS_MGMT equal to 1 is 23.53%.
ISA 570 states that an auditor should issue a qualified opinion instead of an emphasis of matter paragraph if disclosures in the financial statements are inadequate.CONDITIONS_MGMT may be interpreted as the minimum information that is needed to warrant an emphasis of matter opinion, and we next explore its association with the auditor's going concern reporting.Panel A shows that CON-DITIONS_MGMT takes the value 1 for 26 of the 31 bankrupt firms with a going concern opinion.That is, the association between CON-DITIONS_MGMT and GC_AUD is strong but not perfect.The auditor has for one of the five firms with CONDITIONS_MGMT equal to 0 issued a disclaimer of opinion and an emphasis of matter opinion to the remaining four.The management reports of these four firms include no or little information about adverse events and conditions, indicating that some auditors have a very low threshold for concluding that there is sufficient information in the management report to warrant an emphasis of matter opinion.In conclusion, very few firms explicitly refer to going concern uncertainties in the management report, and around one-fourth of the bankrupt firms disclose events and conditions that may adversely affect the firm's ability to continue operations (without explicitly referring to going concern uncertainties).Furthermore, the results reveal that auditors frequently issue an emphasis of matter opinion.
However, there is no explicit mentioning of going concern uncertainties in the management report or notes to the financial statements.Still, most (but not all) management reports include some information about adverse events and conditions when the auditor issues an emphasis of matter opinion.Our interpretation of these results is that auditors mostly adhere to the ISA 570 § § 18-19 but apply a liberal interpretation.Finally, the results show that most bankrupt firms adequately disclose negative information in the management report, indicating that they do not try to mislead users of financial statements by being overly optimistic.The firm has or may have problems paying its debt 1 13

| Descriptive statistics on control variables
Losses have consumed a significant part of the equity, and the firm may have to liquidate unless more equity is injected The firm has large receivables whose collectability is uncertain 0 4 Other unfavourable events or conditions in the management report 1 13 Note: that the mean and median risks are similar for bankrupt and nonbankrupt firms.Furthermore, the table reveals no significant differences in LNSALES between the bankrupt and non-bankrupt firms.
These results suggest that the bankrupt firms and the matched pairs are similar with respect to size and bankruptcy risk.
The main observation with regard to the board variables is that board size (LNBOARDSIZE) is higher for bankrupt firms (p value = 0.053).The table also indicates that the average number of board assignments (LNASSIGNMENTS) is lower for bankrupt firms (p value = 0.071).Furthermore, we can see that parent companies comprise a smaller share of bankrupt firms and that subsidiaries with a domestic parent comprise a larger share of bankrupt firms (p value < 0.001).Finally, the table reveals that bankrupt firms less frequently have a Big 4 auditor (p value = 0.068).
Table 4 presents the correlation matrix.Overall, the correlations between the variables are low.However, an exception is the correlation between GC_AUD and CONDITIONS_MGMT that is 0.783.

| Regression results
A conditional (fixed effect) logistic regression is used to estimate the Equation 1. Conditional logit is a regression with dummy variables for each pair in the data and is designed for matched case-control data (Hosmer & Lemeshow, 2000, pp. 223-243). 15The z values of the logistic regressions are based on Huber-White robust standard errors (Rogers, 1994) using conventional maximum likelihood estimates.
Because GC_MGMT takes the value 0 for all non-bankrupt firms, we report only results with CONDITIONS_MGMT, INFO_MGMT(POS) and INFO_MGMT(NEG) as management report variables.Regressions (1) and ( 2) report results for all observations.In regressions (3) and (4), we drop two firms with a disclaimer of opinion and their matched pairs, and therefore, those regressions report results for 246 observations for which the auditor has issued an emphasis of matter going concern opinion or no going concern at all.ISA 570 stipulates that the auditor should issue an emphasis of matter opinion only if the disclosure of material uncertainty related to the going concern issue in the financial reports is adequate.Therefore, GC_AUD and CON-DITIONS_MGMT will be highly correlated if auditors adhere to the requirements in ISA 570, and CONDITIONS_MGMT adequately measures the quality and quantity of information in the management report that the auditor requires to warrant an emphasis of matter opinion.An implication of this for the empirical analyses is that regressions ( 1) and (3) may suffer from problems with multicollinearity.To examine that issue, higher than 5. Multicollinearity may increase the variance of coefficient estimates and may produce coefficient estimates of an implausible magnitude. 16 Regressions ( 1) and (3) in Table 5 show that CONDITIONS_ MGMT has a positive coefficient significant at the 0.01 level.Regression ( 2) and ( 4) report results with INFO_MGMT(POS) and INFO_MGMT(NEG).They show that negative information in the management report is significantly positively associated with bankruptcy, whereas positive information is not significantly associated with bankruptcy.Furthermore, GC_AUD has a positive coefficient significant at the 0.01 level in all regressions.The results indicate that disclosures in the management and auditor reports have an informative value in private firms in a less strictly regulated environment.
Further, based on Table 5, the following conclusions can be made in relation to the control variables.Firms audited by Big 4 auditors seem less likely to go bankrupt, whereas firms that have switched audit partners are more likely to fail.An explanation of the former result is that a higher quality auditing reduces information asymmetry problems between the firm and creditors, which reduces the likelihood of liquidating bankruptcy as the resolution of the financial difficulties (Sundgren, 2009).A possible reason for audit partner switches being associated with bankruptcy is that auditor switches indicate a higher business risk for the client and, thereby, a bankruptcy risk (cf.Ghosh & Tang, 2015).Table 5 also shows that LNBOARDSIZE is positively (at the 0.05 level) related to bankruptcy risk.The result is consistent with some prior evidence (Fich & Slezak, 2008).Indeed, the findings in the literature are ambiguous, as some research also suggests that there is no significant association between board size and the likelihood of failure (Wang & Deng, 2006;Schultz et al., 2017).
The LNASSIGNMENTS is negatively (at the 0.10 level) related to bankruptcy risk.This finding corresponds with Platt and Platt (2012), who found that non-bankrupt firms have a significantly higher proportion of interlocked directors than bankrupt firms.They suggest that the advantage of interlocked boards is enhanced information flows between firms and industries.

| Robustness tests
We conduct several additional analyses to test the robustness of the results.The tests and their results (untabulated) are described below.First, we exclude the bankrupt firms and matched pairs with the difference in PROBZ in the 90th percentile (i.e., ΔPROBZ exceeding 2.5%).Results are qualitatively similar.Then, we replace PROBZ with NITA, CACL and TLTA, that is, the variables in Shumway's (2001) bankruptcy prediction model.The reason for this is the possibility of fairly large differences in these ratios between the groups of bankrupt and non-bankrupt firms (even if ΔPROBZ is low).These results are also qualitatively similar to the ones presented in Table 5.
We also attempted to exclude bankrupt firms if the time between the balance sheet date and bankruptcy is more than 365 days.After these omissions, 34 bankrupt and 125 non-bankrupt firms remain in the sample.We estimate regressions (1)-(4) without the fixed effects for each matched pair.The results show that INFO_MGMT(NEG) has a positive coefficient significant at the 0.01 level, but CON-DITIONS_MGMT has insignificant coefficients.Thus, these results only partially support our finding in Table 5.Further, the result shows that GC_AUD has positive coefficients significant at the 0.05 level in three of the four regressions.
The identification of events and conditions that may cast significant doubt on a firm's possibilities to continue its operations involves some judgement, and we, therefore, study an alternative management report measure.Panel C in Table 2 reports the kind of events and conditions motivating the coding of CONDITIONS_MGMT with 1.
Although we believe all conditions included are of a kind that indicates a firm may face problems in continuing its operations, the least obvious is the firm's 'large receivables whose collectability is uncertain', and the most indistinct the 'other unfavourable events or conditions in the management report'.The exclusion of these statements results in a measure taking the value one for 2 non-bankrupt firms and 25 bankrupt firms.This measure is positively significant at the 0.01 level in (1) and (3).

| Discriminative power
One way of evaluating the economic significance of the variables is to study the proportions of correctly classified bankrupt and nonbankrupt firms.We compare the proportions of correctly classified firms using ordinary binary logit models without dummy variables for the matched pairs (not reported in the tables).We firstly regress Equation 1 without the test variables, and this model classifies 71.20% of the bankrupt and non-bankrupt firms correctly using 0.50 as the cutoff point.
We then estimate models with only (i) CONDITIONS_MGMT, (ii) INFO_MGMT(POS) and INFO_MGMT(NEG) or (iii) GC_AUD.These models classify 74.80%, 74.00% and 77.20%, respectively.We can, therefore, conclude that the accuracy improves more when the model includes the audit report variable than the management report variables.Finally, the models classify 77.60% of the bankrupt and nonbankrupt firms correctly when we include the audit report and management report variables, irrespective of the management report variables used.In conclusion, the improvement in classification is small compared with a model with only GC_AUD.An alternative way to study the discriminative power is to compare the area under the receiver operating characteristic (ROC) curve.

Dependent variable
We use the method suggested by Cleves (2002)

| Roles of managements and auditors opinions
Following Mayew, Sethuraman, and Venkatachalam (2015), we next study whether the auditor's going concern opinion complements or substitutes for the managerial going concern disclosure using mediation analysis.In mediation, the purpose is to analyse if an independent variable X influences a mediator variable M, which, in turn, influences Y.
Indeed, the X variable may directly influence Y and influence Y through M. The first effect is the direct effect of X, and the second effect is the indirect effect (e.g., MacKinnon et al., 2007;Zhao, Lynch, & Chen, 2010).
We assume management's reporting can influence the probability of bankruptcy directly and indirectly through the auditor's reporting.
More explicitly, we study whether CONDITIONS_MGMT à BRUPT and CONDITIONS_MGMT à GC_AUD à BRUPT.The assumption underlying these paths is that the management's reporting influences the auditor's reporting, not vice versa.This assumption follows Mayew, Sethuraman, and Venkatachalam (2015) and is also based on the principle that the auditor decides its reporting based on information in the financial statements.We use structural linear equations (SEMs) to estimate the following equations 17 : Coefficient β 1 is the direct effect of CONDITIONS_MGMT on the probability of bankruptcy (after controlling for GC_AUD).Assuming CONDITIONS_MGMT causes GC_AUD, the indirect effect can be calculated as α 1 β 2 and the total effect as α 1 β 2 + β 1 (e.g., Zhao, Lynch, & Chen, 2010).Furthermore, CONDITIONS_MGMT has a positive coefficient significant at the 0.10 level in regression (2).However, the total effect of CONDITIONS_MGMT on the probability of bankruptcy is much higher than what is revealed from regression (2) in Table 6 because a part of the total effect of CONDITIONS_MGMT on the probability of bankruptcy is mediated through GC_AUD.The bottom part of Table 6 reports the direct, indirect and total effects of CONDITIONS_MGMT on the probability of bankruptcy, and the results show that the total effect of CONDITIONS_MGMT is significant at the 0.01 level and that 63.1% of the total effect is mediated through GC_AUD.

Regression (1) in
In sum, the regressions support the view that the causal direction goes from management's reporting to the auditor's reporting, that is, that the auditor considers the management's report when (s)he decides whether to issue a going concern opinion or not.Therefore, the total impact of management's reporting when assessing a firm's bankruptcy probability may be much greater than what is indicated from the direct effect reported in the right-hand column in Table 6.
Indeed, we cannot rule out that the causal association is reversed, that is, that management takes the auditor's intended going concern reporting into account when it decides what information to include about the going concern issue in the audit report. 18

| CONCLUSIONS
Prior studies suggest the auditor's and management's opinions have explanatory power in models predicting the likelihood of bankruptcy.
However, these analyses (mostly in Anglo-Saxon studies) are based on public companies operating in strict litigation settings.Our study is carried out in Sweden, which represents a considerably different institutional environment characterized by low litigation risk.Also, contrary to prior studies, we focus on privately held companies that make up an essential part of the European economy.Given this different study setting, we investigate the informativeness of going concern disclosures in the management and audit reports in relation to subsequent bankruptcy.To measure the impact of the management report, we use the following variables: (1) management going concern disclosure, (2) unfavourable conditions disclosures and (3) disclosures of positive/negative events.The effect of the auditor report is measured using the variable auditor going concern remark.
We use a matched pair design with 125 bankrupt and 125 nonbankrupt firms.First, we test the informational value of management reports and find that the disclosures very seldom explicitly refer to the term going concern or express in other ways doubts about the firm's ability to continue its operations.This finding differs from prior studies of publicly listed firms in more litigious environments.The expression of going concern doubts implies a kind of trade-off between potential self-fulfilling prophecy effect and litigation risk, and it likely differs in the less litigious Swedish setting from that in an Anglo-American setting.A further factor arguably deterring Swedish firms from disclosing going concern-related problems is that accounting standards do not include an explicit requirement to disclose those circumstances during the period we study.However, the results indicate that managers yet report the important conditions for assessing the firm development and other events that are of material importance to the firm.We find that the disclosure of unfavourable firm conditions and negative events is significantly associated with the probability of bankruptcy.Second, we test the relation between auditors going concern opinions and subsequent bankruptcy and find a positive and significant association.Our results, therefore, indicate that both the management report and auditor report have an informational value in terms of bankruptcy prediction.These findings are consistent with the study of Mayew, Sethuraman, and Venkatachalam (2015) and demonstrate that the results hold in a less litigious environment and on data comprising privately held firms.
However, the proportion of firms where managers disclose significant information and auditors issue a going concern audit opinion is considerably lower in our study compared to Mayew, Sethuraman, and Venkatachalam (2015).
A care because the wrong classifications do not necessarily imply an audit failure.Finally, there are most likely differences in how exactly the management's going concern reporting in privately held firms is regulated around the world.The Swedish rules are relatively imprecise, and it would be interesting to know if the information contents of management's going concern reporting depend on regulation.
cern reporting in the later part of the years we study in the empirical section.Sundgren and Svanström (2014a) report that there were 193 sanctions against auditors in the 2009-2012 period, and of these, 21 (10.88%) were related to the firm's going concern reporting.In the 2004-2008 period, there were 283 sanctions against auditors, and of these, only one was related to the going concern reporting.
7 Notice that a matched pair design is not equal to propensity score matching.In a matched pair design, the participants in the samples should be chosen so they are similar on all characteristics except the ones under investigation.In our case, the bankrupt and non-bankrupt firms should be similar on all characteristics than on their going concern and management report variables.However, because a large number of factors may affect the probability of bankruptcy, we are not able to control on every aspect, and we, therefore (in addition to the effects we control through matching), add some control variables (see (GC_MGMT = 1).
• 'Since the earnings trend has been weak during the first months of the financial year, the company has not been able to meet the conditions (covenants) that are associated with the external financing.
An agreement has been reached between the parties, which, under certain given conditions, ensures the company's financing'.
• 'Renegotiation of the loan terms with … Corporation was underway in 2012, and an oral agreement was reached in 2013 on the extension of existing loans that were due'.
• 'As a short-term loan debt, the company reports 8,484,000 SEK on financing an aircraft.The loan is due for payment in August 2008 and has not yet been renegotiated or refinanced'.
• 'During the financial year, it has been found that the company does not currently meet all the conditions set out in the principal's loan terms.A discussion is ongoing between the company and the lender about how the conditions should be met again.No loans are cancelled from the lender'.
• 'The company is undergoing reconstruction from 2010-06-24.The Board sees good opportunities for this process to lead to the

Abbreviation Definition
Dependent variable: BRUPT An indicator variable taking the value 1 if the firm files for liquidation bankruptcy within 730 days after the balance sheet date.
Test variables: GC_MGMT An indicator variable taking the value 1 if management makes a statement that the firm may be unable to continue as a going concern.
CONDITIONS_MGMT An indicator variable taking the value 1 if the management describes adverse events and conditions that are of a nature that the firm may become unable to continue as a going concern.Different from the coding of GC_MGMT, we do not require that the firm explicitly mention that events and conditions may threaten the firm's ability to continue as a going concern.

INFO_MGMT(POS)
An indicator taking the value 1 if the firm conveys positive news in the management report and 0 if it communicates neutral news.

INFO_MGMT(NEG)
An indicator taking the value 1 if the firm conveys negative news in the management report and 0 if it communicates neutral news.

RISK_MGMT
An indicator taking the value 1 if the firm discloses serious problems (e.g., a poor liquidity, an uncertain future, that the firm reorganizes or that it should liquidate according to the rules in the Company Act).

GC_AUD
An indicator variable set to 1 if auditors expressed substantial doubt about the firm being able to continue as a going concern in the audit report.PARENT An indicator taking the value 1 if the firm is a parent in a group.

SUBS_DOM
An indicator variable set to 1 if the firm is a subsidiary to a Swedish parent.

SUBS_FOREIGN
An indicator variable set to 1 if the firm is a subsidiary to a foreign parent.
company being able to continue in the future as far-reaching negotiations are underway with new stakeholders.These financiers intend to raise new capital for the company'.
• 'During a corporate restructuring, a large number of negotiations are conducted to reduce costs and improve the living conditions of the company in question … The company's definitive conditions following a corporate restructuring are, of course, uncertain until a potential agreement on proposals gains legal force.With the negotiation results that can be assumed to be achieved internally, [the company] seems to regain its profitability after the reconstruction has been completed.In light of this, the annual report has been prepared assuming continued operations'.
• 'In connection with this work, a question has arisen about the company's solvency, which is why the Board of Directors decided at the beginning of 2010 to establish a control balance sheet and reduced share capital to 1 00,000 SEK.In connection with this, another action plan is developed to seek to secure the company's operations'.
• 'Going concern: The … Group has a capital need that is under discussion with owners and new investors.A decision on capital injection has not yet been made'.
• In May … applied for company restructuring … as the company's work to reach a long-term financing solution could not be completed in time to avoid a liquidity shortage.The application was approved, and the company is in a restructuring process … The company's equity is still consumed when this annual report is signed.
• 'In the agreement with the bank consortium, a large part of the borrowing matures at the end of 2015.Since it is obvious that the company cannot repay the loan at this time, negotiations with the credit institution have already begun in autumn 2014.Since the banking consortium consists of several banks, the negotiations have been difficult and time-consuming.It is necessary for continued operations that a new agreement is reached in these negotiations.The Board is of the opinion that the company has sufficient resources for continued operations, which is why the balance sheet and income statement have been prepared under the assumption of continued operations under normal business conditions'.
• 'In parallel with the implementation of the savings programme, the company has extensive negotiations with a financier and a term sheet has been established.The purpose of the transaction is to restore capital and create good conditions to continue the business.The conditions described above constitute uncertainty factors that can lead to significant doubts about the company's ability to continue operations.Therefore, the company may not be able to realize its assets and pay its debts in the normal course of business'.
• 'During the year, hardware and consultancy sales were lower than expected.This has resulted in a lower level of earnings.The outlook for the future is uncertain in the current economic climate'.
• 'The company's largest customer reduced significantly order volumes during the year.To meet the reduced demand, the company decided to discontinue operations in … The company's earnings for the year were affected by these closure costs.Some of the closure costs have been reported as items affecting comparability'.
• 'The trend during the year has not been the desired one.Already at the start of the year, there was an awareness of the need for development and change as last year's results were negative.
Therefore, several efficiency measures and cost-saving measures were implemented during the year …' • 'During the year, the company experienced a stronger than expected decline in the total expansion of district heating networks in Sweden.This has, of course, affected the company to a great extent, and the consequences are that sales have halved since 2008.The parent company will raise capital in 2011 if it proves necessary to ensure continued operation of the company'.
• 'The Group's largest customers reduced significantly on ordered volumes during the year.In order to cope with the reduced demand, the unit in … was closed … The result this year is charged with closure costs and other restructuring costs'.
On the other hand, one may argue that auditors nevertheless have some incentives to disclose going concern uncertainties in a less strictly regulated environment.There are different mechanisms in the Swedish institutional setting for ensuring auditor independence.The Swedish Accountants Act ( §21) includes rules prohibiting auditors from conducting the audit in situations that could impair auditor independence.Swedish laws and practice with regard to independence are, in general, consistent with the directives of the EU (2014/56/EU; 2006/46/EC), the EU recommendation on auditor independence (2002/590/EC) and the IFAC's Code of Ethics for professional accountants (2006).Furthermore, the Supervisory Board of Public Accountants (SBPA) issues disciplinary sanctions against auditors.
missing annual report of the bankrupt firm, one pair due to missing control variables, 15 pairs of public firms have a sample with exclusively privately held firms and five pairs that were part of the same bankrupt group.Thus, 125 bankrupt firms and 125 matched non-bankrupt pairs are left after these omissions.The mean (median) assets of the bankrupt firms are SEK 321.2 (93.2) million.The mean and median assets of the non-bankrupt firms are SEK 1068.5 million and SEK 108.0 million, respectively. 11For comparison, the mean and median assets of large privately held firms in Sweden over the 2004-2014 period are SEK 1190.3 million and SEK 112.1 million, respectively.Although the mean assets of the nonbankrupt firms are significantly smaller than those of all firms, we cannot reject the null hypothesis that the median assets of the bankrupt firms are significantly different from the median assets of all bankrupt firms (p-value = 0.48 for a Wilcoxon signed-rank test), suggesting that the bankrupt firms in the sample mirror the entire population of large firms fairly well.The mean and median assets for all privately held Swedish firms reported above are based on 29,470 observations for 6341 firms over the 2004-2014 period.The data are taken from Serrano.

CONDITIONS_MGMT with 1 .
Panel C shows the distribution of different kinds of the information included in CONDITIONS_MGMT for bankrupt and non-bankrupt firms.It reveals a considerable difference in all kinds of disclosed information between bankrupt and nonbankrupt firms.Panel A shows CONDITIONS_MGMT takes the value 1 for 26.40% (33/125) of the bankrupt firms, and Panel B shows that the corresponding percentage for non-bankrupt firms is 2.4% (3/125).

Finally,
Panel A shows that 79 out of the 125 bankrupt firms' management reports include negative information and only two positive information, suggesting that most bankrupt firms properly disclose negative information.Panel B shows that the number of nonbankrupt firms with positive information is 17, and the number with negative information is 32, indicating that the non-bankrupt firms' management reports are more optimistic.
The variables are explained in Appendix A.
to evaluate the significance of the difference in the area between models.Compared with the model without the test variables, the inclusion of (i) CONDI-TIONS_MGMT, (ii) INFO_MGMT(POS) and INFO_MGMT(NEG) or (iii) GC_AUD significantly increases the area under the ROC-curve (p value < 0.05).Furthermore, the area under the ROC curve is significantly greater when the logit regression includes INFO_MGMT(POS), INFO_MGMT(NEG) and GC_AUD than when the model includes only GC_AUD and control variables.However, the inclusion of CON-DITIONS_MGMT to a model with GC_AUD does not significantly improve the area under the ROC curve.In sum, the results show that management report variables or going concern paragraphs are important predictors of bankruptcy.Still, we find only limited support for going concern paragraphs and management report variables being important predictors.
the value 1 if at least one board member has resigned.BIG4An indicator taking the value 1 if the firm's auditor is PwC, KPMG, EY or Deloitte.AUD_SWITCHAn indicator taking the value 1 if the auditor partner has changed.

Table 1
displays the industry and year distributions.The largest industries are the manufacturing of industrial goods (14.80%), shopping goods (20.00%), finance and real estate (14.40%) and corporate services (24.40%).The years with the largest number of observations are2007, 2010 and 2011.
Carey, Kortum, & Moroney, 2012;Carson et al., 2013;we study firms filing for bankruptcy within 1 year.Panel B shows that 1.60% (2/125) of the non-bankrupt firms receive a going concern opinion.Overall, the Type II error rate appears high, and the Type I error rate appears low.These Type II (Type I) error rates are also higher (lower) than in prior studies (e.g.,Carey, Kortum, & Moroney, 2012;Carson et al., 2013; MGMT(NEG)).Overall, the chi-square tests reported inTable 2 reveal significantly positive associations between the auditor's and management's reporting.The only insignificant association is between GC_AUD and INFO_MGMT for non-bankrupt firms.Panel A in Table 2 also shows that 24.80% (31/125) of the bankrupt firms receive a going concern opinion before bankruptcy.
T A B L E 2 The association between disclosures in the management and audit reports (Continues) (Continues)

Table 3
presents the mean and median values of the control variables.Using PROBZ as the measure of the bankruptcy risk, the table showsT A B L E 2 (Continued) The Pearson chi-square test in panels A and B examines the null hypothesis that the auditor's going concern opinions are unrelated with management's going concern reporting in the management report.Variables are explained in Appendix A. *Denotes two-tailed statistical significance at the 0.10 level.**Denotes two-tailed statistical significance at the 0.05 level.***Denotes two-tailed statistical significance at the 0.01 level.
Table 5 reports variance inflation factors (VIFs).Descriptive statistics on control variables The p values are for a two-tailed t test for the continuous variables and for a Pearson chi-square test for the dichotomous variables.The variables are explained in Appendix A.
for a conditional logit model.The VIFs of GC_AUD are 5.61 and 5.79 in regressions (1) and (3), respectively.The VIF of GC_MGMT is 5.79 in regression (1) and 6.02 in regression (3).The VIFs of INFO_MGMT (POS), INFO_MGMT(NEG) and GC_AUD in regressions (2) and (4) are lower.There are various rules of thumb for VIFs in the literature, but one benchmark is that variables are highly correlated if the VIF is T A B L E 3 The table reports conditional logit regressions estimates.Regressions (1) and (2) report results for the full sample and regressions.In regressions (3) and (4), we only include observations with an emphasis of matter going concern opinion.Z values based on robust standard errors are in parentheses.The VIF values are from an OLS regression with fixed effects for the matched pairs.The dependent and independent variables are explained in Appendix A.

Table 6
further noteworthy finding in the study is that the Type II error rate is high.A possible reason is that in the private firm setting, audi-The relations between management disclosures, auditor disclosures and bankruptcy emphasis of matter opinion only if the management report includes adequate disclosures).A further limitation is that our sample includes firms going bankrupt within 2 years from the balance sheet date because many privately held firms overlook the requirement to publish their financial statements before bankruptcy.This means that the result concerning Type II audit error should be interpreted with great T A B L E 6 Note: The table reports linear structural equation model results.t values based on robust standard errors are in parentheses.The dependent and independent variables are explained in Appendix A. *Denotes two-tailed statistical significance at the 0.10 level.**Denotes two-tailed statistical significance at the 0.05 level.***Denotes two-tailed statistical significance at the 0.01 level.
Shipman, Swanquist, & Whited, 2017eview of methodological issues related to the use of matched samples).In propensity score matching, the matching is on the treatment variable (seeShipman, Swanquist, & Whited, 2017for a review of the use of propensity score matching in accounting research).In our case, we have several treatment variables (i.e., auditors' going concern opinions and management report variables) and propensity score matching, therefore, not a feasible alternative.