What drives the consequences of intentional misstatements? Evidence from rating analysts reactions. Martin Schmidt * Martin Bierey **

Transcription

1 What drives the consequences of intentional misstatements? Evidence from rating analysts reactions Martin Schmidt * Martin Bierey ** * ESCP Europe, Berlin, Heubnerweg 8 10, D Berlin, Germany mschmidt@escpeurope.eu ** ESCP Europe, Berlin, Heubnerweg 8 10, D Berlin, Germany mbierey@escpeurope.eu (corresponding author) Acknowledgements: We thank Ulf Brüggemann, Joachim Gassen, David Hillier, Mark Nelson, Beatriz Garcia Osma, William Rees, Edward Riedl, Thorsten Sellhorn, Florin P. Vasvari, Teri L. Yohn and the participants of the 2015 AAA/Deloitte/J. Michael Cook Doctoral Consortium, the 2015 EAA Annual Congress, the 2015 EAA Doctoral Colloquium, the 2014 IAFD Symposium in Trondheim, the research seminar at the Humboldt University of Berlin, and the ESCP research workshop for their comments. Access to restatement data was provided by Karen M. Hennes, Andrew J. Leone and Brian P. Miller, and is gratefully acknowledged.

2 What drives the consequences of intentional misstatements? Evidence from rating analysts reactions This version: April 2016 Abstract This paper analyzes the adverse consequences that firms face after an intentional misstatement becomes publicly known. Using a mixed methods approach, we examine (i) the mechanisms through which misstatements adversely affect misreporting firms, and (ii) to what extent and how persistently these mechanisms adversely affect firms credit ratings. A content analysis of rating reports reveals seven mechanisms. Rating analysts are most concerned about misstatement-related violations of debt covenants that increase a firm s liquidity risk. We find that the identified mechanisms have a persistent adverse effect on firms credit ratings. After a misstatement becomes publicly known, we find a significantly negative effect on credit ratings in the subsequent six years. The adverse effect of the misstatement on firms credit ratings is particularly pronounced in cases where rating analysts express concern about misstatement-related violations of debt covenants in their rating reports. Our results suggest that misstatement-related violations of debt covenants appear to be the most severe mechanism through which misstatements adversely affect firms creditworthiness. JEL Classification: M40, M41, G24 Keywords: credit ratings, financial reporting quality, misstatements, restatements, liquidity risk, debt covenants, consequences of misreporting

3 1 1. INTRODUCTION Firms that misstate earnings face severe capital market consequences. After a misstatement becomes publicly known, firms inter alia face higher cost of equity capital (Hribar and Jenkins 2004), larger bid-ask spreads (Dechow et al., 1996; Anderson and Yohn 2002) and a decline in the credibility of their accounting information (Anderson and Yohn, 2002; Wilson, 2008; Chen et al., 2014). Most visible is the substantial market value loss, which ranges from 20% to 38% after a fraud-related misstatement becomes publicly known (Palmrose et al., 2004; Karpoff et al., 2008a). However, there is considerable controversy about the mechanisms that cause the documented capital market effects. 1 A simple framework can be used to classify these mechanisms by splitting the observable market value loss into two components, a mean effect and a variance effect (Hribar and Jenkins, 2004). Investors will revise expected future cash flows when the misstatement becomes publicly known (mean effect), because their cash flow projections have to incorporate a firm s restated financial situation and expected cash outflows due to legal penalties. However, the documented consequences of misreporting cannot be explained by investors incorporating the firm s restated financial situation and expected legal penalties alone. In fact, Karpoff et al. (2008a) estimate that only a third of the misstatement-related market value loss is attributable to these two mechanisms. It appears that investors also perceive additional risks to which the firm is exposed subsequent to a misstatement becoming publicly known (variance effect). The firm is perceived as being more risky and consequently investors demand higher cost of capital (Hribar and Jenkins, 2004). The additional risks that investors perceive include those that arise from the misstatement becoming publicly known, and risks that existed before but are revealed by the misstatement becoming publicly known. In this regard, prior studies have primarily focused on one misstatement-related risk, namely the reduced credibility of the firm s accounting information which adversely affects the firm for up to three years after an intentional misstatement becomes publicly known (Anderson and Yohn, 2002; Wu, 2002, Wilson, 2008; Chen et al., 2014). To the best of our knowledge, this is the only misstatementrelated mechanism for which the persistence has been documented. 1 In our context, we define mechanisms as the factors that market participants are concerned about subsequently to a misstatement becoming publicly known, and which eventually cause the adverse consequences that are observable.

4 2 Previous studies have focused on the observable consequences of misreporting, but not on the mechanisms that cause these adverse consequences. In fact, from the findings on the aggregate market reaction to misstatements (e.g., cost of capital increase, market value loss) one can hardly infer which mechanisms caused the documented consequences. Thus, there is limited knowledge with respect to (i) which mechanisms cause the adverse consequences that are observable after misstatements, (ii) which of these mechanisms have the most severe impact on misstatement firms, and (iii) how long misstatement firms are adversely affected by these mechanisms. Using a mixed-methods approach, our study examines these questions by focusing on the adverse impact that these mechanisms have on firms creditworthiness. We first examine credit rating analysts reactions to intentional misstatements qualitatively by conducting a content analysis of the rating reports that are released after a firm s intentional misstatement becomes publicly known. If rating analysts incorporate misstatement-related mechanisms into their rating decisions, they will specify them in the underlying rating report. Thus, the content analysis allows us to examine the adverse mechanisms that rating analysts are concerned about after an intentional misstatement becomes publicly known. Seven mechanisms are mentioned by rating analysts. Subsequent to a misstatement becoming publicly known, rating analysts are most concerned about misstatement-related covenant violations that lead to an increase in firm s liquidity risk. In the majority of these cases, the misstatement causes a substantial delay in the firm s 10-K/10-Q filings, and thereby the firm violates debt covenants under existing loans that specify when financial statements must be provided to the creditor. This, in turn, entitles creditors to demand accelerated repayment, or to re-negotiate the terms of the debt agreement, which puts substantial pressure on a firm s liquidity situation. Rating analysts are concerned about further misstatement-related mechanisms, namely: (i) the incorporation of a firm s restated financial situation, (ii) internal control weaknesses, (iii) legal penalties, (iv) misstatement firms distraction from the operating business, (v) firm s loss of reputation among investors and other stakeholders, and (vii) the inability to assess a firm s financial condition due to the absence of financial statements. In the second part of our study we analyze the impact that these mechanisms have on firms credit rating. Our sample consists of 3,619 US firms that have a credit rating assigned by Standard & Poor (S&P). The sample is comprised of 199 firms with intentional misstatements, 251 firms with unintentional misstatements, and 3,170 firms without alleged misstatements according to the Governmental Accountability Office (GAO) Financial

5 3 Restatement Database, and the Securities Exchange Commission (SEC) enforcement releases. We regress firms S&P credit rating on 11 rating determinants (e.g., firm size, leverage, profitability) and control for year-fixed and firm-fixed effects. This design allows us to empirically model a firm s rating ( predicted rating ) and to analyze the deviation of the actual rating from the predicted rating. We include indicator variables for the firm-years that lie after a firm s misstatement becomes publicly known. These indicator variables capture the deviation of misstatement firm s actual credit rating from the predicted rating in the years subsequently to the misstatement becoming publicly known. If misstatement-related mechanisms adversely affect credit ratings, we should observe a negative deviation of a misstatement firm s actual rating from its predicted rating when the misstatement becomes publicly known. In subsequent years, this negative deviation should gradually decrease if the adverse impact of misstatement-related mechanisms fades away during the years following the misstatement becoming publicly known. This research design enables us to analyze the within-firm variation of credit ratings and the adverse effect of misstatement-related mechanisms while controlling for the real consequences of misstatements (e.g., declining profitability, restated financial situation). Our findings show that, in the year of the intentional misstatement becoming publicly known and the following six years firm s credit rating is significantly lower than the rating that is predicted by rating determinants. This indicates that the full range of misstatementrelated mechanisms has a persistent adverse effect on firms credit ratings. The effect is most pronounced in cases where rating analysts are concerned about an increased liquidity risk resulting from misstatement-related debt covenant violations. In the year of the misstatement becoming publicly known, the mean rating of these firms is four rating notches lower than their predicted rating. Our paper contributes to the literature in three respects. First, we shed light on the types of misstatement-related mechanisms that credit rating analysts (creditors) are concerned about. In fact, our findings show that they are most concerned about the increased liquidity risk that results from debt covenant violations. While this mechanism has not received attention in the misreporting literature, it appears to be the most severe mechanism through which misstatements adversely affect firms creditworthiness. Equity investors are assumed to be equally or even more concerned about this mechanism given that covenant violations transfer control rights to creditors (Chava and Roberts, 2008).

6 4 Second, we contribute to the misreporting literature by documenting that intentional misstatements have a persistent adverse effect on a firm s creditworthiness. While Chen et al. (2014) document that equity investors perceive misstatement firms accounting information as less credible for three years, our findings illustrate that firms endure adverse consequences from misstatement-related mechanisms for a much longer period. In fact, the persistent adverse effect on a firm s creditworthiness might be one of the most severe consequences of misreporting; this is the case, because after an intentional misstatement becomes publicly known, firms have only restricted access to equity financing and heavily rely on debt financing (Chen et al., 2013). Third, we contribute to the literature on debt covenants by illustrating a situation in which covenant violations are extremely costly. Covenant violations are common and creditors will not necessarily impose high costs on violating firms (Dichev and Skinner, 2002). However, the costs of covenant violations increase with the information asymmetry between creditors and violating firms (Gao et al., 2015). Subsequent to an intentional misstatement becoming publicly known, firms are adversely affected by a number of mechanisms that impede creditors ability to assess the creditworthiness of the misstatement firm. This increases the information asymmetry between creditors and misstatement firms and decreases creditors willingness to waive their right to accelerated repayment. Our findings show that covenant violations in this situation can substantially endanger firms liquidity situation and, in a number of cases, even lead to bankruptcy. 2. MOTIVATION The literature on the capital market consequences of corporate misreporting is broad (for an overview, see Dechow et al., 2010). Studies have typically analyzed observable consequences resulting from misreporting rather than the mechanisms that cause these observable consequences (i.e., the factors that market participants are concerned about subsequently to a misstatement becoming publicly known). There is extensive empirical evidence on the adverse capital market consequences, such as a substantial market value decline (Dechow et al., 1996; Anderson and Yohn, 2002; Hribar and Jenkins, 2004; Palmrose et al., 2004), a decline in analyst earnings forecast accuracy and greater forecast dispersion (Palmrose et al., 2004), a decline in the information content of future earnings announcements (Anderson and Yohn, 2002; Wu, 2002; Wilson, 2008; Chen et al. 2014), an increase in the cost of equity capital (Hribar and Jenkins, 2004) and larger bid-ask spreads (Dechow et al., 1996; Anderson

7 5 and Yohn, 2002; Palmrose et al., 2004). These studies have focused on equity investors, and only a few have analyzed the consequences of misreporting from the perspective of creditors. Graham et al. (2008) document that banks require higher spreads and use tighter contract terms (e.g., stricter covenants and/or more collateral) for loans that are granted to firms after an intentional misstatement becomes publicly known. Cornil (2009) analyzed the bond market reaction and documents a significant increase firm s cost of debt immediately after the restatement announcement. Similarly, Park and Wu (2009) analyzed the secondary loan market and show that the misstatement announcement leads to abnormally negative bond returns and higher bid-ask spreads. Leng et al. (2011) analyze the long-term consequences of misstatement firms that were cited in the SEC s Accounting and Auditing Enforcement Releases (AAERs). Their findings show that firms exhibit poorer operating performance, negative abnormal stock returns and higher default probability in the years after the SEC issued an AAER. 2 While the aforementioned studies provide evidence on the observable consequences of misreporting, there is controversy about the mechanisms through which misstatements adversely affect firms. These mechanisms could either contribute to (i) investors revising expected cash flows downwards (mean effect) or (ii) investors requiring a higher cost of capital because they perceive additional risks that the misstatement firm is exposed to (variance effect). The mean effect is caused by mechanisms that will directly affect a misstatement firm s expected future cash flows. There are two main mechanisms that underlie investor revisions of expected future cash flows. First, after a misstatement becomes publicly known, market participants base their expectations of future cash flows on the revised financial situation, as presented in the (restated) financial statements, whereas the original expectation was based on the (erroneous) financial situation, as presented in the misstated financial statements. 3 Second, the misstatement firm might be subject to legal penalties triggered by SEC enforcement actions or shareholder lawsuits. Karpoff et al. (2008a) document that incorporating the firm s restated financial situation and legal penalties (i.e., the mean effect) only accounts for a third of the market value loss that is observable after a fraud-related misstatement becomes publicly known. This indicates that there have to be additional 2 3 However, the authors do not examine the specific duration over which the consequences of the misstatement adversely affect firms. Karpoff et al. (2008a) refer to this effect as a readjustment effect.

8 6 mechanisms that contribute considerably to the documented capital market consequences of misreporting. The variance effect is caused by investors perceiving that the firm is exposed to additional risks (misstatement-related risks) subsequent to a misstatement becoming publicly known. These risks are associated with a higher degree of uncertainty compared to the mechanisms underlying the mean effect. The amount of the restatement and expected legal penalties will directly translate into investors future cash-flow projections. In contrast, predicting the impact of misstatement-related risks and whether they will have an impact on future cash flows is associated with a higher degree of uncertainty. Investors are also unaware as to whether these risks are persistent, or will resolve themselves in the short term. Previous studies have not examined which specific mechanisms cause the observable consequences of misreporting. Besides obvious mechanisms such as legal penalties or the incorporation of the firm s restated financial situation, one can infer from the findings on a firm s bid-ask ask spread increase (Dechow, 1996; Anderson and Yohn, 2002; Palmrose et al., 2004; Park and Wu, 2009) and increased analyst forecast dispersion (Palmrose et al., 2004) that investors perceive greater information asymmetry. In addition, the decline of the firm s earnings response coefficient (Wu, 2002; Wilson, 2008; Chen et al., 2014) suggests that, from the perspective of investors, the misstatement erodes the credibility of firm s accounting information. In fact, this loss of credibility is the most frequently listed misstatement-related mechanism (e.g., Dechow, 1996; Wu, 2002; Anderson and Yohn, 2002; Hribar and Jenkins, 2004; Wilson, 2008; Graham et al., 2008; Park and Wu, 2009; Chen et al., 2013; Chen et al., 2014). 4 In addition, it is the only mechanism for which the persistence has been documented by prior studies. Chen et al. (2014) document that subsequent to an intentional misstatement becoming publicly known, equity investors perceive firms earnings announcements as less credible for a period of three years. However, is the reduced credibility of accounting information the most severe mechanism through which intentional misstatements adversely affect misreporting firms? In addition, the question arises whether there are further mechanisms (besides the incorporation 4 Some studies use different wording to describe the decline in the credibility of a firm s accounting information, such as the credibility of firm s financial disclosures (Dechow, 1996), decrease in investors perceptions about the reliability of the accounting information (Anderson and Yohn, 2002) and uncertainty with respect to [ ] overall perception of earnings quality (Hribar and Jenkins, 2004).

9 7 of a firm s restated situation, legal penalties and the reduced credibility of accounting information) that adversely affect misreporting firms. From the studies that analyze misstatement firms changes in market value, the cost of capital, bank-loan contracting, bond returns, operating performance and default risk, it is difficult to infer which mechanisms cause the consequences documented. These studies offer mechanisms such as uncertainty with respect to management credibility and managerial competence (Hribar and Jenkins, 2004), increased risk and uncertainty (Palmrose et al., 2004), negative signal of credibility issues sent to a firm s contractual parties (Leng et al., 2011) a loss of reputation 5 (Karpoff et al., 2008a), and a decline in the credibility of accounting information as reasons for the observable consequences that were documented in the respective studies. However, these studies only analyzed the aggregate reaction of the market (banks), and thus can hardly be used to infer which mechanisms caused the reaction documented (i.e., the factors that market participants (banks) were concerned about after the misstatement became publicly known, and which eventually caused the documented consequence). Thus, after reviewing the prior literature it remains unclear (i) which mechanisms cause the adverse consequences that are observable after misstatements, (ii) which mechanism has the most severe impact on misstatement firms, and (iii) how long misstatement firms are adversely affected by these mechanisms. Examining credit rating analysts and their reaction to the misstatement is a suitable means of analyzing these research questions for three reasons: First, we aim to analyze a consequence that is highly relevant to misstatement firms. Subsequent to a misstatement becoming publicly known, firms have restricted access to equity financing and must rely heavily on debt financing (Chen et al., 2013). Thus, the question as to what extent a misstatement adversely affects a firm s creditworthiness (credit rating) is crucial for the firm s opportunity to obtain external financing. Second, analyzing firms credit ratings subsequent to a misstatement becoming publicly known has methodological advantages. In their rating reports, analysts specify the mechanisms through which the misstatement adversely affects a firm s creditworthiness if 5 We note that it is debatable as to whether a loss in reputation should lead to a downward revision of expected future cash flows (mean effect) or higher cost of capital (variance effect). The former classification is more appropriate when the loss of reputation primarily concerns a firm s reputation with customers, whereas the latter is appropriate if it concerns the reputation loss with capital providers. In this light, Karpoff et al. (2008b) and Murphy et al. (2009) note that lost reputation could contribute to both mean and variance effect.

10 8 these mechanisms are considered in the rating decision. Thus, analyzing the content of rating analysts reports enables us to identify the mechanisms that adversely affect misstatement firms. Furthermore, our multivariate analysis allows us to analyze the extent to which these mechanisms affect the firm s creditworthiness, and our research design is suited to capturing their impact on a firm s creditworthiness even in the long-term. In addition, this mixed method approach helps us to address potential endogeneity concerns. Our content analysis allows us to identify those cases where rating analysts are particularly concerned about the impact of the misstatement. If our variables of interest capture the impact of the misstatement on credit ratings, they should reflect the concerns that rating analysts have mentioned in the analyzed reports. Third, credit rating analysts (creditors) have access to private (superior) information channels and should be less concerned about the credibility of public accounting information than equity investors. Focusing on the perspective of credit rating analysts and their perception of the consequences of the misstatement will shed light on mechanisms that might have been neglected in prior studies, which mostly focused on equity investors. 3. CONTENT ANALYSIS OF RATING REPORTS (a) Sample selection and classification We focus on the largest credit rating agency (S&P s) and its corporate credit ratings, which S&P defines as our opinion of the general creditworthiness of an obligor (S&P, 2008). Our sample includes all active and inactive Compustat firms, for which S&P s Long Term Domestic Issuer Credit Rating is available, for at least one year. We exclude all firms that operate in financial industries and firms without the required data to calculate all of the independent variables. This leaves us with a sample of 3,620 unique firms and 30,340 firmyears between 1988 and [Insert Table 1 here] We obtain information on firms misstatement events from two sources: The GAO and the SEC. Restatement information is taken from the GAO Financial Restatement Databases which cover the period between 1997 until We follow the classification of Hennes et al. (2008) to distinguish between restatements that are related to intentional misstatements ( irregularities ) and restatements that are related to unintentional misstatements ( errors ). We include misstatement events for which the SEC subsequently issued one or more Accounting and Auditing Enforcement Releases (AAERs). Misstatements that cause the SEC

11 9 to initiate an investigation and subsequently to release an AAER can be considered severe (Karpoff, 2008a; Files, 2012; Schrand and Zechman, 2012). Following Schrand and Zechman (2012) we assume that misstatements which lead to one (or multiple) AAERs have been intentional, and thus we classify them as intentional misstatements. Data on AAERs between 1982 and 2012 are obtained from the Center for Financial Reporting and Management (University of California, Berkeley). This dataset provides a reconciliation of each AAER to the underlying misstatement event. Based on the GVKEY (Global Vantage Key) mentioned in the GAO restatement reports, we identify misstatements associated with a restatement (GAO database) and a subsequent AAER. We match the GVKEYs of the misstatement firms with the GVKEYs of the 3,619 rating sample firms. 3,170 firms are associated neither with a restatement (GAO database) nor an AAER. We include these firms in our sample and assume that they did not misstate their financial statements during the period investigated (Non-Misstatement Firms). 6 The remaining 450 firms in our sample are classified as misstatement firms, and subclassified according to the severity of their misstatements. 251 of these firms have restatements that, according to Hennes et al. (2008), are related to unintentional misstatements (Error Firms). The remaining 199 firms intentionally misstated their financial statements (Fraud Firms). Fraud Firms can further be sub-classified based on whether the intentional misstatements lead to an AAER (Fraud Firms SEC, 126 firms) or did not lead to an AAER (Fraud Firms No SEC, 73 firms). (b) Misstatement-related mechanisms that rating analysts are concerned about We conduct a content analysis of rating reports that were released by S&P subsequent to an intentional misstatement becoming publicly known. From the full sample of 204 firms with intentional misstatements, 7 we have to exclude (i) firms that according to Compustat, do not have a S&P credit rating at the end of the year in which the misstatement becomes publicly 6 7 We acknowledge that not all misstatements might be included in the GAO or AAER datasets (Karpoff et al., 2014). Thus, our analysis might suffer from a Type II-error, since we potentially classify firms as nonmisstatement firms (control group), even though there might have been a misstatement. Similar to previous studies, this poses a limitation. However, this limitation would bias against finding a persistent impact of misstatement-induced risks on a firm s credit rating. We do not include the 5 firms with multiple intentional misstatements that lie in different years in our regression analyses, because we cannot distinguish between the effects of the individual misstatements (see page 16). However, we do include these firms in our content analysis and focus on rating analysts reaction to the first misstatement. This should not introduce a bias given that, at the time when the rating analyst discussed the consequences of the first misstatement, the second misstatement had not become publicly known (in one case, we had to focus on the second misstatement, because that firm was not assigned with a credit rating at the time when the first misstatement became publicly known).

12 10 known (71 firms), (ii) firms for which we do not find a rating report released by S&P during the first year subsequent to the misstatement becoming publicly known (24 firms) 8, and (iii) firms for which S&P withdraws the rating shortly after the misstatement becoming publicly known without any reference to the misstatement (2 firms). For the remaining 107 firms, we read the rating reports that are released within a year of the misstatement becoming publicly known, and analyze whether the misstatement is mentioned or not. Rating analysts do not mention the misstatement or potential consequences in the reports of 49 firms. For the remaining 58 firms, we analyze the content of the rating report in which the misstatement is first mentioned. We collect all text passages from the 58 reports in which rating analysts mention the misstatement and the potential consequences; this results in 76 text passages (i.e., in some rating reports, analysts mention multiple misstatement-related concerns). 9 We then inductively develop categories from this material, thereby following the procedures for qualitative content analysis as suggested by Mayring (2004). We revise the categories as we are processing the material and finally derive 9 categories which summarize rating analysts perception of the consequences resulting from the misstatement (we present details on the formation of the categories and two representative anchor examples for each category in Appendix A). 10 Our 9 categories are as follows: (i) Increased liquidity risk Most analysts are concerned about a firm s increased liquidity risk resulting from misstatement-related violation of debt covenants. In the majority of these cases, the misstatement caused a substantial delay in the firm s 10-Q/10-K filings. In fact, the company is only able to file its current 10-Q/10-K after internal (and external) investigations into the accounting matter have been completed and historical financial statements have been restated and re-audited. A delay in the firm s 10- Q/10-K filing thereby violates a general covenant under the existing debt agreement, in accordance with which the firm must provide financial statements to the creditor within a Rating reports are obtained from the S&P Ratings Direct. Each rating report is related to a respective rating action. Regarding the reports of the 58 firms that mentioned the respective misstatements, firms were placed on Credit Watch Negative in 29 cases, downgraded in 12 cases, and their rating outlook was revised to negative in 4 cases. In 4 cases, the ratings were affirmed. In 3 cases, the firms were downgraded and placed on Credit Watch Negative. In 3 cases, firms remained on Credit Watch Negative. In 1 case, the firm was placed on Credit Watch Positive. In 1 case, the firm was placed on Credit Watch Developing. In 1 case, the rating was withdrawn. An intercoder reliability check that assesses if our classification of rating analysts statements to one of the 9 categories is objective, yielded an intercoder agreement of 80.3%.

13 11 specified period. A breach of this covenant triggers a technical default. If the company is not able to file the 10-Q/10-K within the grace period (usually 30/60 days), then creditors are entitled to demand accelerated repayment of the debt or to renegotiate the terms of the credit facility. Creditors are entitled to the same right if the restatement causes a violation of financial covenants, which require the borrower to maintain a specified minimum or maximum level of a financial ratio. Even if covenants under only one existing debt agreement are violated, the creditors of other debt agreements may also be entitled to demand accelerated debt repayment or contract modification if these other debt agreements contain cross-default clauses. The risk of accelerated debt repayment in this situation could lead to a liquidity crisis, because misstatement firms have only restricted access to external capital (Chen et al., 2013). While Dichev and Skinner (2002) illustrate that covenant violations are not associated with costly consequences per se, misstatement-related violations of covenants have more severe consequences. Creditors face a higher level of uncertainty about firms future prospects after a misstatement becomes publicly known and the revelation of the intentional misstatement decreases firm s creditworthiness. An increase in information asymmetry and higher default probability reduce creditors willingness to waive the right of accelerated repayment (Chen and Wei, 1993; HassabElnaby, 2006; Chava and Roberts, 2008). In cases where creditors are willing to waive this right, they will often renegotiate the terms of the facility and will, for instance, require additional collateral, stricter covenants or shorter maturity, which restricts firms financial leeway under the existing facilities (see Beneish and Press, 1993; Chava and Roberts, 2008). Another consequence of misstatement-related covenant violations was that a number of firms were prevented from drawing on existing credit facilities. In addition, firms are not able to issue public securities for as long as they do not file their 10-K/10-Q reports with the SEC. Rating analysts statements illustrate that misstatement-related covenant violations can endanger firm's liquidity situation and might be one of the most severe mechanisms through which misstatements adversely affect firms creditworthiness. (ii) + (iii) Adverse effect of the restatement on financial ratios + legal penalties Subsequent to a misstatement becoming publicly known, rating analysts assess the impact that the restatement has (will have) on firms financial ratios. If the restated financial statements are available when the misstatement becomes publicly known, rating analysts are

14 12 able to assess the impact of the restatement on firms financial ratios with certainty. Otherwise, they will form expectations on the restatement amount based on the companies misstatement-related disclosures (e.g., reason for restatement, quarters affected and accounts). Rating analysts are concerned in such cases where the restatement has (or is expected to have) a considerable negative impact on firm s earnings and/or expected future cash flows. In addition, rating analysts are concerned about the legal penalties that could arise from SEC enforcement actions and shareholder lawsuits. Both concerns, legal penalties and the adverse effect of the restatement on financial ratios, were the main mechanisms underlying the mean effect and have been thoroughly analyzed in prior literature (e.g., Karpoff 2008a). (iv) Internal control weaknesses Rating analysts are concerned about misstatement firms internal control weaknesses. The misstatement becoming publicly known reveals internal control deficiencies and lack of management oversight. Rating analysts concerns about internal control deficiencies refer, but are not limited to, the accuracy and credibility of firm s accounting information. Internal control weaknesses might also lead to excessive risk taking, which debt holders are especially afraid of due to their asymmetric payoff structure. (v) Distraction from operating business In three cases, rating analysts are concerned that misstatement-related consequences distract the firm from its core business. The main reasons for that are internal and external (e.g., SEC) investigations into the accounting matter that demand managerial attention. In addition, rating analysts consider management turnover to be an additional challenge on the misstatement firm in the short-term, further shifting its focus away from core business activities. (vi) Loss of reputation Of further concern is the loss of reputation among the financial community and other stakeholders, which results from the misstatement becoming publicly known. For one, firm s credibility among investors might erode and investors are less willing to provide capital to firms that intentionally misstate earnings (which should lead to higher cost of capital (variance effect)). For another, firm s reputation among customers might erode and translate into decreasing sales volume (which should lead to a downward revision of expected future cash flows (mean effect)). (vii) Inability to assess firm s financial condition due to delay in financial statements

15 13 Two rating analysts state that they face considerable difficulties in assessing firm s creditworthiness, since it had not filed annual or quarterly financial statements. (viii) General concern with respect to potential consequences of the misstatement Rating analysts mention the misstatement and are concerned that it could lead to adverse consequences. However, in these eleven cases, they do not specify the mechanisms they are concerned about. (ix) Rating analyst is not concerned about the consequences of the misstatement In ten cases, rating analysts state that they are not concerned about the consequences of the misstatement. In these cases, it appears that the misstatement did not trigger a violation of financial covenants or rating analysts are not concerned about it. In addition, the magnitude of these misstatements or expected penalties do not adversely affect rating ratios to an extent that would require a downgrade to the next lower rating level. In the remaining 49 cases rating analysts do not mention the misstatement in the analyzed rating reports. 4. RESEARCH DESIGN (a) Model and variables In the following we aim to examine the impact of the misstatement on firm s creditworthiness, and use the credit rating as a proxy for firm s creditworthiness. We use the following regression model to analyze the extent (duration) to which (over which) the misstatement adversely affects the misstatement firm s credit rating: 10 RATING i,t = α 1 + β n YEAR [n] AFTER MISSTATEMENT DISCLOSURE + γ k RATING DETERMINANTS i,t 1 n= 1 + δ FIRM FIXED EFFECTS + φ YEAR FIXED EFFECTS + ε it Our variables of interest are indicator variables for the year before the misstatement becoming publicly known (-1), the year in which the misstatement becomes publicly known (0) and each of the following ten firm-years (1 10). This design is similar to the one used by Wilson (2008) and Chen et al. (2014). The indicator variable YEAR 1 AFTER MISSTATEMENT will take the value 1 for every firm-year that lies one year after the misstatement becomes publicly known. Thus, the β n coefficients capture the adverse impact of the misstatement on firm s credit rating in the respective year. For example, if the misstatement adversely affects credit ratings in the year of the misstatement becoming publicly known and the following three years, then the

16 14 coefficients of β 0, β 1, β 2 and β 3 would be negative and significant. If the misstatement does not adversely impact the firm s credit rating in year four after the misstatement becomes publicly known and the following years, then β 4 to β 10 would be insignificant. To ensure that our indicator variables only capture the adverse impact of the misstatement, we control for the following aspects that could otherwise bias our β n coefficients. First, we include 11 rating determinants that, according to prior literature, are assumed to predict firm s credit rating. In particular, we follow the rating prediction models in prior literature and control for firm s size, profitability, leverage, liquidity, retained earnings, capital expenditures and the number of employees (see Matthies (2013) for an overview of rating prediction models used in prior literature). 11 We additionally include a firm s decile rank with respect to size, profitability and interest coverage, given that S&P s rating assessments are based on other factors besides absolute credit metrics. Rating assessments are also based on the relative rank of a firm s credit metrics within its peer group (Metz, 2006; S&P, 2008). The 11 RATING DETERMINANTS capture changes in a firm s financial statement information and control for indirect consequences arising from the misstatement becoming publicly known that have materialized in, or before, the firm-year under investigation. For instance, if a firm s profitability, liquidity or leverage deteriorates during the years following a misstatement becoming publicly known, the related impact would be captured by our RATING DETERMINANTS. Conversely, the adverse impact of misstatement-related mechanisms that have not yet been reflected in RATING DETERMINANTS (e.g., increased liquidity/refinancing risk) would be captured by our indicator variables (β n coefficients). This design enables us to analyze the within-firm variation of credit ratings and the adverse effect of misstatementrelated mechanisms, while controlling for any changes in firm s RATING DETERMINANTS that occur after the misstatement becoming publicly known (e.g., declining profitability, restated financial situation) We use the following determinants: Log of Total Assets (Size), Return On Assets (Profitability), Operating Cash Flow Ratio (Profitability), Total Debt Ratio (Leverage), Current Ratio (Liquidity), Retained Earnings Ratio, Capital Expenditure Ratio, and Number of Employees. Appendix C includes a definition of the determinants. For our study, it is crucial that the RATING DETERMINANTS explain a large part of the variation in a firm s credit rating because we aim to capture the effect of additional, misstatement-induced risks on a firm s credit rating that are not attributable to a change in RATING DETERMINANTS. Minimizing the variation in a firms credit ratings that cannot be explained by RATING DETERMINANTS allows us to better capture the effect of the additional, misstatement-induced risks.

17 15 Second, we control for any time-invariant factors that impact firm s credit rating, but are not captured by our RATING DETERMINANTS (e.g., a firm s control environment). Our sample includes non-misstatement and misstatement firms. Both types of firms have different firm characteristics, which causes their credit ratings to differ. 13 If our RATING DETERMINANTS do not absorb all of these differences in the firm characteristics, then our β n coefficients would capture not only the consequences of the misstatement but also the differences in characteristics of non-misstatement and misstatement firms. Therefore, we include firm-fixed effects to control for these differences. Third, we expect that credit rating agencies have gradually increased their requirements to achieve a certain rating over time (Blume et al., 1998; Baghai et al., 2014). Indeed, we find that credit ratings have decreased over time when RATING DETERMINANTS are held constant. We include year-fixed effects to control for this trend in rating requirements. Our dependent variable is S&P's Long Term Domestic Issuer Credit Rating as of 31 December of the respective year. We transform S&P s credit ratings into numerical values between 1 and 23 (similar to Anderson et al., 2003). [Insert Table 2 here] We multiply the assigned rating scores by minus one, so that higher values indicate a stronger credit rating. Given that we use credit rating data as of 31 December, we have to match a firm s credit rating with lagged RATING DETERMINANTS to ensure that this information was available to rating analysts (i.e., firms accounting information as of 31 December 2002 will be disclosed in the first months of 2003 and can be incorporated into the rating decisions that follow). In all of our analyses, we cluster standard errors at the firm level. Appendix C provides details on all of the variables that are included in our analysis. Our analyses would suffer from an endogeneity bias if we did not control for factors that are correlated with both (i) a firm s credit rating and (ii) its propensity to misstate earnings. In our setting, it is plausible that unobservable factors (e.g., a firm s control environment) cause an endogeneity bias. Thus, we do not use a propensity-score-matched control sample, since it does not alleviate an endogeneity bias that arises from unobservable factors (Lennox et al., 2012). Instead, we include firms without a misstatement (Non- Misstatement Firms) in all of our analyses and include firm-fixed effects. Including firmfixed effects helps us to address the aforementioned endogeneity concern, since firm-fixed 13 Considering the complete sample period, Non-Misstatement Firms have a mean credit rating of , as compared to (Error Firms) and (Fraud Firms).

18 16 effects absorb unobservable time-invariant factors that affect a firm s credit rating and the likelihood of an earnings misstatement. In addition, including a large sample of nonmisstatement firms contributes to a better estimation of the coefficients of our RATING DETERMINANTS and year-fixed effects, which ultimately leads to a better rating prediction model. Although the scale of credit ratings is ordinal in nature, we use a linear regression to estimate our model, rather than an ordered probit model, because our analyses require controlling for firm-fixed and time-fixed effects. Including firm-fixed effects in an ordered probit model would introduce an incidental parameter problem because our analysis is based on a small number of observations per firm (Baghai et al., 2014). Consequently, we use a linear regression model with firm-fixed effects for our main analyses and additionally test whether our results remain robust when using an ordered probit model with industry-fixed effects (Table 8). (b) Coding of the misstatement indicators We collect data on misstatements from two databases (i.e., GAO restatement database and the database SEC AAER releases). Error Firms (unintentional misstatement) and Fraud Firms No SEC (intentional misstatement without subsequent AAER) only appear in the GAO database, but are not associated with AAER. For these firms, we assume that the misstatement becomes publicly known on the day of the restatement. For Fraud Firms SEC (intentional misstatement with subsequent AAER), the AAER dataset provides information on the date when the AAER is released. However, we cannot assume that the underlying misstatement becomes publicly known when the AAER is released. Therefore, we identify the dates on which the misstatement underlying the subsequent AAER becomes publicly known (see Appendix B for details). On average, we find that the misstatement became publicly known 1,038 days before the respective AAER was released, which is very similar to the time lag reported in Karpoff et al. (2014). Presumably, credit ratings were already impacted shortly after the misstatement became publicly known (and not only when the AAER was released), which is why we focus on the date of the misstatement becoming publicly known. We assume that credit rating agencies are concerned about the risks resulting from the most severe type of misstatement. Thus, we classify firms into one of the three types according to their most severe misstatement. 55 sample firms have both unintentional and

19 17 intentional misstatements during our sample period. 14 We classify these firms as Fraud Firms, because we focus on the more severe (intentional) misstatement. In addition, for every firm type, there are firms with multiple misstatements of the same category. If a firm has multiple misstatements of the same category that become publicly known in different years, it is unclear whether the impact of the misstatement on firm s credit rating in a given year is attributable to the first misstatement, the most recent misstatement, or both. In our sample there are 62 Error Firms with multiple restatements (related to unintentional misstatements) that lie in different years, 2 Fraud Firms No SEC with multiple restatements (related to intentional misstatements) that lie in different years, and 3 Fraud Firms SEC with multiple misstatement events, with each of these events leading to at least one AAER. 15 We exclude all of these firms with multiple misstatements that lie in different years, because we cannot distinguish between the effects of their individual misstatements. Our coding strategy is described in Appendix B. (c) Descriptive statistics Table 3 provides descriptive statistics on the sample firms, sub-classified by firm type, for the period before and after the misstatement becomes publicly known. [Insert Table 3 here] The findings show that misstatement firms credit ratings deteriorate after the misstatement becomes publicly known (comparison of mean credit rating in pre- vs. post-misstatement period). While the difference is also driven by a downward time trend in the credit ratings, we find that firms with intentional misstatements experience a significantly (1%-level) larger decline in credit ratings as compared to firms with unintentional misstatements. When comparing the differences between the pre- and post-misstatement periods, intentional misstatements firms development in terms of size, liquidity and profitability show a more pronounced increase in default risk, than the development of these ratios for unintentional misstatement firms. 5. MULTIVARIATE ANALYSIS (a) The impact of the misstatement on firm s credit rating of these firms are Fraud Firms No SEC, while the remaining 33 are Fraud Firms SEC. Multiple restatement announcements that are related to the same misstatement (Hennes et al., 2008 coded them as duplicate restatements ) are not classified as multiple misstatements. For these firms, the first restatement indicates when the misstatement becomes publicly known and these firms are not excluded from our analyses.

20 18 To date, Liu et al. (2012) is the only study that has investigated the impact of misstatements on credit ratings. Their findings indicate that, on average, restatements do not have a negative effect on credit ratings. However, the study s design does not distinguish between restatements that are related to intentional and unintentional misstatements. In addition, the study only analyzes the year of the restatement and the sample does not include misstatements with subsequent AAERs. In contrast to Liu et al. (2012), our study distinguishes between intentional and unintentional misstatements, additionally includes misstatements associated with subsequent AAERs and analyzes a number of years that follow subsequent to the misstatement becoming publicly known. In a first step, we visually illustrate how firms credit ratings develop after a misstatement becomes publicly known. We use all sample firms, regress their credit rating on all RATING DETERMINANTS and include firm-fixed, as well as year-fixed effects. We use the fitted values from the regression, which can be interpreted as firm s empirically modeled credit rating ( predicted rating ), without controlling for any misstatements. We calculate the difference between the firm s actual and predicted rating, and plot the mean differences separately for every N th year after the misstatement becomes publicly known (Figure 1). [Insert Figure 1 here] On average, the actual credit ratings of Non-Misstatement Firms exceed their predicted ratings. This is the case since misstatement firms (Error Firms and Fraud Firms) that are included in the rating prediction model have, on average, lower credit ratings than Non- Misstatement Firms, even after controlling for all RATING DETERMINANTS. The figure illustrates that, even in firm-years before the misstatement becomes publicly known, misstatement firms actual credit rating is lower than the predicted rating. In the case of firms with intentional misstatements (Fraud Firms), the negative deviation from the predicted rating becomes more pronounced in the year when the misstatement becomes publicly known, which reflects the adverse impact of the misstatement on firm s creditworthiness. In the years that follow the misstatement becoming publicly known, the deviation from the predicted rating declines and returns to the pre-misstatement level. This pattern suggests that the adverse impact of the misstatement fades away in the years that follow subsequent to the misstatement becoming publicly known. In the next step, we test if the adverse effect on credit ratings is significant and how persistent it is. Table 4 presents the results of our main analysis. The coefficients of 10 (out of 11) RATING DETERMINANTS are significant. RETURN ON ASSETS is insignificant, which might be explained by the fact that firms profitability is already captured by DECILE RETURN ON