School of Accounting Seminar Series. The role of audit verification in debt contracting: Evidence from covenant violations.

Transcription

1 Australian School of Business Accounting School of Accounting Seminar Series Semester 1, 2014 The role of audit verification in debt contracting: Evidence from covenant violations Hui Zhou Melbourne Business School Date: Friday 4 th April, 2014 Time: 3.00pm 4.30pm Venue: ASB Last updated: 29/03/14 CRICOS Code: 00098G

2 The Role of Audit Verification in Debt Contracting: Evidence from Covenant Violations Liangliang Jiang Department of Economics Lingnan University 8 Castle Peak Road Tuen Mun, New Territories, Hong Kong Tel: liangliangjiang@ln.edu.hk Hui Zhou* Melbourne Business School University of Melbourne 200 Leicester Street Carlton, VIC 3053 Australia Tel: h.zhou@mbs.edu * Corresponding author. We thank Paul Coram, Jim Frederickson, Ganapathi Narayanamoorthy, Padmakumar Sivadasan, participants at the 2012 New Accounting Faculty Conference in Melbourne (especially the discussant Bin Ke), and participants at the 2012 International Symposium on Audit Research for helpful comments. Arbitor Ma provided excellent research assistance. We gratefully acknowledge the financial support from Lingnan University and Melbourne Business School. An earlier version of this paper was circulated under the title Do Auditors Play a Positive Role in the Resolution of Debt Covenant Violations?.

3 The Role of Audit Verification in Debt Contracting: Evidence from Covenant Violations Abstract We examine whether debt covenant violations result in a demand for differentially higher levels of audit verification. Using audit fees as the proxy for the level of audit verification, we find that violating firms pay for higher-priced audit services following the violation event than otherwise similar non-violating firms. Additional tests show that auditor litigation risk is unlikely to be the primary driver of the audit fee response to debt covenant violations. Consist with the notion that the higher audit fees arise from the demand for differentially higher levels of audit following covenant violations, the audit fee response to covenant violations is more pronounced for firms with more intensive lender and shareholder monitoring. Moreover, audit committees respond to covenant violations by including more independent directors and meeting more frequently. Collectively, our findings help shed light on the role of audit verification in the resolution process following debt covenant violations. Keywords: audit verification; debt covenant violations; audit committee 1

4 1. Introduction Financial reporting and audit verification of financial statements play an important role in contracting and monitoring (Armstrong et al., 2010; Bushman and Smith, 2001; Lambert, 2001; Shivakumar, 2013; Watts and Zimmerman, 1983, 1986). In the context of debt contracts, theory posits that accounting-based debt covenants help mitigate the agency cost arising from the conflict of interest between managers and creditors. While previous literature documents substantial evidence that accounting-based debt covenants provide ex-ante contracting efficiency incentives, little is known about the role of accounting and auditing following debt covenant violations. This study contributes to the literature by investigating whether covenant violations result in a demand for differentially higher levels of audit verification in the resolution process. Previous research has established that covenant violations occur frequently, affect a large number of firms, are usually observed well outside of financial distress, and rarely lead to acceleration of the loan (Dichev and Skinner, 2002; Gopalakrishnan and Parkash, 1995; Nini et al., 2012). Moreover, recent empirical evidence suggests an improvement in both the operating and stock market performance following covenant violations due to increased lender monitoring that serves to reduce managerial slack (Chava and Roberts, 2008; Nini et al., 2009, 2012). These findings highlight the importance of the resolution of covenant violations in debt contracting and raise questions about the role of accounting and auditing in the resolution process. Dichev and Skinner (2002) conclude that lenders use debt covenant violations as early warning signals that allow them to review and renegotiate debt agreements. As accounting information plays an important role in informing the negotiation process that determines the key parameters of debt contracts (Shivakumar, 2013), an enhanced credibility of accounting information can alleviate the information uncertainty faced by the lender and thus help reduce 2

5 the renegotiation costs following covenant violations. Previous research provides evidence that the verification of financial statements performed by independent auditors serves as a mechanism for improving the credibility of accounting information and mitigating borrowing costs. For example, Kim et al. (2011) and Minnis (2011) document that voluntary external audits are associated with lower costs of debt using samples of private firms not subject to mandatory audit. We propose that debt covenant violations will result in a heightened demand for higher levels of audit verification as a mechanism to mitigate the renegotiation costs in the resolution process. 1 We investigate the demand for audit verification following covenant violations by examining both the change in the pricing of audit services and the response of audit committees. Following prior literature that uses higher audit fees to capture the differential demand for higher levels of audit services (e.g. Ball et al., 2012), our first set of analysis examines the audit fee response to covenant violations. Using a comprehensive data set that covers U.S. public firms with outstanding loans, we examine the incremental impact of covenant violations after controlling for other determinants of audit fees in the literature. We document that firms that have recently violated covenant violations pay for higher priced audit services than their non-violating peers during the violation year, but not in the year immediately before the violation event. While the above results are consistent with the notion that covenant violations lead to increased demand for higher levels of audit verification, an alternative explanation attributes the higher audit fees to potential auditor litigation risk associated with covenant violations. It is worth to note that a covenant violation is not a violation of financial reporting or auditing requirements. Thus, the link between a covenant violation and the probability of the auditor 1 Here we implicitly assume that the level of audit verification reflects the choice of the client in an environment where external audit is mandatory. This view is consistent with the existing literature that treats audit as a differentiated product with a private demand rather than a standardized commodity mandated by regulation (e.g. Ball et al., 2012; Watts and Zimmerman 1983, 1986). 3

6 being sued for negligence is not clear unless the violation is associated with other events that can result in increased auditor litigation risk. For example, a stylized fact is that lawsuits against auditors and other types of class action lawsuits tend to follow sharp declines in stock prices (Shu, 2000). If covenant violations tend to trigger large drops in stock prices, auditors will face higher litigation risk in the post violation years. To directly investigate the effect of auditor litigation risk, we compare the audit fee pattern for violating firms of different levels of auditor litigation risk. Specifically, we use firm stock returns, net profit, and operating cash flows to gauge auditor litigation risk. We do not find statistical difference in the audit fee responses to covenant violations for violating firms that experience positive vs. negative stock returns during the violation year, that report net profit vs. loss during the violation year, or that report positive vs. negative operating cash flows during the violation year. These results show that the higher audit fees following covenant violations are unlikely due to the violation event being a signal of deteriorating financial prospect that significantly increases auditor litigation risk. To address the concern that audit fee differences between the violating and non-violating firms are driven by other omitted variables, we use the propensity score matching approach to construct an optimal comparison group consisting of firms that share similar characteristics but have not reported a covenant violation. We estimate the average audit fee differences between the violating firms and firms in this comparison group. Unlike multivariate regression models that require the control variables to be linearly related to audit fees, the propensity score matching approach has the advantage of not restricting the specific relationship between audit fees and the firm characteristics affecting audit pricing. The results from the propensity score 4

7 matching analysis confirm that violating firms pay for higher priced audit services than otherwise similar firms that have not reported a debt covenant violation. After documenting the immediate impact of covenant violations, we examine whether the audit fees for violating firms stay at higher levels following the violation year. As we suspect that the demand for higher levels of audit services may persist for several years after the initial violation, we use dynamic models to investigate how the audit fees evolve following the events of violation. The results show that violating firms pay higher audit fees than non-violating firms for at least three years after the initial violation. The documented differences in audit fees in post violation years are both statistically and economically significant, ranging from 5% to 10%. Having found evidence on the impact of covenant violations on audit fees, we further investigate whether the audit fee response to covenant violations varies with the level of lender and shareholder monitoring. If the observed increase in audit fees arises from heightened demand for higher-level audit verification following covenant violations, the audit fee response should be more pronounced for firms with more intensive lender and shareholder monitoring. We find supporting evidence by using leverage as the proxy for lender monitoring (Klein, 2002) and institutional ownership as the proxy for shareholder monitoring (Bhojraj and Sengupta, 2003). Our second set of analysis examines the corporate governance mechanisms underlying the observed audit fee response to debt covenant violations. To this end, we focus on the response of the audit committee to covenant violations as the audit committee is the most important oversight element in the financial reporting process. A primary channel for the audit committee to exercise this oversight role is by selecting and monitoring the external auditor to ensure the quality of the financial reporting verification process performed by the external auditor. In the context of 5

8 covenant violations, the heightened demand for differentially higher-quality audit services is likely to manifest through the response of audit committee to the violation event. We examine two audit committee characteristics in our analysis of audit committee response to covenant violations: the number of independent directors on the audit committee and the frequency of audit committee meetings. Anderson et al. (2004) provide evidence that lenders price more effective monitoring of the financial reporting process in debt agreements by documenting that more independent and active audit committees are associated with lower cost of debt. Focusing on the same audit committee characteristics, we find evidence that audit committees respond to covenant violations by including more independent directors and meeting more frequently in the post violation years. Our results extend the findings in Anderson et al. (2004) by shedding light on the role of audit committee characteristics in the resolution process following covenant violations. This study provides new evidence that covenant violations result in a heightened demand for higher-level audit verification. The findings extend the evidence that the financial statement verification process performed by external auditors helps mitigate contracting costs in debt financing (Kim et al., 2011; Minnis, 2011). We show that this role does not stop at a debt covenant violation but extends to the resolution process, thus providing novel evidence highlighting the importance of audit verification in debt contracting. Our paper complements previous research documenting that creditors start to play a more active corporate governance role following debt covenant violations (Chava and Roberts, 2008; Nini et al., 2009, 2012). We find evidence that other stakeholders in the corporate governance system (i.e. the audit committee and the external auditor) also respond to the event of covenant violation. The findings add to evidence in the literature showing that the heightened demand for 6

9 monitoring from the corporate governance system leads to the acquiring of a differentially higher level of audit services (Carcello et al., 2002; Fan and Wong, 2005). This study is also related to the line of research that investigates the consequences of debt covenant violations (e.g. Beneish and Press, 1993; Beneish and Press, 1995; DeFond and Jiambalvo, 1994; Sweeney, 1994). While we document that covenant violations result in significantly higher audit fees, we caution against interpreting the increase in audit fees as a cost of covenant violations that invariantly yields a negative outcome. Notably, recent empirical evidence suggests that penalties imposed by creditors in response to covenant violations (e.g. in the form of restrictions on capital expenditures) yield positive externalities to equity holders by constraining value-reducing managerial behavior (Nini et al., 2009, 2012). In similar vein, increased auditor monitoring in response to covenant violations, while resulting in higher out-ofpocket audit expense, represents a governance mechanism for mitigating managerial agency problems and thus may positively impact firm value. The remainder of this paper proceeds as follows. The next section describes the sample construction process and our data. Section 3 presents the empirical tests and our findings. We summarize and conclude in Section Data 2.1. Sample construction process We start our sample collection from the merged COMPUSTAT-Audit Analytics dataset. For each observation in the merged sample, we collect accounting information from COMPUSTAT and auditor/audit fee information from Audit Analytics. 2 We then match the merged dataset with 2 It is important to note that fees for due diligence performed by external auditors related to loan initiation are part of non-audit fees and therefore not included in audit fees. 7

10 the Reuters Loan Pricing Corporation s Dealscan database using the DealScan-COMPUSTAT link file constructed by Chava and Roberts (2008). This ensures that our sample excludes firms that have never borrowed loans since year 1987 as Dealscan started recording bank loans in We then merge this sample with the comprehensive dataset of covenant violations used in Roberts and Sufi (2009) and Nini et al. (2012). 3 Finally, we exclude all financial firms and utility firms (SIC codes between 6,000 and 6,999, and between 4,900 and 4,999, respectively) in the sample. Our sample period starts from 2000 because audit fee data became available starting year Nonetheless, we still keep the covenant violations recorded between year 1996 and 1999 in order to minimize the truncation problem. The sample construction process yields 2,103 new covenant violations by 1,682 firms out of the total of 4,561 sample firms between year 2000 to We follow Roberts and Sufi (2009) and consider a violation new if there is no financial covenant violation by the same firm in the previous four quarters. 4 The events of covenant violation in our sample are not clustered in any particular years Violation variables As our sample includes both the violating and non-violating firms, we use an array of dummy variables to identify 1) whether the firm-year observation represents a violating firm, and 2) the distance to violation event if the observation represents a violating firm. We use an approach similar to that employed by Bogart and Chaudhary (2012) in order to track both the immediate and long term impact of covenant violations. 3 The dataset is compiled by Professor Amir Sufi and can be downloaded from 4 Our main results do not change if we change the definition of new violation from four quarters to eight quarters. 8

11 Specifically, we create a sequence of dummy variables (Vio_Pre, Vio_Event, Vio_PostY1, Vio_PostY2, and Vio_PostY3) to capture the distance to the event of violation for violating firms. For firm-year observations that represent a violating firm, Vio_Event indicates the violation year (the year that the firm reported a covenant violation). Vio_Pre indicates the year immediately before the violation year. Vio_PostY1, Vio_PostY2, and Vio_PostY3 indicates the year immediately following the violation year, two years after the violation year, and three years after the violation year respectively. Accordingly, a firm-year observation that represents a nonviolating firm will have zero values for all the dummy variables. The mean of the dummy variable Vio_Event is 0.09, indicating that about 9% of the firm-year observations included our sample represent a newly reported covenant violation Control variables We include factors that have been shown to affect audit fess and audit committee in the literature as control variables in our analysis (See Appendix for variable definitions). Consistent with previous literature, the control variables cover auditee size, complexity, operating performance, growth potential, leverage level, liquidity status, accounting fraud risk, auditor characteristics, and audit outcomes (Bell et al., 2001; Doogar et al., 2010; Francis et al., 2005; Hay, et al., 2006; O Keefe et al., 1994; Simunic, 1980). As our sample period spans from 2000 to 2008, we include an additional indicator variable to control for the effect of Section 404(b) of the Sarbanes-Oxley Act. It is set to one if the firm is an accelerated filer (filer status reported in Audit Analytics) and the year is 2004 or later and set to zero otherwise. 5 The inclusion of this variable is important as Section 404(b) of the Sarbanes Oxley Act expands the scope of auditing 5 Filing both the management report on internal controls (required under Section 404(a) )and the auditor attestation of this management report (required under Section 404(b)) went into effect starting fiscal year 2004 for accelerated filers while non-accelerated filers were never subject to the requirement of Section 404 (b). 9

12 by requiring auditors to evaluate the effectiveness of the internal controls and attest to the assessment made by management, which can significantly impact the overall level of audit fees. Table 1 reports the descriptive statistics for audit fees and control variables used in our analysis. All the dollar value variables in Table 1 have been converted to constant 2006 U.S. dollars using the U.S. Bureau of Labor Statistics (BLS) CPI series as deflator. The mean and median of audit fees is $1.015 million and $0.337 million respectively. The mean firm size is about $2.14 billion and the median firm in our sample has total assets of $259 million. The median and mean of FOREIGN is 0 and 0.1 respectively, suggesting that about 10% of our sample firms have significant foreign operations. [Insert Table 1 here] As reported in Table 1, about 73% of our sample firms hire a Big 4 auditor (the mean of BIG4 is 0.73) and about 23% of our sample firms hire an auditor with industry specialization (the mean of SPECIALIST is 0.23). The mean of FRSK is 0.31, indicating that about 31% of the sample observations are associated with higher audit risk based on the measure adopted by Doogar et al. (2010). The percentage is lower but still comparable to the 42 percent reported by Doogar et al. (2010). Finally, about 31% of the sample observations represent accelerated filers that are subject to both Section 404 (a) and 404(b) of the Sarbanes-Oxley Act since 2004 (the mean of SOX is 0.31). 3. Empirical Analysis and Results 3.1. Audit fees before and after covenant violations: Graphical analysis Our empirical tests start with the univariate analysis that explores the trend of audit fees around covenant violations for the violating firms. Figure 1 presents the unconditional mean of 10

13 audit fees for the violating firms in our sample along the covenant violation timeline. The vertical axis represents the mean of audit fees per dollar of firm assets (audit fees scaled by the firm s total assets). The horizontal line represents the timeline around the events of covenant violations (three years before the violation year to three years after the violation year), with year 0 being the year of violation. [Insert Figure 1 here] Figure 1 indicates that the violating firms experience a significant increase in audit fees following the event of covenant violation while there is no discernible trend in audit fees during the pre-violation period. Specifically, the unconditional mean of audit fees per dollar of firm assets for violating firms rises from around cents in the year immediately before the violation year to more than 0.26 cents in the violation year. This increase in audit fees also shows persistence as Figure 1 suggests that audit fees continue to stay at a high level in the postviolation period. Overall, the trend of the unconditional average of audit fees around covenant violations for the violating firms presented in Figure 1 is consistent with our expectation that auditors exercise closer monitoring and thus charge higher fees following debt covenant violations. The magnitude of the increase in the unconditional mean of audit fees following covenant violations is both statistically and economically significant (an increase of about 40 percent from year -1 to year 0). An alternative explanation for the increase in audit fees shown in Figure 1 is that the trend is driven by economy-wide changes during the sample period that affect the overall level of audit fees regardless of whether a covenant violation has occurred or not. We address this issue by examining the audit fee trend for a subsample of the non-violating firms in our sample that are similar to the violating firms in terms of industry and size. Specifically, for each firm-year 11

14 observation for each violating firm in our sample, we select matching firm-year observations from our sample that meet the following criteria: 1) from the same calendar year, 2) the matching firm has the same 2-digit SIC code as the violating firm and, 3) the total asset of the matching firm is within +/-25% of the violating firm s total asset, 6 and 4) the matching firm has no record of covenant violations. [Insert Figure 2 here] Figure 2 presents the mean audit fee (scaled by firm size) for the constructed subsample of non-violating firms along the covenant violation timeline. The graph shows no significant difference in audit fees between the pre- and post-violation period for these matching firms during the same period that we observe a significant increase in audit fees for the violating firms in our sample. The results thus help rule out the alternative explanation that the audit fee response to covenant violations following covenant violations is driven by changes in the auditing environment unrelated to the violation event. However, we recognize that it is difficult to accurately estimate the effect of covenant violations using univariate analysis given the existing confounding factors. In the next section, we perform multivariate analysis to investigate the effect of covenant violations after taking into consideration other determinants of audit fees Covenant violations and audit fees: Immediate impact This section describes regression analysis that examines the immediate impact of covenant violations on audit fees. We first run a regression with the natural logarithm of audit fees as the dependent variable, Vio_Pre as our main variable of interest, and all the control variables (Column 1 Table 2). The purpose of this regression is to account for the potential differences in audit fees between violation and non-violating firms in absence of the violation event. In the next 6 This is consistent with the cutoff value for size matching in Farber (2005). 12

15 step, we estimate the differences in audit fees between violating firms and non-violating firms during the violation year by running a regression with the natural logarithm of audit fees as the dependent variable, Vio_Event as our main variable of interest, and the control variables (Column 2 Table 2). [Insert Table 2 here] Table 2 reports the regression results on the immediate impact of covenant violations. Notably, the coefficient on Vio_Pre in column (1) is , indicating that violating firms actually pay lower audit fees than non-violating firms before the violation event occurs. In contrast, the coefficients on the violation event dummy variables is positive (0.0410) and statistically significant. The results show that violating firms pay higher audit fees during the violation year but not before the violation event occurs, suggesting that events of covenant violations lead to higher audit fees. As reported in Table 2, the overall goodness of fit for each of the regression models is comparable to previous research (adjusted R 2 is around 0.6 in each case). The results for the control variables are largely consistent with those reported in the extant literature. Most notably, audit fees are positively associated with control variables that proxy for size (LNTA), complexity (LNSEG), and delay in the financial reporting process (DELAY). Consistent with our expectation, the coefficient on SOX is around 0.4 and statistically significant (p<0.01) in each model, suggesting that accelerated filers in the post-sox period face significantly higher audit fees. The results are consistent with the findings complied by Audit Analytics that accelerated filers experienced a spike in audit fees in 2004 due to the implementation of Section 404 (b) of the Sarbanes-Oxley Act (Audit Analytics, 2011). 13

16 3.3 Litigation risk and audit fee response to covenant violations An alternative explanation attributes the observed association between audit fees and covenant violations to heightened auditor litigation risk as covenant violation may represent a signal of the deteriorating financial prospect that significantly increases litigation risk for the engaged auditor. While previous research using large samples has established that covenant violations rarely lead to loan acceleration or financial distress (Dichev and Skinner, 2002; Gopalakrishnan and Parkash, 1995; Nini et al., 2012), we further address this concern by examining the cross sectional differences in audit fees for violating firms with varying level of auditor litigation risk. As both anecdotal evidence and findings from empirical research show that lawsuits against auditors tend to follow sharp declines in stock prices (Shu, 2000), auditors may face higher litigation risk in the post violation years if covenant violations tend to trigger large drops in stock prices. To address this concern, we partition the violating firms into two groups: violating firms with positive annual stock return during the violation year vs. violating firms with negative annual stock return during the violation year. The statistics (untabulated) show that about 43% of the new covenant violations in our sample represent violating firms that experienced positive annual stock returns during the violation year while the remaining 57% represent those that experienced negative annual stock returns during the violation year. If the audit fee response to covenant violations is driven by litigation risk, we should observe higher audit fees for the violating firms with negative stock returns during the violation year than for violating firms with positive stock returns during the violation year, using non-violating firms in the same period as the base group. 14

17 Similar to the analysis based on stock price performance, we examine cross sectional differences in audit fees by partitioning the violating firms into loss vs. profit and negative operating cash flows vs. positive operating cash flows groups, respectively. The statistics show that about 55% (63%) of the new covenant violations in our sample represent violating firms that reported a profit (positive operating cash flows) during the violation year while the remaining 45% (37%) represent those that reported a loss (negative operating cash flows) during the violation year. If higher audit fees following covenant violations arise from heightened auditor litigation risk as covenant violations signal deteriorating financial prospect, we should observe higher audit fees for violating firms with poor operating performance and cash flow generation. Therefore, we compare the audit fees for the violating firms that report a loss (negative operating cash flows) with those for the violating firms that report a profit (positive operating cash flows), using the non-violating firms in the same period as the baseline group. [Insert Table 3 here] Table 3 presents the results for the cross sectional analysis of audit fees across different levels of auditor litigation risk. The dependent variable for all the regression models in Table 3 is the natural logarithm of audit fees and all the control variables are the same as defined earlier. Column 1 compares the audit fees for violating firms with positive vs. negative stock returns during the violation year, using the non-violating firms as the baseline group. As reported in Table 3, the coefficients on Positive_Return and Negative_Return are and respectively, both being statistically significant. The results show that audit fees for both groups of violating firms are significantly higher than the baseline group of non-violating firms. However, the audit fees for violating firms with poor stock market performance during the violation year are no higher (actually lower) than those for violating firms with positive stock 15

18 returns during the violation year (0.0338<0.0508), although the difference between audit fees across the two groups of violating firms is not statistically different. As lawsuits against auditors typically follow sharp declines in stock prices, the results show that covenant violations result in higher priced audit services even when the violation event is unlikely to be associated with increased auditor litigation risk. Column 2 of Table 3 compares the audit fees for violating firms that report a profit vs. loss during the violation year. The results show that the difference in audit fees across the two groups of violating firms is not statistically significant, although both groups face significantly higher audit fees than the baseline group of non-violating firms. Column 3 reports similar results from the comparison of audit fees for violating firms that report positive vs negative operating cash flows. Again, we find that the audit fees for both positive- and negative-cash-flow-violatingfirms are higher compared to those of non-violating firms, but the difference across the two groups of violating firms is not statistically significant. Overall, the findings suggest that the higher audit fees following covenant violations are unlikely due to the violation event being a signal of deteriorating financial prospect that significantly increases auditor litigation risk. 3.4 Propensity score matching To further address the concern that audit fee differences between the violating and nonviolating firms are driven by other omitted variables, we use the propensity score matching approach to construct an optimal control group of non-violating firms and estimate the average audit fee differences between the violating firms and this control group. Unlike multivariate regression models that require the control variables to be linearly related to audit fees, the propensity score matching approach has the advantage of not restricting the specific relationship 16

19 between audit fees and the firm characteristics affecting audit pricing. Therefore, this approach helps alleviate the concern that the results from our multivariate regression analysis are biased as some non-linear terms of the firm characteristics may affect both the likelihood of a covenant violation occurrence and the level of audit pricing. The propensity score matching approach has been increasingly used in finance and accounting literature (see, for example, Drucker and Puri, 2005;, Minnis, 2011; Houston, et al., 2014) as econometric studies have shown that propensity score matching can allow for a more accurate analysis (e.g. Conniffe et al., 2000). Using the propensity score matching approach, we match violating firms with non-violating firms based on propensity scores and then calculate the average audit fee differences between the two groups of the firms. 7 The propensity scores are estimated via a probit model. The dependent variable for the probit model is the dummy variable indicating whether the firm reported a new covenant violation during the current year. The independent variables are the same as the control variables in the audit fee regression model described earlier, including firm size, profitability, leverage, growth, segments, foreign assets ratio, auditee loss indicator for net income, inventory and receivable over total assets, current and quick ratio, credit rating, fraud risk, big-four auditor dummy variable, specialist, busy, and delay indicator, going concern opinion, SOX indicator, firm and year fixed effects. [Insert Table 4 here] Table 4 presents the results from the propensity score matching analysis using different specifications (nearest neighbor matching, radius matching, Epanechnikov kernel-based functions, and stratification matching). The results show that violating firms consistently pay higher audit fees than similar non-violating firms during the violation year. The differential estimates range from 3.9% to 8.5% of the sample mean, which are not only statistically 7 The Appendix of Drucker and Puri (2005) has provided detailed explanation of this estimation method. 17

20 significant but also economically comparable to the results from our multiple regression analysis. This result alleviates the concern that endogeneity biases our main findings Covenant violations and audit fees: Long-term impact In this section, we examine whether the audit fees for violating firms stay at higher levels following the violation year after controlling for other determinants of audit fees. We first run regressions similar to those reported in Table 2 but with Vio_PostY1, Vio_PostY2, and Vio_PostY3 as the main variable of interest respectively (Columns 1-3 Table 5). Finally, we use a dynamic model to examine the long term impact of covenant violations on audit fees. Specifically, we run a regression with all the violation event dummies Vio_PostY1, Vio_PostY2, and Vio_PostY3 jointly included in the model (Column 4 Table 5). The purpose of this regression is to examine how audit fees for violating firms change in the post-violation years, after controlling for the trend of audit fees for non-violating firms during the same period. [Insert Table 5 here] Table 5 reports the regression results on the long-term audit fee response to covenant violations. Notably, the coefficients on the violation event dummy variables are consistently positive and statistically significant across Column 1 to Column 3, indicating that violating firms pay higher audit fees than non-violating firms until three years after the initial violation. As the dependent variable is the natural logarithm of audit fees, the results reported in Column 4 provide an estimate of the percentage magnitude of the long term impact of covenant violations on audit fees. For the average violating firm, after controlling the trend of audit fees for nonviolating firms during the same period, audit fees are 8.34%, 10.86%, 8.91%, and 5.2% higher during the violation year, the year immediately following the violation, two years after the 18

21 violation, and three years after the violation respectively compared with the non-event years. Overall, the results suggest that the increases in audit fees in response to debt covenant violations persist for at least three years after the initial violation Effect of lender and shareholder monitoring In this section, we investigate whether the increase in audit fees following debt covenant violations varies with the level of lender and shareholder monitoring. If the observed increase in audit fees arises from heightened demand for higher-level audit verification following covenant violations, we expect the audit fee response to be more pronounced for firms with greater level of lender monitoring. Similar to Klein (2002), we use leverage to capture the variation in the level of lender monitoring as we expect covenant violations to be more consequential events for high leverage firms. In addition to lender monitoring, previous research suggests that shareholder monitoring also plays an important role in debt contracting. In particular, Bhojraj and Sengupta (2003) demonstrate that greater institutional ownership is associated with lower borrowing costs as more effective monitoring by the shareholders helps reduce potential conflicts of interests between the management and providers of capital, including debt holders. Consistent with this notion, we expect the increased demand for higher-level of audit verification following covenant violations to be more pronounced for firms that have greater institutional ownership. We test this prediction by examining whether the increase in audit fees following covenant violations varies with the level of institutional ownership. As we are interested in how lender and shareholder monitoring moderates both the immediate and long term audit fee response to violation events, we use the dynamic model described in the 19

22 last section. Specifically, we create interaction terms for each of the post-violation dummy variables in the dynamic model (Vio_Event, Vio_PostY1, Vio_PostY2, and Vio_PostY3) with leverage and institutional ownership respectively and add these interaction terms to the dynamic model from the last section to examine how the audit fee response to covenant violations varies with the level of leverage and institutional ownership. [Insert Table 6 here] Table 6 reports the results on the effect of lender and shareholder monitoring. Consistent with our expectation, regression coefficients for the interaction terms of leverage and the violation dummy variables are consistently positive, with the coefficients for Leverage*Event, Leverage*PostY2, and Leverage*PostY3 being statistically significant. The results indicate that higher leverage level is associated with steeper increase in audit fees following covenant violations. Similarly, the regression coefficients for the interaction terms of institutional ownership and the violation dummy variables (are positive and statistically significant. The results suggest that the increase in audit fees in response to covenant violations is more pronounced for firms with greater level of shareholder monitoring. Overall, the results reported in Table 6 support the notion that the increase in audit fees following covenant violations is driven by heightened demand for higher-level audit verification Audit committee response to covenant violations To better understand the mechanisms underlying audit fee response to covenant violations, we examine the change in audit committee characteristics following debt covenant violations. As one of the key committees of the board of directors, the audit committee selects the external auditor and serves as the ultimate monitor of the financial accounting reporting process (NYSE 20

23 and NASD 1999, page 7). Consistent with the Blue Ribbon Committee Report and findings from previous research, we focus on two audit committee characteristics: audit committee independence and the frequency of audit committee meetings. Advocacy for placing more outside directors on the audit committee has a long tradition based on the notion that only outside directors who are independent of the management can perform effective monitoring. Klein (2002) provides evidence that firms choose the level of audit committee independence to suit their specific economic environments. However, in contrast to her expectation, she finds no association between audit committee independence and the proxy for the level of monitoring demanded by creditors (asset-to-debt ratio). We revisit this question in the setting of covenant violations by examining whether creditors demand for increased monitoring following covenant violations results in higher level of audit committee independence. In addition to audit committee composition, how frequently the audit committee meets is also important as it signals the diligence level of the audit committee (Anderson et al., 2004; Farber, 2005; Sommer, 1991). Thus, we examine whether the audit committee meets more frequently as the demand for monitoring increases following a debt covenant violation. Klein (2002) measures audit committee independence as the percentage of outside directors on the audit committee (the number of outside directors on the audit committee/audit committee size). This measure is not applicable to studies using data after 2000 due to the regulatory change. In response to SEC s call for better oversight of the financial reporting process, NYSE and NASDAQ modified their listing requirements in December The new standards require firms to maintain audit committees composed solely of outside directors. As a result, the majority of firms in our sample (with the sample period of 2000 to 2008) have audit committees composed of 100% outside directors. However, this does not mean that there is no variation in 21

24 the level of audit committee independence after In particular, the absolute number of outside directors on the audit committee is also an important determinant of audit committee independence. This notion is behind the modified NYSE & NASDAQ listing standards requiring that all firms must maintain audit committees with at least three independent directors. Therefore, we use the absolute number of outside directors on the audit committee to capture the level of audit committee independence. The audit committee s membership information is obtained from RiskMetrics. When such information is not directly available from RiskMetrics, we manually searched the firm s proxy statement. We have also hand collected information on the number of audit committee annual meetings from the firm s proxy statement. We lose some observations due to lack of audit committee membership/meeting data. As a result, the sample size is reduced to 19,204 firm-year observations for the audit committee independence regression and to 17,878 observations for the audit committee meeting regression. Similar to the dynamic model we use for audit fee analysis, we use five dummy variables (VIO_pre, Vio_Event, Vio_PostY1, Vio-PostY2, and Yio-PostY3) to indicate the distance to the event of violation. When constructing the baseline model for audit committee independence and the frequency of audit committee meetings, we retain the control variables in the audit fee model discussed earlier. [Insert Table 7 here] Table 7 reports estimation results on the response of audit committee to debt covenant violations. Klein (2002) hypothesizes a positive association between audit committee independence and creditors demand for monitoring, but finds no relation between audit committee independence and the proxy for the level of monitoring demanded by creditors (debt- 22

25 to-assets ratio). Consistent with this finding, our regression results on audit committee independence (Column 1 Table 5) also indicate that leverage does not have explanatory power over audit committee independence. However, we find some evidence that audit committee independence improves in response to creditors increased demand for monitoring following debt covenant violations (the coefficient on Vio_PostY1 being positive and statistically significant). The results on audit committee meetings (Column 2 Table 5) indicate that audit committees meet more frequently in response to debt covenant violations, with the coefficient on Vio_Event, Vio_PostY1, and Vio_PostY2 all being statistically significant. Overall, the results reported in Table 7 indicate that debt covenant violations lead to improved audit committee independence and more frequent audit committee meetings in the post-violation period. As the audit committee has the responsibility to review with the external auditor the scope of audit work and audit fee, the findings suggest a link between the demand for audit verification and the audit fee response following debt covenant violations. The findings are consistent with previous research showing that firms with more independent/diligent boards are more likely to purchase higher-quality audit services and thus pay higher audit fees (Carcello et al., 2002). 4. Conclusion Accounting theory posits that accounting-based debt covenants help mitigate the agency cost in debt contracting by providing a basis for benchmarking and monitoring managerial behavior. We show that this role does not stop at a debt covenant violation and extends to the resolution process by documenting that the external auditor and audit committee exercise closer monitoring in response to debt covenant violations. Using audit fee as the proxy to capture increased auditor 23

26 monitoring, we find that the audit fees for firms that reported a new debt covenant violation experience a significant increase in the violation year and continue to stay at a high level until at least three years after the initial violation. Additional tests show that audit fees are not statistically different for violating firms of different levels of auditor litigation risk, suggesting that auditor litigation risk is unlikely to be the primary driver of the audit fee response to debt covenant violations. Further analysis shows that the audit fee response to covenant violations is more pronounced for firms with more intensive lender and shareholder monitoring. Moreover, we find evidence that audit committees respond to covenant violations by including more independent directors and meeting more frequently in the post violation years. These results support the notion that the observed audit increase following covenant violations is driven by the heightened demand for differentially higher-quality audit services in response to the violation event. Our findings provide new evidence that covenant violations result in a heightened demand for higher-level audit verification. The findings complement the evidence that financial statement verification process performed by external auditors helps mitigate contracting costs in debt financing (Kim et al., 2011; Minnis, 2011). We show that this role does not stop at a debt covenant violation but extends to the resolution process, thus providing novel evidence highlighting the importance of audit verification in debt contracting. Additionally, the implications of our results are consistent with the broader view of corporate governance in the literature that models the corporate governance system as a dynamic web of various stakeholders (Triantis and Daniels, 1995). External auditors represent an important yet understudied stakeholder in the corporate governance system. Although auditors do not have direct authority for corporate governance, they actively participate in the governance process by 24

27 assuming the responsibility to reveal internal control weaknesses as well as to check on the quality of financial reporting outputs. Through its interaction with the board (particularly the audit committee), the auditor both responds to the level of monitoring demanded by the governance body and provides feedback to the governance body regarding the level of monitoring required. Despite this important role that external auditors play, there is relatively little (albeit growing) research that investigates auditing phenomena from the governance perspective. More research is warranted to provide insight in this area. 25

28 References Anderson, R.C., Mansi, S.A., Reeb, D.M., Board characteristics, accounting report integrity, and the cost of debt. Journal of Accounting and Economics 37, Armstrong, C.S., Guay, W.R., Weber, J., The role of information and financial reporting in corporate governance and debt contracting. Journal of Accounting and Economics 50, Audit Analytics, Audit fees and non-audit fees: An eight year trend. Ball, R., Jayaraman, S., Shivakumar, L., Audited financial reporting and voluntary disclosure as compliments: A test of the confirmation hypothesis. Journal of Accounting and Economics 53, Bell, T.B., Landsman, W.R., Shackelford, D.A., Auditors perceived business risk and audit fees: Analysis and evidence. Journal of Accounting Research 39, Beneish, M., Press, E., Costs of technical violation of accounting-based debt covenants, The Accounting Review 68, Beneish, M., Press, E., The resolution of technical default. The Accounting Review 70, Bhojraj, S., Sengupta, P., Effects of corporate governance on bond ratings and yields: The role of institutional investors and outside directors. Journal of Business 76, Bogart, D., Chaudhary, L., Regulation, ownership, and costs: A historical perspective from Indian railways. American Economic Journal: Economic Policy 4: Bushman, R., Smith, A.J., Financial accounting information and corporate governance. Journal of Accounting and Economics 32, Carcello, J.V., Hermanson, D.R., Neal, T.L., Riley, R.A., Jr., Board characteristics and audit fees. Contemporary Accounting Research 19, Chava, S., Roberts, M., How does financing impact investment? The role of debt covenants. Journal of Finance 63, Conniffe, D., Gash, V., O Connell, P.J., 2000, Evaluating state programmes: Natural experiments and propensity scores, The Economic and Social Review 31, DeFond, M., Jiambalvo, J., Debt covenant violation and manipulation of accruals. Journal of Accounting and Economics 17, Dichev, I., Skinner, D., Large-sample evidence on the debt covenant hypothesis. Journal of Accounting Research 40, Doogar, R., Sivadasan, P., Solomon, I., The regulation of public company auditing: evidence from the transition to AS5. Journal of Accounting Research 48, Drucker, S., Puri, M., On the benefits of concurrent lending and underwriting. Journal of Finance 60, Fan, J.P.H., Wong, T.J Do external auditors perform a corporate governance role in emerging markets? Evidence from East Asia. Journal of Accounting Research 43, Farber, D., Restoring trust after fraud: Does corporate governance matter? Accounting Review 80, Francis, J., Reichelt, K., Wang, D., The pricing of national and city-specific reputations for industry expertise in the U.S. audit market. The Accounting Review 80, Gopalakrishnan, V., Parkash, M., Borrower and lender perceptions of accounting information in corporate lending agreements, Accounting Horizons 9, Hay, D., Knechel, C.R.,Wong, N., Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research 23, Houston, J., Jiang, L., Lin, C., Ma, Y., Political connections and the cost of bank loans. Journal of Accounting Research 52,