Unintended consequences of the increased asset threshold for FDICIA internal controls: Evidence from U.S. private banks

Journal of Banking & Finance 37 (2013) 4879–4892

Contents lists available at ScienceDirect

Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf

Unintended consequences of the increased asset threshold for FDICIA internal controls: Evidence from U.S. private banks Justin Yiqiang Jin a,1, Kiridaran Kanagaretnam b,2, Gerald J. Lobo c,⇑ a

DeGroote School of Business, McMaster University, Hamilton, Ontario L8S 4M4, Canada Schulich School of Business, York University, Toronto ON M3J 1P3, Canada c C.T. Bauer College of Business, University of Houston, Houston, TX 77204, United States b

a r t i c l e

i n f o

Article history: Received 14 January 2013 Accepted 24 August 2013 Available online 5 September 2013 JEL classification: G14 G21 M41 M42 Keywords: FDICIA Internal controls Bank failure Bank financial trouble Audit quality

a b s t r a c t We examine the unintended consequences of the 2005 increase from $500 million to $1 billion in the asset threshold for the Federal Deposit Insurance Corporation Improvement Act (FDICIA) internal control reporting requirements. We focus on a test sample of banks that increased their total assets from between $100 million and $500 million prior to the change in regulation to between $500 million and $1 billion within two years following the change. These ‘‘affected’’ banks are no longer subject to the internal control requirements but would have been had the regulation not been changed. We hypothesize that these affected banks are likely to make riskier loans, which will increase the likelihood of failure during the crisis period. We find evidence consistent with this hypothesis. Affected banks have higher likelihood of failure during the crisis period than banks from two different control samples. We also find that auditor reputation (i.e., whether the bank is audited by a Big 4 auditor or an industry specialist auditor) has a moderating effect on the likelihood of failure for these affected banks. Ó 2013 Published by Elsevier B.V.

1. Introduction In 2005, the Federal Deposit Insurance Corporation (FDIC) increased the asset threshold exempting U.S. banks from internal control assessment by management and external auditors from $500 million to $1 billion. The timing of this relaxation of the regulation (which had been in place since 1993) is of particular interest because it coincided with the height of the real estate bubble and the banking sector boom, which were followed by the banking crisis of 2007–2009. In this study, we examine an unintended consequence of the 2005 regulatory change  increased bank failure during the recent financial crisis  for a sample of U.S. private banks that would have been subject to the internal control regulation but no longer are as a result of the regulatory change. In response to the savings and loan debacle of the 1980s, the United States Congress enacted the Federal Deposit Insurance Corporation Improvement Act (FDICIA) in 1991 to strengthen the financial condition of the banking and thrift industries. While

FDICIA contains much more than deposit insurance reform, of particular interest are the requirements for annual audit and reporting of management’s and auditors’ assessments of the effectiveness of internal control for banks with $500 million or more in total assets. More specifically, FDICIA requires the management of these institutions to evaluate the internal control over financial reporting and the auditor to attest to the report on the effectiveness of internal controls over financial reporting. These regulations, especially the regulations related to internal control requirements for financial reporting, were passed ostensibly to enhance the transparency of reported financial information. In particular, these provisions were intended to aid in the early detection of problems in the financial management of insured banks, since early warning systems depend on reliable accounting information. U.S. regulators use the Uniform Financial Rating System, informally known as CAMELS ratings, to assess the health of individual banks.3 Following an on-site examination, bank examiners assign a score from 1 (best) to 6 (worst) for each of the six CAMELS components as well as a single summary measure, known as the composite

⇑ Corresponding author. Tel.: +1 713 743 4838. E-mail addresses: [email protected] (J.Y. Jin), kkanagaretnam@schulich. yorku.ca (K. Kanagaretnam), [email protected] (G.J. Lobo). 1 Tel.: +1 905 525 9140. 2 Tel.: +1 416 736 2100. 0378-4266/$ - see front matter Ó 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.jbankfin.2013.08.024

3 The most widely known rating system for banks is the CAMELS system, which stands for Capital Adequacy, Asset Quality, Management, Earnings, Liquidity, and Systematic Risk.

4880

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

rating. In most cases, the intensity of regulatory monitoring and supervision is based on the composite CAMELS rating. Given that the CAMELS rating system is primarily based on accounting numbers from regulatory filings (commonly known as call reports), a reliable financial reporting system is critical to the effectiveness of the regulatory process. This is especially true for private banks that do not file audited financial statements for broader public consumption. In addition to increasing the asset size threshold for internal control assessments by management and external auditors in 2005, the FDIC relaxed the requirement for independent outside directors to serve on the audit committee but retained the financial statement audit and other reporting requirements for all institutions with $500 million or more in total assets.4 This exemption is partly motivated by the exemption for small firms from Sarbanes–Oxley Act (SOX) Section 404 internal control reporting requirements and is intended to grant relief from the high burden of compliance to smaller banks. The FDIC’s intentions are clearly expressed in the following excerpt from an FDIC report: As the environment has changed and continues to change since the enactment of the Sarbanes–Oxley Act, the FDIC has observed that compliance with the audit and reporting requirements of Part 363 has and will continue to become more burdensome and costly, particularly for smaller non-public covered institutions. Thus, the FDIC reviewed the current asset size threshold for compliance with Part 363 in light of the discretion granted by Section 36 that permits the FDIC to determine the appropriate size threshold (at or above $150 million) at which insured institutions should be subject to the various provisions of Section 36. Based on this review, the FDIC proposed to amend Part 363 to increase the asset size threshold for internal control assessments by management and external auditors from $500 million to $1 billion (FDIC, 2008). Furthermore, the FDIC believed that raising the threshold to $1 billion would achieve meaningful burden reduction without sacrificing safety and soundness. In reaching this decision, the FDIC concluded that raising the $500 million asset size threshold to $1 billion and exempting all institutions below this higher size level from all of the reporting requirements of Part 363 would be inconsistent with the objective of the underlying statute (i.e., early identification of needed improvements in financial management). However, the FDIC believed that relieving smaller covered institutions from the burden of internal control assessments, while retaining the financial statement audit and other reporting requirements for institutions with $500 million or more in total assets, struck an appropriate balance in accomplishing this objective. Our main focus in this study is to examine whether relaxation of the rules on sound financial management through effective monitoring of internal controls (i.e., some of the requirements of FDIC Part 363) just prior to the banking crisis led to increased failure for the group of affected banks, presumably because they engaged in riskier behavior. We reason that these affected banks would not have engaged in such risky behavior and hence would not have experienced a higher incidence of failure had they not been exempted from the earlier, more stringent financial reporting and internal control requirements. We also examine whether auditing and auditor reputation reduce these banks’ likelihood of failure. We obtain data for our empirical analysis from the Federal Reserve Bank of Chicago’s Commercial Bank Data quarterly call reports, which covers all private banks insured by the FDIC. Our research focuses on private banks that grew from less than $500 4 Please see Section 2 on institutional background for a more complete discussion of FDICIA and FDIC Part 363, which stipulates the requirements on independent audit and internal controls.

million in assets prior to 2005 (banks that were not subject to FDIC Part 363) to between $500 million and $1 billion in assets in 2007 (banks that were exempted from compliance with key requirements of FDIC Part 363 as a result of the increase in minimum asset threshold). More specifically, our test sample consists of 194 private banks with assets between $100 million and $500 million on January 1, 2005, that grew their assets to between $500 million and $1 billion by January 1, 2007.5 Our objective in choosing these banks is to identify a sample of private U.S. banks that would otherwise have been subject to the FDICIA internal control requirements just prior to the banking crisis. We compare this affected bank sample to a growth-matched control sample of 194 private banks. This control sample includes private banks with assets between $100 million and $500 million on January 1, 2005, and that remained between $100 million and $500 million on January 1, 2007, that are matched with affected banks based on the percentage growth of total assets from January 1, 2005, to January 1, 2007. This ensures that both the affected banks and the control banks have similar growth rates after the 2005 change in minimum asset threshold for FDICIA internal controls. We refer to this control sample as the growthmatched control sample. Because this control sample includes smaller private banks that were not subject to FDICIA internal control requirements both before and after the 2005 rule change, we also test our main predictions using an additional control sample that includes 65 banks with total assets greater than $500 million in the period 2000–2004 and more than $1 billion in the period 2005– 2006.6 These banks were subject to FDICIA internal control requirements in the pre- and post-2005 FDICIA asset threshold rule change and are somewhat similar in size to the affected banks. We refer to this group of banks as the FDICIA control sample. We report several key findings. First, affected banks have a reliably higher likelihood of failure than control sample banks during the crisis period. This result is obtained from comparisons with each of the two control samples (i.e., the growth-matched control sample and the FDICIA control sample). In additional tests, we document that affected banks also have a higher likelihood of financial trouble, as reflected in large losses (poor performance) and large loan loss provisions (low asset quality), and overall bank trouble (a combined measure that includes banks with one or more of the following: a large loss, a large loan loss provision, and a low capital). Again, this result is obtained for both control groups. For the tests examining the effects of auditor reputation, using the growthmatched control sample, we find that auditor reputation (i.e., whether the bank has a Big 4 auditor) has a moderating effect on the likelihood of bank failure for the test affected banks. Our study adds to prior literature that examines the unintended consequences of government regulations. For example, Gao et al. (2009) study the effect of exempting small firms (firms with public float less than $75 million) from SOX Section 404 internal control requirements. They find that these exempted firms remained small through deliberate actions such as making fewer investments and more bad news disclosures than a set of control firms. In contrast, we provide evidence of a lack of proper financial management and its negative consequences as a result of lax internal controls for a sample of private U.S. banks that would otherwise have been subject to the internal control requirements. Our study documents the implications of lax internal controls on bank performance during the recent financial crisis. This crisis provides a unique environment in which to study the benefits of internal controls, especially since it appears that the widespread

5 We focus on asset growth by January 1, 2007, because it is generally accepted that the recent financial crisis started in the latter half of 2007 (Ryan 2008; Erkens et al. 2012). 6 We exclude FDICIA banks that had total assets over $1.5 billion in 2006 to limit the differences in bank size between affected banks and control banks in this sample.

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

bank failures coupled with the recent, massive write-downs of loans in the banking industry came as a surprise to regulators and the financial community.7 Since the FDIC has considerable latitude in choosing the minimum asset threshold for auditing and internal control requirements, our results will serve as useful input to policy makers in their future deliberations on auditing requirements and asset thresholds. In a related study, Jin et al. (2013) examine the impact of FDICIA internal control requirements on banks’ risk-taking behavior prior to the recent financial crisis and the consequent implications for bank failure and financial trouble during the crisis period. Our study differs from Jin et al. (2013) in two major respects: target sample and research questions. Although both studies examine the FDICIA requirements for annual audit and reporting of management’s and auditors’ assessment of the effectiveness of internal control, the target samples differ between the two studies. Jin et al. (2013) compare the risk-taking behaviors of two groups of banks (target sample banks that are required to comply with the FDICIA internal control requirements and control sample banks that are not required to comply with the FDICIA internal control requirements). Their target sample banks are subject to the FDICIA internal control requirements both before and after the 2005 asset threshold change. Our study, however, focuses on a target sample of banks that would have been subject to the FDICIA internal control requirements but no longer are because of the increased asset threshold. We believe that studying this group of banks is important, especially given the subsequent changes by FDICIA in 2009 (please see Section 2.1 for details) related to this specific group, presumably because of the resulting increase in internal control risks. Jin et al.’s (2013) research focus is on whether banks required to comply with the FDICIA internal control requirements have lower risk-taking behaviors in the pre-crisis period and are less likely to experience failure and financial trouble during the crisis period. By contrast, our study focuses on whether the affected banks have a higher likelihood of failure than the control banks. Furthermore, our study examines whether auditor reputation mitigates the likelihood of bank failure for these affected banks, whereas Jin et al. (2013) do not test the impact of auditor reputation on bank failure. We present the institutional background on FDICIA, particularly the 2005 changes, in the next section. We also discuss the effect of FDICIA internal controls on bank failure and the effect of auditing on bank failure. We describe the sample in Section 3, discuss the results in Section 4, and present the conclusions of our study in Section 5. 2. Institutional and research background 2.1. Institutional background In response to the breakdown of thrifts and banks during the late 1980s and early 1990s, the United States Congress enacted FDICIA in 1991 to strengthen the financial condition of the banking and thrift industries. FDICIA is the most important banking legislation in the U.S. since the Banking (Glass-Steagall) Act of 1933 (Benston and Kaufman, 1997). Although FDICIA contains much more than deposit insurance reform, of particular interest are the requirements for annual audit and reporting of management’s assessment of the effectiveness of internal control that are intended to aid early detection of problems in the financial management of insured banks. In June 1993, the FDIC enacted Part 363, titled ‘‘Annual Independent Audits and Reporting Requirements,’’ 7 The FDIC quarterly review of banks for the first quarter of 2007 highlights the total net income of $36.0 billion for insured commercial banks and savings institutions as the fourth-highest amount ever reported without any mention of an impending banking crisis.

4881

which requires each insured depository institution with $500 million or more in total assets at the beginning of the year to prepare the following (see Sections 363.2 and 363.3 of FDIC 1993):8,9 (1) Audited annual financial statements; (2) A management report signed by the chief executive officer (CEO) and chief financial officer (CFO) that contains the following: (a) a statement of management’s responsibilities for i. preparing financial statements, ii. establishing and maintaining an adequate internal control structure over financial reporting, and iii. complying with the laws and regulations that are designated by the FDIC and other federal banking agencies; (b) assessments by management of i. the effectiveness of the internal control structure over financial reporting and ii. the institution’s compliance with the designated laws and regulations during the year; (3) The independent public accountant’s report on the audited financial statements; and (4) The independent public accountant’s attestation report on the assertion of management concerning the institution’s internal control structure and procedures for financial reporting. Effective in 2005, the threshold for requirements (2b) i and 4 was increased from $500 million to $1 billion in total assets (FDIC, 2005b). Although the FDIC increased the asset size threshold for internal control assessments performed by management and external auditors, it retained the financial statement audit and other reporting requirements for all institutions with $500 million or more in total assets to preserve safety and soundness in financial management (FDIC, 2005a). Additionally, it relaxed the requirement that the audit committee include independent directors who are outside directors and independent of management. The FDIC noted, ‘‘FDIC has observed that a number of smaller covered institutions, particularly those with few shareholders that have recently exceeded $500 million in total assets and become subject to Part 363, have encountered difficulty in satisfying the independent audit committee requirement’’ (FDIC, 2008). These internal control reporting requirements and audit committee independent-director exemptions are partly motivated by the exemption for small firms from SOX Section 404 internal control reporting requirements and partly to grant smaller banks relief from the higher burden of compliance. Although most of the comments the FDIC received on its request for comments on this proposal to change the asset threshold were positive, the FDIC’s Office of Inspector General (FDIC-OIG) alerted the FDIC to the proposal’s potential adverse effects (FDIC, 2008). According to the FDIC, ‘‘the FDIC-OIG indicated that, in reviewing past failures of insured institutions, it had observed that weak corporate governance, including financial reporting problems and the lack of independence of the board of directors from institution management, was often a factor in the failure of these 8 Section 112 of FDICIA added Section 36, ‘‘Early Identification of Needed Improvements in Financial Management,’’ to the FDI Act. Section 36 gives the FDIC discretion to select an asset-size threshold of $150 million or more to exempt smaller depository institutions from the above reporting requirements. This exemption was intended to ease the compliance cost for small and community banks. 9 In addition, Section 363.5 requires each depository institution to establish an independent audit committee, comprising outside directors who are independent of its management.

4882

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

institutions and contributed to material losses ($25 million or more) to the deposit insurance funds administered by the FDIC.’’ The FDIC-OIG also stated that maintaining the full requirements of Part 363 for less-than-satisfactory institutions would address potential concerns about deficiencies in the board of directors and internal control, internal audit, and external audit, and thereby mitigate the possibility of institution failure. Notwithstanding these concerns, the FDIC went ahead with the asset threshold increase in 2005. In June 2009, the FDIC made additional amendments to Part 363 regarding its regulation, which sets forth annual independent audit and reporting requirements for insured institutions with $500 million or more in total assets. These amendments were adopted in light of the aftermath of the severe banking crisis experienced in the U.S. to facilitate the incorporation of certain audit, reporting, and audit committee practices that belong to SOX Section 404, which do not apply to private banks and small public banks. According to the FDIC, these amendments also provide clearer and more complete guidance for compliance with Part 363. Highlights of these changes include the following (FDIC, 2009):  Annual Reporting Requirement As amended, Part 363 requires disclosure of the internal control framework and identified material weaknesses, requires management’s assessment of compliance with laws and regulations to disclose any noncompliance, and provides illustrative management reports.  Independent Public Accountants As amended, Part 363 clarifies the independence standards applicable to accountants, requires certain communications to audit committees, and establishes a uniform retention requirement for audit working papers.  Filing and Notice Requirements The amendments extend the annual-report-filing deadline for non-public institutions and include a late filing notification requirement.  Audit Committee The amendments specify that the audit committee’s duties regarding the independent public accountant require audit committees to ensure that audit engagement letters do not contain unsafe and unsound limitation of liability provisions and require boards of directors to develop and apply written criteria for evaluating audit committee members’ independence. These changes reflect the importance of effective internal controls and independent audit functions, which are key to a reliable financial reporting process and early identification of trouble in the banking sector. However, it is important to note that these additional changes were made only after the recent banking crisis. 2.2. FDICIA internal controls Recent research has examined the effects of FDICIA internal control requirements on bank earnings quality and auditor independence. For example, Altamuro and Beatty (2010) examine several earnings quality measures prior to and following FDICIA and find that the mandated internal control requirements increased the validity of loan loss provisions (LLP), increased earnings persistence and cash flow predictability, and reduced benchmark-beating and accounting conservatism for affected as opposed to unaffected banks. Kanagaretnam et al. (2010a) study the relationship between earnings management via loan loss provisions and fees paid to the auditor, particularly for banks that are exempt from FDICIA and SOX Section 404 internal control regulations. They find a strong negative association between income-increasing (negative) abnormal LLP and both unexpected total fees and unexpected non-audit fees for small banks. In other words, higher auditor fees are associated with higher income-increasing earnings

management for small banks. These authors also document a negative association between unexpected non-audit fees and incomedecreasing (positive) abnormal LLP for small banks. Their results indicate greater earnings management by small banks that are exempted from internal control regulations and that pay higher abnormal fees to the auditor. Taken together, Altamuro and Beatty (2010) and Kanagaretnam et al. (2010a) provide consistent evidence that FDICIA-mandated internal control requirements are associated with better bank earnings quality. How important are internal control requirements for a bank’s financial health? Jin et al. (2013) find that banks required to comply with the FDICIA internal control requirements exhibit lower risk-taking behavior in the pre-crisis period. Specifically, the volatility of net interest margin, volatility of earnings, and Z score show less risk-taking behavior. Furthermore, these banks are less likely to experience failure and financial trouble during the crisis period. A growing body of literature has studied internal control weaknesses in industrial firms. Several studies examine the impact of material internal control weakness on information uncertainty and the cost of capital, and the evidence provided is mixed (e.g., Ogneva et al., 2007; Beneish et al., 2008; Hammersley et al., 2008; Ashbaugh-Skaife et al., 2009). Zhang et al. (2007) find that firms are more likely to be identified with an internal control weakness if their audit committees have less financial expertise or their auditors are more independent. Bronson et al. (2009) find that the benefits of audit committee independence are consistently achieved only when the audit committee is completely independent. Schneider and Church (2008) find that an adverse internal control opinion can undermine the assurance provided by an unqualified opinion on financial statements and have a negative effect on lenders’ assessments. Lopez et al. (2009) suggest that an adverse audit opinion on internal control over financial reporting provides incremental value-relevant information to investors beyond that contained in the financial statement audit opinion alone. According to a Government Accountability Office report (GAO, 1991), accounting, internal control, and auditing elements of the system of corporate governance are essential for a successful and effective regulatory process. Following the savings and loan debacle, the GAO in the late 1980s conducted a detailed study of accounting and auditing reforms needed to strengthen the regulatory environment. It analyzed 39 banks that failed in 1988 and 1989, with a focus on the impact of accounting and internal control weaknesses associated with those failures. GAO (1991) indicates that ‘‘Of the 39 banks, 33 had serious internal control problems which regulators cited as contributing significantly to their failure. Had these problems been corrected, the banks might not have failed or their failure could have been less expensive to the Fund.’’ GAO (1991, p. 6) goes onto recommend that the roles of both management and the auditors would be strengthened if they were required to assume responsibility for assessing and reporting on the condition of internal controls. The above discussion suggests that both Congress and regulators were aware of the critical role that internal control reporting requirements play in the health of a bank. However, for the test sample of affected banks (which is the focus of this study), internal control assessment by management and external auditors was relaxed for the period leading up to the banking crisis. 2.3. Audit quality and failed banks We also compare the effect of audit quality on the likelihood of a bank failing across our test and control banks. Given that auditing is an important external monitoring mechanism, auditing, and especially higher quality auditing, likely reduces the probability of a bank getting into trouble and subsequent failing. In addition, auditing banks is more complex than auditing industrial firms. This

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

claim is supported by the American Institute of Certified Public Accountants (AICPA, 2006) Center for Public Company Audit Firms in its May 2006 report on large-firm Public Company Accounting Oversight Board (PCAOB) inspection deficiency analysis, which reported that bank loan loss allowance ranks number one among the various deficiencies found by inspectors. The findings of this report indicate that auditing of the loan loss allowance and the related loan loss provision is a challenging task for auditors in general. In addition to studying the role of the external auditor in whether a bank has failed, we also examine the impact of auditor reputation on a bank’s health. High-reputation auditors have incentives to provide high-quality audits to avoid jeopardizing their reputation capital. Consequently, banks audited by high-reputation auditors are less likely to face trouble. We examine two aspects of auditor reputation. First, we investigate the implications of auditor type (Big 4 vs. non-Big 4 auditors) for bank failure. A large body of empirical research documents that higher audit quality is associated with Big 4 auditors for industrial firms (Dopuch and Simunic, 1982; Palmrose, 1988; Teoh and Wong, 1993; Craswell et al., 1995; Becker et al., 1998; Khurana and Raman, 2006; Behn et al., 2008). Research focusing on restatements draws similar conclusions. For example, Lennox and Pittman (2010) report a strong negative relationship between Big 5 auditors and accounting frauds. Relative to non-Big 4 auditors, Big 4 auditors have greater expertise, resources, and (more important) market-based incentives (e.g., mitigating the risk of litigation and protecting their reputation capital) to constrain the tendency of their audit clients to engage in aggressive reporting or fraud. Although empirical evidence on auditor reputation and audit quality in the banking industry is limited, the economic incentives faced by Big 4 auditors of banks are similar to those faced by Big 4 auditors in other industries (i.e., preserving reputation capital and mitigating the risk of litigation). In addition, auditor type may be more important for industries such as banking, where information uncertainty is higher than industrial firms due to the greater complexity of banking operations and the difficulty of assessing risk on a large portfolio of loans (Autore et al., 2009). Given the above, we predict that the probability of bank failure is lower for banks audited by Big 4 auditors. Second, we examine the relationship between auditor industry specialization and the likelihood of a bank’s failure. We measure auditor industry specialization/expertise by an auditor’s industry market share. Auditors who are specialists in the banking industry can better assess the adequacy of the loan loss provision and loan loss allowance than non-specialist auditors. Several studies examine the benefits of auditor industry specialization or expertise on audit effectiveness for industrial firms (Bedard and Biggs, 1991; Carcello and Nagy, 2004; Krishnan, 2003). Overall, the evidence from this stream of research indicates that the ability to detect material misstatements in financial reporting is associated with auditor industry specialization. In the banking industry, Kanagaretnam et al. (2009) find that once auditor type and industry expertise are separated, only auditor industry expertise has a significant impact on the valuation of discretionary LLP. In summary, the collective evidence indicates that there are benefits to auditor industry specialization in terms of enhanced audit effectiveness and credibility of financial statements. Additionally, in an international banking setting, Kanagaretnam et al. (2010b) find that auditor industry specialization constrains earnings management. Therefore, we predict that the probability of failure is lower for banks audited by industry specialists. In related research employing a large sample of FDICIA and nonFDICIA banks (covering both public and private banks), Jin et al. (2011) document a lower probability of failure for banks audited by more reputable auditors. Overall, our prediction that higher auditor reputation lowers the probability of failure for non-FDI-

4883

CIA-affected banks is consistent with the findings of Jin et al. (2011). 3. Sample selection and model 3.1. Sample selection We obtain accounting data on private banks from the Commercial Bank Data quarterly call reports available from the Federal Reserve Bank of Chicago’s website.10 We construct annual measures of the variables of interest from quarterly data. As indicated earlier, we identify a sample of affected banks and compare changes in the characteristics of these banks to corresponding changes in a control sample of banks. The affected bank sample includes private banks that had assets between $100 million and $500 million on January 1, 2005, and grew their assets to between $500 million and $1 billion by January 1, 2007. Because these banks had assets greater than $500 million on January 1, 2007, they would have been subject to the FDICIA internal control requirements if the regulation had not changed. We identify 194 private banks that meet these criteria. We obtain the auditor name variable TEXTC703 from the Bank Holding Company Data quarterly files. To merge the Commercial Bank Data quarterly call reports and Bank Holding Company Data quarterly files, we use the variable RSSD9348 of the Commercial Bank Data to match the variable RSSD9001 of the Bank Holding Company Data. We employ two groups of control banks in our analyses because each group of control banks can account for specific confounding factors. Our first control group (referred to as the growth-matched control sample) includes a sample of 194 private banks that had assets between $100 million and $500 million on January 1, 2005, and remained between $100 million and $500 million on January 1, 2007. We match the affected banks with the control banks using the percentage growth of total assets from January 1, 2005, to January 1, 2007, plus or minus 1%. If multiple control banks meet this asset growth rate matching criterion, we choose the control bank that is nearest in total assets to the corresponding affected bank. These banks are similar to the affected banks in terms of asset growth; however, despite their growth, these control banks would not have been subject to the internal control regulation, even if the asset threshold had not been raised in 2005. The second control group (referred to as the FDICIA control sample) comprises 65 banks that were required to comply with FDICIA internal control requirements both before and after the 2005 rule change (i.e., banks with total assets greater than $500 million in the period 2000 to 2004 and more than $1 billion in the period 2005 to 2006). These control banks are similar to the affected banks in that they were similar in size both before and after the 2005 rule change. However, they differ from the affected banks because they had to comply with FDICIA internal control requirements both before and after the 2005 rule change, whereas the affected banks did not. We require all banks in the test and control samples to have complete data available from 2006 through 2010. The control variables in our regressions are measured at the end of 2006 (i.e., just prior to the financial crisis that started in 2007), and we use data from the years 2007–2010 to identify financially troubled banks for our additional analysis. For our main analysis using the failed banks, we identify a total of 31 failed banks, 23 from the test sample of affected banks and 8 from the growth-matched control sample, from the failed-banks list published by the FDIC.11,12 In additional analysis, we repeat our tests for financially troubled banks. As discussed earlier, bank 10 We focus on private banks because most of the banks affected by the FDICIA rule change were private banks. 11 http://www.fdic.gov/bank/individual/failed/banklist.html. 12 In the FDICIA control sample, we identify 3 failed banks.

4884

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

Table 1 Descriptive statistics for dependent variables. Variablea

Affected banks mean (1)

Growth-matched control banks mean (2)

FDICIA control banks mean (3)

Difference in means (t-statistic) (1) and (2)

Difference in means (t-statistic) (1)–(3)

FB

0.119

0.041

0.046

LARGE_LOSS

0.098

0.036

0

LARGE_LLP

0.392

0.268

0.200

SMALL_CAP

0.093

0.031

0.062

TB

0.443

0.289

0.231

0.078*** (2.83) 0.062** (2.45) 0.124*** (2.61) 0.062** (2.54) 0.154*** (3.20)

0.073** (2.07) 0.098*** (4.58) 0.192*** (3.14) 0.031 (0.85) 0.212*** (3.34)

Table presents mean differences for the dependent variables between the affected banks and control banks during the financial crisis. The measurement period for FB, LARGE_LOSS, LARGE_LLP, SMALL_CAP and TB is 2007–2010. The affected sample contains 194 private bank annual observations. The growth-matched control sample contains 194 private bank annual observations. The growth-matched control group includes a sample of private banks that had assets between $100 million and $500 million on January 1, 2005, and remained between $100 million and $500 million on January 1, 2007. We match the affected banks with the control banks using the percentage growth of total assets from January 1, 2005 to January 1, 2007 plus or minus 1%. If multiple control banks match with the same affected bank using the asset growth rate, we choose the control bank with the closest total assets of the corresponding affected bank. The FDICIA control banks have 65 annual observations. FDICIA banks include banks with total assets greater than $500 million in the period 2000 to 2004 and more than $1 billion in the period 2005 to 2006. We exclude FDICIA banks that have total assets over $1.5 billion in 2006. We collect the accounting data from the Commercial Bank Data call reports (available at the Federal Reserve Bank of Chicago website http://www.chicagofed.org/).  Indicate significance at the 10% level based on a two-tailed test. ** Indicate significance at the 5% level based on a two-tailed test. *** Indicate significance at the 1% level based on a two-tailed test. a Variable Definitions: FB equals 1 if the bank failed during 2007–2010, and 0 otherwise. LARGE_LOSS equals 1 if ROA is less than 5% in any one of the four years during 2007–2010, and 0 otherwise. LARGE_LLP equals 1 if the ratio of loan loss provision to total loans is greater than 4% in any one of the four years during 2007–2010, and 0 otherwise. SMALL_CAP equals 1 if Tier 1 capital ratio is less than 8% in any one of the four years during 2007–2010, and 0 otherwise. TB equals 1 if any one of LARGE_LOSS, LARGE_LLP and SMALL_CAP equals 1 in any one of the four years during 2007–2010, and 0 otherwise.

examiners use the CAMELS rating system, which is based on several financial ratios and management characteristics, to identify banks that are in financial trouble. Because this rating is not publicly available, we classify banks as troubled using publicly available data that reflect profitability, asset quality, and capital adequacy if they met any of the following three criteria in any year during 2007–2010: (1) return on assets (ROA) is less than 5%, labeled as LARGE_LOSS (a proxy for performance); (2) loan loss provision divided by total loans is greater than 4%, labeled as LARGE_LLP (a proxy for asset quality); and (3) Tier 1 capital ratio is less than 8%, labeled as SMALL_CAP (a proxy for balance sheet strength).13 Our fourth measure of troubled banks, labeled as TB, is an overall measure of bank trouble. It classifies a bank as troubled if the bank meets any one (or more) of the above three criteria. We employ a simple univariate analysis (mean differences between the affected and control samples) and a more detailed multivariate analysis (using the two control samples) to test our predictions. Table 1 presents descriptive statistics for the dependent variables, the proportion of failed banks (FB), and the proportion of troubled banks (TB) for the affected bank sample and for the two control samples during the financial crisis. The results are consistent with our prediction that banks that grew without proper internal controls have higher likelihood of failing (or getting into financial trouble). The means of FB, LARGE_LOSS, LARGE_LLP, and TB for the affected bank sample are each significantly higher than the corresponding means for the growth-matched control sample and the FDICIA control sample. Panel A of Table 2 presents descriptive statistics for the independent variables, including AUDITED, BIG4, and SPECIALIST and other control variables, for the affected bank sample and the growth-matched control sample in 2006 (just prior to the financial crisis). The proportion of banks audited by independent auditors is significantly higher for the affected banks than for the growthmatched control banks. There are no significant differences between the affected banks and the growth-matched control banks in the proportion of banks audited by Big 4 auditors or specialist

13

We winsorize observations in the top 1% of the distribution of each continuous variable to reduce the effects of extreme values on our results.

auditors. The affected banks have a significantly lower Tier 1 capital ratio, a lower percentage of non-performing loans over total loans, and a lower ratio of interest-bearing balances to total assets than do the growth-matched control banks. Panel B of Table 2 presents Pearson correlations among the variables used in the multivariate regressions for the combined affected banks and growth-matched control banks. The dependent variables (FB, LARGE_LOSS, LARGE_LLP, SMALL_CAP, and TB) are significantly positively correlated with AFFECTED (a binary variable that equals 1 if the bank belongs to the affected sample and 0 otherwise), indicating that the affected banks are more likely to fail and have financial trouble in the crisis period than the growthmatched control banks. LARGE_LLP is significantly positively correlated with CI, indicating that banks with large loan loss provisions have higher growth in commercial and industrial loans. TB is significantly positively correlated with ALLOWANCE and CI, indicating that banks with a large allowance for loan losses or a large percentage growth of commercial and industrial loans are more likely to have financial trouble. 3.2. Model for testing probability of bank failure during the crisis period As our main test, we examine the prediction that the probability of bank failure is higher for the affected bank sample than for the control sample. Our test specifications closely follow Lel and Miller (2008), Beltratti and Stulz (2010) and Erkens et al. (2012). We estimate the following model to assess whether affected banks have a higher probability of failure/trouble than banks in the control sample:

FB ¼ b0 þ b1 AFFECTED þ b2 CAP þ b3 NPL þ b4 LIQUIDITY þ b5 ALLOWANCE þ b6 RER14 þ b7 CI þ b8 G ASSETS þ b9 SIZE R þ b1 0LEV þ b1 1REGION2 þ b1 2REGION3 þ b1 3REGION4 þ e

ð1Þ

The dependent variable in Eq. (1), FB, is an indicator variable that equals 1 if the bank failed during 2007–2010 and equals 0 otherwise. AFFECTED is an indicator variable that equals 1 if a bank had assets

Table 2 Descriptive statistics for independent variables and pearson correlation. Variablea

Affected banks mean

Difference in means (t-statistic)

0.325

BIG4

0.041

0.031

SPECIALIST

0.031

0.026

CAP

8.807

9.412

NPL

0.096

0.158

LIQUIDITY

0.029

0.036

ALLOWANCE

0.032

0.033

RER14

0.160

0.119

CI

0.175

0.103

G_ASSETS

0.168

0.174

SIZE_R

0.0005

0.0004

LEV

0.095

0.098

REGION2

0.237

0.289

REGION3

0.428

0.418

REGION4

0.165

0.108

0.190*** (3.87) 0.010 (0.54) 0.005 (0.31) 0.605*** (2.74) 0.062** (2.26) 0.007*** (2.78) 0.001 (1.32) 0.041 (1.01) 0.072* (1.88) 0.006 (0.40) 0.0001 (0.06) 0.003 (1.25) 0.052 (1.15) 0.010 (0.21) 0.057 (1.63)

2 Panel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

3

4

5

6

7

8

9

10

11

12

B: Pearson correlations between dependent and independent variables for affected banks and growth-matched control banks FB 0.45 0.30 0.48 0.39 0.14 0.01 0.06 0.05 0.05 0.02 0.02 LARGE_LOSS 1.00 0.27 0.57 0.35 0.12 0.01 0.05 0.05 0.02 0.03 0.06 LARGE_LLP 1.00 0.23 0.92 0.13 0.12 0.01 0.01 0.06 0.07 0.09 SMALL_CAP 1.00 0.34 0.13 0.01 0.05 0.04 0.05 0.07 0.07 TB 1.00 0.16 0.08 0.03 0.01 0.05 0.06 0.09 AFFECTED 1.00 0.19 0.03 0.02 0.13 0.11 0.14 AUDITED 1.00 0.23 0.20 0.04 0.02 0.03 BIG4 1.00 0.80 0.04 0.01 0.06 SPECIALIST 1.00 0.04 0.02 0.04 CAP 1.00 0.13 0.01 NPL 1.00 0.04 LIQUIDITY 1.00 ALLOWANCE RER14 CI

13

14

15

16

17

18

19

20

21

0.08 0.01 0.17 0.03 0.17 0.07 0.02 0.07 0.03 0.22 0.03 0.03 1.00

0.04 0.01 0.05 0.01 0.06 0.02 0.04 0.003 0.01 0.03 0.04 0.24 0.03 1.00

0.02 0.01 0.12 0.02 0.13 0.07 0.06 0.02 0.01 0.03 0.04 0.22 0.08 0.02 1.00

0.18 0.10 0.04 0.15 0.08 0.12 0.03 0.01 0.04 0.04 0.21 0.04 0.06 0.09 0.16

0.02 0.02 0.02 0.03 0.05 0.006 0.07 0.05 0.04 0.03 0.01 0.03 0.06 0.04 0.01

0.06 0.001 0.04 0.09 0.03 0.06 0.06 0.01 0.05 0.38 0.10 0.02 0.12 0.08 0.03

0.13 0.11 0.03 0.13 0.04 0.06 0.05 0.01 0.07 0.04 0.03 0.03 0.01 0.08 0.12

0.09 0.06 0.05 0.02 0.04 0.01 0.05 0.08 0.11 0.09 0.02 0.09 0.09 0.01 0.06

0.13 0.13 0.09 0.15 0.13 0.08 0.01 0.04 0.02 0.13 0.04 0.06 0.02 0.13 0.01

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

Growth-matched control banks mean

Panel A: Descriptive statistics for independent variables in 2006 (Measured during the Pre-Crisis Period) AUDITED 0.515

(continued on next page)

4885

Panel A of Table presents the mean differences for the independent variables between the affected banks and growth-matched control banks using the year 2006 annual data. Panel B of Table presents Pearson correlation matrix for the variables. Bold numbers are significant at less than the 5% level. The combined sample and control banks has 388 private bank annual observations. The affected sample contains 194 private bank annual observations. The growth-matched control sample contains 194 private bank annual observations. We collect the year 2006 accounting data from the Commercial Bank Data call reports and auditor data from the Bank Holding Company database. We winsorize the top and bottom 1% of each continuous independent variable. * Indicate significance at the 10% level based on a two-tailed test. ** Indicate significance at the 5% level based on a two-tailed test. *** Indicate significance at the 1% level based on a two-tailed test. a Variable Definitions: FB equals 1 if the bank failed during 2007–2010, and 0 otherwise. LARGE_LOSS equals 1 if ROA is less than 5% in any one of the four years during 2007–2010, and 0 otherwise. LARGE_LLP equals 1 if the ratio of loan loss provision to total loans is greater than 4% in any one of the four years during 2007–2010, and 0 otherwise. SMALL_CAP equals 1 if Tier 1 capital ratio is less than 8% in any one of the four years during 2007–2010, and 0 otherwise. TB equals 1 if any one of LARGE_LOSS, LARGE_LLP and SMALL_CAP equals 1 in any one of the four years during 2007–2010, and 0 otherwise. AFFECTED equals 1 if the bank observation belongs to the test sample, and 0 otherwise. AUDITED equals 1 if the bank is audited by an independent auditing firm in 2006, and 0 otherwise. BIG4 equals 1 if the auditor is a Big4 auditor in 2006, and 0 otherwise. SPECIALIST equals 1 if the auditor is PricewaterhouseCoopers or KPMG in 2006, and 0 otherwise. CAP is the ratio of Tier 1 capital to total assets times 100. NPL is the ratio of non-performing loans to total loans times 100. LIQUIDITY is the ratio of interest-bearing balances to total assets. ALLOWANCE is the ratio of allowance for loan and leases losses to total assets. RER14 is annual percentage growth of real estate residential single-family 1–4 mortgages. CI is annual percentage growth of commercial and industrial loans. G_ASSETS is annual percentage growth in total assets. SIZE_R is the residual of the natural logarithm of total assets in millions of dollars. We run an OLS regression of each firm’s natural logarithm of total assets on ten variables (AFFECTED, BIG4, CAP, NPL, LIQUIDITY, ALLOWANCE, RER14, CI, G_ASSETS, and LEV). The orthogonal residual component of the natural logarithm of total assets with respect to the ten variables is the residual from this regression. LEV is the ratio of bank’s equity to its total assets. REGION2 equals 1 if a bank is in the Midwest region, and 0 otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western regions respectively.

0.12 0.11 0.09 0.24 0.34 1.00 0.08 0.07 0.11 0.51 1.00

20 19

0.15 0.15 0.04 1.00 0.02 0.03 1.00

18 17

0.05 1.00 G_ASSETS SIZE_R LEV REGION2 REGION3 REGION4 16 17 18 19 20 21

Table 2 (continued)

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

1.00

21

4886

between $100 million and $500 million on January 1, 2005, and grew its assets to between $500 million and $1 billion by January 1, 2007, and equals 0 otherwise. Eq. (1) includes the following control variables: CAP, the ratio of Tier 1 capital to total assets; NPL, the percentage of non-performing loans over total loans; LIQUIDITY, the ratio of interest-bearing balances to total assets; ALLOWANCE, the ratio of allowance for loan and leases losses to total assets; PER14, the annual percentage growth of real estate residential single-family 1–4 mortgages; CI, the annual percentage growth in commercial and industrial loans; G_ASSETS, the annual percentage growth rate of total assets; SIZE_R, the orthogonal residual component of the natural logarithm of total assets in millions of dollars14; and LEV, the ratio of bank’s equity to total assets. Following Ng and Roychowdhury (2011), we include three geographic indicator variables to mitigate concerns that the empirical results are driven by heterogeneous regional characteristics. REGION2 is an indicator variable equaling 1 if a bank is in the Midwest region and 0 otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western regions, respectively. By construction, the Northeast serves as the benchmark region. We include the Tier 1 capital ratio (CAP) as a proxy for balance sheet strength. A higher capital ratio provides a bigger cushion for banks to write off bad loans in the future (Berger et al., 1995; Kim and Kross, 1998). Additionally, banks with higher capital likely suffer less from debt overhang problems (Myers, 1977) and have more flexibility to respond to adverse shocks (Beltratti and Stulz, 2010). Therefore, we expect CAP to be negatively related to bank failure and financial trouble. Using data from the last major banking crisis in 1985–1992, Cole and Gunther (1995, 1998) document that banks with more liquidity are less likely to fail. Similarly, Cole and White (2012) find a negative relationship between bank liquidity and bank failure during 2009. Based on these findings, we expect LIQUIDITY to be negatively related to bank failure and financial trouble. We do not offer directional predictions for nonperforming loans (NPL) and loan loss allowance (ALLOWANCE). The effect of growth in individual loan categories and total loans on future bank failure is dependent on the quality of incremental loans. However, large growth in risky loans likely increases the probability of a bank’s failure. In particular, incremental loans associated with inflated assets during asset bubble periods may pose higher risk than growth in other categories of loans (Bhattacharyya and Purnanandam, 2010). Consistent with this argument, we are more likely to observe a positive association between growth in residential mortgages (i.e., 1–4 family residential mortgages) and commercial and industrial loans and future bank failure and financial trouble. We also control for differences in size, growth in total assets, and financial leverage in the regressions. As an additional test, we also examine the probability of bank financial trouble for affected banks. We do so by estimating a model similar to Eq. (1), except that we replace the dependent variable FB with the dependent variable TB, where TB is an indicator variable that equals 1 if the bank is in financial trouble during 2007–2010 and equals 0 otherwise. We use four proxies to represent troubled banks: LARGE_LOSS, LARGE_LLP, SMALL_CAP, and TB. LARGE_LOSS equals 1 if ROA is less than 5% in any year during 2007–2010 and equals 0 otherwise; LARGE_LLP equals 1 if LLP/TOTAL LOANS is larger than 4% in any year during 2007–2010 and equals 0 otherwise; SMALL_CAP equals 1 if the Tier 1 capital ratio is less than 8% in any 14 Since total assets is significantly correlated with AFFECTED and BIG4 (Pearson correlation = 0.86 and 0.79, respectively), we follow Hou and Moskowitz (2005) and Huang et al. (2011) and use the residual component of total assets, SIZE_R, as the independent variable in equation (1). This approach allows us to effectively reduce the multi-collinearity between total assets and the AFFECTED and BIG4 variables in equation (1). We run an OLS regression of each firm’s natural logarithm of total assets on 10 variables (AFFECTED, BIG4, CAP, NPL, LIQUIDITY, ALLOWANCE, RER14, CI, G_ASSETS, and LEV). The orthogonal residual component of the natural logarithm of total assets with respect to these 10 variables is the residual from this regression.

4887

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892 Table 3 Difference in probability of bank failure during the crisis period between affected and control banks. Variablea

Predicted sign

Using growth-matched control banks Dependent variable = FB coefficient (Wald Chi-Square) (1)

Using FDICIA control banks Dependent variable = FB coefficient (Wald Chi-Square) (2)

Intercept AFFECTED

? +

CAP



NPL

?

LIQUIDITY



ALLOWANCE

?

RER14

+

CI

+

G_ASSETS

+

SIZE_R



LEV



REGION2

+

REGION3

+

REGION4

+

1.949 (1.51) 0.364** (4.86) [3.56%] 0.484*** (8.92) 0.653* (2.85) 4.314 (0.53) 9.945** (4.46) 1.142** (3.96) 0.153 (0.28) 1.793*** (6.92) 0.682 (2.52) 3.137 (0.19) 0.128 (0.08) 0.839*** (7.26) 1.031*** (8.00) 117.9 0.307 84.4 388

0.162 (0.01) 1.078*** (7.56) [10.55%] 0.461** (6.61) 1.234*** (8.12) 3.592 (0.11) 43.067** (5.32) 0.828 (1.73) 0.183 (1.01) 1.643** (5.01) 0.600 (1.82) 2.287 (0.06) 0.233 (0.22) 0.468 (1.91) 1.120*** (8.04) 93.3 0.376 86.6 259

Log-Likelihood Pseudo-R2 Percent Concordant # of Observations

Table reports estimation results for the logistic regression. Sample 1 includes 194 affected banks and 194 growth-matched control banks. Sample 2 includes 194 affected banks and 65 FDICIA control banks. The independent variables are measured in 2006. We winsorize the top and bottom 1% of each continuous independent variable. For the AFFECTED variable, we also report the marginal effect (in percent) in the square brackets. The marginal effect indicates the change in the probability of bank failure for affected banks compared to control banks (holding other variables constant). The marginal effect of failure on affected banks is computed as p  (1  p)  b  1, where p is the base rate (0.11) and b is the estimated coefficient from the logistic regression (Liao, 1994). * Indicate significance at the 10% level based on a two-tailed test. ** Indicate significance at the 5% level based on a two-tailed test. *** Indicate significance at the 1% level based on a two-tailed test. a Variable definitions: FB equals 1 if the bank failed during 2007–2010, and 0 otherwise. AFFECTED equals 1 if the bank observation belongs to the test sample, and 0 otherwise. CAP is the ratio of Tier 1 capital to total assets times 100. NPL is the ratio of non-performing loans to total loans times 100. LIQUIDITY is the ratio of interest-bearing balances to total assets. ALLOWANCE is the ratio of allowance for loan and leases losses to total assets. RER14 is the annual percentage growth of real estate residential singlefamily 1–4 mortgages. CI is annual percentage growth of commercial and industrial loans. G_ASSETS is annual percentage growth in total assets. SIZE_R is the residual of the natural logarithm of total assets in millions of dollars. We run an OLS regression of each firm’s natural logarithm of total assets on 10 variables (AFFECTED, BIG4, CAP, NPL, LIQUIDITY, ALLOWANCE, RER14, CI, G_ASSETS, and LEV). The orthogonal residual component of the natural logarithm of total assets with respect to the 10 variables is the residual from this regression. LEV is the ratio of bank’s equity to its total assets. REGION2 equals 1 if a bank is in the Midwest region, and 0 otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western regions respectively.

year during 2007–2010 and 0 otherwise.15 TB equals 1 if any one (or more) of LARGE_LOSS, LARGE_LLP, and SMALL_CAP equals 1 in any of the four years during 2007–2010 and 0 otherwise.

4. Results 4.1. Probability of bank failure during the crisis period Table 3 reports the estimation results of Eq. (1). We estimate the regression using each of the two control groups. Of interest is the coefficient on AFFECTED, which indicates the difference in probability of failure between banks in the affected and control 15 During the pre-crisis period 2005-2006, the mean ROA of the combined affected and control banks is 3.3% and the standard deviation of ROA is 2.5%, the mean LLP/ TOTAL LOANS is 0.5% and the standard deviation of LLP/TOTAL LOANS is 0.6%, and the mean Tier 1 capital ratio is 12.7% and the standard deviation of Tier 1 capital ratio is 3.9%. We choose our cutoff points for the crisis period such that they are at least one standard deviation above/below the mean value of the pre-crisis period distribution. We also test the sensitivity of our results to alternate cutoff points.

samples. For this main variable of interest, we report the regression coefficient, followed by the Wald statistic in parentheses and the marginal effect (in percentage) in square brackets. The marginal effect indicates the change in the probability of bank failure for affected banks as compared to control banks (holding other variables constant).16 The results show that the coefficient on AFFECTED is positive and significant at the 5% and 1% level (Wald chi-square = 4.86 and 7.56, respectively) in the two regressions using the growth-matched control banks and FDICIA control banks, indicating that the affected banks are significantly more likely than the control banks to fail. The marginal effect on bank failure for affected banks indicates that the economic significance of being affected is nontrivial. For example, Columns (1) and (2) show that the difference in propensity to fail between affected banks and growthmatched control banks is 3.56% and the corresponding difference 16 We thank an anonymous reviewer for pointing us in this direction. The marginal effect of failure on affected banks is computed as p  (1  p)  b  1, where p is the base rate (0.11) and b is the estimated coefficient from the logistic regression (Liao, 1994).

4888

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

Table 4 Difference in probability of bank troubles during the crisis period between affected and control banks. Variablea

Predicted sign

Dependent variable = TB Coefficient (Wald Chi-Square) (1)

Panel A: Affected banks and growth-matched control banks Intercept ? 2.012 (2.23) AFFECTED + 0.290** (4.94) [2.84%] CAP  0.300** (4.60) NPL ? 0.509** (3.94) LIQUIDITY  5.277* (3.33) ALLOWANCE ? 15.094** (6.27) RER14 + 0.340* (2.83) CI + 0.368** (4.50) G_ASSETS + 0.303 (0.81) SIZE_R  0.684 (1.51) LEV  11.095* (3.03) REGION2 + 0.544** (4.96) REGION3 + 0.538** (5.62) REGION4 + 0.847*** (11.23) Log-Likelihood 258.8 Pseudo-R2 0.220 Percent 72.3 Concordant # of Observations 388 Panel B: Affected banks and FDICIA control banks Intercept ? 1.653 (0.88) AFFECTED + 0.823*** (9.67) [8.06%] CAP  0.287 (2.14) NPL ? 1.110** (6.61) LIQUIDITY  5.575 (1.25) ALLOWANCE ? 8.590 (0.66) RER14 + 0.167 (0.15) CI + 0.231 (2.64) G_ASSETS + 0.182 (0.15) SIZE_R  0.095 (1.25) LEV  9.009 (1.65) REGION2 + 0.423 (1.86) REGION3 + 0.513* (3.30) REGION4 + 1.221*** (13.43) Log-Likelihood 176.9

Dependent variable = LARGE_LOSS Coefficient (Wald Chi-Square) (2)

Dependent variable = LARGE_LLP Coefficient (Wald Chi-Square) (3)

Dependent variable = SMALL_CAP Coefficient (Wald Chi-Square) (4)

3.661*** (8.07) 0.538** (4.63) [5.27%] 0.119 (0.83) 0.048 (0.01) 9.428 (0.42) 3.846 (1.30) 0.067 (0.04) 0.114 (0.62) 0.703 (1.85) 0.283 (1.95) 3.820 (0.62) 0.106 (0.06) 0.657* (3.76) 0.947** (5.67) 100.8 0.158 76.6

3.494* (3.55) 0.545*** (7.31) [5.34%] 0.338* (3.00) 0.488* (3.48) 4.036 (1.82) 22.012*** (9.31) 0.196 (2.07) 0.400** (3.85) 0.143 (0.18) 0.231 (0.59) 19.794*** (8.45) 0.705*** (7.53) 0.619** (6.05) 0.821*** (8.89) 249.0 0.193 71.2

7.194** (4.15) 0.002 (0.01) [0.02%] 0.303 (0.72) 1.152 (1.93) 29.430* (3.77) 2.855 (0.22) 0.087 (0.03) 0.146 (0.66) 1.826** (5.86) 1.341 (1.94) 60.634*** (9.48) 2.037** (4.12) 0.037 (0.01) 0.542 (1.86) 100.8 0.437 91.4

388

388

388

10.116*** (14.87) 5.038*** (82.11) [49.53%] 0.371 (1.78) 0.568 (0.51) 59.842*** (8.98) 25.429 (2.26) 0.409 (0.53) 0.171 (1.11) 0.254 (0.09) 0.078 (2.47) 5.510 (0.62) 0.005 (0.01) 0.502 (1.48) 1.341*** (6.66) 73.2

5.250*** (6.92) 0.348** (5.66) [3.41%] 0.573** (5.78) 0.393 (1.16) 7.736 (1.78) 23.541** (4.56) 0.145 (0.15) 0.172 (1.65) 0.053 (0.01) 0.052 (1.93) 19.061** (4.90) 0.711** (4.86) 0.758** (5.88) 1.272*** (13.88) 169.6

5.633** (6.51) 0.534 (1.24) [5.23%] 0.401* (2.94) 0.849*** (7.15) 23.065 (1.96) 8.520 (0.23) 0.107 (0.03) 0.010 (0.04) 1.880** (6.34) 1.172 (1.17) 41.529*** (6.97) 0.464 (0.48) 0.151 (0.17) 0.894** (5.56) 84.8

4889

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892 Table 4 (continued) Variablea

Pseudo-R2 Percent Concordant # of Observations

Predicted sign

Dependent variable = TB Coefficient (Wald Chi-Square) (1)

Dependent variable = LARGE_LOSS Coefficient (Wald Chi-Square) (2)

Dependent variable = LARGE_LLP Coefficient (Wald Chi-Square) (3)

Dependent variable = SMALL_CAP Coefficient (Wald Chi-Square) (4)

0.264 75.6

0.348 87.0

0.245 75.0

0.402 88.1

259

259

259

259

Table reports estimation results for the logistic regression. Panel A includes 194 affected banks and 194 growth-matched control banks. Panel B includes 194 affected banks and 65 FDICIA control banks. We winsorize the top and bottom 1% of each continuous independent variable. For the AFFECTED variable, we also report the marginal effect (in percent) in the square brackets. The marginal effect indicates the change in the probability of bank trouble for affected banks compared to control banks (holding other variables constant). The marginal effect of bank trouble on affected banks is computed as p  (1  p)  b  1, where p is the base rate (0.11) and b is the estimated coefficient from the logistic regression (Liao, 1994). * Indicate significance at the 10% level based on a two-tailed test. ** Indicate significance at the 5% level based on a two-tailed test. *** Indicate significance at the 1% level based on a two-tailed test. a Variable definitions: TB equals 1 if any one of LARGE_LOSS, LARGE_LLP and SMALL_CAP equals 1 in any one of the four years during 2007–2010, and 0 otherwise. LARGE_LOSS equals 1 if ROA is less than 5% in any one of the four years during 2007–2010, and 0 otherwise. LARGE_LLP equals 1 if the ratio of loan loss provision to total loans is greater than 4% in any one of the four years during 2007–2010, and 0 otherwise. SMALL_CAP equals 1 if Tier 1 capital ratio is less than 8% in any one of the four years during 2007–2010, and 0 otherwise. AFFECTED equals 1 if the bank observation belongs to the test sample, and 0 otherwise. CAP is the ratio of Tier 1 capital to total assets times 100. NPL is the ratio of non-performing loans to total loans times 100. LIQUIDITY is the ratio of interest-bearing balances to total assets. ALLOWANCE is the ratio of allowance for loan and leases losses to total assets. RER14 is the percentage growth of real estate residential single-family 1–4 mortgages. CI is the percentage growth of commercial and industrial loans. G_ASSETS is the percentage growth in total assets. SIZE_R is the orthogonal residual of the natural logarithm of total assets in millions of dollars with respect to ten variables (AFFECTED, BIG4, CAP, NPL, LIQUIDITY, ALLOWANCE, RER14, CI, G_ASSETS, and LEV). LEV is the ratio of bank’s equity to its total assets. REGION2 equals 1 if a bank is in the Midwest region, and 0 otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western regions respectively.

between affected banks and FDICIA control banks is 10.55%. Overall, the evidence indicates that the affected banks have a higher probability of failing during the crisis than do the control banks. With regard to bank-level controls, the relationship between bank failure and the control variables is generally consistent with expectations. In the regression results shown in Table 3, Tier 1 capital ratios (CAP) are negatively related to bank failures, and their coefficients are statistically significant at least at the 5% level, indicating that a strong Tier 1 capital ratio prevents banks from failing in subsequent years, and non-performing loans (NPL) are positively related to bank failures, indicating that troubled loans in 2006 are a good predictor of bank failure in subsequent years. As an additional analysis, we also examine the probability of affected banks being in financial trouble during the crisis period. To do so, we re-estimate Eq. (1) for troubled blanks. Panel A of Table 4 presents the estimation results using the growth-matched control banks, and Panel B presents the results using the FDICIA control banks. The results in Panel A indicate that the coefficient on AFFECTED is positive and significant at the 5% level when TB and LARGE_LOSS are the dependent variables and at the 1% level when LARGE_LLP is the dependent variable. The results in Panel B indicate that the coefficient on AFFECTED is positive and significant at the 1% level when TB and LARGE_LOSS are the dependent variables and at the 5% level when LARGE_LLP is the dependent variable. Once again, these results imply that, regardless of the control sample used, the affected sample banks are more likely to be financially troubled than the control sample banks during the crisis period. These results confirm our prediction that banks that grew without proper internal control monitoring are more likely to experience financial trouble during the crisis period. The marginal effect of affected banks getting into financial trouble is economically significant.17 For example, Column (1) of Panel A shows that switching from being a control bank to being an affected bank increases a bank’s propensity to have financial trouble by 2.84% when we use the growth-matched control sample. Column (1) of Panel B shows that switching from being a control bank to being

17 We thank an anonymous reviewer for pointing us in this direction. The marginal effect of financial trouble on affected banks is computed as p  (1  p)  b  1, where p is the base rate (0.11) and b is the estimated coefficient from the logistic regression (Liao, 1994).

an affected bank increases a bank’s propensity to have financial trouble by 8.06% when we use the FDICIA control sample. Overall, the evidence demonstrates that the affected banks have a higher probability to have financial trouble during the crisis than do the control banks. 4.2. Effects of auditing and audit quality on bank failure We test the prediction that the positive relationship between FB and AFFECTED reported in Table 3 is weaker when the bank uses an independent external auditor, a Big 4 auditor, or an industry specialist auditor. GAO (2003) identified PricewaterhouseCoopers and KPMG as the top two market leaders in the banking industry after 2002; we therefore classify these two audit firms as industry specialist auditors. We estimate the following equation to investigate the impact of auditing, auditor type, and auditor specialization on the difference in probability of failure across the affected and growth-matched control banks18:

FB ¼ c0 þ c1 AFFECTED þ c2 AUDITVar þ c3 AFFECTED  AUDITVar þ c4 CAP þ c5 NPL þ c6 LIQUIDITY þ c7 ALLOWANCE þ c8 RER14 þ c9 CI þ c10 G ASSETS þ c11 SIZER þ c12 LEV þ c13 REGION2 þ c14 REGION3 þ c15 REGION4 þ e

ð2Þ

where AUDITVar is defined as AUDITED, which equals 1 if the bank is audited by an independent auditing firm in 2006 and 0 otherwise; as BIG4, which equals 1 if the bank is audited by a Big 4 auditor in 2006 and 0 otherwise; or as SPECIALIST, which equals 1 if the bank is audited by PricewaterhouseCoopers or KPMG in 2006 and 0 otherwise. Table 5 reports the estimation results of Eq. (2). For the main variables of interest (i.e., BIG4 and SPECIALIST), we report the regression coefficient, followed by the Wald statistic in parentheses. Of interest is the coefficient on AFFECTED  AUDITVar, which reflects the incremental effect of auditors on the relationship 18 We also check the robustness of our results using the FDICIA control sample. We do not tabulate these to save space. The untabulated results are qualitatively similar to those reported in Table 5, using our growth-matched control sample.

4890

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

Table 5 Impact of auditors on difference in probability of bank failure during the crisis period between affected and growth-matched control banks. Variable

a

Predicted sign

Dependent variable = FB Coefficient (Wald Chi-Square) (1)

Dependent variable = FB Coefficient (Wald Chi-Square) (2)

Dependent variable = FB Coefficient (Wald Chi-Square) (3)

Intercept

?

AFFECTED

+

2.038*** (12.97) 0.627** (6.16)

2.082*** (13.93) 0.619** (5.93)

AUDITED



2.075*** (12.99) 0.644** (4.34) 0.208 (0.23)

BIG4



SPECIALIST



AFFECTED  AUDITED



AFFECTED  BIG4



AFFECTED  SPECIALIST



CAP



NPL

?

LIQUIDITY



ALLOWANCE

?

RER14

+

CI

+

G_ASSETS

+

SIZE_R



LEV



REGION2

+

REGION3

+

REGION4

+

Log-Likelihood Pseudo-R2 Percent Concordant # of Observations

3.235*** (95.05) 3.272*** (71.38) 0.025 (0.02) 1.209*** (6.90)

0.686 (0.92) 0.617 (2.61) 5.518 (0.79) 12.314** (6.29) 1.177** (4.13) 0.145 (0.23) 1.917*** (8.04) 0.831 (1.39) 2.278 (0.11) 0.059 (0.02) 0.876*** (9.04) 1.082*** (8.98) 117.9 0.315 84.9 388

0.643 (0.76) 0.632 (2.63) 5.701 (0.86) 12.393** (6.56) 1.199** (4.06) 0.112 (0.19) 1.826*** (7.41) 0.836 (1.52) 3.178 (0.19) 0.060 (0.02) 0.827*** (7.45) 1.088*** (8.81) 117.9 0.322 85.1 388

1.147** (4.56) 0.638 (0.78) 0.639* (2.72) 5.348 (0.77) 12.327*** (6.68) 1.199** (4.07) 0.130 (0.22) 1.869*** (7.71) 0.841 (1.80) 2.909 (0.16) 0.046 (0.01) 0.827*** (7.43) 1.073*** (8.70) 117.9 0.318 85.0 388

Table reports estimation results for the logistic regression. The dependent variable is FB. The sample includes 194 affected banks and 194 growth-matched control banks. The independent variables are measured in 2006. We collect the 2006 accounting data from the Commercial Bank Data call reports and auditor data from the Bank Holding Company database (available at the Federal Reserve Bank of Chicago website http://www.chicagofed.org/). We winsorize the top and bottom 1% of each continuous independent variable. * Indicate significance at the 10% level based on a two-tailed test. ** Indicate significance at the 5% level based on a two-tailed test. *** Indicate significance at the 1% level based on a two-tailed test. a Variable definitions: FB equals 1 if the bank failed during 2007–2010, and 0 otherwise. AFFECTED equals 1 if the bank observation belongs to the test sample, and 0 otherwise. AUDITED equals 1 if the bank is audited by an independent auditing firm in 2006, and 0 otherwise. BIG4 equals 1 if the auditor is a Big4 auditor in 2006, and 0 otherwise. SPECIALIST equals 1 if the auditor is PricewaterhouseCoopers or KPMG in 2006, and 0 otherwise. CAP is the ratio of Tier 1 capital to total assets times 100. NPL is the ratio of non-performing loans to total loans times 100. LIQUIDITY is the ratio of interest-bearing balances to total assets. ALLOWANCE is the ratio of allowance for loan and leases losses to total assets. RER14 is the annual percentage growth of real estate residential single-family 1–4 mortgages. CI is annual percentage growth of commercial and industrial loans. G_ASSETS is annual percentage growth in total assets. SIZE_R is the orthogonal residual of the natural logarithm of total assets in millions of dollars with respect to 10 variables (AFFECTED, BIG4, CAP, NPL, LIQUIDITY, ALLOWANCE, RER14, CI, G_ASSETS, and LEV). LEV is the ratio of bank’s equity to its total assets. REGION2 equals 1 if a bank is in the Midwest region, and 0 otherwise; REGION3 and REGION4 are defined analogously for the Southern and Western regions respectively.

between FB and AFFECTED. Columns (1) through (3) present the results for the effects of three variables (AUDITED, BIG4, and SPECIALIST) on the propensity for bank failure, after controlling for bank characteristics and regional fixed effects. A significant and negative coefficient on AFFECTED  BIG4 indicates that the affected banks are less likely to fail when they are audited by a Big 4 auditor. A significant and negative coefficient on AFFECTED  SPECIALIST indicates that the affected banks are less likely to fail when they are audited by an industry specialist auditor. Columns (1) through (3) of Table 5 show that the coefficients on AFFECTED are positive and significant at the 5% level, suggesting

that the affected sample banks have a higher likelihood of failure than the control sample banks. Column (1) shows that the coefficient on AFFECTED  AUDITED is not significantly different from zero, indicating that the probability of failure for affected banks does not differ across audited and unaudited banks. Column (2) shows that the coefficient on AFFECTED  BIG4 is negative and significant at the 1% level (Wald chi-square = 6.90). This is consistent with our prediction that the probability of bank failure for affected banks is significantly reduced when the affected banks hire Big 4 auditors. The results in Column (3) show that the coefficient on AFFECTED  SPECIALIST is negative and significant at the 5% level

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

(Wald chi-square = 4.56), indicating that the industry specialist auditors significantly reduce the probability of failure for the affected sample banks. Interestingly, the coefficients on both the BIG4 and SPECIALIST variables are negative and significant at the 1% level, suggesting that banks audited by Big 4 auditors and/or industry specialist auditors have a lower likelihood of failure. Table 5 examines the mitigating effects when affected firms are audited, and particularly when audited by the Big 4 or the two specialist firms. The interaction terms of AFFECTED  BIG4 and AFFECTED  SPECIALIST are negative and significant. The difference in probability of bank failure between affected banks and control banks is significantly reduced when affected banks hire Big 4 auditors.19 This result implies that Big 4 auditors have a significant impact on the financial reporting quality and financial health of affected banks. After the 2005 increase in the asset threshold for FDICIA internal control requirements, the affected banks are no longer subject to the internal control requirements but would have been if the regulation had not been changed. With weak internal control reporting, the affected banks are more likely to benefit from better audit procedures of Big 4 auditors. Therefore, auditor reputation (i.e., whether the bank is audited by a Big 4 auditor or an industry specialist auditor) has a moderating effect on the likelihood of bank failure for the affected banks.

5. Conclusions We document the unintended consequences of the 2005 increased asset threshold for FDICIA internal controls for a sample of U.S. private banks. In particular, during the crisis period, we document that the affected banks had a higher likelihood of failure. They also had a higher likelihood of financial trouble during the crisis period, as reflected in large losses (poor performance) and large loan loss provisions (low asset quality). Interestingly, auditor reputation (i.e., whether the bank had a Big 4 auditor or whether the bank was audited by a bank audit expert – KPMG and PricewaterhouseCoopers) had a moderating effect on the likelihood of bank failure for the sample of affected banks. The FDIC’s decision to increase the asset threshold for internal control reporting requirements and to exempt these banks from audit committee independent director requirements was partly motivated by the exemption for small firms from SOX Section 404 internal control reporting requirements and partly to grant relief from the higher burden of compliance on smaller banks. Our results provide evidence of the unintended costs arising from these exemptions. Our evidence also suggests the usefulness of effective internal controls for a group of small private banks that were exempted from internal control requirements after 2005. This is consistent with evidence provided by the GAO (1991) on failed banks from the earlier savings and loan crisis, which was partly caused by severe internal control weaknesses. Our study can be seen as an attempt to extend the growing literature on the usefulness of effective internal controls for the banking industry.

References AICPA, 2006. Large Firm PCAOB Inspection Deficiency Analysis. . Altamuro, J., Beatty, A., 2010. How does internal control regulation affect financial reporting? Journal of Accounting and Economics 49, 58–74. 19 We add together the coefficients of AFFECTED and AFFECTED  BIG4 and check whether the sum of the coefficients of AFFECTED and AFFECTED  BIG4 is different from zero. The Wald Chi-Square for the sum of the coefficients of AFFECTED and AFFECTED  BIG4 is 1.91, indicating that the sum of the two coefficients is not significantly different from zero.

4891

Ashbaugh-Skaife, H., Collins, D.W., Kinney, W.R., LaFond, R., 2009. The effect of SOX internal control deficiencies on firm risk and cost of equity. Journal of Accounting Research 47, 1–34. Autore, D.M., Billingsley, R.S., Schneller, M.I., 2009. Information uncertainty and auditor reputation. Journal of Banking and Finance 33, 183–192. Becker, C., DeFond, M., Jiambalvo, J., Subramanyam, K., 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15, 1–24. Bedard, J.C., Biggs, S.F., 1991. The effect of domain-specific experience on evaluation of management representations in analytical procedures. Auditing: A Journal of Practice & Theory 10, 77–90. Behn, B.K., Choi, J.H., Kang, T., 2008. Audit quality and properties of analyst earnings forecasts. The Accounting Review 83, 327–349. Beltratti, A., Stulz, R.M., 2010. The Credit Crisis Around the Globe: Why Did Some Banks Perform Better? Working Paper. Bocconi University and Ohio State University. Beneish, M.D., Billings, M.B., Hodder, L.D., 2008. Internal control weakness and information uncertainty. The Accounting Review 83 (3), 665–703. Benston, G.J., Kaufman, G.G., 1997. FDICIA after five years. Journal of Economic Perspectives 11, 139–158. Berger, A.N., Herring, R.J., Szego, G.P., 1995. The role of capital in financial institutions. Journal of Banking and Finance 19, 393–430. Bhattacharyya, S., Purnanandam, A., 2010. Risk-Taking by Banks: What Did We Know and When Did We Know It? Working Paper. University of Michigan. Bronson, S.N., Carcello, J.V., Hollingsworth, C.W., Neal, T., 2009. Are fully independent audit committees really necessary? Journal of Accounting and Public Policy 28, 265–280. Carcello, J., Nagy, A., 2004. Client size, auditor specialization and fraudulent financial reporting. Managerial Auditing Journal 19, 651–668. Cole, R.A., Gunther, J.W., 1995. Separating the likelihood and timing of bank failure. Journal of Banking and Finance 19, 1073–1089. Cole, R.A., Gunther, J.W., 1998. Predicting bank failure: a comparison of on- and offsite monitoring systems. Journal of Financial Services Research 13, 103–117. Cole, R.A., White, L.J., 2012. Déjà vu all over again: the causes of U.S. commercial bank failures this time around. Journal of Financial Services Research 42, 5–29. Craswell, A.T., Francis, J.R., Taylor, S.L., 1995. Auditor brand name reputations and industry specializations. Journal of Accounting and Economics 20, 297–322. Dopuch, N., Simunic, D., 1982. Competition in auditing: an assessment. In: Paper Presented at Symposium on Auditing Research IV. University of Illinois at Urbana-Champaign. Erkens, D., Hung, M., Matos, P., 2012. Corporate governance in the 2007–2008 financial crisis: evidence from financial institutions worldwide. Journal of Corporate Finance 18 (2), 389–411. FDIC. 2005a. Proposed Rules. . FDIC. 2005b. Financial Institution Letter. . FDIC. 2008. 12 CFR Part 363: Independent Audits and Reporting Requirements. FDIC. 2009. Financial Institution Letters. FIL-33-2009. GAO. 1991. Failed Banks: Accounting and Auditing Reforms Urgently Needed. Report to the Congressional Committees. GAO. 2003. Public Accounting Firms, Mandated Study on Consolidation and Competition. Report to the Senate Committee on Banking, Housing, and Urban Affairs and the House Committee on Financial Services. Gao, F., Wu, J.S., Zimmerman, J., 2009. Unintended consequences of granting small firms exemptions from securities regulation: evidence from the Sarbanes–Oxley act. Journal of Accounting Research 47, 459–506. Hammersley, J.S., Myers, L.A., Shakespeare, C., 2008. Market reactions to the disclosure of internal control weaknesses and to the characteristics of those weaknesses under section 302 of the Sarbanes–Oxley act of 2002. Review of Accounting Studies 13, 141–165. Hou, K., Moskowitz, T., 2005. Market frictions, price delay, and the cross-section of expected returns. Review of Financial Studies 18, 981–1020. Huang, L.Y., Lai, G.C., McNamara, M., Wang, J., 2011. Corporate governance and efficiency: evidence from U.S. property-liability insurance industry. Journal of Risk and Insurance 78, 519–550. Jin, J.Y., Kanagaretnam, K., Lobo, G.J., 2011. Ability of accounting and audit quality variables to predict bank failures during the recent financial crisis. Journal of Banking and Finance 35, 2811–2819. Jin, J.Y., Kanagaretnam, K., Lobo, G.J., Mathieu, R., 2013. Impact of FDICIA internal controls on bank risk taking. Journal of Banking and Finance 37 (2), 614–624. Kanagaretnam, K., Krishnan, G., Lobo, G.J., 2009. Is the market valuation of banks’ loan loss provision conditional on auditor reputation? Journal of Banking and Finance 33, 1039–1047. Kanagaretnam, K., Krishnan, G., Lobo, G.J., 2010a. An empirical analysis of auditor independence in the banking industry. The Accounting Review 85, 2011–2046. Kanagaretnam, K., Lim, C.Y., Lobo, G.J., 2010b. Auditor reputation and earnings management: international evidence from the banking industry. Journal of Banking and Finance 34, 2318–2327. Khurana, I.K., Raman, K.K., 2006. Do investors care about the auditor’s economic dependence on the client? Contemporary Accounting Research 23, 977–1016. Kim, M., Kross, W., 1998. The impact of the 1989 change in bank capital standards on loan loss provisions and loan write-offs. Journal of Accounting and Economics 25, 69–99. Krishnan, G., 2003. Does Big 6 auditor industry expertise constrain earnings management? Accounting Horizons 17, 1–16.

4892

J.Y. Jin et al. / Journal of Banking & Finance 37 (2013) 4879–4892

Lel, U., Miller, D., 2008. International cross-listing, firm performance and top management turnover: a test of the bonding hypothesis. Journal of Finance 63, 1897–1937. Lennox, C., Pittman, J.A., 2010. Big five audits and accounting fraud. Contemporary Accounting Research 27, 209–247. Liao, T. 1994. Interpreting Probability Models: Logit, Probit, and other Generalized Linear Models. Sage University Paper Series on Quantitative Applications in the Social Sciences, Series No. 07-101. Lopez, T., Vandervelde, S.D., Wu, Y.J., 2009. Investor perceptions of an auditor’s adverse internal control opinion. Journal of Accounting and Public Policy 28, 231–250. Myers, S.C., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147–175. Ng, J., Roychowdhury, S., 2011. Loan Loss Reserves, Regulatory Capital, and Bank Failures: Evidence from the Recent Economic Crisis. Working Paper. Singapore Management University and Massachusetts Institute of Technology.

Ogneva, M., Subramanyam, K.R., Raghunandan, K., 2007. Internal control weakness and cost of equity: evidence from SOX section 404 disclosures. The Accounting Review 82, 1255–1297. Palmrose, Z., 1988. An analysis of auditor litigation and audit service quality. The Accounting Review 63, 55–73. Ryan, S.G., 2008. Accounting in and for the subprime crisis. The Accounting Review 83, 1605–1638. Schneider, A., Church, B.K., 2008. The effect of auditors’ internal control opinions on loan decisions. Journal of Accounting and Public Policy 27, 1–18. Teoh, S., Wong, T., 1993. Perceived auditor quality and earnings response coefficient. The Accounting Review 68, 346–367. Zhang, Y., Zhou, J., Zhou, N., 2007. Audit committee quality, auditor independence, and internal control weakness. Journal of Accounting and Public Policy 26 (3), 300–327.