当前位置:首页 >> 公共/行政管理 >>

Internal Control Weaknesses and Accounting Conservatism Evidence From the Post Sarbanes Oxley Period


Internal Control Weaknesses and Accounting Conservatism: Evidence From the Post–Sarbanes– Oxley Period

Journal of Accounting, Auditing & Finance 28(2) 152–191 ?The Author(s) 2013 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/0148558X13479057 jaf.sagepub.com

Santanu Mitra1, Bikki Jaggi2, and Mahmud Hossain3

Abstract This study examines the relationship between accounting conservatism and internal control weaknesses (ICW) in the post–Sarbanes–Oxley Act of 2002 (SOX) period when the U.S. firms have been subject to higher regulations and enhanced corporate oversight and scrutiny. Our multivariate analyses show that the firms having ICW, especially the firms with company-level ICW, have significantly changed their conservative reporting practice from the pre- to the post-SOX period. The analyses further show that the ICW firms exhibit greater accounting conservatism in the post-SOX period compared with the firms with effective internal controls (non-ICW). The result is mostly driven by increased conditional conservatism by the firms having company-level ICW that are more pervasive in effect, less auditable, and more difficult to detect and prevent. Furthermore, we find that the difference in conservatism between ICW and non-ICW firms is more prominent in the first 3 post-SOX years than in the last 3 post-SOX years of the sample period. These findings suggest that enhanced corporate oversight and scrutiny have induced the ICW firms to use more accounting conservatism in an effort to reduce reporting uncertainty, enhance information reliability, and promote contracting efficiency. Our findings are consistent with prior studies that demonstrate a shift in the U.S. firms’ financial reporting strategies in response to stringent regulations and governance in the post-SOX period. Keywords accounting conservatism, internal control weaknesses, pre- and post-SOX period, company-level and account-specific internal control weaknesses, higher regulations, corporate oversight and scrutiny

1 2

Wayne State University, Detroit, MI, USA Rutgers, The State University of New Jersey, Piscataway, USA 3 American University of Sharjah, Sharjah, UAE Corresponding Author: Santanu Mitra, Department of Accounting, School of Business Administration, Wayne State University, 5229 Cass Avenue, 136 Rands House, Detroit, MI 48374, USA. Email: smitra@wayne.edu

Mitra et al.

153

Introduction
In a recent study, Goh and Li (2011) demonstrate that the firms having material internal control weaknesses (ICW) exhibit lower conservatism compared with the firms having no such weaknesses, and that the subsequent remediation of ICW leads to greater accounting conservatism. Their findings suggest that effective internal controls lead to the adoption of a reporting strategy that is based on more accounting conservatism. Internal control quality facilitates conservatism in financial reporting. Their findings are based on data prior to the implementation of the Sarbanes–Oxley Act of 2002 (SOX). We extend their line of research and examine the empirical link between accounting conservatism and ICW in the post-SOX period when the U.S. corporations are subject to enhanced regulatory oversight, higher corporate scrutiny, higher penalties for financial misstatements, stringent audit standards, and rigorous audit quality inspections by the Public Company Accounting Oversight Board (PCAOB) that significantly affect firms’ financial reporting process and investors’ expectation about the quality and reliability of reported financial information. Specifically, we evaluate how the changed regulatory environment and corporate oversight in the post-SOX period impact the choice of conservative reporting strategy by ICW ` -vis non-ICW firms. We argue that inadequate internal monitoring in the absence firms vis-a of effective internal controls is likely to result in a higher agency problem because it provides an opportunity to managers to make operating and financial reporting decisions that are likely to serve their own interests at the cost of other stakeholders. Thus, weak internal controls are deemed to exacerbate managers’ aggressive risk-taking behavior and their tendency to misreport financial information. Consequently, the very existence of weak internal controls would aggravate the risk perception among various contracting parties. AshbaughSkaife, Collins, Kinney, and LaFond (2008) and Doyle, Ge, and McVay (2007a) document that ICW firms are associated with lower quality of accounting accruals as a result of intentional and unintentional errors or biases and aggressive accounting practices. We conjecture that enhanced and stricter regulatory requirements in the post-SOX period, which are likely to result in higher expectation for better-quality financial reports, encourage the ICW firms to adopt a more conservative reporting strategy that would send positive signals on the reliability of reported information to various contracting parties. Prior studies suggest that accounting conservatism, which is considered to be a part of an efficient reporting strategy, benefits the financial statement users and the parties in various contracting situations, and constrains opportunistic payments to managers and other contracting parties (e.g., Watts, 2003a). Moreover, accounting conservatism is considered as a means of addressing moral hazard in the firms having asymmetric information, asymmetric payoffs, limited horizons, and limited liability (e.g., LaFond & Roychowdhury, 2008; Watts, 2003a). It constrains managerial opportunistic behavior in financial reporting and offsets managerial biases with its asymmetrical verifiability requirement (Watts, 2003a), which reduces uncertainty and enhances the reliability of reported information. Ball and Shivakumar (2005) argue that ‘‘while unconditional conservatism seems inefficient or at best neutral in contracting, conditional conservatism (timely loss recognition) can enhance contracting efficiency (p. 91).’’1 It, however, needs to be pointed out that some prior studies argue that conservatism may increase earnings management (i.e., Levitt, 1998; Penman & Zhang, 2002) and exacerbate rather than mitigate agency conflicts.2 The above discussion suggests that a lack of adequate monitoring and control over executive behavior in the ICW firms gives managers an incentive to make aggressive accounting policy choices. But enhanced and stricter post-SOX regulatory oversight, consequent

154

Journal of Accounting, Auditing & Finance

increase in creditor/lender expectation for a higher quality financial reporting, and stringent audit standards and audit quality inspection by PCAOB might collectively provide counter weight for weak internal controls and constrain managerial discretion to make opportunistic accounting choices. We, therefore, argue that this changed regulatory environment is more likely to induce the ICW firms to change their reporting strategy and adopt more accounting conservatism to reduce the risk and uncertainty associated with ICW. Mandatory postSOX disclosure requirements are likely to increase the flow of firm-specific information to outside stakeholders and potentially restrict managerial ability to conceal bad news, which leads to more asymmetrically timely loss recognition in earnings.3 Previous studies contend that increased flow of firm-specific information limits managers’ ability to conceal bad news (Dhaliwal et al., 2008). Thus, the expected fallout in enhanced regulatory environment is an increase of accounting conservatism in financial reporting by the ICW firms in the post-SOX years.4 We first examine the difference in conservative reporting practice of the ICW and the non-ICW firms between the pre- and the post-SOX periods. Specifically, we investigate the effect of SOX on these two groups of firms by classifying the ICW firms into the firms with company-level ICW and account-specific ICW. We evaluate whether accounting conservatism varies with the severity and pervasiveness of ICW. This leads us to our second research question to evaluate the difference in the post-SOX conservatism between the firms with company-level ICW and the firms with account-specific ICW relative to the firms with effective internal controls (non-ICW) for a 6-year period from 2004 through 2009.5 In addition, we compare accounting conservatism between the ICW and the nonICW firms separately for the first 3 years (i.e., 2004-2006) and the last 3 years (i.e., 20072009) of our post-SOX sample period. We use a sample of test firms whose internal controls are reported as ineffective by their external auditors and a sample of control firms whose internal controls are reported as effective in their auditors’ attestation reports under SOX Section 404. On the basis of propensity score matching process that controls for any potential effect of endogeneity and self-selection biases, we develop a matched-pair sample of ICW and non-ICW firms. We use the following three measures of conditional conservatism: Timeliness of Earnings to News (Basu, 1997), Persistence of Earnings Changes (Basu, 1997), and Accrual-Based Loss Recognition (Ball & Shivakumar, 2005). The results for the pre-SOX versus post-SOX period indicate that though the non-ICW firms have higher conservatism than the ICW firms in the pre-SOX period, the difference in conservatism decreases in the post-SOX period. Additional analyses show that SOX, in general, has a larger effect on the change in conservative reporting behavior of ICW firms, especially the firms with company-level ICW. We find a significantly greater change in their post-SOX accounting conservatism compared with the non-ICW firms. The result is primarily driven by greater conservatism of the firms with company-level ICW. We, however, do not find any difference in conservatism between the account-specific ICW and the non-ICW firms. We further document that the ICW firms, in general, have more accounting conservatism than the non-ICW firms in the post-SOX period, a result that is mainly attributed to greater conservative reporting practices by the company-level ICW firms relative to the other firms in the sample. Our results suggest that higher post-SOX regulatory scrutiny and enhanced corporate oversight collectively influence the ICW firms, especially the firms with pervasive company-level ICW to adopt more conservative reporting strategy. The use of more accounting conservatism could be a part of their overall effort to reduce reporting risk,

Mitra et al.

155

enhance reporting reliability, and promote contracting efficiency when faced with higher regulations and corporate scrutiny. Because the non-ICW firms are likely to face smaller agency problems in financial reporting, they are relatively less impacted by enhanced postSOX regulatory environment in pursuing their reporting strategy. Our analyses further show that the differences in post-SOX conservatism between the ICW and non-ICW firms and between the company-level ICW and the other two groups of firms are not uniform over the test period. The difference is more robust in an earlier postSOX period (i.e., 2004-2006) than in the later post-SOX period (i.e., 2007-2009). We attribute this difference to the immediately stronger reaction by the ICW firms to the post-SOX stricter regulations and corporate oversight, and the resulting increase in demand for conservatism from contracting parties. But with the passage of time, a greater safety net for investors might have developed as a result of SOX, and persistently superior flow of firm information mitigates the intensity of demand for conservative reporting by the ICW firms. This is accompanied by the non-ICW firms’ probable adjustments to their conservative reporting practices over time. The combined effect is reflected in narrowing gap in their reporting strategies. Finally, our tests show that subsequent remediation of control weaknesses in the post-SOX period does not have any additional effect on the ICW firms’ accounting conservatism probably as they already made significant adjustments to their conservative reporting practices. The study makes the following contributions to the literature on ICW and accounting conservatism. First, the study shows that accounting conservatism of the firms having higher internal monitoring problems due to ineffective internal controls are likely to be more impacted by the changed post-SOX regulatory environment. Second, the ICW firms exhibit greater accounting conservatism in terms of asymmetric loss recognition by earnings than the non-ICW firms in the post-SOX period. The pre-SOX contracting inefficiency is mitigated through stricter verification standards for recognizing good news as gains than recognizing bad news as losses. Third, asymmetrically timely loss recognition is more prominent in the firms where serious and more pervasive ICW are likely to create significant uncertainty about managerial incentives to report reliable accounting information. These findings add to the discussion initiated by previous studies on accounting conservatism (e.g., Ahmed & Duellman, 2007; Ball & Shivakumar, 2005; Basu, 1997; LaFond & Roychowdhury, 2008). Finally, our study complements concurrent research on internal controls over financial reporting (e.g., Ashbaugh-Skaife et al., 2008; Doyle et al., 2007a, 2007b; Goh & Li, 2011). The remainder of the article is organized as follows. The section titled ‘‘Background and Hypotheses’’ contains background and discussion on the development of hypotheses. The ‘‘Research Design’’ section discusses the research design. The section titled ‘‘Sample, Descriptive Data, and Correlations’’ discusses the sample, descriptive data, and correlations. The results are discussed in the ‘‘Results and Discussion’’ section, and additional analyses are included in the ‘‘Additional Analyses’’ section. Concluding remarks are contained in ‘‘Conclusions’’ section.

Background and Hypotheses Internal Controls and Financial Reporting
Internal control over financial reporting (ICFR) has received special emphasis in SOX (Sections 302 and 404). Under the new SOX requirements, management is required to

156

Journal of Accounting, Auditing & Finance

provide an assessment on the effectiveness of internal controls, and external auditors are required to conduct an audit of internal controls and provide certification about the management’s assessment of these internal controls. Even though effective internal controls may not fully eliminate all potential intentional and unintentional accounting errors and adjustments, they can limit managers’ ability to manage reported earnings or accruals (Jiambalvo, 1996) and minimize the probability of financial misstatements. Recent studies document that ineffective internal controls lead to a greater risk of financial misreporting (e.g., Ashbaugh-Skaife et al., 2008; Doyle et al., 2007a). It however needs to be recognized that despite strong internal controls, managers may still be able to use their flexibility to make accounting choices to achieve certain reporting objectives. The recent literature distinguishes between company-level ICW and account-specific ICW (e.g., Doyle et al., 2007a, 2007b; Raghunandan & Rama, 2006).6 The company-level ICW are all-pervasive in nature and are less auditable and detectable, whereas the accountspecific ICW relate to specific accounts and/or transactions that are more auditable and are likely to be more detectable by external auditors. Doyle et al. (2007a) find that the company-level ICW are significantly associated with lower accruals quality, but they do not find any association between account-specific ICW and accruals quality. They argue that company-level ICW that are less auditable are likely to result in more erroneous financial reporting. In view of the differential impact of these two types of ICW on earnings quality, the bond-rating agencies also respond differently to them.7 Moody’s Investors Service (2004) further suggests that company-level control weaknesses call into question not only the management’s ability to prepare accurate financial reports but also its ability to control the business. Consistent with Doyle et al. (2007a, 2007b) and Raghunandan and Rama (2006), we classify the ICW firms as having company-level ICW when they have either companylevel pervasive control weaknesses or both company-level pervasive control weaknesses and transaction/account-specific control weaknesses. Furthermore, if an ICW firm has at least three transaction/account-specific control weaknesses but no company-level weaknesses, we classify the firm as having company-level ICW. The remaining ICW firms are classified as having account-specific control weaknesses. We explain the selection criteria in the appendix to this article. In classifying the ICW firms into company-level and account-specific groups, we consider the approaches adopted by Raghunandan and Rama and Doyle et al. (2007a, 2007b) to make the classification more reasonable and precise. The association between ICW and monitoring of managerial behavior has been discussed by Jensen (1993), who argues that if strong internal controls provide effective monitoring of managerial behavior, they are likely to mitigate the agency problems and consequently may not provide any managerial incentive to adopt an accounting strategy, including accounting conservatism. Thus, if strong internal controls constrain managerial tendency to manipulate or smooth earnings, they are not expected to create any incentive or disincentive for firms to report conservative earnings. In this respect, Goh and Li (2011) also recognize that Moving away from the agency framework of conservatism as a beneficial governance mechanism, it becomes unclear why one would expect a positive link between internal controls and conservatism . . . it is questionable whether conservatism is beneficial to firms and shareholders . . . regulators and standard setters do not necessarily view conservatism as a desirable attribute of financial reporting (p. 976).

Mitra et al.

157

Accounting Conservatism and Financial Reporting
Accounting conservatism involves a higher degree of verification for recognizing good news than bad news in the financial statements. Basu (1997) argues that a firm’s accounting is more conservative when earnings reflect bad news more quickly than good news, implying an asymmetric timeliness of loss recognition by earnings. The conservative earnings are also deemed to be more reliable and less risky from the standpoint of contracting parties. The higher reliability is especially supported by the fact that higher verification standards are required to recognize gains, which reduce the risk of aggressive financial reporting. When there is an agency conflict, firms use more conservative reporting to minimize its adverse effect. Furthermore, conservatism reduces managers’ ability to loosen or avoid dividend restrictions and mitigates the bondholder–shareholder conflict over dividend policy (Ahmed et al., 2002). Watts (2003a) asserts that conservatism reduces litigation costs for the firm, whereas LaFond and Roychowdhury (2008) suggest that conservatism is viewed as a financial reporting practice that minimizes agency costs. They document that demand for accounting conservatism increases with decreases in managerial stock ownership, implying that a higher agency problem arising from the separation of ownership and control leads to a greater demand for conservative reporting. Another strand of research looks at the link between agency problem and reporting strategy based on accounting conservatism, and resultant reporting quality. Their findings show that overall there is a positive relationship between agency problem and accounting conservatism. Ahmed et al. (2002) find that the firms with more severe bondholder–shareholder dividend policy conflicts (i.e., an indicator of greater agency problems), on average, use more conservative accounting. LaFond and Roychowdhury (2008), who address the issue of conditional conservatism from the standpoint of agency framework, show that conservatism plays a critical role in addressing agency problem between managers and shareholders. Watts (2003a) suggests that agency problems arise when interests of managers and shareholders are not properly aligned. In support of this contention, LaFond and Roychowdhury (2008) find that as the managerial ownership decreases, earnings become less timely in recognizing good news and more asymmetrically timely in recognizing bad news. Their evidence suggests that in low managerial ownership firms (with greater separation of ownership and control), there is greater asymmetry in verifying standards for recognizing good news as gains than bad news as losses. With an increase in managerial ownership, the agency problem diminishes, and the demand for accounting conservatism declines. These results thus show that accounting conservatism is viewed as a compensating mechanism against agency conflicts and weaker governance. Several other studies focus on the role of accounting conservatism in promoting contracting efficiency (i.e., by reducing the perceived risk and uncertainty associated with aggressive financial reporting) within the agency framework.8 Their results suggest that, among other things, a firm’s risk characteristics and internal and external governance mechanisms are likely to impact managerial decisions to report conservative earnings to achieve contracting efficiency. Dhaliwal et al. (2008) show that greater firm-specific information flow to outsiders limits managers’ ability to conceal bad news and increases the level of accounting conservatism.

Hypotheses
The previous discussion of the extant research on accounting conservatism and financial reporting shows that, as a part of reporting strategy, managers use accounting conservatism

158

Journal of Accounting, Auditing & Finance

to mitigate the potential adverse effect of agency problems in an uncertain (or a risky) corporate environment. Based on this evidence, we argue that firms’ exposure to stricter regulatory environment and corporate oversight is likely to encourage them to use more accounting conservatism in financial reporting. The SOX that introduced several important regulations to strengthen corporate governance, auditor independence, and enhance financial reporting quality, attempts to restore investor confidence in reported information by improving the internal control mechanism, providing a more effective monitoring of managerial behavior through a better corporate governance, ensuring higher audit quality by supervising the work of audit firms, and consequently higher reliability of reported information. Various provisions in SOX are intended to strengthen both internal and external governance of the publicly traded corporations. It increases management’s responsibility by requiring the company’s top executives to certify the effectiveness of internal controls, and stiffens penalties for certifying fraudulent misstatements. It established the PCAOB to monitor audit of public companies through periodic inspection of audit firms. Nelson (2006) finds that the SOX-mandated reforms have increased both clients’ and auditors’ incentives for accurate financial reporting. He further states that the continuing threat of PCAOB annual inspections and sanctions encourages auditors to detect the financial misstatements more effectively and resist client pressure to allow opportunistic reporting.9 The mandated ICFR audits under SOX 404, which provides additional monitoring and scrutiny over the financial reporting process, is especially more relevant to ensure reliability of reported information by firms with ICW. If auditors’ attestation reports identify that the firm’s internal controls are ineffective, the firm’s financial reporting process is likely to be viewed with skepticism by various stakeholders and contracting parties, which will have a negative impact on the quality of reported information (e.g., LaFond & Roychowdhury, 2008). This adverse sentiment about the reliability of reported information is even more pertinent for the ICW firms that have company-wide pervasive control weaknesses, because various stakeholders and contracting parties are likely to view those firms as having serious risks in their control and financial reporting process. We argue that to mitigate this negative sentiment, the ICW firms are likely to adopt more conservative reporting strategy, especially in the period when the control weaknesses are publicly reported. This argument is consistent with the view that enhanced governmental scrutiny and higher regulations for financial reporting contribute to more conservatism (Watts, 2003a). We, therefore, conjecture that enhanced regulatory requirements and stricter monitoring of managerial behavior are likely to affect the firms’ reporting strategy, especially when firms suffer from pervasive and serious control weaknesses. In view of high-risk situation in financial reporting and the associated agency and contracting problems, the ICW firms, especially the firms having all-pervasive company-level control problems, may have a stronger incentive to make changes in their reporting practices in the post-SOX period to produce more reliable information and improve contracting efficiency.10 Our argument is supported by the views expressed in prior literature that regulations significantly influence management’s reporting practices. Watts (2003a), for example, argues that government regulation of financial reporting contributes to conservatism, whereas Ribstein (2002) contends that lower probability of being held accountable for conservative reporting encourages executives to underreport earnings. Browning (2002) suggests that the new SOX rules requiring CEO/CFO certification about the fairness of financial statements in all material respects may significantly induce firms to report conservative financial numbers. Watts (2003b) further argues that courts generally punish overstatements more than

Mitra et al.

159

understatements because relevant stakeholders are more likely to suffer losses when earnings and assets are overstated rather than understated. Prior studies also demonstrate that when faced with higher regulations and governance in the post-SOX period, the U.S. firms are likely to recalibrate their reporting strategies, which potentially involve, among other things, simultaneous balancing among alternative earnings management vehicles.11 We specifically argue that enhanced regulatory and stricter monitoring environment in the post-SOX period encourage firms especially the ICW firms to change their reporting strategy toward more conservative accounting to counter the financial statement users’ perception about higher risk of financial misreporting due to weak internal controls and thus send positive signals on the reliability of their reported information. We express this argument in the form of the following two hypotheses to examine the impact of SOX on ` -vis non-ICW firms. accounting conservatism of the ICW firms vis-a Hypothesis 1 (H1): Ceteris paribus, the ICW firms have greater accounting conservatism in the post-SOX period compared with the non-ICW firms. Hypothesis 2 (H2): Ceteris paribus, the company-level ICW firms have greater accounting conservatism in the post-SOX period than the account-specific ICW firms and the non-ICW firms.

Research Design Timeliness of Earnings to News
Our first measure of accounting conservatism is the asymmetrically timely loss recognition in earnings, as developed by Basu (1997). NI = b0 1 b1 NEG 1 b2 RET 1 b3 NEG 3 RET1 e, (1)

where NI = net income before extraordinary items scaled by the beginning equity market value, RET = market-adjusted annual stock return, and NEG = a dichotomous variable of 1 if RET is negative and 0 otherwise.12 A higher positive association between negative stock returns and earnings than between positive stock returns and earnings indicates timely reporting of bad news than good news, that is, earnings are reported more conservatively. In Equation 1, b2 measures the response of earnings to returns when returns are positive (i.e., good news), whereas b3 measures the incremental response of earnings to returns when returns are negative (i.e., bad news). b2 1 b3 capture the earnings response to stock returns when the returns are negative. The asymmetric timeliness coefficient, b3, is used to measure the degree of conservatism indicating whether earnings reflect bad news more quickly than good news. If earnings are conservative, b3 . 0 and b2 1 b3 . b2. Based on the prior literature (e.g., Basu, 1997; Dhaliwal et al., 2008; Goh & Li, 2011; Jenkins & Velury, 2008; LaFond & Roychowdhury, 2008), we use the following expanded Basu’s (1997) regression model with some modifications and examine the difference in accounting conservatism in terms of asymmetric timely loss recognition by earnings in the following four tests: between the firms with ICW and the firms without ICW, between the firms with systematic control weaknesses and the firms with nonsystematic control weaknesses, between the firms with systematic control weaknesses and the firms with no control weaknesses, and between the firms with nonsystematic control weaknesses and the firms with no control weaknesses.

160

Journal of Accounting, Auditing & Finance

NI = b0 1 b1 NEG 1 b2 RET 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 NEG 3 RET 1 b9 NEG 3 X 1 b10 NEG 3 MB 1 b11NEG 3 LEV 1 b12 NEG 3 SIZE 1 b13 NEG 3 LIT 1 b14 RET 3 X 1 b15 RET 3 MB 1 b16 RET 3 LEV 1 b17 RET 3 SIZE 1b18RET 3 LIT1 b19 RET 3 NEG 3 X 1b20 RET 3 NEG 3 MB 1 b21 RET 3 NEG 3 LEV 1 b22 RET 3 NEG 3 SIZE 1 b23 RET 3 NEG 3 LIT 1 b24 IND 1 b25 NEG 3 IND 1 b26 RET 3 IND 1 b27 NEG 3 RET 3 IND 1 e,

(2)

where X = a categorical variable of 1 for ICW firms in the test between the ICW and nonICW firms; 1 for the firms with company-level ICW in the test between the company-level and account-specific ICW firms; 1 for the firms with company-level ICW in the test between the company-level ICW and non-ICW firms; 1 for the firms with account-specific ICW in the test between the account-specific ICW and non-ICW firms. NI = net income before extraordinary items scaled by the beginning total assets; RET = market-adjusted annual stock returns; NEG = a dichotomous variable of 1 if RET is negative and 0 otherwise; MB = market-to-book ratio; LEV = leverage ratio computed as total debts divided by total assets; SIZE = natural log of total assets; LIT = a dichotomous variable of 1 for observations belonging to litigious industries (Standard Industrial Classification [SIC] Codes 2833-2836, 3570-3577, 3600-3674, 5200-5961, and 7370-7374); 0 otherwise (Francis, Philbrick, & Schipper, 1994); IND = industry dummy variables following Frankel, Johnson, and Nelson (2002) classification.13 The coefficient of RET 3 NEG 3 X, b19, reflects the difference in conservatism (measured in terms of asymmetric timely loss recognition by earnings) between the categories of firms in the four regression analyses.

Persistence of Earnings Changes
Our second measure of conservatism is the persistence of earnings changes in the subsequent fiscal period. Basu (1997) argues that under conservative accounting, bad news impacts earnings immediately but the effect does not persist. Good news takes longer to be recognized in earnings, but once it is factored in the earnings, it is likely to persist in future periods. So the decrease in earnings in the current period due to bad news is more likely to reverse in the future periods, whereas an increase in earnings due to good news recognition is less likely to reverse in the following period. As a result, the reversal of bad news in earnings occurs more quickly than the reversal of good news. This situation leads to a negative correlation between earnings changes in the current and future periods. Basu expresses this asymmetric persistence of earnings changes in response to bad versus good news in the following model. DNIt = b0 1 b1 NEG 1 b2 DNIt
– 1

1 b3 NEG 3 DNIt

– 1,

(3)

where DNIt (DNIt – 1) = change in net income before extraordinary items in year t (t – 1) scaled by the beginning equity market value; NEG = a categorical variable of 1 if DNIt – 1 \ 0, and 0 otherwise. The reversal of transitory gain is indicated by b2 \ 0, while the reversal of transitory loss is indicated by b2 1 b3. If losses are asymmetrically recognized in a more timely manner than gains, we expect that b3 \ 0 and b2 1 b3 \ b2.

Mitra et al.

161

Similar to Equation 2, the above regression model (Equation 3) is expanded by including the variables of interest and their interactions with the accounting conservatism measure. Five control variables and their interactions with conservatism measures are included to control for the effect of firm factor on the level of accounting conservatism. The expanded regression equation is described below: DNIt = b0 1 b1 NEG 1 b2 DNIt – 1 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 NEG 3 DNIt – 1 1 b9 NEG 3 X 1 b10 NEG 3 MB 1 b11NEG 3 LEV 1 b12NEG 3 SIZE 1 b13 NEG 3 LIT 1 b14 DNIt – 1 3 X 1b15 DNIt – 1 3 MB 1 b16 DNIt – 1 3 LEV 1 b17 DNIt – 1 3 SIZE 1 b18 DNIt – 1 3 LIT 1 b19 NEG 3 DNIt – 1 3 X 1 b20 NEG 3 DNIt – 1 3 MB 1 b21 NEG 3 DNIt – 1 3 LEV 1 b22 NEG 3 DNIt – 1 3 SIZE 1 b23 NEG 3 DNIt – 1 3 LIT 1 b24 IND 1 b25 NEG 3 IND 1 b26 DNIt – 1 3 IND 1 b27 NEG 3 DNIt – 1 3 IND 1 e.

(4)

All variables are defined in previous sections. The coefficient of NEG 3 DNIt – 1 3 X, b19 reflects the difference in conservatism (measured in terms of reversal of transitory loss in future earnings) between the categories of firms in the four regression analyses.

Accrual-Based Loss Recognition
Our third measure of accounting conservatism is accrual-based asymmetric loss recognition developed by Ball and Shivakumar (2005). According to Dechow, Kothari, and Watts (1998), the role of accruals is to mitigate noise in cash flows and make earnings more informative and less noisy. One implication is that accruals and cash flows are negatively correlated. Ball and Shivakumar envision a second role of accruals, which is timely recognition of gains and losses. They develop a model that relies on the notion that asymmetry in accruals adjustments occurs because economic losses are more likely to be recognized on a timely basis as accrued charges against income, whereas economic gains are likely to be recognized when realized. This situation implies that the positive correlation between accruals and cash flows from the second role of accruals is greater in case of losses. Ball and Shivakumar estimate the following model of accounting conservatism, which reflects accruals’ asymmetrically timelier recognition of economic losses than of economic gains. ACC = b0 1 b1 NEG_CFO 1 b2 CFO 1 b3 CFO 3 NEG_CFO, (5)

where ACC = net income before extraordinary items minus cash flow from operations scaled by the beginning total assets; CFO = cash flow from operations scaled by the beginning total assets; NEG_CFO is a dichotomous variable equal to 1 if CFO \ 0, and 0 otherwise. We expect b2 to be negative, indicating a generally negative relationship between accruals and cash flows where accruals mitigate the noise in cash flows (Dechow, 1994). The second role of accruals, as suggested by Ball and Shivakumar (2005), is the timely recognition of economic losses than gains. The coefficient b3 indicates asymmetric timeliness of loss or gain recognition by accruals, loss being recognized more quickly than gains. So we expect b3 to be significantly positive, which is the measure of accounting conservatism; thus, b2 1 b3 . b2.

162

Journal of Accounting, Auditing & Finance

The above regression model (Equation 5) is extended by including the variables of interest and their interactions with accounting conservatism measure. Five control variables and their interactions with conservatism measures are included to control for the effect of several firm factors on the level of accounting conservatism. The expanded regression equation is described below. ACC = b0 1 b1 NEG_CFO 1 b2 CFO 1 b3 CFO 3 NEG_CFO 1 b4 X 1 b5 NEG_CFO 3 X 1 b6 CFO 3 X 1 b7 CFO 3 NEG_CFO 3 X 1 b8 MB 1 b9 NEG_CFO 3 MB 1 b10 CFO 3 MB 1 b11 CFO 3 NEG_CFO 3 MB 1 b12 LEV 1 b13 NEG_CFO 3 LEV 1 b14 CFO 3 LEV 1 b15 CFO 3 NEG_CFO 3 LEV 1 b16 SIZE 1 b17 NEG_CFO 3 SIZE 1 b18 CFO 3 SIZE 1 b19 CFO 3 NEG_CFO 3 SIZE 1 b20 LIT 1 b21 NEG_CFO 3 LIT 1 b22 CFO 3 LIT 1 b23 CFO 3 NEG_CFO 3 LIT 1 b24 IND 1 b25 NEG_CFO 3 IND (6) 1 b26 CFO 3 IND 1 b27 CFO 3 NEG_CFO 3 IND 1 e.14 All variables are defined in the previous sections. The coefficient of CFO 3 NEG_CFO 3 X, b7, reflects the difference in conservatism (measured in terms of asymmetric timely loss recognition by accruals) between the categories of firms in the four regression analyses.

Sample, Descriptive Data, and Correlations
The 2004-2009 Audit Analytics database is used to select firms for the study. Initially, we select 2,314 firm observations that are associated with ineffective internal controls as per their auditors’ attestation reports for the fiscal years from 2004 through 2009 (i.e., ICW firms).15 Those selected firms have both ticker symbol and SIC codes available from Audit Analytics, making it convenient for us to match them with other databases like Compustat and Center for Research in Security Prices (CRSP). From the initial sample, we exclude 406 firm observations for which the required data for analyses are not available from the Compustat and CRSP databases. Next, we remove 162 outlier observations for all continuous variables at the top and bottom 0.5% levels. This filtering process results in a final sample of 1,746 ICW firm observations, of which 756 relate to company-level ICW and 990 relate to account-specific ICW. The sample selection process is summarized in Panel A of Table 1. Finally, to control for potential endogeneity and self-selection bias in the sample, we develop a matched-pair sample of non-ICW firms for the ICW firms on the basis of a propensity score matching process from estimating the following first-stage probit regression for the ICW determinants.16 ICW = b0 1 b1 MVE 1 b2 AGE 1 b3 LOSS 1 b4 SEGMENT 1 b5 FOREIGN 1 b6 M&A 1 b7 RESTRUCT 1 b8 EXTR_SALES 1 b9 BIG 4 1 b10 RESTATE 1 b11 AUD_CHANGE 1 e.17

(7)

Using the estimated coefficients of the first-stage regression, we determine the propensity score (predicted probability) of being an ICW firm for both the ICW and non-ICW firms. For each ICW firm, we choose a non-ICW firm whose predicted probability of being an ICW firm is closest to that of the ICW firm. This process enables us to select a set of

Mitra et al.
Table 1. Sample Details. Panel A: Sample Selection. ICW firms initially selected from Audit Analytics per auditors’ attestation reports with appropriate ticker symbols and SIC for the years from 2004 to 2009 Less: Firms for which data are not available in Compustat and CRSP for conducting analyses Less: Outlier observations for all continuous variables at the top and bottom 0.5% levels Final sample of ICW firms Total firm observations for post-SOX analyses ICW firm observations Non-ICW firm observations Out of 1,746 ICW firms Firms with company-level ICW Firms with account-specific ICW Panel B: Year-Wise Sample Distribution. ICW observations Years 2004 2005 2006 2007 2008 2009 Matched control non-ICW firms (Based on propensity score matching process) Total 374 396 289 263 245 179 1,746 1,746 Company level 166 133 141 124 115 77 756 756 2,314 (406) (162) 1,746 3,492 1,746 1,746 756 990

163

Account specific 208 263 148 139 130 102 990 990

Panel C: Industry Distribution of the Sample Firms (Based on Frankel, Johnson, and Nelson, 2002, Classification). Industry classification Agriculture Mining and construction Food Textile and printing/publishing Chemicals Pharmaceuticals Extraction Durable manufacturers Transportation Utilities Retail Financial Services Computers Total Total firm observations 30 98 220 174 264 550 278 552 108 124 332 156 280 326 3,492 ICW firm observations 15 49 110 87 132 275 139 276 54 62 166 78 140 163 1,746 % 0.86 2.81 6.30 4.98 7.56 15.75 7.96 15.81 3.09 3.55 9.51 4.47 8.02 9.34 100 Non-ICW firm observations 15 49 110 87 132 275 139 276 54 62 166 78 140 163 1,746 % 0.86 2.81 6.30 4.98 7.56 15.75 7.96 15.81 3.09 3.55 9.51 4.47 8.02 9.34 100

Note: ICW = internal control weaknesses; SIC = Standard Industrial Classification; CRSP = Center for Research in Security Prices; SOX = Sarbanes–Oxley Act of 2002.

164

Journal of Accounting, Auditing & Finance

benchmark non-ICW firms that are characteristically similar to the test ICW firms but do not have ICW. From the Audit Analytics database, we select a sample of 10,542 non-ICW firm observations for the period from 2004 through 2009 using the following filters: appropriate ticker symbols and SIC codes required to match the data with other sources, and data availability from Compustat and CRSP to perform analyses and removal of outlier observations for all continuous data at the top and bottom 0.5% levels. We estimate the regression equation (Equation 7) for a combined sample of the ICW firms and non-ICW firms for each industry based on Frankel et al. (2002) industry classification criteria. Within each industry, each non-ICW firm is matched with an ICW firm on the basis of closest propensity score. In the process, for the 1,746 ICW observations, we arrive at 1,746 propensity score matched non-ICW observations, resulting in the final sample of 3,492 observations for the main analyses. The industry distribution of the sample firms is provided in Panel B of Table 1. Durable manufacturing firms are the most highly represented industry (with 552 firms), followed by 550 firms belonging to the pharmaceuticals industry in our sample; the agricultural sector has the least number of firms, with a total of 30 firms. Panel A of Table 2 reports the descriptive data for the sample. Some data are noteworthy. Except for MB and LEV, we do not find any significant mean difference for other variables between the ICW and non-ICW firms. The non-ICW firms have, on average, higher market-to-book ratio (MB) and use significantly more leverage (LEV) in the capital structure than the ICW firms. Panel B of Table 2 presents Pearson correlation statistics. Though several independent variables are correlated with each other, the multicollinearity problem in the main analysis is of inconsequential magnitude as the variance inflation factors, in any case, do not exceed 4.

Results and Discussion Pre-SOX Versus Post-SOX Accounting Conservatism
First, we conduct test on the immediate pre- and post-SOX period to evaluate difference in accounting conservatism between the two periods. We use 2000-2001 as the pre-SOX period and 2003-2004 as the immediate post-SOX period. We leave out the period of 2002 from the analyses because it is a transition period. We use the sample firms from the year 2004 for this test. There are 374 ICW and 374 propensity score matched non-ICW firms from year 2004 (the first year of SOX 404 reporting). Consistent with Goh and Li (2011), we assume that the status of having either effective or ineffective internal controls remain the same before their mandatory disclosure in 2004. Based on this assumption, we use 2,992 firm observations over a 4-year time period (2000, 2001, 2003, and 2004) with both the ICW firms and non-ICW firms each having 1,496 (374 3 4) observations. Out of 374 ICW firms, 166 firms have company-level ICW and 208 firms have account-specific ICW. Thus, 664 (166 3 4) and 832 (208 3 4) observations relate to the company-level and account-specific ICW, respectively, over the 4-year test period. The results are contained in Table 3. Panels A, B, and C of Table 3, respectively, present the results with timeliness of earnings to news, persistence of earnings changes, and accruals-based loss recognition as different conservatism measures. In all cases except for account-specific ICW versus non-ICW firms, we find that the ICW firms in general and the firms with company-level ICW in particular have lower conservatism than the non-ICW firms. But the interactions between the

Mitra et al.
Table 2. Descriptive Data and Correlation Statistics.
Panel A: Descriptive Data for the Full Sample. Firms with effective internal controls (n = 1,746) Variables NI DNIt DNIt – 1 ACC CFO RET MB LEV SIZE (US$ in millions) LIT M 0.028 0.004 0.010 20.016 0.033 0.059 4.042 0.375 2,554.611 0.285 Median 0.016 0.008 0.002 20.01 0.029 0.029 2.039 0.416 1,816.905 0.000 SD 0.852 0.645 0.724 0.489 0.511 0.495 2.801 0.487 2,104.572 0.451 Firms with ineffective internal controls (n = 1,746) M 0.021 20.006 0.002 20.01 0.04 0.048 3.715 0.302 2,439.025 0.285 Median 0.011 20.002 20.004 20.009 0.022 0.015 2.892 0.246 1,525.392 0.000 SD 1.016 0.528 0.921 0.673 0.696 0.784 1.506 0.708 2,587.324 0.451

165

t statistics for mean difference 0.288 0.501 0.285 20.301 20.339 0.495 4.714*** 3.551*** 1.448

Panel B: Correlation Statistics. Variables NI DNIt DNIt – ACC CFO RET MB LEV SIZE LIT NI DNIt DNIt – 1 ACC CFO RET MB LEV SIZE LIT

1

1.000 0.074** 1.000 0.059* 20.095** 1.000 0.112*** 0.053* 0.014 1.000 0.120*** 0.059* 0.038* 20.135*** 0.205*** 0.189*** 0.110*** 0.098*** 0.046* 0.065* 0.077** 0.048* 0.062* 0.019 0.025 0.079** 0.069* 0.043* 0.010 0.124*** 20.010 20.015 20.009 0.011

1.000 0.153*** 1.000 0.044* 0.120*** 1.000 0.089** 0.039* 20.045* 1.000 0.166*** 0.101*** 0.092** 0.105*** 1.000 0.022 20.040* 0.085** 0.049* 20.030 1.000

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses; SIC = Standard Industrial Classification. Variable definitions: NI = net income before extraordinary items scaled by the beginning total assets; DNIt = change in net income before extraordinary items in year t scaled by the beginning total assets; DNIt – 1 = change in net income before extraordinary items in year t – 1 scaled by the beginning total assets; ACC = income before extraordinary items minus cash flow from operations scaled by the beginning total assets; CFO = operating cash flow scaled by the beginning total assets; RET = market-adjusted annual stock returns; MB = market-to-book ratio; LEV = leverage ratio calculated as total debts divided by total assets; SIZE = natural log of total assets; LIT = a dummy variable of 1 if a firm operates in a litigious industry and 0 otherwise. Consistent with Francis, Philbrick, and Schipper (1994), the firms with primary SIC codes of 2833-2836 (biotechnology), 3570-3577 (computer equipment), 3600-3674 (electronics), 5200-5961 (retailing), and 7370-7374 (computer services) are considered operating in a litigious industry. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests.

relative conservatism measures and SOX variable indicate that the ICW firms, especially the firms with company-level ICW, increase their accounting conservatism in the post-SOX period. As a result, the relatively higher pre-SOX conservatism by the non-ICW firms than the ICW firms declines in the first 2 years of the post-SOX period. We also conduct regression tests separately for the three categories of firms, that is, the firms with company-level ICW, the firms with account-specific ICW, and the non-ICW

166
4 Company level vs. account specific X = Company level Coefficient .022 .398 2.018 .519 .095 .008*** .112 .000*** 2.003 .910 .025 .231 2.010 .409 .042 .053* .038 .072* 2.014 .175 2.060 .042** .058 .044** .047 .052* Included .276 1,496 Company level: 664 Account specific: 832 .033 .295 2.006 .770 .122 .000*** .135 .000*** 2.012 .591 .016 .195 2.008 .662 .040 .058* .019 .516 2.020 .446 2.071 .035** .065 .040** .042 .061* Included .265 1,328 Company level: 664 Non-ICW: 664 p value Coefficient value Coefficient X = Company level Company-level vs. non-ICW 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value .045 .065* 2.019 .411 .116 .000*** .121 .000*** 2.016 .209 .015 .187 2.002 .981 .018 .215 .024 .149 2.005 .806 2.022 .173 .045 .053* .009 .841 Included .269 1,664 Account specific: 832 Non-ICW: 832

Table 3. Pre-SOX Versus Post-SOX Accounting Conservatism (Years 2000 and 2001 Versus Years 2003 and 2004).

Panel A: Timeliness of Earnings to News of Accounting Conservatism.

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

Intercept NEG RET NEG 3 RET X SOX NEG 3 X RET 3 X NEG 3 SOX RET 3 SOX NEG 3 RET 3 X NEG 3 RET 3 SOX NEG 3 RET 3 X 3 SOX Industry effects Adjusted R2 n

.039 .080* 2.015 .442 .119 .000*** .128 .000*** 2.011 .604 .020 .179 2.005 .708 .036 .092* .029 .208 2.010 .368 2.065 .042** .050 .062* .039 .068* Included .268 2,992 ICW: 1,496 Non-ICW: 1,496

Note: The above table represents the results for the main variables of interest from estimating the following expanded Basu (1997) regression model (e.g., Goh & Li, 2011; LaFond & Roychowdhury, 2008): NI = b0 1 b1 NEG 1 b2 RET 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 SOX 1 b9 NEG 3 RET 1 b10 NEG 3 X 1 b11 NEG 3 MB 1 b12NEG 3 LEV 1 b13 NEG 3 SIZE 1 b14 NEG 3 LIT 1 b15 NEG 3 SOX 1b16 RET 3 X 1 b17 RET 3 MB 1 b18 RET 3 LEV 1 b19 RET 3 SIZE 1b20 RET 3 LIT1 b21 RET 3 SOX 1 b22 NEG 3 RET 3 X 1 b23 NEG 3 RET 3 MB 1 b24 NEG 3 RET 3 LEV 1 b25 NEG 3 RET 3 SIZE 1 b26 NEG 3 RET 3 LIT 1 b27 NEG 3 RET 3 SOX 1 b28 NEG 3 RET 3 X 3 SOX 1 Industry Effects 1 e. The variables are defined in Table 2. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias.

Table 3. (continued)

Panel B: Persistence of Earnings Changes Measure of Accounting Conservatism. 3 Company level vs. account specific X = Company level Coefficient 2.077 .026** 2.010 .306 2.099 .015** 2.110 .000*** .015 .285 .008 .612 .020 .139 2.077 .028** .002 .919 .005 .710 .048 .059* 2.068 .035** 20.032 0.089* Included .235 1,496 Company level: 664 Account specific: 832 2.055 .060* 2.016 .385 2.086 .019** 2.129 .000*** .006 .592 .011 .347 .017 .206 2.065 .040** .010 .405 .018 .201 .059 .040** 2.056 .044** 20.040 .058* Included .210 1,328 Company level: 664 Non-ICW: 664 p value Coefficient p value Coefficient X = Company level Company-level vs. non-ICW 4 5 6 7 8 9

1

2

ICW vs. non-ICW

Account-specific vs. non-ICW X = Account specific p value

X = ICW

Variables

Coefficient

p value

Intercept NEG DNIt – 1 NEG 3 DNIt – 1 X SOX NEG 3 X DNIt – 1 3 X NEG 3 SOX DNIt – 1 3 SOX NEG 3 DNIt – 1 3 X NEG 3 DNIt – 1 3 SOX NEG 3 DNIt –1 3 X 3 SOX Industry effects Adjusted R2 n

2.068 .040** 2.012 .296 2.095 .029** 2.122 .000*** .010 .419 .019 .185 .016 .191 2.082 .025** 2.005 .609 .011 .193 .044 .060* 2.062 .040** 2.039 .075* Included .229 2,992 ICW: 1,496 Non-ICW: 1,496

2.070 .030** 2.011 .477 2.109 .000*** 2.135 .000*** .004 .718 .018 .202 .005 .809 2.070 .035** .012 .283 .008 .616 .025 .119 2.055 .048** 2.014 .272 Included .218 1,664 Account specific: 832 Non-ICW: 832

167

Note: The above table reports the regression results for the main variables of interest from estimating the following expanded Basu (1997) regression model (e.g., Goh & Li, 2011): DNIt = b0 1 b1 NEG 1 b2 DNIt – 1 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 SOX 1 b9 NEG 3 DNIt – 1 1 b10 NEG X 1 b11 NEG 3 MB 1 b12 NEG 3 LEV 1 b13 NEG 3 SIZE 1 b14 NEG 3 LIT 1 b15 NEG 3 SOX 1 b16 DNIt – 1 3 X 1b17 DNIt – 1 3 MB 1 b18 DNIt – 1 3 LEV 1 b19 DNIt – 1 3 SIZE 1 b20 DNIt – 1 3 LIT 1 b21 DNIt – 1 3 SOX 1 b22 NEG 3 DNIt – 1 3 X 1 b23 NEG 3 DNIt – 1 3 MB 1 b24 NEG 3 DNIt – 1 3 LEV 1 b25 NEG 3 DNIt – 1 3 SIZE 1 b26 NEG 3 DNIt – 1 3 LIT 1 b27 NEG 3 DNIt – 1 3 SOX 1 b28 NEG 3 DNIt – 1 3 X 3 SOX 1 industry effects 1 e. The variables are defined in Table 2. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias.

Table 3. (continued)

168
4 Company level vs. account specific X = Company level Coefficient .020 .185 .004 .719 2.385 .000*** .189 .000*** 2.018 .152 .025 .116 .012 .319 .003 .874 .016 .210 2.006 .342 2.075 .022** .061 .040** .042 .060* Included .305 1,496 Company level: 664 Account specific: 832 .069 .040** .012 .392 2.306 .000*** .234 .000*** 2.004 .722 .015 .289 .009 .558 .014 .395 .008 .377 2.015 .210 2.067 .035** .045 .055** .052 .049** Included .311 1,328 Company level: 664 Non-ICW: 664 p value Coefficient p value Coefficient X = Company level Company-level vs. non-ICW Account-specific vs. non-ICW X = Account specific p value 5 6 7 8 9 .055 .068* .010 .411 2.275 .000*** .195 .000*** 2.011 .303 .022 .185 .015 .232 .018 .215 .010 .298 2.003 .672 2.016 .285 .048 .052* .004 .869 Included .292 1,664 Account specific: 832 Non-ICW: 832

Panel C: Accruals-Based Loss Recognition Measure of Accounting Conservatism.

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

Intercept NEG CFO NEG 3 CFO X SOX NEG 3 X CFO 3 X NEG 3 SOX CFO 3 SOX NEG 3 CFO 3 X NEG 3 CFO 3 SOX NEG 3 CFO 3 X 3 SOX Industry effects Adjusted R2 n

.030 .105 .016 .380 2.442 .000*** .304 .000*** 2.006 .315 .019 .148 .005 .804 .010 .409 .011 .350 2.004 .593 2.070 .032** .049 .059* .045 .056* Included .295 2,992 ICW: 1,496 Non-ICW: 1,496

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses. The above table reports regression results for the main variables of interest from estimating the following expanded Ball and Shivakumar (2005) model (e.g., Goh & Li, 2011): ACC = b0 1 b1 NEG1 b2 CFO 1 b3 NEG 3 CFO 1 b4 X 1 b5 NEG 3 X 1 b6 CFO 3 X 1 b7 NEG 3 CFO 3 X 1 b8 MB 1 b9 NEG 3 MB 1 b10 CFO 3 MB 1 b11 NEG 3 CFO 3 MB 1 b12 LEV 1 b13 NEG 3 LEV 1 b14 CFO 3 LEV 1 b15 NEG 3 LEV 1 b16 SIZE 1 b17 NEG 3 SIZE 1 b18 CFO 3 SIZE 1 b19 NEG 3 CFO 3 SIZE 1 b20 LIT 1 b21 NEG 3 LIT 1 b22 CFO 3 LIT 1 b23 NEG 3 CFO 3 LIT 1 b24 SOX 1 b25 NEG 3 SOX 1 b26 CFO 3 SOX 1 b27 NEG 3 CFO 3 SOX 1 b28 NEG 3 CFO 3 X 3 SOX 1 industry effects 1 e. The variables are defined in Table 2. The analyses use 2,992 firm observations over a 4-year time period (2000, 2001, 2003, and 2004), of which 1,496 (374 3 4) relate to ICW firms and 1,496 relate to non-ICW firms. We select 374 firms with ICW reported for the year 2004, and 374 non-ICW matched firms on the basis of a propensity score matching process. Out of 374 firms, 166 firms have company-level ICW and 208 firms have account-specific ICW. Thus, 664 (166 3 4) observations relate to company-level ICW and 832 (208 3 4) relate to accountspecific ICW over the 4-year test period. Based on the assumption that the firms had the same status of either having or not having ICW before their mandatory disclosures in 2004, we perform the analyses between the pre- and post-SOX periods dividing the total time period into pre-SOX that includes the years 2000 and 2001 (SOX = 0) and postSOX that includes the years 2003 and 2004 (SOX = 1). We consider the year 2002 as the transition year for SOX and thus exclude it from the analysis. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias.

Mitra et al.

169

firms, using the interactions of conservatism measures with SOX as the variables of interest for the 4 fiscal years, 2000, 2001, 2003, and 2004 (SOX = 1 for 2003 and 2004; SOX = 0 for 2000 and 2001). The results (not tabulated here) show that SOX has significant positive effect on accounting conservatism of the company-level ICW firms for all the three conservatism measures, especially the measures of timeliness of earnings to news and accrualsbased loss recognition. It also has significant positive effect on conservatism of the account-specific ICW firms for the accruals-based loss recognition measure. In general, the results suggest that SOX has relatively larger positive effect on the accounting conservatism of the company-level ICW firms, followed by its effect on conservatism of the account-specific ICW and the non-ICW firms. These results complement our findings reported in Table 3 about the relative changes in accounting conservatism of the three categories of firms. The firms with more pervasive ICW make larger changes in their conservative reporting practices (becoming more conservative) than the other two categories of firms when faced with higher regulations, and corporate oversight and scrutiny in the postSOX period.

Accounting Conservatism During the Extended Post-SOX Period Timeliness of Earnings to News Measure
We extend the analyses in the next stage by analyzing the data for 6 post-SOX years (i.e., 2004-2009).The results obtained from estimating the regression equation (Equation 2) are reported in Table 4. Consistent with Basu (1997), the results show that earnings recognize economic losses more quickly than economic gains. The results contained in columns 2 and 3 show that the coefficient (b19) of the interaction term RET 3 NEG 3 X (.045; p = .064) is significantly positive, suggesting that the ICW firms, as a whole, asymmetrically more timely recognize losses than the non-ICW firms. The results support H1. The results contained in columns 4 and 5 show that the coefficient b19 (.050; p = .056) is significantly positive, implying that the firms with company-level ICW asymmetrically more timely recognize losses than gains compared with the firms with account-specific ICW. The results in columns 6 and 7 show that the coefficient b19 (.062; p = .042) is significantly positive, implying that the firms with company-level ICW asymmetrically more timely recognize losses than gains compared with the non-ICW firms. These findings support H2. Finally, columns 8 and 9 show that the coefficient b19 (.015; p = .339) is insignificant, implying that there is no difference in asymmetric loss recognition by earnings between the firms with account-specific ICW and the non-ICW firms. The results suggest that in the post-SOX period, the ICW firms are, in general, more conservative in terms of earlier recognition of losses than gains by earnings than the nonICW firms. Especially, the firms with company-level ICW have significantly greater conservatism than both the account-specific ICW and the non-ICW firms. Furthermore, the significantly negative coefficient of RET 3 X, b14, indicates that the earnings of the ICW firms in general and of the company-level ICW firms in particular are relatively less timely in recognizing good news.

Persistence of Earnings Changes Measure
Table 5 presents the results from estimating the regression equation (Equation 4) with respect to persistence of earnings changes in subsequent year. Consistent with Basu (1997), we find that in general, negative changes in earnings are less persistent in the subsequent period than positive changes in earnings. Columns 2 and 3 (text continues on p. 174)

170
4 Company level vs. account specific X = Company level Coefficient .085 2.012 .118 2.010 .003 2.019 .030 2.040 .126 2.020 .012 .075 2.045 .035 2.044 2.016 .003 .062 2.014 .050 .015** .578 .000*** .483 .692 .181 .094* .065* .000*** .109 .411 .030** .068* .084* .070* .406 .829 .043* .733 .056* .073 2.010 .112 2.018 .006 2.011 .035 2.042 .120 2.022 .009 .055 2.048 .030 2.065 2.008 .002 .050 2.006 .062 .028** .714 .000*** .295 .598 .359 .090* .055* .000*** .205 .770 .053* .060* .092* .040** .526 .904 .059* .810 .042** p value Coefficient p value Coefficient .062 2.014 .121 2.020 .004 2.018 .026 2.035 .132 2.002 .007 .055 2.031 .022 2.019 2.012 .006 .055 2.008 .015 X = Company level Company-level vs. non-ICW 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value .045** .315 .002*** .169 .671 .219 .090* .088* .000*** .904 .685 .061* .099* .191 .151 .125 .505 .052* .544 .339 (continued)

Table 4. Timeliness of Earnings to News Measure of Accounting Conservatism (Post-SOX Period: 2004 to 2009).

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

Intercept NEG RET X MB LEV SIZE LIT NEG 3 RET NEG 3 X NEG 3 MB NEG 3 LEV NEG 3 SIZE NEG 3 LIT RET 3 X RET 3 MB RET 3 LEV RET 3 SIZE RET 3 LIT NEG 3 RET 3 X

.060 2.004 .105 2.013 .009 2.015 .025 2.029 .130 2.010 .022 .068 2.040 .039 2.058 2.005 .011 .045 2.010 .045

.046** .719 .000*** .416 .485 .306 .096* .082* .000*** .298 .179 .040* .077* .086* .052* .818 .510 .069* .612 .064*

Table 4. (continued) 4 Company level vs. account specific X = Company level Coefficient 2.042 .079* .104 .000** 2.020 .116 .019 .138 Included .280 1,746 Company level: 756 Account specific: 990 2.059 .049** .080 .025** 2.026 .108 .005 .862 Included .279 1,512 Company level: 756 Non-ICW: 756 p value Coefficient p value X = Company level Company-level vs. non-ICW 5 6 7 8 9

1

2

3

ICW vs. non-ICW

Account-specific vs. non-ICW X = Account specific Coefficient p value 2.069 .040** .083 .016** 2.040 .058* .010 .419 Included .288 1,980 Account specific: 990 Non-ICW: 990

X = ICW

Variables

Coefficient

p value

NEG 3 RET 3 MB NEG 3 RET 3 LEV NEG 3 RET 3 SIZE NEG 3 RET 3 LIT Industry effects Adjusted R2 n

2.060 .050** .096 .010*** 2.030 .095* .004 .717 Included .285 3,492 ICW: 1,746 Non-ICW: 1,746

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses. The table reports the regression results from estimating the following expanded Basu (1997) regression model (e.g., Goh & Li, 2011; LaFond & Roychowdhury, 2008): NI = b0 1 b1 NEG 1 b2 RET 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 NEG 3 RET 1 b9 NEG 3 X 1 b10 NEG 3 MB 1 b11NEG 3 LEV 1 b12 NEG 3 SIZE 1 b13 NEG 3 LIT 1 b14 RET 3 X 1 b15 RET 3 MB 1 b16 RET 3 LEV 1 b17 RET 3 SIZE 1b18RET 3 LIT1 b19 NEG 3 RET 3 X 1b20 NEG 3 RET 3 MB 1 b21 NEG 3 RET 3 LEV 1 b22 NEG 3 RET 3 SIZE 1 b23 NEG 3 RET 3 LIT 1 b24 IND 1 b25 NEG 3 IND 1 b26 RET 3 IND 1 b27 NEG 3 RET 3 IND 1 e. The variables are defined in Table 2. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias. The sample period includes the post-SOX fiscal years 2004 through 2009. The total firm observations are 3,492 dividing into 1,746 ICW and 1,746 non-ICW observations over the 6-year time period. Out of 1,746 ICW firm observations, 756 relate to company-level and 990 relate to account-specific ICW.

171

172
4 Company level vs. account specific X = Company level Coefficient 2.065 2.020 2.075 .005 .038 .036 .049 2.011 2.131 .014 2.018 2.009 .010 .002 2.019 2.065 .095 .040 2.019 2.053 .036 .819 .000*** .989 .050** .197 .077* .545 .000*** .793 .552 .946 .396 .233 .080* .021** .000*** .131 .682 .044* .042** .149 .028** .409 .071* .084 .054* .316 .000*** .311 .303 .495 .618 .920 .241 .039** .005*** .064** .134 .048** 2.075 2.011 2.105 .001 .048 .025 .039 2.006 2.134 .005 2.013 2.001 .018 .012 2.035 2.080 .118 .026 2.004 2.060 p value Coefficient p value X = Company level Company-level vs. non-ICW 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific Coefficient 2.064 2.008 2.090 .019 .058 .028 .070 2.001 2.144 .010 2.025 2.006 .022 .010 2.028 2.069 .086 .020 2.025 2.009 p value .050 .535 .010*** .115 .044 .215 .032 .992 .000*** .335 .122 .819 .167 .418 .110 .028** .020** .179 .108 .552 (continued)

Table 5. Persistence of Earnings Changes Measure of Accounting Conservatism (Post-SOX Period: 2004 to 2009).

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

Intercept NEG DNIt – 1 X MB LEV SIZE LIT NEG 3 DNIt – 1 NEG 3 X NEG 3 MB NEG 3 LEV NEG 3 SIZE NEG 3 LIT DNIt – 1 3 X DNIt – 1 3 MB DNIt – 1 3 LEV DNIt – 1 3 SIZE DNIt – 1 3 LIT NEG 3 DNIt – 1 3 X

2.077 2.004 2.083 .016 .045 .021 .061 2.008 2.139 .008 2.020 2.004 .019 .006 2.025 2.074 .103 .032 2.011 2.040

.030** .778 .018** .361 .055* .246 .041** .403 .000*** .510 .169 .715 .202 .861 .164 .033** .000*** .098* .158 .069*

Table 5. (continued) 4 Company level vs. account specific X = Company level Coefficient .024 .190 2.096 .011** .018 .292 .005 .498 Included .221 1,746 Company level: 756 Account specific: 990 .039 .116 2.092 .019*** .039 .075* .018 .203 Included .216 1,512 Company level: 756 Non-ICW: 756 p value Coefficient p value X = Company level Company-level vs. non-ICW 5 6 7 8 9

1

2

3

ICW vs. non-ICW

Account-specific vs. non-ICW X = Account specific Coefficient p value .028 .145 2.083 .028** .044 .059* .003 .606 Included .223 1,980 Account specific: 990 Non-ICW: 990

X = ICW

Variables

Coefficient

p value

NEG 3 DNIt – 1 NEG 3 DNIt – 1 NEG 3 DNIt – 1 NEG 3 DNIt – 1 Industry effects Adjusted R2 n

3 3 3 3

MB LEV SIZE LIT

.032 .086** 2.088 .024** .040 .068* .011 .310 Included .218 3,492 ICW: 1,746 Non-ICW: 1,746

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses. The table reports the regression results from estimating the following expanded Basu (1997) regression model (e.g., Goh and Li 2011): DNIt = b0 1 b1 NEG 1 b2 DNIt – 1 1 b3 X 1 b4 MB 1 b5 LEV 1 b6 SIZE 1 b7 LIT 1 b8 NEG 3 DNIt – 1 1 b9 NEG 3 X 1 b10 NEG 3 MB 1 b11NEG 3 LEV 1 b12NEG 3 SIZE 1 b13NEG 3 LIT 1b14 DNIt – 1 3 X 1b15 DNIt – 1 3 MB 1 b16 DNIt – 1 3 LEV 1 b17 DNIt – 1 3 SIZE 1 b18 DNIt – 1 3 LIT 1 b19 NEG 3 DNIt – 1 3 X 1 b20 NEG 3 DNIt – 1 3 MB 1 b21 NEG 3 DNIt – 1 3 LEV 1 b22 NEG 3 DNIt – 1 3 SIZE 1 b23 NEG 3 DNIt – 1 3 LIT 1 b24 IND 1 b25 NEG 3 IND 1 b26 DNIt – 1 3 IND 1 b27 NEG 3 DNIt – 1 3 IND 1 e. The variables are defined in Table 2. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias. The sample period includes the post-SOX fiscal years 2004 through 2009. The total firm observations are 3,492 dividing into 1,746 ICW and 1,746 non-ICW observations over the 6-year time period. Out of 1,746 ICW firm observations, 756 relate to company-level and 990 relate to account-specific ICW.

173

174

Journal of Accounting, Auditing & Finance

show that the coefficient b19 is significantly negative (–.040; p = .069), implying that the reversal of bad news recognition from the previous period for the ICW firms is significantly greater than for the non-ICW firms. The result supports H1, indicating that the ICW firms are more conservative in reporting current-period earnings than are the non-ICW firms. Columns 4 and 5 show that the coefficient b19 is significantly negative (–.053; p = .048), implying that the reversal of bad news recognition from the previous period for the firms with company-level ICW is significantly greater than for the firms with accountspecific ICW. Columns 6 and 7 show that the coefficient b19 is significantly negative (–.060; p = .044), implying the reversal of bad news recognition from the previous period is significantly greater for the company-level ICW firms than for the non-ICW firms. These results support the predictions of H2. Finally, columns 8 and 9 show that the coefficient b19 is insignificant (–.009; p = .552), suggesting that there is no difference in the reversal of bad news recognition in earnings from the previous period between the account-specific ICW firms and the non-ICW firms. The analyses show that the ICW firms in general and the firms with company-level ICW in particular exhibit greater post-SOX accounting conservatism, as indicated by their significantly larger tendency to have negative earnings reversal in the following fiscal period.

Accrual-Based Loss Recognition Measure
Table 6 reports the results from estimating the regression equation (Equation 6) with respect to accrual-based loss recognition measure. In general, we find a significantly negative relationship between ACC and CFO, which is consistent with the negative relationship between accruals and cash flows, and a significantly positive relationship between ACC and CFO 3 NEG_CFO, which indicates an asymmetric timely loss recognition by accruals, a measure of accounting conservatism. Columns 2 and 3 show that the coefficient b7 is significantly positive (.055; p = .046), implying that the loss recognition by accruals is more timely for the ICW firms than for the non-ICW firms. The result supports the predictions of H1. Columns 4 and 5 show that the coefficient b7 is significantly positive (.038; p = .085), implying that the loss recognition by accruals is more timely for the firms with company-level ICW than for the firms with account-specific ICW. Columns 6 and 7 show that the coefficient b7 is significantly positive (.059; p = .040), implying that the loss recognition by accruals is more timely for the company-level ICW firms than for the non-ICW firms. These results together reject the predictions of H2. Finally, columns 8 and 9 show that the coefficient b7 is insignificant (.019; p = .266), indicating that there is no difference between the account-specific ICW firms and the non-ICW firms with respect to asymmetric loss recognition by their accruals.

Summary of Main Results
Our analyses demonstrate that in the post-SOX period, the firms with ineffective internal controls exhibit more accounting conservatism compared with the firms with effective internal controls in terms of the three conditional conservatism measures. This greater conservatism of the ICW firms is, however, mainly attributed to the firms with company-level ICW. No difference in conservatism is found between the firms with account-specific ICW and the non-ICW firms. The results, in general, suggest that the firms with severe internal control problems (with consequent greater agency costs and governance problems) are

Table 6. Accrual-Based Loss Recognition Measure of Accounting Conservatism (Post-SOX Period: 2004 to 2009). 4 Company level vs. account specific X = Company level Coefficient .038 .005 2.329 .296 2.006 .015 2.046 .038 .115 2.006 .011 2.036 .054 2.018 2.116 .110 .089 2.003 .001 .132 .095* .804 .000*** .000*** .510 .289 .053* .085* .000*** .492 .391 .077* .050* .195 .000*** .000*** .019** .818 .945 .006*** .049 .019 2.288 .219 2.012 .019 2.036 .059 .094 2.015 .025 2.026 .062 2.014 2.129 .080 .070 2.022 .011 .120 .068* .285 .000*** .000*** .291 .128 .088* .040** .006*** .309 .108 .155 .045** .201 .000*** .022** .028** .119 .295 .000*** p value Coefficient p value X = Company level Company-level vs. non-ICW 5 6 7 8 9

1

2

3

ICW vs. non-ICW

Account-specific vs. non-ICW X = Account specific Coefficient .043 .015 2.317 .245 2.014 .025 2.010 .019 .089 2.018 .020 2.042 .075 2.009 2.148 .075 .063 2.029 .018 .114 p value .074* .319 .000*** .000*** .218 .124 .171 .266 .012** .225 .119 .069* .032** .285 .000*** .025** .040** .105 .122 .000*** (continued)

X = ICW

Variables

Coefficient

p value

Intercept NEG CFO NEG 3 CFO X NEG 3 X CFO 3 X NEG 3 CFO 3 X MB NEG 3 MB CFO 3 MB NEG 3 CFO 3 MB LEV NEG 3 LEV CFO 3 LEV NEG 3 CFO 3 LEV SIZE NEG 3 SIZE CFO 3 SIZE NEG 3 CFO 3 SIZE

.059 .008 2.352 .279 2.029 .036 2.038 .055 .104 2.010 .018 2.029 .066 2.010 2.139 .088 .077 2.015 .010 .125

.050** .663 .000*** .000*** .176 .085** .081* .046** .000*** .309 .223 .104 .040** .190 .000*** .015** .030** .289 .406 .000***

175

176
4 Company level vs. account specific X = Company level Coefficient 2.026 .144 .018 .166 2.055 .050** .006 .810 Included .316 1,746 Company level: 756 Account specific: 990 2.005 .619 .022 .129 2.069 .038** .015 .222 Included .330 1,512 Company level: 756 Non-ICW: 756 p value Coefficient p value Coefficient X = Company level Company-level vs. non-ICW 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value 2.010 .804 .002 .905 2.028 .112 .001 .958 Included .338 1,980 Account specific: 990 Non-ICW: 990

Table 6. (continued)

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

LIT NEG 3 LIT CFO 3 LIT NEG 3 CFO 3 LIT Industry effects Adjusted R2 n

2.016 .278 .011 .229 2.059 .049** .004 .596 Included .322 3,492 ICW: 1,746 Non-ICW: 1,746

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses. The table reports regression results from estimating the following expanded Ball and Shivakumar (2005) model (e.g., Goh and Li 2011): ACC = b0 1 b1 NEG 1 b2 CFO 1 b3 NEG 3 CFO 1 b4 X 1 b5 NEG 3 X 1 b6 CFO 3 X 1 b7 NEG 3 CFO 3 X 1 b8 MB 1 b9 NEG 3 MB 1 b10 CFO 3 MB 1 b11 NEG 3 CFO 3 MB 1 b12 LEV 1 b13 NEG 3 LEV 1 b14 CFO 3 LEV 1 b15 NEG 3 CFO 3 LEV 1 b16 SIZE 1 b17 NEG 3 SIZE 1 b18 CFO 3 SIZE 1 b19 NEG 3 CFO 3 SIZE 1 b20 LIT 1 b21 NEG 3 LIT 1 b22 CFO 3 LIT 1 b23 NEG 3 CFO 3 LIT 1 b24 IND 1 b25 NEG 3 IND 1 b26 CFO 3 IND 1 b27 NEG 3 CFO 3 IND 1 e . The variables are defined in Table 2. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias. The sample period includes the post-SOX fiscal years 2004 through 2009. The total firm observations are 3,492 dividing into 1,746 ICW and 1,746 non-ICW observations over the 6-year time period. Out of 1,746 ICW firm observations, 756 relate to company-level and 990 relate to account-specific ICW.

Mitra et al.

177

more inclined to change their reporting strategies in the changed regulatory and oversight environment, and exhibit more accounting conservatism.

Additional Analyses Test of Relative Conservatism for the 2004-2006 Period Versus the 2007-2008 Period
In this section, we report the test results separately for earlier and latter part of the postSOX sample period. The objective is to evaluate the relative conditional conservatism between the ICW firms and non-ICW firms in a time 2004-2006 period immediately after the implementation of various SOX provisions when the firms face a dramatic change in corporate oversight and regulations, and in a 2007-2009 period when SOX regulations have been in place for quite some time, and both the ICW firms and non-ICW firms had the opportunity to make necessary adjustments to their reporting strategies. Table 7 reports the partial test results for the variables of interest obtained from estimating regression equations (Equations 2, 4, and 6). Panels A and B of Table 7 present the respective results for the timeliness of earnings to news measure, Panels C and D of Table 7 present the respective results for the persistence of earnings changes measure, and Panels E and F of Table 7 present the respective results for the accruals-based loss recognition measure. The results show an interesting trend. The ICW firms, especially the firms with company-level control weaknesses, exhibit significantly higher conservatism than the non-ICW firms both in the earlier and in the latter part of the SOX period. However, the results are stronger in the early part of the post-SOX period, that is, 2004-2006, than in the latter part of the post-SOX period, that is, 20072009. The results show that the difference in conservatism is moderated during the later post-SOX period. It is noteworthy that the non-ICW firms, in any case, do not exhibit more conservatism than the ICW firms during the two post-SOX periods. We interpret the results to suggest that when firms with serious control problems are subject to significant changes in accounting and audit regulations and corporate oversight in the post-SOX period, they are more inclined to adopt greater conservatism than the firms with no such problems to minimize the potentially negative effect of SOX. They are also likely to be under greater pressure from contracting parties to reduce their aggressive reporting, especially in the initial years when firm-specific information starts flowing to outsiders in greater detail. The non-ICW firms may also become more conservative, but not to the extent of the ICW firms, because the demand for conservatism would not be as strong for those firms. Over time, the intensity of such demand for conservatism may subside as enhanced regulations and oversight make the reporting process more transparent and reliable even for the firms with control weaknesses. Moreover, the non-ICW firms may also make necessary adjustments to their reporting policies over the years as a part of their response to the changing environment. As a result, the difference between reporting conservatism has declined in the latter part of the post-SOX period.

Effect of ICW Remediation on Accounting Conservatism in the Post-SOX Period
Goh and Li (2011) find that the material weakness firms that remediate their ICW as reported in their second SOX 404 report exhibit greater accounting conservatism than the material weakness firms that continue to have control weaknesses at the time of the second (text continues on p. 182)

Table 7. Regression Results for Earlier (2004-2006) versus Latter (2007-2009) Part of SOX Period.

178
3 Company level vs. account specific Company-level vs. non-ICW X = Company level p value 2.026 .104 .135 .000*** .139 .000*** 2.005 .606 2.011 .302 2.049 .052* .060 .040** 1,059 Company level: 440 Account specific: 619 2.011 .295 .128 .000*** .152 .000*** 2.012 .172 2.008 .559 2.073 .031** .066 .035** 880 Company level: 440 Non-ICW: 440 Coefficient p value Coefficient p value Coefficient X = Company level 4 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value 2.016 .402 .110 .000*** .122 .000*** 2.019 .115 2.003 .822 2.025 .104 .015 .129 1,238 Account specific: 619 Non-ICW: 619 4 Company level vs. account specific X = Company level Coefficient 2.002 .810 .133 .000*** .155 .000*** 2.011 .392 2.004 .826 2.018 .195 .019 .144 687 Company level: 316 Account specific: 371 p value 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value 2.015 .209 .149 .000*** .159 .000*** 2.025 .121 2.010 .210 2.006 .332 .042 .060* 632 Company level: 316 Non-ICW: 316 Coefficient p value 2.020 .116 .142 .000*** .136 .000*** 2.008 .514 2.005 .682 2.011 .233 .009 .588 742 Account specific: 371 Non-ICW: 371
(continued)

Panel A: Timeliness of Earnings to News (for the Period 2004 to 2006).

1

2

ICW vs. non-ICW

X = ICW

Variables

Coefficient

NEG RET NEG 3 RET X NEG 3 X RET 3 X NEG 3 RET 3 X n

2.022 .125 .121 .000*** .140 .000*** 2.010 .219 2.004 .781 2.069 .038** .059 .046** 2,118 ICW: 1,059 Non-ICW: 1,059

Panel B: Timeliness of Earnings to News (for the Period 2007 to 2009).

1

2

3

ICW vs. non-ICW

Company-level vs. non-ICW X = Company level Coefficient

X = ICW

Variables

Coefficient

p value

NEG RET NEG 3 RET X NEG 3 X RET 3 X NEG 3 RET 3 X n

2.005 .724 .138 .000*** .148 .000*** 2.019 .168 2.006 .451 2.014 .211 .040 .065* 1,374 ICW: 687 Non-ICW: 687

Table 7. (continued)

Panel C: Persistence of Earnings Changes (for the Period 2004 to 2006). 3 4 5 6 7 8 9

1

2

ICW vs. non-ICW X = Company level p value Coefficient p value Coefficient p value Coefficient X = Company level

Company level vs. account specific Company-level vs. non-ICW Account-specific vs. non-ICW X = Account specific p value

X = ICW

Variables

Coefficient

– 1

NEG DNIt – 1 NEG 3 DNIt X NEG 3 X DNIt – 1 3 X NEG 3 DNIt n

– 1

3X

2.019 .138 2.077 .025** 2.128 .000*** .008 .363 .015 .144 2.036 .080* 2.046 .055* 2,118 ICW: 1,059 Non-ICW: 1,059

2.025 .105 2.092 .004*** 2.119 .000*** .003 .782 .010 .308 2.027 .112 2.065 .042** 1,059 Company level: 440 Account specific: 619

2.018 .120 2.082 .016** 2.136 .000*** .001 .977 .019 .171 2.039 .076* 2.059 .045** 880 Company level: 440 Non-ICW: 440

2.012 .228 2.088 .012** 2.145 .000*** .009 .205 .012 .605 2.033 .095* .018 .179 1,238 Account specific: 619 Non-ICW: 619

Panel D: Persistence of Earnings Changes (for the Period 2007 to 2009). 4 Company level vs. account specific X = Company level Coefficient p value 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value Coefficient p value

1

2

3

ICW vs. non-ICW

Company-level vs. non-ICW X = Company level Coefficient

X = ICW

Variables

Coefficient

p value

– 1

NEG DNIt – 1 NEG 3 DNIt X NEG 3 X DNIt – 1 3 X NEG 3 DNIt n

– 1

3X

2.030 .092* 2.090 .006*** 2.110 .000*** .015 .218 .009 .389 2.021 .156 2.038 .072* 1,374 ICW: 687 Non-ICW: 687 2.016 .201 2.085 .011** 2.102 .000*** .019 .120 .002 .904 2.013 .285 2.024 .161 687 Company level: 316 Account specific: 371

2.025 .177 2.101 .000*** 2.116 .000*** .006 .704 .008 .645 2.005 .691 2.033 .090* 632 Company level: 316 Non-ICW: 316

2.038 .084* 2.093 .004*** 2.105 .000*** .012 .193 .010 .332 2.020 .186 2.002 .955 742 Account specific: 371 Non-ICW: 371
(continued)

179

180
4 Company level vs. account specific X = Company level Coefficient .018 .190 2.355 .000*** .229 .000*** 2.017 .235 .020 .189 2.028 .125 .039 .089* 1,059 Company level: 440 Account specific: 619 .005 .185 2.319 .000*** .196 .000*** 2.010 .331 .040 .069* 2.030 .094* .070 .029** 880 Company level: 440 Non-ICW: 440 p value Coefficient p value Coefficient X = Company level Company-level vs. non-ICW 5 6 7 8 9 Account-specific vs. non-ICW X = Account specific p value .015 .269 2.348 .000*** .208 .000*** 2.008 .344 .010 .212 2.026 .119 .017 .220 1,238 Account specific: 619 Non-ICW: 619 4 Company level vs. account specific X = Company level Coefficient .025 2.298 p value .130 .000*** 5 6 7 Company-level vs. non-ICW X = Company level Coefficient .011 2.267 p value .318 .000*** 8 9 Account-specific vs. non-ICW X = Account specific Coefficient .004 2.311 p value .771 .000*** (continued)

Table 7. (continued)

Panel E: Accruals-Based Loss Recognition (for the Period 2004 to 2006).

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

NEG CFO NEG 3 CFO X NEG 3 X CFO 3 X NEG 3 CFO 3 X n

.011 .318 2.412 .000*** .210 .000*** 2.008 .592 .044 .060* 2.045 .059** .068 .033** 2,118 ICW: 1,059 Non-ICW: 1,059

Table 7. (continued)

Panel F: Accruals-Based Loss Recognition (for the Period 2007 to 2009).

1

2

3

ICW vs. non-ICW

X = ICW

Variables

Coefficient

p value

NEG CFO

.019 2.390

.209 .000***

Table 7. (continued) Panel F: Accruals-Based Loss Recognition (for the Period 2007 to 2009). 4 Company level vs. account specific X = Company level Coefficient .000*** .948 .496 .055* .119 .279 2.016 .003 2.042 .046 632 Company level: 316 Non-ICW: 316 .000*** .190 .818 .068** .052* p value Coefficient p value X = Company level Company-level vs. non-ICW 5 6 7 8 9

1

2

3

ICW vs. non-ICW

Account-specific vs. non-ICW X = Account specific Coefficient p value .000*** .209 .110 .040** .182

X = ICW

Variables

Coefficient

p value

NEG 3 CFO .366 X 2.030 NEG 3 X .018 CFO 3 X 2.035 NEG 3 CFO 3 X .020 n 1,374 ICW: 687 Non-ICW: 687

.326 2.001 .005 .048 .026 687 Company level: 316 Account specific: 371

.000*** .094* .179 .082* .146

.252 2.014 .029 2.063 .020 742 Account specific: 371 Non-ICW: 371

Note: SOX = Sarbanes–Oxley Act of 2002; ICW = internal control weaknesses. The regression equations (Equations 2, 4, and 6) are estimated separately for the earlier part and latter part of the SOX period. The results for the variables on interest are reported for the sake of brevity. The tests are performed on a matched-pair sample obtained from a propensity score matching process that mitigates any potential self-selection bias. The earlier part of the SOX period includes the fiscal years 2004 through 2006, which has 2,118 firm observations divided between 1,059 ICW and 1,059 non-ICW observations. Out of 1,059 ICW observations, 440 relate to company-level and 619 relate to account-specific ICW. The latter part of the SOX period includes the fiscal years 2007 through 2009, which has 1,374 firm observations divided between 687 ICW and 687 non-ICW observations. Out of 687 ICW observations, 316 relate to company-level and 371 relate to account-specific ICW.

181

182

Journal of Accounting, Auditing & Finance

SOX 404 report. As a supplemental analysis, we examine whether remediation of the control weaknesses during the post-SOX period has any further effect on the ICW firms’ accounting conservatism when these firms have already made significant changes in their reporting practices in response to the enhanced oversight and regulatory environment, and exhibited greater accounting conservatism than the non-ICW firms. We identify 945 firms that remediate their control weaknesses in subsequent years after their ICW reporting and do not have recurrence of similar or different control problems later within the sample period (2004-2009).18 Out of 945 ICW remediation firms, 327 relate to company-level ICW and 618 relate to account-specific ICW. We evaluate the change, if any, in conservative reporting practices of those firms upon ICW remediation by comparing their conditional conservatism in the year before and in the year when the remediation takes place. Thus, we have 1,890 firm observations (945 3 2) for the tests divided between 654 observations (327 3 2) for the company-level ICW and 1,236 observations (618 3 2) for the account-specific ICW. We test the effect of ICW remediation action for the two groups having company-level and account-specific ICW by applying Equations 2, 4, and 6, which include the interactions of the respective conservatism measures with the variable REMEDIATE. REMEDIATE is a dummy variable of 1 for the firm observations in the ICW remediation year and 0 for the firm observations in the year before ICW remediation. The interactions of REMEDIATE with RET 3 NEG, that is, RET 3 NEG 3 REMEDIATE in Equation 2, with NEG 3 DNIt – 1, that is, NEG 3 DNIt – 1 3 REMEDIATE in Equation 4 and with CFO 3 NEG_CFO, that is, CFO 3 NEG_CFO 3 REMEDIATE in Equation 6 reflect the change in accounting conservatism of the firms upon the remediation of ICW. The results are reported for the variables of interest in Panels A and B of Table 8. Panel A of Table 8 reports the results on the difference in conservatism of the companylevel ICW firms in the year before remediation and in the remediation year. We find no evidence that the remediation of company-level ICW makes a difference in accounting conservatism. The coefficients of the interaction variables of interest, RET 3 NEG 3 REMEDIATE, NEG 3 DNIt – 1 3 REMEDIATE, and CFO 3 NEG_CFO 3 REMEDIATE, are all insignificant. Panel B of Table 8 reports the remediation results for the accountspecific ICW firms. The results are qualitatively similar in that there is no difference in their conservatism in the year before and in the remediation year. We further repeat our analyses by examining conservatism for the company-level and account-specific ICW firms in the year before and in the year after the remediation year. The results (not tabulated here) again show insignificant difference in their accounting conservatism between the pre- and postremediation periods.

Conclusion
Goh and Li (2011) conclude that there is a positive relationship between internal control quality and conservatism, and that the strong internal controls act as a mechanism that facilitates conservatism. Their findings show that the firms having material control weaknesses exhibit lower conservatism than the firms having effective internal controls. Their study, however, is based on data relating to the pre-SOX period. In this study, we examine the effect of enhanced regulations and corporate oversight in the post-SOX period on the association between internal control quality and accounting conservatism for the firms with or without effective internal controls. Our findings show that the ICW firms have lower conservatism than the non-ICW firms in the pre-SOX years, as argued by Goh and Li. But

Mitra et al.
Table 8. Effect of ICW Remediation on Accounting Conservatism. Panel A: Firms With Company-Level ICW (n = 654). Asymmetric loss recognition Variables REMEDIATE NEG 3 REMEDIATE RET 3 REMEDIATE RET 3 NEG 3 REMEDIATE REMEDIATE NEG 3 REMEDIATE DNIt – 1 3 REMEDIATE NEG 3 DNIt – 1 3 REMEDIATE REMEDIATE NEG_CFO 3 REMEDIATE CFO 3 REMEDIATE CFO 3 NEG_CFO 3 REMEDIATE Persistence of earnings changes

183

Accrual-based loss recognition

Coefficient p value Coefficient p value Coefficient p value .024 2.010 .008 .029 .119 .404 .815 .109 2.043 .005 .019 2.012 .066* .910 .225 .514 .050 2.011 .038 .006 .048** .612 .089 .895

Panel B: Firms With Account-Specific ICW (n = 1,236). Asymmetric loss recognition Variables REMEDIATE NEG 3 REMEDIATE RET 3 REMEDIATE RET 3 NEG 3 REMEDIATE REMEDIATE NEG 3 REMEDIATE DNIt – 1 3 REMEDIATE NEG 3 DNIt – 1 3 REMEDIATE REMEDIATE NEG_CFO 3 REMEDIATE CFO 3 REMEDIATE CFO 3 NEG_CFO 3 REMEDIATE Persistence of earnings changes Accrual-based loss recognition

Coefficient p value Coefficient p value Coefficient p value .003 .010 2.036 .015 .958 .314 .085* .231 .022 .007 2.028 2.022 .190 .811 .105 .126 .011 2.005 2.045 .010 .262 .816 .052* .285

Note: ICW = internal control weaknesses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-tailed tests using the heteroscedasticity-adjusted White’s (1980) t statistic. The tables present regression results with REMEDIATE and its interactions with conservatism measures as the variables of interest. REMEDIATE is a categorical variable of 1 for the firm observations associated with the remediation of ICW and 0 otherwise. For the company-level ICW firms, the comparative analyses are made using 327 pre-remediation and 327 postremediation firm observations, and for the account-specific ICW firms, the comparative analyses are made using 618 pre-remediation and 618 post-remediation firm observations. Though the full models are estimated, the results for the variables of interest are reported for the sake of brevity.

the relationship between accounting conservatism and internal control quality significantly changes in the post-SOX period of enhanced regulations and corporate oversight. The ICW firms exhibit greater conservatism than the non-ICW firms in the post-SOX period. Our results are more robust for the firms having broader and pervasive company-level ICW

184

Journal of Accounting, Auditing & Finance

than for the firms having account-specific localized ICW. We attribute this change in the reporting practices to enhanced regulations, increased corporate scrutiny and oversight, and greater awareness of various contracting parties, leading to a larger demand for conservatism in the firms suffering from ICW and the associated governance and agency problems. We do not find any difference in conservatism between the firms having account-specific control weaknesses that are more localized in nature and the firms having effective internal controls. Additional analyses show that the difference in conservatism between the ICW firms and non-ICW firms is larger and more significant in the earlier part than in the latter part of the post-SOX sample period. We conjecture that in the earlier post-SOX period, the reaction of the ICW firms might be stronger with immediate effect on their conservative reporting policy in view of enhanced regulations and oversight, and dramatic changes in governance mechanisms. The firms with higher control and related agency problems are expected to be more affected by a changed environment than the firms with few or no problems in that respect. The enhanced regulatory oversight and scrutiny pursuant to SOX are most likely to pressure those firms to respond to greater need for conservative information and ensure efficient contracting. However, over time, both the ICW and non-ICW firms may have made necessary adjustments to their reporting practices and the demand for conservatism may also change due to greater flow of firm-specific information to outsiders. As a result, the difference in conservatism between those firms is attenuated during the latter part of the SOX period. However, in any case, the non-ICW firms do not exhibit greater conservatism than the ICW firms. The analyses further show that the remediation of ICW in the years subsequent to the ICW reporting has insignificant effect on accounting conservatism of the ICW firms, probably because of their prior action to become more conservative in response to the changed SOX environment. The caveat is that though conservatism plays an important role in resolving agency conflicts and improving contracting efficiency, it does not necessarily imply higher financial reporting quality. Prior studies find that firms with more conservative accounting have less reliable accruals (Ashbaugh-Skaife et al., 2008); conservatism leads to doubtful earnings quality (Penman & Zhang, 2002); and earnings management and conservatism can coexist within a firm (Givoly et al., 2010). Our results should thus be interpreted keeping in mind that the ICW firms, though exhibiting higher post-SOX conservatism, may still continue to make greater use of their reporting flexibility and manage financial information in convenient forms (e.g., accruals vs. real activities) to meet their reporting thresholds.

Appendix Classification of Internal Control Weaknesses (ICW)
We consider the criteria applied by Doyle et al. (2007a, 2007b) and Raghunandan and Rama (2006) to classify ICW reported under SOX 404 into company-level and accountspecific ICW. Company-level weaknesses are less auditable and more pervasive at the company level, whereas account-specific weaknesses are more auditable and relate to specific accounts and/or transactions. We read both management’s and auditor’s reports on internal control over financial reporting (ICFR) to develop a better sense of the nature of the weaknesses so that they could properly be classified as company-level and accountspecific ICW as far as practicable.

Mitra et al. Examples of company-level ICW are as follows: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

185

Quality and training of accounting personnel Segregation of duties Reconciliation of accounts and financial statement preparation Information systems–related problems Quality of internal audit or audit committee Inconsistencies in the application of company policies among business units and segments Material weaknesses in the interpretation and application of complex accounting standards, such as standards related to hedge transactions Weak internal controls related to contracting parties Deficiencies related to the design of policies and execution of processes relating to accounting for transactions Deficiencies in the period-end reporting process

In addition to the above, we consider the following two additional broad criteria to classify ICW as systematic and having company-wide pervasive effect: 1. Override by senior management 2. Ineffective control environment However, disclosures of ICW are not very clear-cut for many sample firms. Hence, in addition to evaluating whether ICW relate to one or more of the above areas, we consider a cutoff point of at least three account-specific internal control problems to classify a firm as having company-level ICW (see Doyle et al., 2007b). Account-specific ICW relate to specific accounts or transactions, namely, 1. Inadequate controls for income tax accounting including deferred income taxes with proper reconciliations between book and tax income 2. Inadequate controls for accounting for loss contingencies 3. Inadequate controls for accounting for receivables including bad debts 4. Revenue recognition problems 5. Deficiencies in the documentation of a receivables securitization program 6. Inadequate internal controls over the application of new accounting principles or existing accounting principles to new transactions If the number of control problems is restricted to one or two without any control weaknesses at the company level, we classify a firm as having account-specific ICW. However, if a firm has both account-specific and company-level ICW like segregation of duties or override by senior management as reported by the external auditor, it is classified as having company-level ICW. So when a firm has both company-level and account-specific control weaknesses, or has ICW at the company level, it is considered to have companylevel ICW. The remaining ICW firms are classified as having account-specific control weaknesses. Authors’ Note
Data used in the analyses are obtained from public sources as described in the text.

186

Journal of Accounting, Auditing & Finance

Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The authors received no financial support for the research, authorship, and/or publication of this article.

Notes 1. Conservatism may be conditional or unconditional. In this article, by accounting conservatism,
we refer to conditional conservatism, which requires a higher degree of verification to recognize good news as gains than to recognize bad news as losses in earnings. Conditional conservatism dictates that earnings reflect bad news more quickly than good news, leading to asymmetrically timely loss recognition (e.g., Ball & Shivakumar, 2005; Basu, 1997; Beaver & Ryan, 2005). This asymmetric variability requirement is deemed to mitigate agency problems associated with aggressive financial reporting. Unconditional conservatism is likely to be news independent or ex-ante where firms report expenses early or defer revenues (Beaver & Ryan, 2005). This is mostly induced by the Generally Accepted Accounting Principles (GAAP) rules and regulations where conservatism means reporting the lowest value among the possible alternative values for assets and the highest alternative value for liabilities. Our study and the related literature examine conditional conservatism, which is more discretionary in nature and is a part of the firm’s reporting strategy to asymmetrically recognize bad news more quickly than good news in response to losses and gains. Penman and Zhang (2002) contend that conservative accounting can create more unrecorded reserves that provide managers with the flexibility to report more income in the future. Levitt (1998) suggests that three out of five financial reporting problems deal with understatement of assets, for example, big bath reserves, creative acquisition accounting, and cookie jar reserves, which is part of the conservative accounting process. Though it is hard to identify any specific factor(s) that are more responsible than others to influence managers’ reporting decisions because the post–Sarbanes–Oxley Act of 2002 (SOX) environment encompasses so many simultaneous changes in regulations, corporate oversight, and governance, we expect that the combined effect of all the post-SOX regulatory and governance changes, the resulting public awareness, and expectation of a higher quality of reported information would largely impact the firms’ conservative reporting decisions. We conjecture that the non–internal control weaknesses (ICW) firms have effective control systems and thus relatively smaller agency and governance problems than the ICW firms. So the impact of SOX regulations and resulting changed oversight environment would have relatively less impact on the conservative reporting practices of those firms. If accounting conservatism is viewed as a reporting strategy to reduce agency problems and promote contracting efficiency, the ICW firms are expected to adopt more conservatism than the non-ICW firms when faced with higher regulations, corporate oversight, and governance. Because ineffective internal controls potentially increase the reporting risk, exacerbate agency problems in financial reporting, and reduce contracting efficiency, prior studies show that improved internal controls reduce uncertainty surrounding the financial reporting process. Doyle, Ge, and McVay (2007a, 2007b) and Ashbaugh-Skaife, Collins, Kinney, and LaFond (2008) document a positive relationship between internal control quality and accruals quality as measured by the extent to which accruals are realized in cash flows or by the reduction of abnormal accruals. Furthermore, two types of ICW have been identified in the literature, that is, company-level and account-specific ICW (e.g., Doyle et al., 2007a, 2007b; Raghunandan & Rama, 2006).

2.

3.

4.

5.

Mitra et al.

187

6. 7.

8.

9.

10.

11.

12.

13.

Company-level ICW are less auditable and more pervasive in effect, whereas account-specific ICW are relatively more identifiable and auditable and more localized in effect. By virtue of their broader and more pervasive negative effect on financial reporting, company-level ICW are more likely to have a stronger impact on accounting conservatism than account-specific ICW. For more discussion on and examples of the types of the ICW, see Raghunandan and Rama (2006, pp. 109-110) and Appendix A of Doyle et al. (2007a). Fitch Ratings (2005) classifies certain control weaknesses as pervasive/systematic. It notes, ‘‘Certain material weaknesses might constitute pervasive risk, such as problems with ‘tone at the top’ or the quality of personnel in charge of the financial reporting functions. . . . Rating for companies not previously identified as having such pervasive weaknesses by Fitch will need to be looked at carefully, and negative action is likely. . . . Material weaknesses in internal controls can also occur at the transaction level, potentially affecting information such as specific account balances (p. 3).’’ Similarly, Moody’s Investors Service (2004) states that material weaknesses will be classified as Type A for account/transaction weaknesses and Type B for systematic weaknesses, and their impact on bond ratings will be different. The adverse consequences are more likely to follow for the Type B ICW. For example, Watts (2003a) suggests that in debt, compensation, and other contract explanations, conservatism emerges almost naturally as an efficient contracting mechanism because it is optimal for contracts’ measures to have more stringent verification standards for gains than for losses. Kellog (1984) suggests that reduction of aggressive reporting through more stringent verification standards for gains compared with losses makes conservatism an efficient contracting mechanism. In response to a survey conducted by the Office of Economic Analysis of the U.S. Securities and Exchange Commission (SEC) in 2009, the participating company executives indicated the following benefits derived from the implementation of Section 404: (a) quality of internal control structure, (b) audit committee’s confidence in the company’s internal control over financial reporting (ICFR), (c) quality of company’s financial reporting, (d) company’s ability to prevent and detect fraud, and (e) confidence in the financial reports of other companies complying with Section 404. Because the company-level ICW firms are deemed to have higher risk in their reporting environment caused by all-pervasive control weaknesses, we expect the effect of SOX to be larger on those firms’ accounting conservatism relative to the account-specific ICW firms and the nonICW firms. For example, in a survey of top executives, Graham, Harvey, and Rajgopal (2005) document that managers prefer real earnings management compared with accruals-based earnings management because real activity manipulations are less likely to be scrutinized by auditors and regulators and are less likely to be detected, although the consequences of such activities can be economically significant to the firms. Consistent with this finding, Cohen, Dey, and Lys (2008) show that the firms have largely switched from accrual-based to real activity-based earnings management strategies after the adoption of SOX, whereas Lobo and Zhou (2006) demonstrate that, in general, management becomes more conservative in financial reporting after SOX. The model is based on the premise that stock price is adjusted on the basis of information obtained from sources other than financial statements. Information content of earnings is impounded in the stock prices before earnings are officially announced, implying that the stock price leads earnings information. Conservatism dictates that recognition of good news requires a higher degree of verification than recognition of bad news. So earnings incorporate bad news as measured by losses more quickly than good news measured by gains. Because accounting conservatism is also impacted by various firm characteristics, we include four control variables and their interactions with the variables of interest in the model. Roychowdhury and Watts (2007) point out that the beginning composition of equity value is determined by both the investment opportunity set of a company and the effect of past

188

Journal of Accounting, Auditing & Finance
asymmetric timeliness in loss recognition. Because both, in turn, affect future asymmetric timeliness of earnings, we include MB as a control variable. LEV controls for debtholders’ demand for accounting conservatism, as prior studies find evidence that debt contracting creates a demand for financial reporting conservatism (e.g., Ball, Robin, & Sadka, 2008; Beatty, Weber, & Yu, 2007; Frankel & Roychowdhury, 2008; Zhang, 2008). SIZE controls for its negative relationship with asymmetric timeliness of earnings (LaFond & Watts, 2008). As greater litigation risk induces managers to recognize bad news earlier than good news (Basu, 1997; Watts, 2003a), we include LIT to control for such effect for the firms operating in a high litigation risk industry as identified by Francis, Philbrick, and Schipper (1994). Finally, following Givoly, Hayn, and Natarajan (2007) that conservatism varies across different industries, we include an industry dummy variable, IND, on the basis of Frankel, Johnson, and Nelson (2002) industry classification. In time-series models like the ones used in our article, there is a possibility that the errors or disturbances ui do not have the same variance across all observations, which means that there is a possibility of the existence of heteroscedasticity, which could result in biased parameter estimation. To address such a potential existence of heteroscedasticity, we use White’s (1980) t statistic, which corrects heteroscedasticity-related standard errors (Greene, 2008). Furthermore, to be doubly sure we also calculate Durbin–Watson Statistics or d statistics to test for autocorrelation, and do not find any d statistics greater than 2, which indicate insignificant correlations among disturbance terms (Verbeek, 2008). We collect information about internal control quality from auditors’ attestation reports under SOX Section 404 because that information provides an unambiguous signal from an independent third party about the effectiveness of internal controls (see Ashbaugh-Skaife et al., 2008, for more discussion). Auditors’ attestation reports for the non-ICW firms indicate that those firms have effective internal controls over financial reporting. Their reports for the ICW firms indicate that those firms have ineffective ICFR. The reports also identify the nature and types of material weaknesses and the number of weaknesses existing in each area. We read the 10-K annual reports, including management’s reports and auditors’ separate ICFR audit reports, to develop a better sense of the nature of weaknesses so that systematic and nonsystematic control weaknesses could be precisely identified as far as possible. Some sample firms remediate their control problems in 1 year but have different types of control problems surface in subsequent years within the sample period. Those firms are still classified as ICW firms for the purpose of the analysis. When there is a possibility of potential endogeneity and self-selection bias, it is recommended that researchers report ordinary least squares (OLS) results based on a matched-pair sample technique using a propensity score matching process (Lennox, Francis, & Wang, 2012). We utilize a propensity score matching methodology to obtain a matched control sample, a methodology that is increasingly being used in the accounting and finance literature (Lawrence, Minutti-Meza, & Zhang, 2011; Armstrong, Jagolinzer, & Larcker, 2010). The propensity score matching technique results in a set of firms that do not have control weaknesses (non-ICW) to obtain a directly and observationally comparable match to ICW firms. By matching on a number of predictive variables, we econometrically mitigate potential selection bias (Iskandar-Datta & Jia, 2013; LaLonde, 1986). The determinant variables for ICW are obtained from Goh and Li (2011), Doyle et al. (2007a, 2007b), and Ashbaugh, Collins, and Kinney (2007). ICW = a dummy variable of 1 for a firm with ICW, 0 otherwise; MVE = log of market value of equity; AGE = firm age measured by the number of years the firm appears in Center for Research in Security Prices (CRSP) database; LOSS = a dummy variable of 1 if the net income before extraordinary items is negative, 0 otherwise; SEGMENT = log of the number of operating and geographic segments; FOREIGN = a dummy variable of 1 if the firm has a nonzero foreign currency translation, 0 otherwise; M&A = a dummy variable of 1 if the firm has a nonzero merger and acquisition activity, 0 otherwise; RESTRUCT = restructuring charges divided by equity market capitalization; EXTR_SALES = a dummy variable of 1 if the year-to-year industry-adjusted sales growth falls into the top quintile,

14.

15.

16.

17.

Mitra et al.

189

0 otherwise; BIG 4 = a dummy variable of 1 if the firm is audited by one of the Big 4 auditors, 0 otherwise; RESTATE = a dummy variable of 1 if the firm has a restatement in the 12-month period and 0 otherwise; AUD_CHANGE = a dummy variable of 1 if there is an auditor change in the 12-month period, 0 otherwise. 18. We exclude the firms that remediate their control problems in 1 year but have different types of control weaknesses surfaced in subsequent years within the sample period. Those firms are still considered as having ICW at the end of the sample period.

References
Ahmed, A. S., Billings, B., Morton, R. M., & Stanford-Harris, M. (2002). The role of accounting conservatism in mitigating bondholder-shareholder conflicts over dividend policy and in reducing debt costs. Accounting Review, 77, 867-890. Ahmed, A. S., & Duellman, S. (2007). Accounting conservatism and board of director characteristics: An empirical analysis. Journal of Accounting & Economics, 43, 411-437. Armstrong, C. S., Jagolinzer, A. D., & Larcker, D. F. (2010). Chief executive officer equity incentives and accounting irregularities. Journal of Accounting Research, 48, 225-271. Ashbaugh, H., Collins, D. W., & Kinney, W. R. (2007). The discovery and reporting of internal control deficiencies prior to SOX-mandated audits. Journal of Accounting and Economics, 44, 166192. Ashbaugh-Skaife, H., Collins, D., Kinney, W., & LaFond, R. (2008). The effect of SOX internal control deficiencies and their remediation on accruals quality. Accounting Review, 83, 217-250. Ball, R., Robin, A., & Sadka, G. (2008). Is financial reporting shaped by equity markets or by debt markets? An international study of timeliness and conservatism. Review of Accounting Studies, 13, 168-205. Ball, R., & Shivakumar, L. (2005). Earnings quality in UK private firms: Comparative loss recognition timeliness. Journal of Accounting & Economics, 39, 83-128. Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting & Economics, 24, 3-37. Beatty, A., Weber, J., & Yu, J. (2007, January). Conservatism and debt (Working paper). Ohio State University and MIT. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=956367 Beaver, W. H., & Ryan, S. G. (2005). Conditional and unconditional conservatism: Concepts and modeling. Review of Accounting Studies, 10, 269-309. Browning, E. S. (2002, July 9). Dow Industrials give back nearly one-third of Friday’s gain (Eastern Edition). The Wall Street Journal, p. C1. Cohen, D., Dey, A., & Lys, T. (2008). Real and accrual-based earnings management in the pre- and post-Sarbanes Oxley periods. Accounting Review, 83, 757-787. Dechow, P. M. (1994). Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals. Journal of Accounting & Economics, 18, 3-42. Dechow, P. M., Kothari, S. P., & Watts, R. L. (1998). The relation between earnings and cash flows. Journal of Accounting & Economics, 25, 133-168. Dhaliwal, D., Huang, S., Khurana, I. K., & Pereira, R. (2008, December). Product market competition and accounting conservatism (Working paper). University of Arizona. Retrieved from http:// papers.ssrn.com/sol3/papers.cfm?abstract_id=1266754 Doyle, J., Ge, W., & McVay, S. (2007a). Accruals quality and internal control over financial reporting. Accounting Review, 82, 1141-1170. Doyle, J., Ge, W., & McVay, S. (2007b). Determinants of weaknesses in internal control over financial reporting. Journal of Accounting & Economics, 44, 193-223. Fitch Ratings. (2005). Sarbanes-Oxley Section 404: Fitch’s approach to evaluating management and auditor assessments of internal controls. New York. NY. Fitch (Special Report, January 19). Francis, J. R., Philbrick, D. R., & Schipper, K. (1994). Shareholder litigation and corporate disclosures. Journal of Accounting Research, 32, 137-164.

190

Journal of Accounting, Auditing & Finance

Frankel, R. M., Johnson, M. F., & Nelson, K. (2002). The relation between auditors’ fees for nonaudit services and earnings management. Accounting Review, 77(Suppl.), 71-105. Frankel, R. M., & Roychowdhury, S. (2008, December). Are all special items equally special? The predictive role of conservatism (Working paper). MIT. Retrieved from http://papers.ssrn.com/sol3/ papers.cfm?abstract_id=1001434 Givoly, D., Hayn, C., & Katz, S. P. (2010). Does public ownership of equity improve earnings quality? Accounting Review, 85, 195-225. Givoly, D., Hayn, C., & Natarajan, A. (2007). Measuring reporting conservatism. Accounting Review, 82, 65-106. Goh, B. W., & Li, D. (2011). Internal controls and conditional conservatism. Accounting Review, 86, 975-1005. Graham, J. R., Harvey, C. R., & Rajgopal, S. (2005). The economic implications of corporate financial reporting. Journal of Accounting & Economics, 40, 3-73. Greene, W. H. (2008). Econometric analysis (6th ed.). Upper Saddle River, NJ: Pearson. Iskandar-Datta, M., & Jia, Y. (2013). Valuation consequences of clawback provisions. Accounting Review, 88, 171-198. Jenkins, D. S., & Velury, U. (2008). Does auditor tenure influence the reporting of conservative earnings? Journal of Accounting and Public Policy, 27, 115-132. Jensen, M. C. (1993). The modern industrial revolution, exit and the failure of internal control system. Journal of Finance, 48, 831-880. Jiambalvo, J. (1996). Discussion of causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research, 13, 37-47. Kellog, R. L. (1984). Accounting activities, security prices and class action lawsuits. Journal of Accounting & Economics, 6, 185-204. LaFond, R., & Roychowdhury, S. (2008). Managerial ownership and accounting conservatism. Journal of Accounting Research, 46, 101-135. LaFond, R., & Watts, R. L. (2008). The information role of conservatism. Accounting Review, 83, 447-478. LaLonde, R. (1986). Evaluating the econometric evaluations of training programs with experimental data. American Economic Review, 76, 604-620. Lawrence, A., Minutti-Meza, M., & Zhang, P. (2011). Can Big 4 versus non-Big 4 differences in audit quality proxies be attributed to client characteristics? Accounting Review, 86, 259-286. Lennox, C., Francis, J. R., & Wang, Z. (2012). Selection models in accounting research. Accounting Review, 87, 589-616. Levitt, A. (1998, September 28). The numbers game. Speech delivered at New York University Center for Law and Business, New York City. Lobo, G. J., & Zhou, J. (2006). Did conservatism in financial reporting increase after the Sarbanes– Oxley Act? Initial evidence. Accounting Horizons, 20, 57-73. Moody’s Investors Service. (2004). Special comment: Section 404 reports on internal control. New York, NY: Author. Nelson, M. W. (2006). Ameliorating conflicts of interest in auditing: Effects of recent reforms on auditors and their clients. Academy of Management Review, 31, 30-42. Office of Economic Analysis of U.S. Securities and Exchange Commission. (2009). Study of the Sarbanes–Oxley Act of 2002 Section 404 internal control over financial reporting requirements. Retrieved from http://www.sec.gov/news/studies/2009/sox-404_study.pdf Penman, S. H., & Zhang, X. J. (2002). Accounting conservatism, the quality of earnings, and stock returns. Accounting Review, 77, 237-264. Raghunandan, K., & Rama, D. V. (2006). SOX Section 404 material weakness disclosures and audit fees. Auditing: A Journal of Practice & Theory, 25(1), 99-114. Ribstein, L. E. (2002). Market and regulatory responses to corporate fraud: A critique of the Sarbanes–Oxley Act of 2002. Journal of Corporation Law, 28, 1-67.

Mitra et al.

191

Roychowdhury, S., & Watts, R. L. (2007). Asymmetric timeliness of earnings, market-to-book and conservatism in financial reporting. Journal of Accounting & Economics, 44, 2-31. Verbeek. (2008). A guide to modern econometrics (3rd ed.). UK: John Wiley. Watts, R. L. (2003a). Conservatism in accounting Part I: Explanations and implications. Accounting Horizons, 17, 207-221. Watts, R. L. (2003b). Conservatism in accounting Part II: Explanations and implications. Accounting Horizons, 17, 287-301. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817-838. Zhang, J. (2008). The contracting benefits of accounting conservatism to lenders and borrowers. Journal of Accounting & Economics, 45, 27-54.

Copyright of Journal of Accounting, Auditing & Finance is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.


相关文章:
更多相关标签: