In uence of the Sarbanes-Oxley Act on Financial Reporting Quality: An Overview of EU Firms Cross-Listed in the USA


 This paper investigate whether compliance with the Sarbanes–Oxley Act of 2002 (SOX) Sect. 302 (financial reporting) and 404 (internal controls) enhances financial reporting quality (FRQ). This study focuses on EU publicly traded companies that are cross-listed in the US markets. Using a novel approach with respect to operationalization of the SOX, the empirical research integrated into this paper advances the understanding of financial reporting quality for both practitioners and policymakers. The study argues that financial reporting quality increased after SOX entered into force but, notably, we find that FRQ improves with compliance with SOX302 but not with SOX404. Examination of the latter relationship at the subsection level also reveals that compliance with certain SOX requirements is not satisfactory. We find that three out of six subsections of SOX302 are directly associated with financial reporting, while subsections (1), (5) and (6) of SOX302 are not related with FRQ, indicating that the management team, albeit not entirely, provides a reliable financial reporting systems. We also find that compliance with some SOX404’s subsections has been relatively low (i.e. subsections (1) and (3) of SOX404)), suggesting that corporations have not established and are not maintaining suitable internal control systems over financial reporting.


Introduction
The nancial scandals that occurred in the last decade led to the Sarbanes-Oxley Act (SOX) of 2002 being passed. The Act aimed to recon gure the corporate governance mosaic and restore investors' con dence in nancial markets. Indeed, SOX not only applies to a speci c concern (e.g. earnings management, nancial reporting quality etc.), but seeks to move corporate governance in speci c directions to enhance the reliability and accuracy of nancial reporting, the functionality of internal controls, information-security activities, and risk management, in the process creating new responsibilities for the audit committee and external auditing services, among others (Berninger et al., 2018;Drogalas et al., 2020).
The quality of nancial reporting may be in uenced by managerial manipulations of accounting statements, such as overstating revenues and expenditures or understating other accounting categories using a range of accounting practices (Zhou et al., 2017). Accordingly, two key mechanisms designed to prevent such manipulations and ensure high quality nancial reporting are the e ciency of corporate governance and regulatory policies (Bradbury and Mear, 2017;Melis and Carta, 2010;Pereira and Alves, 2017).
The relationship between regulatory policies and corporate governance as a whole has been investigated extensively (Chang et al., 2017). Yet several issues are still controversial and warrant further inquiry. The rst issue is the operationalization of SOX compliance. Many studies use individual corporate mechanisms (i.e. audit committee expertise, board size, board independence) to proxy for compliance with SOX provisions (Fischer et al., 2014;Kang and Kim, 2012;Khan et al., 2020;Kinney and Shepardson, 2011), but few studies have operationalized compliance with SOX as a direct measure of compliance with particular sections and provisions. The second underexplored issue is how SOX affects non-US rms cross-listed in the USA, with especially those belonging to European countries yet to be su ciently addressed. These rms are interesting to study because their cost of compliance with regulatory policies is higher as they are subject not only to SOX, but also to their domestic and/or international legislation. This is not a trivial issue as the costs of regulatory compliance are not marginal (Krishnan et al., 2008;Li, 2014).
From the corporate governance perspective, the most contentious parts of SOX concern nancial reporting (section 302) and internal controls (section 404), which continue to attract widespread attention from scholars because they determine the quality of nancial reporting and corporate effectiveness as a whole (Ettredge et al. 2018;Fischer et al. 2014). In effect, SOX302 makes senior executives focus on their responsibility to ensure a reliable nancial reporting system, while SOX404 adds further emphasis to SOX302 by requiring management to annually assess the internal controls as well as an external audit or opinion on their reliability. Investigating compliance at the cumulative (section) and constituent (subsection) levels, this paper advances our knowledge about changes in corporate governance and nancial reporting quality following the passing of SOX. The study ndings are thus of interest to both practitioners and policymakers. Another distinctive feature of the study is the idiosyncratic setting examined given that most prior studies in the area focused on US rms. EU rms are a particularly interesting object of inquiry because they face higher costs of compliance with regulatory policies since they are not only subject to SOX but also to the eighth Directive on Statutory Auditing (8thCLD) and the International Financial Reporting Standards (IFRS).
The study makes the following contributions to the literature. First, it shows whether the quality of corporate governance with respect to nancial reporting and internal controls has increased since SOX went into effect. This issue is important because regulators are constantly urging rms to eliminate weaknesses in their corporate governance. To con rm this, we run regression analyses which show that not all SOX provisions achieve the same effect in terms of ensuring nancial reporting quality. Second, our ndings hold implications for policymakers regarding whether they should focus on regulatory provisions used to increase corporate responsibility for nancial reporting, on internal controls at the composite level (sections), or on the more detailed component level (subsections). Moreover, this study reveals there is no monotonic association between SOX and FRQ, at least at the level of components. Our third contribution is by highlighting the effects of SOX on cross-listed EU rms, thereby also providing guidance for rms which intend to become listed in the US market in the future.
The next section provides the background for the study and develops our hypotheses. The research design and methodology are outlined in the second section. The third section presents the basic empirical analysis, while the nal section discusses the ndings and draws conclusions along with some policy implications of our ndings.

Literature Review And Hypothesis Development
Literature Review A review of the literature does not reveal a common concept of corporate governance, particularly because of the political, economic, ethical, environmental, cultural and religious systems that states have put in place over the years (Franks and Mayer 2017;Hylton Meier and Meier 2013;Im and Nam 2019). For example, Jackson (2010), Franks et al. (2009) emphasise that in the Anglo-American system the governance concept is shareholder-oriented and market-based, while the ownership of corporations is scattered. This governance concept shows that half the board or audit committee members are independent. Contrary to this practice, in EU corporate governance, reform is an ongoing process and only in the last decade has the EU successfully pursued a policy for advancing governance practices (e.g. board of directors, audit committees), protecting shareholder rights, increasing the ow of information, establishing CSR indicators and so on (Gamerschlag et al., 2011;Gull et al., 2020;Hopt, 2015;Hylton Meier and Meier, 2013;Maama and Mkhize, 2020). Further, the quality of corporate governance relates to: (1) better functioning of corporate mechanisms and the monitoring and maintenance of the internal controls system for nancial reporting; and (2) rms' internal procedures and adaptability in response to regulatory policy. Bajra and Cadez (2018) nd the independence of the board of directors, nancial experience, size, and the frequency of meetings play an active role in increasing the quality of governance (Cho et al., 2017). Likewise, Jeanjean and Stolowy (2008) document that the effect of corporate mechanisms depends on the implementation of corporate policies.
Anecdotal evidence supports the notion the requirements of Sections 302 and 404 have established a common practice regarding corporate governance for rms that are cross-listed and represented by any of the American Depository Receipts (ADR) programmes. Theoretically, this means that rms have already established good corporate governance management which promotes high-quality nancial reporting systems and effective internal control systems (Linck et al., 2009;Veldman and Willmott, 2016). Accordingly, current practices show that credible corporate governance requires the establishment of clear regulatory policies and accurate nancial reporting (disclosure) (Hoitash et al., 2009;Kumar et al., 2012;Venturelli et al., 2017). On the other hand, external factors like the global nancial crisis of 2008 in the USA and the debt crisis of 2010 in Europe triggered the dilemma of whether the corporate governance mosaic was being maintained and updated with relevant policies. With this rationale, the quality of nancial reporting may have been adversely affected by the nancial crises, causing the governance mosaic to shrink (Asel et al., 2011;García-Benau et al., 2013).
The primary reason for nancial reporting is to keep shareholders and other interested parties regularly and transparently informed. In the interest of comparability and objectivity, regulatory policies set standards on how nancial statements or reports should be prepared (Taipaleenmäki and Ikäheimo, 2013). Indeed, the standard methods rms use to report their nancial results determine the quality of their nancial reporting (Sunder, 2016). In this sense, nancial statements are considered well prepared when they present the true and fair nancial position and performance of a company in line with the legal framework (Bradley and Chen, 2011).
In contrast, the discretionary component identi es management choices which often interchangeably denote earnings management, nancial reporting quality, income smoothing, accounting disclosures, earnings quality etc. In contrast to the sound conceptual de nition of management choices, no direct methods exist to empirically measure these accounting categories (Givoly et al., 2010). A commonly used proxy is discretionary accruals (Badertscher et al., 2012), which signal the degree to which management in uences nancial reporting as per their disclosures (Caskey and Laux, 2017;Jorgensen and Kirshenheiter, 2012). As noted by Iatridis and Kadorinis (Iatridis and Kadorinis, 2009), fewer accounting disclosures are followed by the occurrence of accrual earnings (low quality nancial reporting), which is linked to rms' tendency to exceed nancial analysts' forecasts and to report a loss or low pro tability and high debt levels (high leverage measures). Further, rms that tend to make more accounting disclosures are less likely to provide low nancial reporting quality, respectively earnings management. In fact, earnings management proxied by discretionary accruals is driven by the personal interests of the management team (Ayers et al., 2006;Degeorge et al., 1999). This is particularly apparent when their compensation depends on the rm's targets or objectives (sales growth, market share growth, etc.). Although the evidence shows that many accounting techniques are used to report nancial statements (Iatridis and Kadorinis, 2009), nancial records (earnings or expenditures) are typically manipulated using alternative accounting methods with respect to earnings quality.
The rst method is to increase earnings in the current period on the income statement by arti cially in ating revenues or by decreasing expenses in the current period (Trompeter et al., 2013). This accounting approach makes a company's nancial condition look better than it is in order to meet nancial analysts' forecasts in their own interest. The second method for manipulating nancial statements is to decrease current earnings by de ating revenues or by in ating expenses in that period (Jo and Kim, 2007;Lo, 2008). In sum, a direct method for calculating earnings management or nancial reporting quality has not been used in the literature so far, although many proxies are employed, such as a residual of the difference between total discretionary components and nondiscretionary components (Givoly et al., 2010).

Hypothesis development
Alternative regulatory policies impose certain requirements, restrictions and guidelines on corporations, with the aim of maintaining the integrity of their nancial reporting and internal controls. Although it is di cult to conclude whether SOX alone is responsible for increasing the quality of nancial reporting, the endorsement of alternative regulatory policies has increased FRQ (Fischer et al., 2014). Using this rationale, Lobo and Zhou (2010) show that, in the post-SOX period, rms appear to be more conservative as they also demonstrate higher quality nancial reporting. They also claim this is consistent with SOX provision-related bene ts from better nancial reporting and improved internal control systems. In contrast to US companies, EU rms lack a comprehensive governance concept since they do not have a singular governance concept and generally have a concentrated ownership structure (e.g., in Germany and France) (Dai and Helfrich 2016;Franks and Mayer 2017). Speci cally, the agency problem, independent boards or audit committees, information ow disclosures (internal control de ciency [ICD]), and ownership structure-voting rights are some issues still needing to be addressed by EU rms (Böcking et al., 2015;Gong et al., 2013;. Therefore, this reveals the incomplete framework of the EU corporate mosaic (Coen and Richardson, 2009), and these cross-cutting aspects are easily identi able because the implementation of SOX provisions signi cantly affects the quality of corporate governance in EU rms.
EU corporations are of course not required to adhere to SOX, but SOX still in uences all those that operate overseas and are listed in the US market, regardless of their origin (Litvak, 2007). In addition, EU rms cross-listed in the US must meet other legal requirements, such as the 8th CLD and the International Financial Reporting Standards (IAS/IFRS). Evidence shows that IFRS and similar adoption standards improve rm performance and enhance disclosure quality.
In order to consider whether compliance with SOX sections/provisions improves nancial reporting quality, Section 302 (SOX302, with six subsections) and Section 404 (SOX404, with three subsections) are examined. Apart from looking at FRQ's relationship with the pre-and post-SOX period, we examined whether the quality of reporting was affected by the global nancial crisis (US) and the debt crisis in Europe.
The current literature does not provide much evidence on whether alternative regulatory policies can either substitute for or complement each other, although SOX is more likely to be stringent and in some cases act in both roles, as a substitute for and complement to other alternative regulatory policies. For example, while SOX explicitly tasks CEOs to strengthen shareholder con dence in nancial statements, IFRS and 8th CLD do not require such speci c involvement by CEOs or for CFOs and often only serve as a guide, without imposing legal obligations/sanctions. Con icting opinions are expressed about the application of SOX and European companies' commitment to meeting its requirements. As noted by Stanberry (2010), only 43% of EUCs believe the legislation's bene ts outweigh its costs (Litvak, 2008). Another argument in support is provided by Li (2014), who examines the short-and long-term impact of SOX on cross-listed foreign private issuers, suggesting the costs of SOX compliance signi cantly exceed its bene ts in both respects. As a result, some question whether SOX provisions should be amended to better re ect actual business operations (i.e., due to the cost of implementing it and doubts concerning whether it protects shareholders' interests) (Coates and Srinivasan, 2014;Doidge et al., 2010).
Another distinct view is that SOX does not distinguish between rms' size and nancial position. Therefore, measurement of its effect is challenging. For example, although the literature uses the auditing fee as a proxy for Section 404, this approach is not accurate when SOX provisions have been implemented (e.g., whether an external auditor attests to the internal control system). Hence, our study expands on previous studies, most of which take an indirect and one-dimensional approach (e.g., Fischer et al. 2014;Lobo and Zhou 2010), by capturing the direct effects of Sections 302 and 404 of SOX without using a proxy measure.
The principles set out in Sections 302 and 404 of SOX have been designed to promote the quality of corporate governance, indicating they do not allow discrepancies between participants in governance as the tasks and commitment of each are subject to veri cation and attestation by third-party external auditors.
SOX302 requires the senior management (CEOs and CFOs) of reporting companies to certify nancial and other information in their reports. Further, it requires that any de ciencies and material weaknesses in internal controls over nancial reporting and any agency problems (i.e., con icts of interest) be disclosed to the auditors and the audit committee. With this rationale, we therefore hypothesize that compliance with SOX302 affects the quality of nancial reporting of EU rms cross-listed in US markets.
Hypothesis 1: Compliance with SOX302 improves the quality of nancial reporting of EU rms cross-listed in US markets.
In addition, we assume that corporate efforts to establish good governance practices preceded SOX. Hence, SOX404 reinforces SOX302 by requiring management to assess and report on the effectiveness of internal controls over nancial reporting. Apart from direct control, SOX provides indirect control over nancial reporting systems, which is not covered by its counterpart (i.e., the 8th CLD). While senior management remains accountable for ful lling corporate responsibilities, SOX404 requires inter alia that an independent opinion by external auditors (attestation) con rms the management's assertions regarding the reliability of the internal control systems over nancial reporting. Such actions also provide indirect protection of shareholder rights. Despite the general belief that SOX404 has helped improve nancial reporting, some scholars disagree. For example, Singer and You (2011) claim that SOX404 helped protect investors and restored their con dence in the stock market by improving the accuracy and reliability of corporate disclosure, whereas Li (2014) suggests the costs of SOX compliance signi cantly exceed its bene ts. Regardless of the cost of compliance, SOX establishes additional standards that make the internal control environment and nancial reporting more credible and comparable with those in the European Union. While SOX404 indirectly improves the protection of investors' rights, on other hand it is expected that compliance with SOX404 further enhances the quality of nancial reporting and we therefore hypothesize that: Hypothesis 2: Compliance with SOX404 improves the quality of nancial reporting of EU rms cross-listed in US markets.
Taking a time dynamic perspective, we expect that SOX requirements have led to an improvement in corporate governance and the avoidance of earnings management (Bajra and Čadež 2018;Ettredge et al. 2006;Goh and Li 2013). Indeed, corporations have been observed establishing adequate and reliable internal controls systems over nancial reporting and, in this regard, we expect the quality of nancial reporting to increase over time, while the promulgation of SOX (2002) adds a further positive impact by advancing the whole corporate governance con guration and not simply the quality of nancial reporting. Consequently, since greater compliance with the SOX provisions supposedly increases the quality of corporate governance, we therefore posit our third hypothesis: Hypothesis 3: The quality of nancial reporting is higher in the post-SOX period than in the pre-SOX period.
In addition to our main hypothesis, we complement this study by examining the effect of the nancial and debt crisis in 2008 and 2010 on corporate governance (Reguera-Alvarado et al., 2019; Rivera et al., 2017;Trombetta and Imperatore, 2014). In particular, the nancial crisis forced many rms into bankruptcy or caused them to delist from the main nancial markets (e.g. NASDAQ and NYSE), and a contingent of them have taken appropriate measures imposed by regulators (Daugherty and Georgieva, 2011;Martinez and Serve, 2017). Moreover, a signi cant drop also occurred in initial public offerings (IPOs), leading current issuers to seek delisting from these capital markets to nd alternatives for funding (Gao et al., 2013;Rose and Solomon, 2016). Therefore, our last hypothesis is that both the global nancial and the debt crises had a negative effect on the quality of nancial reporting.
Hypothesis 4: The global nancial and the debt crises are negatively associated with nancial reporting quality.

Sample And Methodology
Sample description Since SOX went into effect, all rms listed on US markets (regardless of their origin), including EU rms, which have highly penetrated these markets, are subject to its provisions. Therefore, we sample all EU rms cross-listed in the US, proxied by ADRs for a period of 14 years (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). From the initial population (around 130), we excluded all rms without at least six years of consecutive data for the variables of interest. This procedure reduced the sample to 118 EU cross-listed rms, in part due to the declining number of EU rms operating in US markets over time, as noted in reports published by the SEC or others that are accessible (e.g., NASDAQ, NYSE, AMEX, or OTC market), listing all non-US rms on US markets.
Although SOX is mandatory for every rm listed on US markets, the sample data are homogeneous and treated as a single EU market, not as a single country. Figure 1 shows that UK-based rms dominate our sample with 36% (or 42 rms), followed by France with 17% (20 rms), and Germany with 14% (17 rms), while the remaining 33% (39 rms) originate in other EU countries.Unlike the nancial records (provided by Amadeus and Bloomberg), the data that relate to SOX302 and SOX404 and their components were collected manually from thousands of annual nancial reports and corporate governance reports (EDGAR's annual proxy statement). The sample period is from 2000 to 2013, su cient to cover the period before and after implementation of SOX, allowing a meaningful investigation of its impact on nancial reporting quality and corporate governance as a whole.

Measurement of dependent variables
As shown in prior studies, great attention has been given to whether discretionary components are a good proxy for de ning the quality of nancial reporting. Although nancial reporting quality cannot be measured directly, in this study we employ several alternative estimation procedures used extensively in earlier research (Enomoto et al., 2015;Keung and Shih, 2014), which achieve an almost identical level of discretionary accruals. In particular, we use the Modi ed Jones model (M-Jones model), widely supported as the most powerful model for capturing the discretionary component (Givoly et al., 2010)[1]. Accordingly, by using industry-year cross-sectional data for several EU rms, we calculate nancial reporting quality via the discretionary accruals. Thus, the calculation of discrete components is conducted using a two-stage procedure.
In the rst stage, we estimate total accruals (TA) as the change in current assets (Δca) minus the change in current liabilities (Δcl), minus the change in cash ow (Δcc), plus the change in interest-bearing liabilities (Δipaid), and minus depreciation and amortization (da) for rm i in year t.
In the second stage, the discretionary accruals (DA) component (in absolute value) is estimated as the difference between total accruals and nondiscretionary accruals as follows: DA it = (TA it /Tas i,t-1 )-α1(1/Tas i,t-1 )+α2(∆Rev it -∆Rec it )/Toas i,t-1 + α3PPE i,t /Tas i,t-1 +ε it (1) where DA is discretionary accrual components (FRQ), TA is total accruals for rm i in year t, Tas i,t-1 is total assets for rm i in year t-1, ΔRev it is the change in revenue for rm i between years t and t-1; ΔRec it is changes in accounts receivable for rm i between years t and t-1; and PPE it is gross property, plant and equipment for rm i in year t. In our case, the difference between the dependent variable FRQ (y i ) and the estimated systemic in uence of independent variables on FRQ (predicted value ŷ i ) represent the residual (e i ), where e i = (y i -ŷ i ) or (e i = FRQ i +u i ). e i is the model deviation

Measurement of independent variables
We used a novel method to collect data on the SOX provisions, and the dataset captures all rm activities related to SOX302 and SOX404, including external audit reports, corporate governance reports, and nancial statements for rm i in year t. This implies the disclosures made by the management team or by the corporation toward ful lling the SOX legal requirements are all obtained. The disclosure and opinion provided by external auditors on the nancial statements serves as a benchmark for whether rms comply with the SOX requirements and have established good corporate governance.
With this methodology, Corporate Responsibility for Financial Reports (SOX302) for rm i in year t was measured in terms of compliance with its six subsections (Bajra and Čadež, 2020). On this basis, a composite score was constructed such that SOX302 represents a summary score for the six subsections, with values falling within a theoretical range of 0 to 6. A value of 6 denotes high compliance with SOX302 and greater disclosures by senior management, whereas a value of 0 indicates low (or no) compliance with SOX302. In the rst stage, with respect to subsection 1, SOX302(1), a value of 1 is assigned if the signing o cers reviewed the report, and 0 otherwise. With respect to subsection 2, SOX302(2), a value of 1 is assigned if the management team disclosed that the report does not contain any materially untrue statements, and 0 otherwise. In addition, a value of 1 is assigned to SOX302(3) if the audit of nancial statements con rms the assertions of the signing o cers that nancial statements and related information fairly present the nancial condition and the results in all material respects, and 0 otherwise.
With respect to subsection 4, SOX302(4), a value of 1 is assigned if the signing o cers are responsible for internal controls and evaluated these internal controls, and 0 otherwise. Concerning subsection 5, SOX302(5), a value of 1 was assigned if a list of de ciencies in the internal controls and information on any fraud was disclosed by the management team, and 0 otherwise. Finally, with respect to subsection 6, SOX302(6), we control for whether the management team disclosed any signi cant changes in internal controls or related factors and how often their internal control procedures were revised. If such changes seem to be made within a year or in the past two years, as required by SOX302(6), a value of 1 was assigned, and 0 otherwise.
As per the above method, we assessed Management Assessment of Internal Controls (SOX404) for rm i in year t along three dimensions. Hence, a composite score was constructed, with SOX404 representing the summary score of the three subsections, with values falling within a theoretical range of 0 to 3. A value of 3 denotes effective implementation of SOX404 (full compliance), whereas a value of 0 indicates ineffective implementation of SOX404 (no compliance). Regarding subsection SOX404(1), a value of 1 was assigned if the management team had established and maintained an adequate internal control structure and procedures for nancial reporting (i.e., accounting policies, board remuneration policy, policy on auditing and non-auditing fees), and 0 otherwise. With respect to subsection SOX404(2), if an assessment of the effectiveness of the internal control structure and procedures of the issuer for nancial reporting was issued, a value of 1 was assigned, and 0 otherwise. In relation to subsection SOX404(3), a value of 1 was assigned if an attestation (by external auditors) concerning the internal control was provided, and 0 otherwise.
SOXd is a proxy for corporate responsibilities for nancial reporting and internal control systems before (P1) and after (P2) the introduction of SOX, assuming that rms' efforts to provide qualitative nancial reporting and effective internal control were visible much earlier than SOX itself. In addition, the US and EU crises are measured as a dummy variable, representing the period before and after the US nancial crisis (2008) and the European sovereign debt crisis (2010), respectively.

Research model
This study tests the effects of SOX302 and SOX404 on the quality of nancial reporting using the following integrated model: FRQ it = b 0 +b 1 SOX302 it +b 2 SOX404 it +b 3 SOXd it + b 4 USCrisis2008 it +b 5 EUCrisis2010 it + +b 6 ROA it +b 7 SIZE it +b 8 GEAR it +b 9 IAS/IFRS it +b 10 AC it +e it +ε it (2) Where: FRQ it = nancial reporting quality for rm i in year t; SOX302 it = compliance with Section 302 for rm i in year t; SOX404 it = compliance with Section 404 for rm i in year t; SOXd it = a dummy variable indicating whether the observed period is before or after the implementation of SOX (before and after 2003); USCrisis 2008 it = a dummy variable indicating whether the observed period is before or after the peak of the US crisis (2008); EUCrisis 2010 it = a dummy variable indicating whether the observed period is before or after the European sovereign debt crisis (2010); ROA it = return on assets for rm i in year t, which is measured by dividing a rm's net income by its total assets; SIZE it = natural log of total assets for rm i in year t; GEAR it = ratio of total debt to total assets for rm i in year t; IAS/IFRS it = a dummy variable denoting the rm's nancial statements were prepared using IFRS for rm i in year t; AC it = a dummy variable denoting the presence (nonexistence) of an audit committee for rm i in year t; e it = control for years xed effects (dummies year) ε it = error term.
Several control variables identi ed in prior research as important determinants of the nondiscretionary component are used as control variables (Bajra and Cadez, 2018;Capkun et al., 2016). The control variables are expected to be negatively associated with the discretionary accruals, thus positively in uencing FRQ. The only exception here is the GEAR variable, which is expected to be negatively related to FRQ.
Next, ROA capture information about management's ine ciency in using the rm's assets to generate earnings.
A rm's size plays an important role in the occurrence of discretionary accruals. Whereas Myers et al. (Myers et al., 2007) note that large rms often manage their earnings to meet analysts' expectations, Siregar and Utama (Siregar and Utama, 2008) nd evidence inconsistent with this study with respect to SIZE.
Further, we include GEAR as a control variable, which is often used as an indicative instrument for the discretionary component. Accordingly, rms with high debt are also likely to have a high degree of earnings management.
We include the IAS/IFRS variable in the model to control for whether the quality of nancial reporting has increased since these rules entry went into effect [e.g. 69]. Setting additional rules on dealing with balance-sheet items increases the possibility of avoidance from nonqualitative reporting, and therefore we expect a positive link between IFRS and FRQ.
Next, we employ the AC variable, indicating that the rms have established an audit committee function. We expect that the presence of AC as a monitoring body increases accounting quality, and hence a positive relationship with FRQ is expected (Chen and Huang, 2013).
[1] In the second stage, an extension of Model 1 is also employed by introducing the adjusted return on assets (ROA), as suggested by Kothari, Leone and Wasley (Kothari et al., 2005). However, due to the consistency and similarity of the results we report the results for the Modi ed Jones model only (Dechow et al., 1995) Empirical Results Empirically, this study cannot test the effects of SOX before it goes into effect but, to overcome possible misunderstanding, within realistic parameters we can test for nancial reporting quality during the pre-and post-SOX periods. As noted above, the rms, especially those listed on the NYSE and NASDAQ, increased their efforts to create a credible system of nancial reporting and internal control, but not necessarily after SOX had been passed.  As shown in Figure 2, we divided the overall period examined into two sub-periods (before -P1 and after -P2 SOX) and calculated the mean values for each sub-period. Figure 2 shows that senior management's responsibilities for ensuring a reliable nancial reporting system as de ned later in SOX302 increased signi cantly across the periods, from period 1 (39%) to period 2 (76%, beginning in 2003, when it became mandatory), while management responsibilities with respect to assessing the internal controls and an external audit or opinion on their reliability as subsequently de ned in SOX404 increased from period 1 (17%) to period 2 (67%, beginning in 2003, when they became mandatory). Similar developments over time can be observed for most subsections of both variables. A notable exception is the SOX404(3) variable, for which the mean value is relatively low throughout the period (around 27.1%).
In addition to descriptive statistics, Table 2 reports Pearson's correlations. Most of the correlations are statistically signi cant. The highest correlation is 0.703 between SOX302 and SOX404. Since high correlations between explanatory variables suggest multicollinearity, a collinearity diagnostic analysis was conducted. The highest recorded variance in ation factor (VIF) was 2.42 for SOX404. However, we assume that multicollinearity is not a serious threat to the validity of the estimated parameters because VIF does not exceed 10 and, in our case, is around 1.82 (Alin, 2010).
We also considered the probability distributions, which are shown through variance (spread of data from the mean), skewness (asymmetry of the distribution) and kurtosis (tailedness). As regards variance, distributions with a coe cient of variation above 1 are considered to be high variance whereas those below 1 are considered to be low variance. As a rule of thumb, the value of skewness is ±2, in contrast, the value of kurtosis is 3 at a maximum. In our case, skewness has the highest value for SOX302(3) (-2.09), but we ignored it as a problem for further treatment. As far as kurtosis is concerned, higher values appear for SOX302(2) (5.10) and SOX302(3) (5.37), although this does not represent a problem for data analysis (Bai and Ng, 2005).

Hypothesis testing and results
In this study, we employed several regressions procedures to test both the integrated model with composite items and the segregated model with components (subsections). In addition to the integrated model, we test regression analyses for four more models at a segregated level for each test variable separately and two additional models mainly related to a robustness check. The model testing used the xed regression estimator.
Hence, in line with our hypothesis, in the rst regression procedures, we examine the association between the ve test variables and FRQ. As seen in Table 3, Model (1) shows that in the presence of four test variables, only two out of four are signi cantly related to FRQ. Accordingly, Table 3 shows that compliance with SOX302 is signi cantly and positively associated with FRQ (p < 0.05), suggesting that the higher a rm's compliance with SOX302, the higher the level of its FRQ will be. Countering our results, Table 3 also shows that a rm's compliance with the requirements of SOX404 is not associated with its FRQ as expected. Further, Table 3 shows that SOXd is positively associated with FRQ, indicating that SOX has an in uence by way of strengthening nancial reporting and internal control systems in corporations. Contrary to SOXd, with respect to nancial crises, none of them, such as the US crisis of 2008 or the EU debt crisis of 2010, are related with FRQ. The signs of the estimated coe cients for the control variables are largely consistent with our expectations. The magnitude of FRQ is related to ROA, SIZE and AC, while a negative association is found with IFRS. The only is GEAR variables, which are not associated with FRQ, as expected. As for the xed effects concern, Table 3 also reports year xed effects. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 To evaluate the direct impact, we tested the integrated model (inclusive model) by including each test variable separately. Model (2) in Table 4 hence shows that compliance with SOX302 is related signi cantly with FRQ. Also, several control variables are associated signi cantly with FRQ, but for some this was not expected (i.e. IAS/IFRS). With respect to SOX404, Model (3) in Table 4 shows no signi cant association with FRQ, while the linkage of the control variables with FRQ is the same as for Model (2). The only difference is that in Model (4) AC is associated with FRQ. Further, Model (4) in Table 4 shows that the post-SOX period has a signi cant impact on FRQ. This signals that corporations have increased their responsibility to nancial reporting and internal control in general. Regarding the control variables, Model (4) in Table 4 shows that ROA, SIZE and AC have a signi cant positive association with FRQ, while IAS/IFRS has a signi cant association with FRQ, but not as expected.
Unlike Model (1), where none of the crises interferes with FRQ, for the last testing variable, Model (5) in Table 4 shows that only the US nancial crisis is related with the quality of nancial reporting, while the EU debt crisis is not related. In addition to nancial crises, we control for xed effects and, because of the homogeneity and the sample's uniqueness, xed effects do have signi cant results. As in the previous models, Table 4 shows that some test variables are positively and signi cantly related to FRQ. In addition to the rst stage regression, we continue by running the model only with the components (segregated model). As seen in Table 5, Model (7), only three of the six SOX302 components (subsections) are positively associated with FRQ at statistically signi cant levels. The remaining three coe cients, respectively compliance with SOX302(1) and SOX302(5) and compliance with SOX302(6) are not related to FRQ. This result is not underestimated with respect to the SOX404 components (subsections). As per Model (8) in Table 5, of the three SOX404 provisions SOX404(2) is positively associated with FRQ at statistically signi cant levels, while SOX404(3) is signi cantly associated with FRQ, but in the opposite direction to what was expected. Contrary to the above, only SOX404 (1) is not related to FRQ. As for the SOXd variable, in both Models (7) and (8) in Table 5, it has a signi cant link with FRQ. We also checked the importance of the US nancial crisis (2008) and the European debt crisis (2010). Our regression results in Table 5 (Model 7 and 8) do not show any signi cant impact of the crises on earlier FRQ results; therefore, the quality of the nancial reporting of cross-listed EU rms was not affected by these crises. In addition to nancial crises, we control for xed effects and, due to the homogeneity and the sample's uniqueness, xed effects do not have any signi cant results.
Another point of difference when comparing the results is the control variables. Although Models (2)-(4) in Table 4 with the composite items support the AC variable, identical results are found even in the segregated model with components in which the AC variable continues to be associated with FRQ. Another important observation is the IFRS variables, which appear to be signi cantly associated with FRQ, but again the direction of this coe cient does not change in the segregated model.
First, the results of regressing the composite model (integrated model) support the direct effect hypothesis that compliance with SOX302 enhances the quality of nancial reporting. Consequently, the ndings suggest that corporate governance responsibilities concerning nancial reporting increased after SOX went into effect.
Second, contrary to our expectations, compliance with SOX404 does not enhance nancial reporting quality. This nding may be explained by the speci c cost of complying with its subsections (Krishnan et al., 2008) and the great impact of its counterpart -probably the EU 8th CLDor corporate resistance to change in rms' governance approach imposed by alternative regulatory policies (Bajra and Čadež, 2018;Christensen et al., 2015). The evidence indicates that implementation of SOX404 is costly and perhaps even unaffordable, which is in line with our results. For example, Piotroski and Srinivasan (2008)  Third, a distinct nding of this study is that shareholders' protection has been enhanced compared to the situation prior to the SOX coming into effect. But this issue remains controversial and requires greater attention in the future because implementation of certain SOX provisions considered key elements for protecting shareholders' rights (e.g. SOX404) was found to be insu cient. The likely reason is it is costly to implement them, making rms hesitant to fully comply. Indeed, our ndings suggest that some SOX provisions should be amended and, unless no other choice is provided by policymakers, companies will need to increase their efforts to establish and maintain appropriate internal controls and demonstrate their effectiveness, while avoiding earnings management and ensuring accurate nancial reporting.
To complement our regression results, we further divided the overall period under study into two sub-periods (before and after the SOX) and calculated the mean values for each sub-period. Our ndings indicate that senior management's responsibility for ensuring a reliable nancial reporting system, as de ned later in SOX302, increased signi cantly during the period, while management responsibilities with respect to assessing internal controls and an external audit or opinion on their reliability, as subsequently de ned in SOX404, increased, yet not satisfactorily. Hence, it is not surprising that our results agree with those provided by Iliev (Iliev, 2010) which reveal delay with respect to Section SOX404 compliance, especially the provisions dealing with the auditor's attestation requirement (i.e., SOX 404 (3)). Further, he claims that SOX compliance has reduced the market value of small rms. This suggests that such a trend is most likely also being followed by EU rms crosslisted in the US markets.
Fourth, our ndings show the quality of nancial reporting is not linked to the global nancial crisis in either the USA or the European context. Moreover, this study does not nd signi cant associations between IFRSs, the presence of the audit committee, and FRQ (Capkun et al., 2016;Hwang et al., 2018;Rainsbury et al., 2009;Zeghal et al., 2012). Only the presence of the audit committee is insu cient to increase the quality of nancial reporting, unless it is based on its attributive characteristics (Bajra and Čadež, 2018). The relationship between IFRS and FRQ is also controversial, suggesting that IFRS does not specify strict criteria for a rm's accounting techniques and, as claimed by Capkun et al. (2016), IFRS provides rms with greater exibility in accounting choices due to the vague criteria that accompany these standards (Verriest et al., 2013).
Consequently, our study argues that the improvement in nancial reporting quality may be largely attributed to SOX because of the comprehensible criteria for all rms listed in the US market (e.g., signatures of nancial statements by CEOs and CFOs, then the responsibility of signing o cers for internal controls and evaluating these internal controls are directly required by SOX, but not by its counterparts) and, while SOX seems to complement other alternative policies (i.e. 8th CLD), it is most likely to also serve as a substitute.
From a broader policy perspective, our evidence indicates that implementation of SOX has positively affected the quality of EU rms' nancial reporting by pointing to the clear nancial reporting responsibility at the management level and responsibility for establishing and maintaining rms' internal controls to maintain an environment with good corporate governance. Therefore, our evidence supports the notion that SOX improves the quality of corporate governance and thereby bene ts the reliability of nancial reporting, which is of great concern to corporate stakeholders.
This study also makes a methodological contribution to the literature. Earlier studies on the measurement of SOX often relied on an indirect quantitative approach (i.e., the cost of implementation, or a binary dimension), whereas this study employs a qualitative approach by checking one by one at the micro level whether rm i at time t adheres to the SOX requirements. In addition, this study applies a novel approach with respect to the operationalization of SOX. Hence, we believe our methodology and empirical approach are useful to both practitioners and policymakers for further understanding the implications of the SOX and nancial reporting quality.
Despite this contribution, like in previous studies this study's ndings have some limitations. A particular limitation concerns measurement of the variables. As noted in the study, discretionary accruals are only a proxy for nancial reporting quality and no perfect measure of nancial reporting quality exists. A similar limitation pertains to measuring compliance with the SOX provisions. SOX consists of many provisions, of which this study considered only two. Although adding more provisions is likely to capture nancial reporting quality more exhaustively, it also increases the likelihood of measurement errors.
[1] A supportive example is the share of equity represented by the NYSE, NASDAQ and AMEX, where the shares of equity dropped from 28.8% in 2002 to 23.0% in 2009 (Woo, 2011). Further, a signi cant drop was also recorded in initial public offerings (IPOs), prompting current issuers to seek delisting from these capital markets as part of nding new alternatives (Gao et al., 2013;Rose and Solomon, 2016). From the EU perspective, it is noted the number of EU rms represented by the ADR programme has also declined since SOX was enacted. According to statistics provided by the SEC (www.sec.gov/divisions/corp n/internatl/companies.shtml), the number of EU rms represented by ADR in December 2004 was 270 (excluding Switzerland), while at the end of December 2013 it had dropped by 107, or around 60% less than in 2004.

Declarations
Con icts of interest: The authors declare no con icts of interest.